Index: Makefile.in ================================================================== --- Makefile.in +++ Makefile.in @@ -165,10 +165,11 @@ backup.lo bitvec.lo btmutex.lo btree.lo build.lo \ callback.lo complete.lo ctime.lo date.lo delete.lo \ expr.lo fault.lo fkey.lo \ fts3.lo fts3_aux.lo fts3_expr.lo fts3_hash.lo fts3_icu.lo \ fts3_porter.lo fts3_snippet.lo fts3_tokenizer.lo fts3_tokenizer1.lo \ + fts3_tokenize_vtab.lo \ fts3_unicode.lo fts3_unicode2.lo fts3_write.lo \ func.lo global.lo hash.lo \ icu.lo insert.lo journal.lo legacy.lo loadext.lo \ main.lo malloc.lo mem0.lo mem1.lo mem2.lo mem3.lo mem5.lo \ memjournal.lo \ @@ -314,10 +315,11 @@ $(TOP)/ext/fts3/fts3_porter.c \ $(TOP)/ext/fts3/fts3_snippet.c \ $(TOP)/ext/fts3/fts3_tokenizer.h \ $(TOP)/ext/fts3/fts3_tokenizer.c \ $(TOP)/ext/fts3/fts3_tokenizer1.c \ + $(TOP)/ext/fts3/fts3_tokenize_vtab.c \ $(TOP)/ext/fts3/fts3_unicode.c \ $(TOP)/ext/fts3/fts3_unicode2.c \ $(TOP)/ext/fts3/fts3_write.c SRC += \ $(TOP)/ext/icu/sqliteicu.h \ @@ -357,11 +359,10 @@ $(TOP)/src/test_config.c \ $(TOP)/src/test_demovfs.c \ $(TOP)/src/test_devsym.c \ $(TOP)/src/test_fs.c \ $(TOP)/src/test_func.c \ - $(TOP)/src/test_fuzzer.c \ $(TOP)/src/test_hexio.c \ $(TOP)/src/test_init.c \ $(TOP)/src/test_intarray.c \ $(TOP)/src/test_journal.c \ $(TOP)/src/test_malloc.c \ @@ -369,25 +370,35 @@ $(TOP)/src/test_mutex.c \ $(TOP)/src/test_onefile.c \ $(TOP)/src/test_osinst.c \ $(TOP)/src/test_pcache.c \ $(TOP)/src/test_quota.c \ - $(TOP)/src/test_regexp.c \ $(TOP)/src/test_rtree.c \ $(TOP)/src/test_schema.c \ $(TOP)/src/test_server.c \ $(TOP)/src/test_superlock.c \ $(TOP)/src/test_syscall.c \ $(TOP)/src/test_stat.c \ $(TOP)/src/test_tclvar.c \ $(TOP)/src/test_thread.c \ $(TOP)/src/test_vfs.c \ - $(TOP)/src/test_wholenumber.c \ $(TOP)/src/test_wsd.c \ $(TOP)/ext/fts3/fts3_term.c \ $(TOP)/ext/fts3/fts3_test.c +# Statically linked extensions +# +TESTSRC += \ + $(TOP)/ext/misc/amatch.c \ + $(TOP)/ext/misc/closure.c \ + $(TOP)/ext/misc/fuzzer.c \ + $(TOP)/ext/misc/ieee754.c \ + $(TOP)/ext/misc/nextchar.c \ + $(TOP)/ext/misc/regexp.c \ + $(TOP)/ext/misc/spellfix.c \ + $(TOP)/ext/misc/wholenumber.c + # Source code to the library files needed by the test fixture # TESTSRC2 = \ $(TOP)/src/attach.c \ $(TOP)/src/backup.c \ @@ -499,10 +510,15 @@ sqlite3$(TEXE): $(TOP)/src/shell.c libsqlite3.la sqlite3.h $(LTLINK) $(READLINE_FLAGS) \ -o $@ $(TOP)/src/shell.c libsqlite3.la \ $(LIBREADLINE) $(TLIBS) -rpath "$(libdir)" + +mptester$(EXE): sqlite3.c $(TOP)/mptest/mptest.c + $(LTLINK) -o $@ -I. $(TOP)/mptest/mptest.c sqlite3.c \ + $(TLIBS) -rpath "$(libdir)" + # This target creates a directory named "tsrc" and fills it with # copies of all of the C source code and header files needed to # build on the target system. Some of the C source code and header # files are automatically generated. This target takes care of @@ -517,10 +533,11 @@ mv vdbe.new tsrc/vdbe.c touch .target_source sqlite3.c: .target_source $(TOP)/tool/mksqlite3c.tcl $(TCLSH_CMD) $(TOP)/tool/mksqlite3c.tcl + cp tsrc/shell.c tsrc/sqlite3ext.h . tclsqlite3.c: sqlite3.c echo '#ifndef USE_SYSTEM_SQLITE' >tclsqlite3.c cat sqlite3.c >>tclsqlite3.c echo '#endif /* USE_SYSTEM_SQLITE */' >>tclsqlite3.c @@ -850,10 +867,13 @@ $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer.c fts3_tokenizer1.lo: $(TOP)/ext/fts3/fts3_tokenizer1.c $(HDR) $(EXTHDR) $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer1.c +fts3_tokenizer_vtab.lo: $(TOP)/ext/fts3/fts3_tokenizer_vtab.c $(HDR) $(EXTHDR) + $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer_vtab.c + fts3_unicode.lo: $(TOP)/ext/fts3/fts3_unicode.c $(HDR) $(EXTHDR) $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_unicode.c fts3_unicode2.lo: $(TOP)/ext/fts3/fts3_unicode2.c $(HDR) $(EXTHDR) $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_unicode2.c @@ -876,11 +896,12 @@ TESTFIXTURE_FLAGS += -DSQLITE_SERVER=1 -DSQLITE_PRIVATE="" -DSQLITE_CORE TESTFIXTURE_FLAGS += -DBUILD_sqlite TESTFIXTURE_SRC0 = $(TESTSRC2) libsqlite3.la TESTFIXTURE_SRC1 = sqlite3.c -TESTFIXTURE_SRC = $(TESTSRC) $(TOP)/src/tclsqlite.c $(TESTFIXTURE_SRC$(USE_AMALGAMATION)) +TESTFIXTURE_SRC = $(TESTSRC) $(TOP)/src/tclsqlite.c +TESTFIXTURE_SRC += $(TESTFIXTURE_SRC$(USE_AMALGAMATION)) testfixture$(TEXE): $(TESTFIXTURE_SRC) $(LTLINK) -DSQLITE_NO_SYNC=1 $(TEMP_STORE) $(TESTFIXTURE_FLAGS) \ -o $@ $(TESTFIXTURE_SRC) $(LIBTCL) $(TLIBS) @@ -942,12 +963,15 @@ rm -rf tsrc .target_source rm -f tclsqlite3$(TEXE) rm -f testfixture$(TEXE) test.db rm -f sqlite3.dll sqlite3.lib sqlite3.exp sqlite3.def rm -f sqlite3.c + rm -f sqlite3rc.h + rm -f shell.c sqlite3ext.h rm -f sqlite3_analyzer$(TEXE) sqlite3_analyzer.c rm -f sqlite-*-output.vsix + rm -f mptester mptester.exe distclean: clean rm -f config.log config.status libtool Makefile sqlite3.pc # Index: Makefile.msc ================================================================== --- Makefile.msc +++ Makefile.msc @@ -476,11 +476,11 @@ backup.lo bitvec.lo btmutex.lo btree.lo build.lo \ callback.lo complete.lo ctime.lo date.lo delete.lo \ expr.lo fault.lo fkey.lo \ fts3.lo fts3_aux.lo fts3_expr.lo fts3_hash.lo fts3_icu.lo \ fts3_porter.lo fts3_snippet.lo fts3_tokenizer.lo fts3_tokenizer1.lo \ - fts3_unicode.lo fts3_unicode2.lo fts3_write.lo \ + fts3_tokenize_vtab.lo fts3_unicode.lo fts3_unicode2.lo fts3_write.lo \ func.lo global.lo hash.lo \ icu.lo insert.lo journal.lo legacy.lo loadext.lo \ main.lo malloc.lo mem0.lo mem1.lo mem2.lo mem3.lo mem5.lo \ memjournal.lo \ mutex.lo mutex_noop.lo mutex_unix.lo mutex_w32.lo \ @@ -636,10 +636,11 @@ $(TOP)\ext\fts3\fts3_porter.c \ $(TOP)\ext\fts3\fts3_snippet.c \ $(TOP)\ext\fts3\fts3_tokenizer.h \ $(TOP)\ext\fts3\fts3_tokenizer.c \ $(TOP)\ext\fts3\fts3_tokenizer1.c \ + $(TOP)\ext\fts3\fts3_tokenize_vtab.c \ $(TOP)\ext\fts3\fts3_unicode.c \ $(TOP)\ext\fts3\fts3_unicode2.c \ $(TOP)\ext\fts3\fts3_write.c SRC = $(SRC) \ $(TOP)\ext\icu\sqliteicu.h \ @@ -678,11 +679,10 @@ $(TOP)\src\test_config.c \ $(TOP)\src\test_demovfs.c \ $(TOP)\src\test_devsym.c \ $(TOP)\src\test_fs.c \ $(TOP)\src\test_func.c \ - $(TOP)\src\test_fuzzer.c \ $(TOP)\src\test_hexio.c \ $(TOP)\src\test_init.c \ $(TOP)\src\test_intarray.c \ $(TOP)\src\test_journal.c \ $(TOP)\src\test_malloc.c \ @@ -690,24 +690,35 @@ $(TOP)\src\test_mutex.c \ $(TOP)\src\test_onefile.c \ $(TOP)\src\test_osinst.c \ $(TOP)\src\test_pcache.c \ $(TOP)\src\test_quota.c \ - $(TOP)\src\test_regexp.c \ $(TOP)\src\test_rtree.c \ $(TOP)\src\test_schema.c \ $(TOP)\src\test_server.c \ $(TOP)\src\test_superlock.c \ $(TOP)\src\test_syscall.c \ $(TOP)\src\test_stat.c \ $(TOP)\src\test_tclvar.c \ $(TOP)\src\test_thread.c \ $(TOP)\src\test_vfs.c \ - $(TOP)\src\test_wholenumber.c \ $(TOP)\src\test_wsd.c \ $(TOP)\ext\fts3\fts3_term.c \ $(TOP)\ext\fts3\fts3_test.c + +# Statically linked extensions +# +TESTEXT = \ + $(TOP)\ext\misc\amatch.c \ + $(TOP)\ext\misc\closure.c \ + $(TOP)\ext\misc\fuzzer.c \ + $(TOP)\ext\misc\ieee754.c \ + $(TOP)\ext\misc\nextchar.c \ + $(TOP)\ext\misc\regexp.c \ + $(TOP)\ext\misc\spellfix.c \ + $(TOP)\ext\misc\wholenumber.c + # Source code to the library files needed by the test fixture # TESTSRC2 = \ $(TOP)\src\attach.c \ @@ -746,10 +757,11 @@ parse.c \ $(TOP)\ext\fts3\fts3.c \ $(TOP)\ext\fts3\fts3_aux.c \ $(TOP)\ext\fts3\fts3_expr.c \ $(TOP)\ext\fts3\fts3_tokenizer.c \ + $(TOP)\ext\fts3\fts3_tokenize_vtab.c \ $(TOP)\ext\fts3\fts3_unicode.c \ $(TOP)\ext\fts3\fts3_unicode2.c \ $(TOP)\ext\fts3\fts3_write.c \ $(TOP)\ext\async\sqlite3async.c @@ -811,10 +823,14 @@ sqlite3.exe: $(TOP)\src\shell.c libsqlite3.lib $(LIBRESOBJS) sqlite3.h $(LTLINK) $(READLINE_FLAGS) \ $(TOP)\src\shell.c \ /link $(LTLINKOPTS) $(LTLIBPATHS) libsqlite3.lib $(LIBRESOBJS) $(LIBREADLINE) $(LTLIBS) $(TLIBS) +mptester.exe: $(TOP)\mptest\mptest.c libsqlite3.lib $(LIBRESOBJS) sqlite3.h + $(LTLINK) $(TOP)\mptest\mptest.c \ + /link $(LTLINKOPTS) $(LTLIBPATHS) libsqlite3.lib $(LIBRESOBJS) $(LIBREADLINE) $(LTLIBS) $(TLIBS) + # This target creates a directory named "tsrc" and fills it with # copies of all of the C source code and header files needed to # build on the target system. Some of the C source code and header # files are automatically generated. This target takes care of # all that automatic generation. @@ -828,10 +844,12 @@ move vdbe.new tsrc\vdbe.c echo > .target_source sqlite3.c: .target_source $(TOP)\tool\mksqlite3c.tcl $(TCLSH_CMD) $(TOP)\tool\mksqlite3c.tcl + copy tsrc\shell.c . + copy tsrc\sqlite3ext.h . sqlite3-all.c: sqlite3.c $(TOP)\tool\split-sqlite3c.tcl $(TCLSH_CMD) $(TOP)\tool\split-sqlite3c.tcl # Rule to build the amalgamation @@ -1168,10 +1186,13 @@ $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)\ext\fts3\fts3_tokenizer.c fts3_tokenizer1.lo: $(TOP)\ext\fts3\fts3_tokenizer1.c $(HDR) $(EXTHDR) $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)\ext\fts3\fts3_tokenizer1.c +fts3_tokenize_vtab.lo: $(TOP)\ext\fts3\fts3_tokenize_vtab.c $(HDR) $(EXTHDR) + $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)\ext\fts3\fts3_tokenize_vtab.c + fts3_unicode.lo: $(TOP)\ext\fts3\fts3_unicode.c $(HDR) $(EXTHDR) $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)\ext\fts3\fts3_unicode.c fts3_unicode2.lo: $(TOP)\ext\fts3\fts3_unicode2.c $(HDR) $(EXTHDR) $(LTCOMPILE) -DSQLITE_CORE -c $(TOP)\ext\fts3\fts3_unicode2.c @@ -1191,12 +1212,12 @@ # hidden when the library is built via the amalgamation). # TESTFIXTURE_FLAGS = -DTCLSH=1 -DSQLITE_TEST=1 -DSQLITE_CRASH_TEST=1 TESTFIXTURE_FLAGS = $(TESTFIXTURE_FLAGS) -DSQLITE_SERVER=1 -DSQLITE_PRIVATE="" -DSQLITE_CORE -TESTFIXTURE_SRC0 = $(TESTSRC2) libsqlite3.lib -TESTFIXTURE_SRC1 = sqlite3.c +TESTFIXTURE_SRC0 = $(TESTEXT) $(TESTSRC2) libsqlite3.lib +TESTFIXTURE_SRC1 = $(TESTEXT) sqlite3.c !IF $(USE_AMALGAMATION)==0 TESTFIXTURE_SRC = $(TESTSRC) $(TOP)\src\tclsqlite.c $(TESTFIXTURE_SRC0) !ELSE TESTFIXTURE_SRC = $(TESTSRC) $(TOP)\src\tclsqlite.c $(TESTFIXTURE_SRC1) !ENDIF @@ -1246,12 +1267,14 @@ del /Q tclsqlite3.exe tclsqlite3.exp del /Q testfixture.exe testfixture.exp test.db del /Q sqlite3.dll sqlite3.lib sqlite3.exp sqlite3.def del /Q sqlite3.c del /Q sqlite3rc.h + del /Q shell.c sqlite3ext.h del /Q sqlite3_analyzer.exe sqlite3_analyzer.exp sqlite3_analyzer.c del /Q sqlite-*-output.vsix + del /Q mptester.exe # Dynamic link library section. # dll: sqlite3.dll Index: Makefile.vxworks ================================================================== --- Makefile.vxworks +++ Makefile.vxworks @@ -660,6 +660,8 @@ rm -f *.da *.bb *.bbg gmon.out rm -rf quota2a quota2b quota2c rm -rf tsrc target_source rm -f testloadext.dll libtestloadext.so rm -f sqlite3.c fts?amal.c tclsqlite3.c + rm -f sqlite3rc.h + rm -f shell.c sqlite3ext.h rm -f $(SHPREFIX)sqlite3.$(SO) Index: VERSION ================================================================== --- VERSION +++ VERSION @@ -1,1 +1,1 @@ -3.7.16 +3.7.17 Index: configure ================================================================== --- configure +++ configure @@ -1,8 +1,8 @@ #! /bin/sh # Guess values for system-dependent variables and create Makefiles. -# Generated by GNU Autoconf 2.62 for sqlite 3.7.16. +# Generated by GNU Autoconf 2.62 for sqlite 3.7.17. # # Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001, # 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc. # This configure script is free software; the Free Software Foundation # gives unlimited permission to copy, distribute and modify it. @@ -741,12 +741,12 @@ SHELL=${CONFIG_SHELL-/bin/sh} # Identity of this package. PACKAGE_NAME='sqlite' PACKAGE_TARNAME='sqlite' -PACKAGE_VERSION='3.7.16' -PACKAGE_STRING='sqlite 3.7.16' +PACKAGE_VERSION='3.7.17' +PACKAGE_STRING='sqlite 3.7.17' PACKAGE_BUGREPORT='' # Factoring default headers for most tests. ac_includes_default="\ #include @@ -1482,11 +1482,11 @@ # if test "$ac_init_help" = "long"; then # Omit some internal or obsolete options to make the list less imposing. # This message is too long to be a string in the A/UX 3.1 sh. cat <<_ACEOF -\`configure' configures sqlite 3.7.16 to adapt to many kinds of systems. +\`configure' configures sqlite 3.7.17 to adapt to many kinds of systems. Usage: $0 [OPTION]... [VAR=VALUE]... To assign environment variables (e.g., CC, CFLAGS...), specify them as VAR=VALUE. See below for descriptions of some of the useful variables. @@ -1547,11 +1547,11 @@ _ACEOF fi if test -n "$ac_init_help"; then case $ac_init_help in - short | recursive ) echo "Configuration of sqlite 3.7.16:";; + short | recursive ) echo "Configuration of sqlite 3.7.17:";; esac cat <<\_ACEOF Optional Features: --disable-option-checking ignore unrecognized --enable/--with options @@ -1663,11 +1663,11 @@ fi test -n "$ac_init_help" && exit $ac_status if $ac_init_version; then cat <<\_ACEOF -sqlite configure 3.7.16 +sqlite configure 3.7.17 generated by GNU Autoconf 2.62 Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc. This configure script is free software; the Free Software Foundation @@ -1677,11 +1677,11 @@ fi cat >config.log <<_ACEOF This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. -It was created by sqlite $as_me 3.7.16, which was +It was created by sqlite $as_me 3.7.17, which was generated by GNU Autoconf 2.62. Invocation command line was $ $0 $@ _ACEOF @@ -14030,11 +14030,11 @@ # Save the log message, to keep $[0] and so on meaningful, and to # report actual input values of CONFIG_FILES etc. instead of their # values after options handling. ac_log=" -This file was extended by sqlite $as_me 3.7.16, which was +This file was extended by sqlite $as_me 3.7.17, which was generated by GNU Autoconf 2.62. Invocation command line was CONFIG_FILES = $CONFIG_FILES CONFIG_HEADERS = $CONFIG_HEADERS CONFIG_LINKS = $CONFIG_LINKS @@ -14083,11 +14083,11 @@ Report bugs to ." _ACEOF cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1 ac_cs_version="\\ -sqlite config.status 3.7.16 +sqlite config.status 3.7.17 configured by $0, generated by GNU Autoconf 2.62, with options \\"`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`\\" Copyright (C) 2008 Free Software Foundation, Inc. This config.status script is free software; the Free Software Foundation Index: ext/fts1/ft_hash.h ================================================================== --- ext/fts1/ft_hash.h +++ ext/fts1/ft_hash.h @@ -7,11 +7,11 @@ ** May you do good and not evil. ** May you find forgiveness for yourself and forgive others. ** May you share freely, never taking more than you give. ** ************************************************************************* -** This is the header file for the generic hash-table implemenation +** This is the header file for the generic hash-table implementation ** used in SQLite. We've modified it slightly to serve as a standalone ** hash table implementation for the full-text indexing module. ** */ #ifndef _HASH_H_ Index: ext/fts1/fts1_hash.h ================================================================== --- ext/fts1/fts1_hash.h +++ ext/fts1/fts1_hash.h @@ -7,11 +7,11 @@ ** May you do good and not evil. ** May you find forgiveness for yourself and forgive others. ** May you share freely, never taking more than you give. ** ************************************************************************* -** This is the header file for the generic hash-table implemenation +** This is the header file for the generic hash-table implementation ** used in SQLite. We've modified it slightly to serve as a standalone ** hash table implementation for the full-text indexing module. ** */ #ifndef _FTS1_HASH_H_ Index: ext/fts2/fts2.c ================================================================== --- ext/fts2/fts2.c +++ ext/fts2/fts2.c @@ -6777,11 +6777,11 @@ void sqlite3Fts2IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule); int sqlite3Fts2InitHashTable(sqlite3 *, fts2Hash *, const char *); /* -** Initialise the fts2 extension. If this extension is built as part +** Initialize the fts2 extension. If this extension is built as part ** of the sqlite library, then this function is called directly by ** SQLite. If fts2 is built as a dynamically loadable extension, this ** function is called by the sqlite3_extension_init() entry point. */ int sqlite3Fts2Init(sqlite3 *db){ @@ -6795,11 +6795,11 @@ sqlite3Fts2PorterTokenizerModule(&pPorter); #ifdef SQLITE_ENABLE_ICU sqlite3Fts2IcuTokenizerModule(&pIcu); #endif - /* Allocate and initialise the hash-table used to store tokenizers. */ + /* Allocate and initialize the hash-table used to store tokenizers. */ pHash = sqlite3_malloc(sizeof(fts2Hash)); if( !pHash ){ rc = SQLITE_NOMEM; }else{ sqlite3Fts2HashInit(pHash, FTS2_HASH_STRING, 1); Index: ext/fts2/fts2_hash.h ================================================================== --- ext/fts2/fts2_hash.h +++ ext/fts2/fts2_hash.h @@ -7,11 +7,11 @@ ** May you do good and not evil. ** May you find forgiveness for yourself and forgive others. ** May you share freely, never taking more than you give. ** ************************************************************************* -** This is the header file for the generic hash-table implemenation +** This is the header file for the generic hash-table implementation ** used in SQLite. We've modified it slightly to serve as a standalone ** hash table implementation for the full-text indexing module. ** */ #ifndef _FTS2_HASH_H_ Index: ext/fts2/fts2_tokenizer.c ================================================================== --- ext/fts2/fts2_tokenizer.c +++ ext/fts2/fts2_tokenizer.c @@ -317,11 +317,11 @@ #endif /* ** Set up SQL objects in database db used to access the contents of ** the hash table pointed to by argument pHash. The hash table must -** been initialised to use string keys, and to take a private copy +** been initialized to use string keys, and to take a private copy ** of the key when a value is inserted. i.e. by a call similar to: ** ** sqlite3Fts2HashInit(pHash, FTS2_HASH_STRING, 1); ** ** This function adds a scalar function (see header comment above Index: ext/fts2/fts2_tokenizer.h ================================================================== --- ext/fts2/fts2_tokenizer.h +++ ext/fts2/fts2_tokenizer.h @@ -68,11 +68,11 @@ ** to the strings "arg1" and "arg2". ** ** This method should return either SQLITE_OK (0), or an SQLite error ** code. If SQLITE_OK is returned, then *ppTokenizer should be set ** to point at the newly created tokenizer structure. The generic - ** sqlite3_tokenizer.pModule variable should not be initialised by + ** sqlite3_tokenizer.pModule variable should not be initialized by ** this callback. The caller will do so. */ int (*xCreate)( int argc, /* Size of argv array */ const char *const*argv, /* Tokenizer argument strings */ Index: ext/fts3/fts3.c ================================================================== --- ext/fts3/fts3.c +++ ext/fts3/fts3.c @@ -1569,11 +1569,11 @@ if( SQLITE_ROW==sqlite3_step(pCsr->pStmt) ){ return SQLITE_OK; }else{ rc = sqlite3_reset(pCsr->pStmt); if( rc==SQLITE_OK && ((Fts3Table *)pCsr->base.pVtab)->zContentTbl==0 ){ - /* If no row was found and no error has occured, then the %_content + /* If no row was found and no error has occurred, then the %_content ** table is missing a row that is present in the full-text index. ** The data structures are corrupt. */ rc = FTS_CORRUPT_VTAB; pCsr->isEof = 1; } @@ -2809,11 +2809,11 @@ sqlite3Fts3SegReaderFinish(pSegcsr); sqlite3_free(pSegcsr); } /* -** This function retreives the doclist for the specified term (or term +** This function retrieves the doclist for the specified term (or term ** prefix) from the database. */ static int fts3TermSelect( Fts3Table *p, /* Virtual table handle */ Fts3PhraseToken *pTok, /* Token to query for */ @@ -2973,18 +2973,16 @@ } pCsr->iLangid = 0; if( nVal==2 ) pCsr->iLangid = sqlite3_value_int(apVal[1]); + assert( p->base.zErrMsg==0 ); rc = sqlite3Fts3ExprParse(p->pTokenizer, pCsr->iLangid, - p->azColumn, p->bFts4, p->nColumn, iCol, zQuery, -1, &pCsr->pExpr + p->azColumn, p->bFts4, p->nColumn, iCol, zQuery, -1, &pCsr->pExpr, + &p->base.zErrMsg ); if( rc!=SQLITE_OK ){ - if( rc==SQLITE_ERROR ){ - static const char *zErr = "malformed MATCH expression: [%s]"; - p->base.zErrMsg = sqlite3_mprintf(zErr, zQuery); - } return rc; } rc = sqlite3Fts3ReadLock(p); if( rc!=SQLITE_OK ) return rc; @@ -3560,11 +3558,11 @@ #ifdef SQLITE_ENABLE_ICU void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule); #endif /* -** Initialise the fts3 extension. If this extension is built as part +** Initialize the fts3 extension. If this extension is built as part ** of the sqlite library, then this function is called directly by ** SQLite. If fts3 is built as a dynamically loadable extension, this ** function is called by the sqlite3_extension_init() entry point. */ int sqlite3Fts3Init(sqlite3 *db){ @@ -3594,11 +3592,11 @@ if( rc!=SQLITE_OK ) return rc; sqlite3Fts3SimpleTokenizerModule(&pSimple); sqlite3Fts3PorterTokenizerModule(&pPorter); - /* Allocate and initialise the hash-table used to store tokenizers. */ + /* Allocate and initialize the hash-table used to store tokenizers. */ pHash = sqlite3_malloc(sizeof(Fts3Hash)); if( !pHash ){ rc = SQLITE_NOMEM; }else{ sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1); @@ -3644,12 +3642,16 @@ if( rc==SQLITE_OK ){ rc = sqlite3_create_module_v2( db, "fts4", &fts3Module, (void *)pHash, 0 ); } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts3InitTok(db, (void *)pHash); + } return rc; } + /* An error has occurred. Delete the hash table and return the error code. */ assert( rc!=SQLITE_OK ); if( pHash ){ sqlite3Fts3HashClear(pHash); @@ -5193,11 +5195,11 @@ ** The returned value is either NULL or a pointer to a buffer containing ** a position-list indicating the occurrences of the phrase in column iCol ** of the current row. ** ** More specifically, the returned buffer contains 1 varint for each -** occurence of the phrase in the column, stored using the normal (delta+2) +** occurrence of the phrase in the column, stored using the normal (delta+2) ** compression and is terminated by either an 0x01 or 0x00 byte. For example, ** if the requested column contains "a b X c d X X" and the position-list ** for 'X' is requested, the buffer returned may contain: ** ** 0x04 0x05 0x03 0x01 or 0x04 0x05 0x03 0x00 Index: ext/fts3/fts3Int.h ================================================================== --- ext/fts3/fts3Int.h +++ ext/fts3/fts3Int.h @@ -522,11 +522,11 @@ ); void sqlite3Fts3Matchinfo(sqlite3_context *, Fts3Cursor *, const char *); /* fts3_expr.c */ int sqlite3Fts3ExprParse(sqlite3_tokenizer *, int, - char **, int, int, int, const char *, int, Fts3Expr ** + char **, int, int, int, const char *, int, Fts3Expr **, char ** ); void sqlite3Fts3ExprFree(Fts3Expr *); #ifdef SQLITE_TEST int sqlite3Fts3ExprInitTestInterface(sqlite3 *db); int sqlite3Fts3InitTerm(sqlite3 *db); @@ -547,14 +547,17 @@ Fts3Table *, Fts3MultiSegReader *, sqlite3_int64 *, char **, int *); int sqlite3Fts3EvalPhrasePoslist(Fts3Cursor *, Fts3Expr *, int iCol, char **); int sqlite3Fts3MsrOvfl(Fts3Cursor *, Fts3MultiSegReader *, int *); int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr); +/* fts3_tokenize_vtab.c */ +int sqlite3Fts3InitTok(sqlite3*, Fts3Hash *); + /* fts3_unicode2.c (functions generated by parsing unicode text files) */ #ifdef SQLITE_ENABLE_FTS4_UNICODE61 int sqlite3FtsUnicodeFold(int, int); int sqlite3FtsUnicodeIsalnum(int); int sqlite3FtsUnicodeIsdiacritic(int); #endif #endif /* !SQLITE_CORE || SQLITE_ENABLE_FTS3 */ #endif /* _FTSINT_H */ Index: ext/fts3/fts3_aux.c ================================================================== --- ext/fts3/fts3_aux.c +++ ext/fts3/fts3_aux.c @@ -68,21 +68,30 @@ int rc; /* value returned by declare_vtab() */ Fts3auxTable *p; /* Virtual table object to return */ UNUSED_PARAMETER(pUnused); - /* The user should specify a single argument - the name of an fts3 table. */ - if( argc!=4 ){ - *pzErr = sqlite3_mprintf( - "wrong number of arguments to fts4aux constructor" - ); - return SQLITE_ERROR; - } + /* The user should invoke this in one of two forms: + ** + ** CREATE VIRTUAL TABLE xxx USING fts4aux(fts4-table); + ** CREATE VIRTUAL TABLE xxx USING fts4aux(fts4-table-db, fts4-table); + */ + if( argc!=4 && argc!=5 ) goto bad_args; zDb = argv[1]; nDb = (int)strlen(zDb); - zFts3 = argv[3]; + if( argc==5 ){ + if( nDb==4 && 0==sqlite3_strnicmp("temp", zDb, 4) ){ + zDb = argv[3]; + nDb = (int)strlen(zDb); + zFts3 = argv[4]; + }else{ + goto bad_args; + } + }else{ + zFts3 = argv[3]; + } nFts3 = (int)strlen(zFts3); rc = sqlite3_declare_vtab(db, FTS3_TERMS_SCHEMA); if( rc!=SQLITE_OK ) return rc; @@ -101,10 +110,14 @@ memcpy((char *)p->pFts3Tab->zName, zFts3, nFts3); sqlite3Fts3Dequote((char *)p->pFts3Tab->zName); *ppVtab = (sqlite3_vtab *)p; return SQLITE_OK; + + bad_args: + *pzErr = sqlite3_mprintf("invalid arguments to fts4aux constructor"); + return SQLITE_ERROR; } /* ** This function does the work for both the xDisconnect and xDestroy methods. ** These tables have no persistent representation of their own, so xDisconnect Index: ext/fts3/fts3_expr.c ================================================================== --- ext/fts3/fts3_expr.c +++ ext/fts3/fts3_expr.c @@ -104,11 +104,11 @@ /* ** This function is equivalent to the standard isspace() function. ** ** The standard isspace() can be awkward to use safely, because although it -** is defined to accept an argument of type int, its behaviour when passed +** is defined to accept an argument of type int, its behavior when passed ** an integer that falls outside of the range of the unsigned char type ** is undefined (and sometimes, "undefined" means segfault). This wrapper ** is defined to accept an argument of type char, and always returns 0 for ** any values that fall outside of the range of the unsigned char type (i.e. ** negative values). @@ -638,12 +638,14 @@ rc = SQLITE_NOMEM; goto exprparse_out; } pNot->eType = FTSQUERY_NOT; pNot->pRight = p; + p->pParent = pNot; if( pNotBranch ){ pNot->pLeft = pNotBranch; + pNotBranch->pParent = pNot; } pNotBranch = pNot; p = pPrev; }else{ int eType = p->eType; @@ -727,10 +729,11 @@ Fts3Expr *pIter = pNotBranch; while( pIter->pLeft ){ pIter = pIter->pLeft; } pIter->pLeft = pRet; + pRet->pParent = pIter; pRet = pNotBranch; } } } *pnConsumed = n - nIn; @@ -742,10 +745,227 @@ pRet = 0; } *ppExpr = pRet; return rc; } + +/* +** Return SQLITE_ERROR if the maximum depth of the expression tree passed +** as the only argument is more than nMaxDepth. +*/ +static int fts3ExprCheckDepth(Fts3Expr *p, int nMaxDepth){ + int rc = SQLITE_OK; + if( p ){ + if( nMaxDepth<0 ){ + rc = SQLITE_TOOBIG; + }else{ + rc = fts3ExprCheckDepth(p->pLeft, nMaxDepth-1); + if( rc==SQLITE_OK ){ + rc = fts3ExprCheckDepth(p->pRight, nMaxDepth-1); + } + } + } + return rc; +} + +/* +** This function attempts to transform the expression tree at (*pp) to +** an equivalent but more balanced form. The tree is modified in place. +** If successful, SQLITE_OK is returned and (*pp) set to point to the +** new root expression node. +** +** nMaxDepth is the maximum allowable depth of the balanced sub-tree. +** +** Otherwise, if an error occurs, an SQLite error code is returned and +** expression (*pp) freed. +*/ +static int fts3ExprBalance(Fts3Expr **pp, int nMaxDepth){ + int rc = SQLITE_OK; /* Return code */ + Fts3Expr *pRoot = *pp; /* Initial root node */ + Fts3Expr *pFree = 0; /* List of free nodes. Linked by pParent. */ + int eType = pRoot->eType; /* Type of node in this tree */ + + if( nMaxDepth==0 ){ + rc = SQLITE_ERROR; + } + + if( rc==SQLITE_OK && (eType==FTSQUERY_AND || eType==FTSQUERY_OR) ){ + Fts3Expr **apLeaf; + apLeaf = (Fts3Expr **)sqlite3_malloc(sizeof(Fts3Expr *) * nMaxDepth); + if( 0==apLeaf ){ + rc = SQLITE_NOMEM; + }else{ + memset(apLeaf, 0, sizeof(Fts3Expr *) * nMaxDepth); + } + + if( rc==SQLITE_OK ){ + int i; + Fts3Expr *p; + + /* Set $p to point to the left-most leaf in the tree of eType nodes. */ + for(p=pRoot; p->eType==eType; p=p->pLeft){ + assert( p->pParent==0 || p->pParent->pLeft==p ); + assert( p->pLeft && p->pRight ); + } + + /* This loop runs once for each leaf in the tree of eType nodes. */ + while( 1 ){ + int iLvl; + Fts3Expr *pParent = p->pParent; /* Current parent of p */ + + assert( pParent==0 || pParent->pLeft==p ); + p->pParent = 0; + if( pParent ){ + pParent->pLeft = 0; + }else{ + pRoot = 0; + } + rc = fts3ExprBalance(&p, nMaxDepth-1); + if( rc!=SQLITE_OK ) break; + + for(iLvl=0; p && iLvlpLeft = apLeaf[iLvl]; + pFree->pRight = p; + pFree->pLeft->pParent = pFree; + pFree->pRight->pParent = pFree; + + p = pFree; + pFree = pFree->pParent; + p->pParent = 0; + apLeaf[iLvl] = 0; + } + } + if( p ){ + sqlite3Fts3ExprFree(p); + rc = SQLITE_TOOBIG; + break; + } + + /* If that was the last leaf node, break out of the loop */ + if( pParent==0 ) break; + + /* Set $p to point to the next leaf in the tree of eType nodes */ + for(p=pParent->pRight; p->eType==eType; p=p->pLeft); + + /* Remove pParent from the original tree. */ + assert( pParent->pParent==0 || pParent->pParent->pLeft==pParent ); + pParent->pRight->pParent = pParent->pParent; + if( pParent->pParent ){ + pParent->pParent->pLeft = pParent->pRight; + }else{ + assert( pParent==pRoot ); + pRoot = pParent->pRight; + } + + /* Link pParent into the free node list. It will be used as an + ** internal node of the new tree. */ + pParent->pParent = pFree; + pFree = pParent; + } + + if( rc==SQLITE_OK ){ + p = 0; + for(i=0; ipParent = 0; + }else{ + assert( pFree!=0 ); + pFree->pRight = p; + pFree->pLeft = apLeaf[i]; + pFree->pLeft->pParent = pFree; + pFree->pRight->pParent = pFree; + + p = pFree; + pFree = pFree->pParent; + p->pParent = 0; + } + } + } + pRoot = p; + }else{ + /* An error occurred. Delete the contents of the apLeaf[] array + ** and pFree list. Everything else is cleaned up by the call to + ** sqlite3Fts3ExprFree(pRoot) below. */ + Fts3Expr *pDel; + for(i=0; ipParent; + sqlite3_free(pDel); + } + } + + assert( pFree==0 ); + sqlite3_free( apLeaf ); + } + } + + if( rc!=SQLITE_OK ){ + sqlite3Fts3ExprFree(pRoot); + pRoot = 0; + } + *pp = pRoot; + return rc; +} + +/* +** This function is similar to sqlite3Fts3ExprParse(), with the following +** differences: +** +** 1. It does not do expression rebalancing. +** 2. It does not check that the expression does not exceed the +** maximum allowable depth. +** 3. Even if it fails, *ppExpr may still be set to point to an +** expression tree. It should be deleted using sqlite3Fts3ExprFree() +** in this case. +*/ +static int fts3ExprParseUnbalanced( + sqlite3_tokenizer *pTokenizer, /* Tokenizer module */ + int iLangid, /* Language id for tokenizer */ + char **azCol, /* Array of column names for fts3 table */ + int bFts4, /* True to allow FTS4-only syntax */ + int nCol, /* Number of entries in azCol[] */ + int iDefaultCol, /* Default column to query */ + const char *z, int n, /* Text of MATCH query */ + Fts3Expr **ppExpr /* OUT: Parsed query structure */ +){ + int nParsed; + int rc; + ParseContext sParse; + + memset(&sParse, 0, sizeof(ParseContext)); + sParse.pTokenizer = pTokenizer; + sParse.iLangid = iLangid; + sParse.azCol = (const char **)azCol; + sParse.nCol = nCol; + sParse.iDefaultCol = iDefaultCol; + sParse.bFts4 = bFts4; + if( z==0 ){ + *ppExpr = 0; + return SQLITE_OK; + } + if( n<0 ){ + n = (int)strlen(z); + } + rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed); + assert( rc==SQLITE_OK || *ppExpr==0 ); + + /* Check for mismatched parenthesis */ + if( rc==SQLITE_OK && sParse.nNest ){ + rc = SQLITE_ERROR; + } + + return rc; +} /* ** Parameters z and n contain a pointer to and length of a buffer containing ** an fts3 query expression, respectively. This function attempts to parse the ** query expression and create a tree of Fts3Expr structures representing the @@ -775,53 +995,78 @@ char **azCol, /* Array of column names for fts3 table */ int bFts4, /* True to allow FTS4-only syntax */ int nCol, /* Number of entries in azCol[] */ int iDefaultCol, /* Default column to query */ const char *z, int n, /* Text of MATCH query */ - Fts3Expr **ppExpr /* OUT: Parsed query structure */ + Fts3Expr **ppExpr, /* OUT: Parsed query structure */ + char **pzErr /* OUT: Error message (sqlite3_malloc) */ ){ - int nParsed; - int rc; - ParseContext sParse; + static const int MAX_EXPR_DEPTH = 12; + int rc = fts3ExprParseUnbalanced( + pTokenizer, iLangid, azCol, bFts4, nCol, iDefaultCol, z, n, ppExpr + ); + + /* Rebalance the expression. And check that its depth does not exceed + ** MAX_EXPR_DEPTH. */ + if( rc==SQLITE_OK && *ppExpr ){ + rc = fts3ExprBalance(ppExpr, MAX_EXPR_DEPTH); + if( rc==SQLITE_OK ){ + rc = fts3ExprCheckDepth(*ppExpr, MAX_EXPR_DEPTH); + } + } - memset(&sParse, 0, sizeof(ParseContext)); - sParse.pTokenizer = pTokenizer; - sParse.iLangid = iLangid; - sParse.azCol = (const char **)azCol; - sParse.nCol = nCol; - sParse.iDefaultCol = iDefaultCol; - sParse.bFts4 = bFts4; - if( z==0 ){ - *ppExpr = 0; - return SQLITE_OK; - } - if( n<0 ){ - n = (int)strlen(z); - } - rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed); - - /* Check for mismatched parenthesis */ - if( rc==SQLITE_OK && sParse.nNest ){ - rc = SQLITE_ERROR; + if( rc!=SQLITE_OK ){ sqlite3Fts3ExprFree(*ppExpr); *ppExpr = 0; + if( rc==SQLITE_TOOBIG ){ + *pzErr = sqlite3_mprintf( + "FTS expression tree is too large (maximum depth %d)", MAX_EXPR_DEPTH + ); + rc = SQLITE_ERROR; + }else if( rc==SQLITE_ERROR ){ + *pzErr = sqlite3_mprintf("malformed MATCH expression: [%s]", z); + } } return rc; } + +/* +** Free a single node of an expression tree. +*/ +static void fts3FreeExprNode(Fts3Expr *p){ + assert( p->eType==FTSQUERY_PHRASE || p->pPhrase==0 ); + sqlite3Fts3EvalPhraseCleanup(p->pPhrase); + sqlite3_free(p->aMI); + sqlite3_free(p); +} /* ** Free a parsed fts3 query expression allocated by sqlite3Fts3ExprParse(). +** +** This function would be simpler if it recursively called itself. But +** that would mean passing a sufficiently large expression to ExprParse() +** could cause a stack overflow. */ -void sqlite3Fts3ExprFree(Fts3Expr *p){ - if( p ){ - assert( p->eType==FTSQUERY_PHRASE || p->pPhrase==0 ); - sqlite3Fts3ExprFree(p->pLeft); - sqlite3Fts3ExprFree(p->pRight); - sqlite3Fts3EvalPhraseCleanup(p->pPhrase); - sqlite3_free(p->aMI); - sqlite3_free(p); +void sqlite3Fts3ExprFree(Fts3Expr *pDel){ + Fts3Expr *p; + assert( pDel==0 || pDel->pParent==0 ); + for(p=pDel; p && (p->pLeft||p->pRight); p=(p->pLeft ? p->pLeft : p->pRight)){ + assert( p->pParent==0 || p==p->pParent->pRight || p==p->pParent->pLeft ); + } + while( p ){ + Fts3Expr *pParent = p->pParent; + fts3FreeExprNode(p); + if( pParent && p==pParent->pLeft && pParent->pRight ){ + p = pParent->pRight; + while( p && (p->pLeft || p->pRight) ){ + assert( p==p->pParent->pRight || p==p->pParent->pLeft ); + p = (p->pLeft ? p->pLeft : p->pRight); + } + }else{ + p = pParent; + } } } /**************************************************************************** ***************************************************************************** @@ -869,10 +1114,13 @@ ** ** If the second argument is not NULL, then its contents are prepended to ** the returned expression text and then freed using sqlite3_free(). */ static char *exprToString(Fts3Expr *pExpr, char *zBuf){ + if( pExpr==0 ){ + return sqlite3_mprintf(""); + } switch( pExpr->eType ){ case FTSQUERY_PHRASE: { Fts3Phrase *pPhrase = pExpr->pPhrase; int i; zBuf = sqlite3_mprintf( @@ -976,14 +1224,25 @@ } for(ii=0; ii USING fts3tokenize( +** , , ... +** ); +** +** The table created has the following schema: +** +** CREATE TABLE (input, token, start, end, position) +** +** When queried, the query must include a WHERE clause of type: +** +** input = +** +** The virtual table module tokenizes this , using the FTS3 +** tokenizer specified by the arguments to the CREATE VIRTUAL TABLE +** statement and returns one row for each token in the result. With +** fields set as follows: +** +** input: Always set to a copy of +** token: A token from the input. +** start: Byte offset of the token within the input . +** end: Byte offset of the byte immediately following the end of the +** token within the input string. +** pos: Token offset of token within input. +** +*/ +#include "fts3Int.h" +#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) + +#include +#include + +typedef struct Fts3tokTable Fts3tokTable; +typedef struct Fts3tokCursor Fts3tokCursor; + +/* +** Virtual table structure. +*/ +struct Fts3tokTable { + sqlite3_vtab base; /* Base class used by SQLite core */ + const sqlite3_tokenizer_module *pMod; + sqlite3_tokenizer *pTok; +}; + +/* +** Virtual table cursor structure. +*/ +struct Fts3tokCursor { + sqlite3_vtab_cursor base; /* Base class used by SQLite core */ + char *zInput; /* Input string */ + sqlite3_tokenizer_cursor *pCsr; /* Cursor to iterate through zInput */ + int iRowid; /* Current 'rowid' value */ + const char *zToken; /* Current 'token' value */ + int nToken; /* Size of zToken in bytes */ + int iStart; /* Current 'start' value */ + int iEnd; /* Current 'end' value */ + int iPos; /* Current 'pos' value */ +}; + +/* +** Query FTS for the tokenizer implementation named zName. +*/ +static int fts3tokQueryTokenizer( + Fts3Hash *pHash, + const char *zName, + const sqlite3_tokenizer_module **pp, + char **pzErr +){ + sqlite3_tokenizer_module *p; + int nName = (int)strlen(zName); + + p = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zName, nName+1); + if( !p ){ + *pzErr = sqlite3_mprintf("unknown tokenizer: %s", zName); + return SQLITE_ERROR; + } + + *pp = p; + return SQLITE_OK; +} + +/* +** The second argument, argv[], is an array of pointers to nul-terminated +** strings. This function makes a copy of the array and strings into a +** single block of memory. It then dequotes any of the strings that appear +** to be quoted. +** +** If successful, output parameter *pazDequote is set to point at the +** array of dequoted strings and SQLITE_OK is returned. The caller is +** responsible for eventually calling sqlite3_free() to free the array +** in this case. Or, if an error occurs, an SQLite error code is returned. +** The final value of *pazDequote is undefined in this case. +*/ +static int fts3tokDequoteArray( + int argc, /* Number of elements in argv[] */ + const char * const *argv, /* Input array */ + char ***pazDequote /* Output array */ +){ + int rc = SQLITE_OK; /* Return code */ + if( argc==0 ){ + *pazDequote = 0; + }else{ + int i; + int nByte = 0; + char **azDequote; + + for(i=0; ixCreate((nDequote>1 ? nDequote-1 : 0), azArg, &pTok); + } + + if( rc==SQLITE_OK ){ + pTab = (Fts3tokTable *)sqlite3_malloc(sizeof(Fts3tokTable)); + if( pTab==0 ){ + rc = SQLITE_NOMEM; + } + } + + if( rc==SQLITE_OK ){ + memset(pTab, 0, sizeof(Fts3tokTable)); + pTab->pMod = pMod; + pTab->pTok = pTok; + *ppVtab = &pTab->base; + }else{ + if( pTok ){ + pMod->xDestroy(pTok); + } + } + + sqlite3_free(azDequote); + return rc; +} + +/* +** This function does the work for both the xDisconnect and xDestroy methods. +** These tables have no persistent representation of their own, so xDisconnect +** and xDestroy are identical operations. +*/ +static int fts3tokDisconnectMethod(sqlite3_vtab *pVtab){ + Fts3tokTable *pTab = (Fts3tokTable *)pVtab; + + pTab->pMod->xDestroy(pTab->pTok); + sqlite3_free(pTab); + return SQLITE_OK; +} + +/* +** xBestIndex - Analyze a WHERE and ORDER BY clause. +*/ +static int fts3tokBestIndexMethod( + sqlite3_vtab *pVTab, + sqlite3_index_info *pInfo +){ + int i; + UNUSED_PARAMETER(pVTab); + + for(i=0; inConstraint; i++){ + if( pInfo->aConstraint[i].usable + && pInfo->aConstraint[i].iColumn==0 + && pInfo->aConstraint[i].op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + pInfo->idxNum = 1; + pInfo->aConstraintUsage[i].argvIndex = 1; + pInfo->aConstraintUsage[i].omit = 1; + pInfo->estimatedCost = 1; + return SQLITE_OK; + } + } + + pInfo->idxNum = 0; + assert( pInfo->estimatedCost>1000000.0 ); + + return SQLITE_OK; +} + +/* +** xOpen - Open a cursor. +*/ +static int fts3tokOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){ + Fts3tokCursor *pCsr; + UNUSED_PARAMETER(pVTab); + + pCsr = (Fts3tokCursor *)sqlite3_malloc(sizeof(Fts3tokCursor)); + if( pCsr==0 ){ + return SQLITE_NOMEM; + } + memset(pCsr, 0, sizeof(Fts3tokCursor)); + + *ppCsr = (sqlite3_vtab_cursor *)pCsr; + return SQLITE_OK; +} + +/* +** Reset the tokenizer cursor passed as the only argument. As if it had +** just been returned by fts3tokOpenMethod(). +*/ +static void fts3tokResetCursor(Fts3tokCursor *pCsr){ + if( pCsr->pCsr ){ + Fts3tokTable *pTab = (Fts3tokTable *)(pCsr->base.pVtab); + pTab->pMod->xClose(pCsr->pCsr); + pCsr->pCsr = 0; + } + sqlite3_free(pCsr->zInput); + pCsr->zInput = 0; + pCsr->zToken = 0; + pCsr->nToken = 0; + pCsr->iStart = 0; + pCsr->iEnd = 0; + pCsr->iPos = 0; + pCsr->iRowid = 0; +} + +/* +** xClose - Close a cursor. +*/ +static int fts3tokCloseMethod(sqlite3_vtab_cursor *pCursor){ + Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor; + + fts3tokResetCursor(pCsr); + sqlite3_free(pCsr); + return SQLITE_OK; +} + +/* +** xNext - Advance the cursor to the next row, if any. +*/ +static int fts3tokNextMethod(sqlite3_vtab_cursor *pCursor){ + Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor; + Fts3tokTable *pTab = (Fts3tokTable *)(pCursor->pVtab); + int rc; /* Return code */ + + pCsr->iRowid++; + rc = pTab->pMod->xNext(pCsr->pCsr, + &pCsr->zToken, &pCsr->nToken, + &pCsr->iStart, &pCsr->iEnd, &pCsr->iPos + ); + + if( rc!=SQLITE_OK ){ + fts3tokResetCursor(pCsr); + if( rc==SQLITE_DONE ) rc = SQLITE_OK; + } + + return rc; +} + +/* +** xFilter - Initialize a cursor to point at the start of its data. +*/ +static int fts3tokFilterMethod( + sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ + int idxNum, /* Strategy index */ + const char *idxStr, /* Unused */ + int nVal, /* Number of elements in apVal */ + sqlite3_value **apVal /* Arguments for the indexing scheme */ +){ + int rc = SQLITE_ERROR; + Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor; + Fts3tokTable *pTab = (Fts3tokTable *)(pCursor->pVtab); + UNUSED_PARAMETER(idxStr); + UNUSED_PARAMETER(nVal); + + fts3tokResetCursor(pCsr); + if( idxNum==1 ){ + const char *zByte = (const char *)sqlite3_value_text(apVal[0]); + int nByte = sqlite3_value_bytes(apVal[0]); + pCsr->zInput = sqlite3_malloc(nByte+1); + if( pCsr->zInput==0 ){ + rc = SQLITE_NOMEM; + }else{ + memcpy(pCsr->zInput, zByte, nByte); + pCsr->zInput[nByte] = 0; + rc = pTab->pMod->xOpen(pTab->pTok, pCsr->zInput, nByte, &pCsr->pCsr); + if( rc==SQLITE_OK ){ + pCsr->pCsr->pTokenizer = pTab->pTok; + } + } + } + + if( rc!=SQLITE_OK ) return rc; + return fts3tokNextMethod(pCursor); +} + +/* +** xEof - Return true if the cursor is at EOF, or false otherwise. +*/ +static int fts3tokEofMethod(sqlite3_vtab_cursor *pCursor){ + Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor; + return (pCsr->zToken==0); +} + +/* +** xColumn - Return a column value. +*/ +static int fts3tokColumnMethod( + sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ + sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */ + int iCol /* Index of column to read value from */ +){ + Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor; + + /* CREATE TABLE x(input, token, start, end, position) */ + switch( iCol ){ + case 0: + sqlite3_result_text(pCtx, pCsr->zInput, -1, SQLITE_TRANSIENT); + break; + case 1: + sqlite3_result_text(pCtx, pCsr->zToken, pCsr->nToken, SQLITE_TRANSIENT); + break; + case 2: + sqlite3_result_int(pCtx, pCsr->iStart); + break; + case 3: + sqlite3_result_int(pCtx, pCsr->iEnd); + break; + default: + assert( iCol==4 ); + sqlite3_result_int(pCtx, pCsr->iPos); + break; + } + return SQLITE_OK; +} + +/* +** xRowid - Return the current rowid for the cursor. +*/ +static int fts3tokRowidMethod( + sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ + sqlite_int64 *pRowid /* OUT: Rowid value */ +){ + Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor; + *pRowid = (sqlite3_int64)pCsr->iRowid; + return SQLITE_OK; +} + +/* +** Register the fts3tok module with database connection db. Return SQLITE_OK +** if successful or an error code if sqlite3_create_module() fails. +*/ +int sqlite3Fts3InitTok(sqlite3 *db, Fts3Hash *pHash){ + static const sqlite3_module fts3tok_module = { + 0, /* iVersion */ + fts3tokConnectMethod, /* xCreate */ + fts3tokConnectMethod, /* xConnect */ + fts3tokBestIndexMethod, /* xBestIndex */ + fts3tokDisconnectMethod, /* xDisconnect */ + fts3tokDisconnectMethod, /* xDestroy */ + fts3tokOpenMethod, /* xOpen */ + fts3tokCloseMethod, /* xClose */ + fts3tokFilterMethod, /* xFilter */ + fts3tokNextMethod, /* xNext */ + fts3tokEofMethod, /* xEof */ + fts3tokColumnMethod, /* xColumn */ + fts3tokRowidMethod, /* xRowid */ + 0, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindFunction */ + 0, /* xRename */ + 0, /* xSavepoint */ + 0, /* xRelease */ + 0 /* xRollbackTo */ + }; + int rc; /* Return code */ + + rc = sqlite3_create_module(db, "fts3tokenize", &fts3tok_module, (void*)pHash); + return rc; +} + +#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ Index: ext/fts3/fts3_tokenizer.c ================================================================== --- ext/fts3/fts3_tokenizer.c +++ ext/fts3/fts3_tokenizer.c @@ -426,11 +426,11 @@ #endif /* ** Set up SQL objects in database db used to access the contents of ** the hash table pointed to by argument pHash. The hash table must -** been initialised to use string keys, and to take a private copy +** been initialized to use string keys, and to take a private copy ** of the key when a value is inserted. i.e. by a call similar to: ** ** sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1); ** ** This function adds a scalar function (see header comment above Index: ext/fts3/fts3_tokenizer.h ================================================================== --- ext/fts3/fts3_tokenizer.h +++ ext/fts3/fts3_tokenizer.h @@ -68,11 +68,11 @@ ** to the strings "arg1" and "arg2". ** ** This method should return either SQLITE_OK (0), or an SQLite error ** code. If SQLITE_OK is returned, then *ppTokenizer should be set ** to point at the newly created tokenizer structure. The generic - ** sqlite3_tokenizer.pModule variable should not be initialised by + ** sqlite3_tokenizer.pModule variable should not be initialized by ** this callback. The caller will do so. */ int (*xCreate)( int argc, /* Size of argv array */ const char *const*argv, /* Tokenizer argument strings */ Index: ext/fts3/fts3_unicode.c ================================================================== --- ext/fts3/fts3_unicode.c +++ ext/fts3/fts3_unicode.c @@ -123,11 +123,11 @@ ** the return value of sqlite3FtsUnicodeIsalnum() is inverted for all ** codepoints in the aiException[] array. ** ** If a standalone diacritic mark (one that sqlite3FtsUnicodeIsdiacritic() ** identifies as a diacritic) occurs in the zIn/nIn string it is ignored. -** It is not possible to change the behaviour of the tokenizer with respect +** It is not possible to change the behavior of the tokenizer with respect ** to these codepoints. */ static int unicodeAddExceptions( unicode_tokenizer *p, /* Tokenizer to add exceptions to */ int bAlnum, /* Replace Isalnum() return value with this */ Index: ext/fts3/fts3_write.c ================================================================== --- ext/fts3/fts3_write.c +++ ext/fts3/fts3_write.c @@ -1480,10 +1480,11 @@ if( ppOffsetList ){ *ppOffsetList = pReader->pOffsetList; *pnOffsetList = (int)(p - pReader->pOffsetList - 1); } + /* List may have been edited in place by fts3EvalNearTrim() */ while( papSegment, nMerge, j, xCmp); + + if( nList>0 && fts3SegReaderIsPending(apSegment[0]) ){ + rc = fts3MsrBufferData(pMsr, pList, nList+1); + if( rc!=SQLITE_OK ) return rc; + assert( (pMsr->aBuffer[nList] & 0xFE)==0x00 ); + pList = pMsr->aBuffer; + } if( pMsr->iColFilter>=0 ){ - fts3ColumnFilter(pMsr->iColFilter, &pList, &nList); + fts3ColumnFilter(pMsr->iColFilter, 1, &pList, &nList); } if( nList>0 ){ - if( fts3SegReaderIsPending(apSegment[0]) ){ - rc = fts3MsrBufferData(pMsr, pList, nList+1); - if( rc!=SQLITE_OK ) return rc; - *paPoslist = pMsr->aBuffer; - assert( (pMsr->aBuffer[nList] & 0xFE)==0x00 ); - }else{ - *paPoslist = pList; - } + *paPoslist = pList; *piDocid = iDocid; *pnPoslist = nList; break; } } @@ -2854,11 +2862,11 @@ fts3SegReaderNextDocid(p, apSegment[j], 0, 0); j++; } if( isColFilter ){ - fts3ColumnFilter(pFilter->iCol, &pList, &nList); + fts3ColumnFilter(pFilter->iCol, 0, &pList, &nList); } if( !isIgnoreEmpty || nList>0 ){ /* Calculate the 'docid' delta value to write into the merged Index: ext/icu/README.txt ================================================================== --- ext/icu/README.txt +++ ext/icu/README.txt @@ -96,11 +96,11 @@ in SQLite documentation: REGEXP This extension uses the ICU defaults for regular expression matching - behaviour. Specifically, this means that: + behavior. Specifically, this means that: * Matching is case-sensitive, * Regular expression comments are not allowed within patterns, and * The '^' and '$' characters match the beginning and end of the argument, not the beginning and end of lines within ADDED ext/misc/amatch.c Index: ext/misc/amatch.c ================================================================== --- ext/misc/amatch.c +++ ext/misc/amatch.c @@ -0,0 +1,1477 @@ +/* +** 2013-03-14 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** This file contains code for a demonstration virtual table that finds +** "approximate matches" - strings from a finite set that are nearly the +** same as a single input string. The virtual table is called "amatch". +** +** A amatch virtual table is created like this: +** +** CREATE VIRTUAL TABLE f USING approximate_match( +** vocabulary_table=, -- V +** vocabulary_word=, -- W +** vocabulary_language=, -- L +** edit_distances= +** ); +** +** When it is created, the new amatch table must be supplied with the +** the name of a table V and columns V.W and V.L such that +** +** SELECT W FROM V WHERE L=$language +** +** returns the allowed vocabulary for the match. If the "vocabulary_language" +** or L columnname is left unspecified or is an empty string, then no +** filtering of the vocabulary by language is performed. +** +** For efficiency, it is essential that the vocabulary table be indexed: +** +** CREATE vocab_index ON V(W) +** +** A separate edit-cost-table provides scoring information that defines +** what it means for one string to be "close" to another. +** +** The edit-cost-table must contain exactly four columns (more precisely, +** the statement "SELECT * FROM " must return records +** that consist of four columns). It does not matter what the columns are +** named. +** +** Each row in the edit-cost-table represents a single character +** transformation going from user input to the vocabulary. The leftmost +** column of the row (column 0) contains an integer identifier of the +** language to which the transformation rule belongs (see "MULTIPLE LANGUAGES" +** below). The second column of the row (column 1) contains the input +** character or characters - the characters of user input. The third +** column contains characters as they appear in the vocabulary table. +** And the fourth column contains the integer cost of making the +** transformation. For example: +** +** CREATE TABLE f_data(iLang, cFrom, cTo, Cost); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '', 'a', 100); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, 'b', '', 87); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, 'o', 'oe', 38); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, 'oe', 'o', 40); +** +** The first row inserted into the edit-cost-table by the SQL script +** above indicates that the cost of having an extra 'a' in the vocabulary +** table that is missing in the user input 100. (All costs are integers. +** Overall cost must not exceed 16777216.) The second INSERT statement +** creates a rule saying that the cost of having a single letter 'b' in +** user input which is missing in the vocabulary table is 87. The third +** INSERT statement mean that the cost of matching an 'o' in user input +** against an 'oe' in the vocabulary table is 38. And so forth. +** +** The following rules are special: +** +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '?', '', 97); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '', '?', 98); +** INSERT INTO f_data(iLang, cFrom, cTo, Cost) VALUES(0, '?', '?', 99); +** +** The '?' to '' rule is the cost of having any single character in the input +** that is not found in the vocabular. The '' to '?' rule is the cost of +** having a character in the vocabulary table that is missing from input. +** And the '?' to '?' rule is the cost of doing an arbitrary character +** substitution. These three generic rules apply across all languages. +** In other words, the iLang field is ignored for the generic substitution +** rules. If more than one cost is given for a generic substitution rule, +** then the lowest cost is used. +** +** Once it has been created, the amatch virtual table can be queried +** as follows: +** +** SELECT word, distance FROM f +** WHERE word MATCH 'abcdefg' +** AND distance<200; +** +** This query outputs the strings contained in the T(F) field that +** are close to "abcdefg" and in order of increasing distance. No string +** is output more than once. If there are multiple ways to transform the +** target string ("abcdefg") into a string in the vocabulary table then +** the lowest cost transform is the one that is returned. In this example, +** the search is limited to strings with a total distance of less than 200. +** +** For efficiency, it is important to put tight bounds on the distance. +** The time and memory space needed to perform this query is exponential +** in the maximum distance. A good rule of thumb is to limit the distance +** to no more than 1.5 or 2 times the maximum cost of any rule in the +** edit-cost-table. +** +** The amatch is a read-only table. Any attempt to DELETE, INSERT, or +** UPDATE on a amatch table will throw an error. +** +** It is important to put some kind of a limit on the amatch output. This +** can be either in the form of a LIMIT clause at the end of the query, +** or better, a "distance +#include +#include +#include +#include + +/* +** Forward declaration of objects used by this implementation +*/ +typedef struct amatch_vtab amatch_vtab; +typedef struct amatch_cursor amatch_cursor; +typedef struct amatch_rule amatch_rule; +typedef struct amatch_word amatch_word; +typedef struct amatch_avl amatch_avl; + + +/***************************************************************************** +** AVL Tree implementation +*/ +/* +** Objects that want to be members of the AVL tree should embedded an +** instance of this structure. +*/ +struct amatch_avl { + amatch_word *pWord; /* Points to the object being stored in the tree */ + char *zKey; /* Key. zero-terminated string. Must be unique */ + amatch_avl *pBefore; /* Other elements less than zKey */ + amatch_avl *pAfter; /* Other elements greater than zKey */ + amatch_avl *pUp; /* Parent element */ + short int height; /* Height of this node. Leaf==1 */ + short int imbalance; /* Height difference between pBefore and pAfter */ +}; + +/* Recompute the amatch_avl.height and amatch_avl.imbalance fields for p. +** Assume that the children of p have correct heights. +*/ +static void amatchAvlRecomputeHeight(amatch_avl *p){ + short int hBefore = p->pBefore ? p->pBefore->height : 0; + short int hAfter = p->pAfter ? p->pAfter->height : 0; + p->imbalance = hBefore - hAfter; /* -: pAfter higher. +: pBefore higher */ + p->height = (hBefore>hAfter ? hBefore : hAfter)+1; +} + +/* +** P B +** / \ / \ +** B Z ==> X P +** / \ / \ +** X Y Y Z +** +*/ +static amatch_avl *amatchAvlRotateBefore(amatch_avl *pP){ + amatch_avl *pB = pP->pBefore; + amatch_avl *pY = pB->pAfter; + pB->pUp = pP->pUp; + pB->pAfter = pP; + pP->pUp = pB; + pP->pBefore = pY; + if( pY ) pY->pUp = pP; + amatchAvlRecomputeHeight(pP); + amatchAvlRecomputeHeight(pB); + return pB; +} + +/* +** P A +** / \ / \ +** X A ==> P Z +** / \ / \ +** Y Z X Y +** +*/ +static amatch_avl *amatchAvlRotateAfter(amatch_avl *pP){ + amatch_avl *pA = pP->pAfter; + amatch_avl *pY = pA->pBefore; + pA->pUp = pP->pUp; + pA->pBefore = pP; + pP->pUp = pA; + pP->pAfter = pY; + if( pY ) pY->pUp = pP; + amatchAvlRecomputeHeight(pP); + amatchAvlRecomputeHeight(pA); + return pA; +} + +/* +** Return a pointer to the pBefore or pAfter pointer in the parent +** of p that points to p. Or if p is the root node, return pp. +*/ +static amatch_avl **amatchAvlFromPtr(amatch_avl *p, amatch_avl **pp){ + amatch_avl *pUp = p->pUp; + if( pUp==0 ) return pp; + if( pUp->pAfter==p ) return &pUp->pAfter; + return &pUp->pBefore; +} + +/* +** Rebalance all nodes starting with p and working up to the root. +** Return the new root. +*/ +static amatch_avl *amatchAvlBalance(amatch_avl *p){ + amatch_avl *pTop = p; + amatch_avl **pp; + while( p ){ + amatchAvlRecomputeHeight(p); + if( p->imbalance>=2 ){ + amatch_avl *pB = p->pBefore; + if( pB->imbalance<0 ) p->pBefore = amatchAvlRotateAfter(pB); + pp = amatchAvlFromPtr(p,&p); + p = *pp = amatchAvlRotateBefore(p); + }else if( p->imbalance<=(-2) ){ + amatch_avl *pA = p->pAfter; + if( pA->imbalance>0 ) p->pAfter = amatchAvlRotateBefore(pA); + pp = amatchAvlFromPtr(p,&p); + p = *pp = amatchAvlRotateAfter(p); + } + pTop = p; + p = p->pUp; + } + return pTop; +} + +/* Search the tree rooted at p for an entry with zKey. Return a pointer +** to the entry or return NULL. +*/ +static amatch_avl *amatchAvlSearch(amatch_avl *p, const char *zKey){ + int c; + while( p && (c = strcmp(zKey, p->zKey))!=0 ){ + p = (c<0) ? p->pBefore : p->pAfter; + } + return p; +} + +/* Find the first node (the one with the smallest key). +*/ +static amatch_avl *amatchAvlFirst(amatch_avl *p){ + if( p ) while( p->pBefore ) p = p->pBefore; + return p; +} + +#if 0 /* NOT USED */ +/* Return the node with the next larger key after p. +*/ +static amatch_avl *amatchAvlNext(amatch_avl *p){ + amatch_avl *pPrev = 0; + while( p && p->pAfter==pPrev ){ + pPrev = p; + p = p->pUp; + } + if( p && pPrev==0 ){ + p = amatchAvlFirst(p->pAfter); + } + return p; +} +#endif + +#if 0 /* NOT USED */ +/* Verify AVL tree integrity +*/ +static int amatchAvlIntegrity(amatch_avl *pHead){ + amatch_avl *p; + if( pHead==0 ) return 1; + if( (p = pHead->pBefore)!=0 ){ + assert( p->pUp==pHead ); + assert( amatchAvlIntegrity(p) ); + assert( strcmp(p->zKey, pHead->zKey)<0 ); + while( p->pAfter ) p = p->pAfter; + assert( strcmp(p->zKey, pHead->zKey)<0 ); + } + if( (p = pHead->pAfter)!=0 ){ + assert( p->pUp==pHead ); + assert( amatchAvlIntegrity(p) ); + assert( strcmp(p->zKey, pHead->zKey)>0 ); + p = amatchAvlFirst(p); + assert( strcmp(p->zKey, pHead->zKey)>0 ); + } + return 1; +} +static int amatchAvlIntegrity2(amatch_avl *pHead){ + amatch_avl *p, *pNext; + for(p=amatchAvlFirst(pHead); p; p=pNext){ + pNext = amatchAvlNext(p); + if( pNext==0 ) break; + assert( strcmp(p->zKey, pNext->zKey)<0 ); + } + return 1; +} +#endif + +/* Insert a new node pNew. Return NULL on success. If the key is not +** unique, then do not perform the insert but instead leave pNew unchanged +** and return a pointer to an existing node with the same key. +*/ +static amatch_avl *amatchAvlInsert(amatch_avl **ppHead, amatch_avl *pNew){ + int c; + amatch_avl *p = *ppHead; + if( p==0 ){ + p = pNew; + pNew->pUp = 0; + }else{ + while( p ){ + c = strcmp(pNew->zKey, p->zKey); + if( c<0 ){ + if( p->pBefore ){ + p = p->pBefore; + }else{ + p->pBefore = pNew; + pNew->pUp = p; + break; + } + }else if( c>0 ){ + if( p->pAfter ){ + p = p->pAfter; + }else{ + p->pAfter = pNew; + pNew->pUp = p; + break; + } + }else{ + return p; + } + } + } + pNew->pBefore = 0; + pNew->pAfter = 0; + pNew->height = 1; + pNew->imbalance = 0; + *ppHead = amatchAvlBalance(p); + /* assert( amatchAvlIntegrity(*ppHead) ); */ + /* assert( amatchAvlIntegrity2(*ppHead) ); */ + return 0; +} + +/* Remove node pOld from the tree. pOld must be an element of the tree or +** the AVL tree will become corrupt. +*/ +static void amatchAvlRemove(amatch_avl **ppHead, amatch_avl *pOld){ + amatch_avl **ppParent; + amatch_avl *pBalance; + /* assert( amatchAvlSearch(*ppHead, pOld->zKey)==pOld ); */ + ppParent = amatchAvlFromPtr(pOld, ppHead); + if( pOld->pBefore==0 && pOld->pAfter==0 ){ + *ppParent = 0; + pBalance = pOld->pUp; + }else if( pOld->pBefore && pOld->pAfter ){ + amatch_avl *pX, *pY; + pX = amatchAvlFirst(pOld->pAfter); + *amatchAvlFromPtr(pX, 0) = pX->pAfter; + if( pX->pAfter ) pX->pAfter->pUp = pX->pUp; + pBalance = pX->pUp; + pX->pAfter = pOld->pAfter; + if( pX->pAfter ){ + pX->pAfter->pUp = pX; + }else{ + assert( pBalance==pOld ); + pBalance = pX; + } + pX->pBefore = pY = pOld->pBefore; + if( pY ) pY->pUp = pX; + pX->pUp = pOld->pUp; + *ppParent = pX; + }else if( pOld->pBefore==0 ){ + *ppParent = pBalance = pOld->pAfter; + pBalance->pUp = pOld->pUp; + }else if( pOld->pAfter==0 ){ + *ppParent = pBalance = pOld->pBefore; + pBalance->pUp = pOld->pUp; + } + *ppHead = amatchAvlBalance(pBalance); + pOld->pUp = 0; + pOld->pBefore = 0; + pOld->pAfter = 0; + /* assert( amatchAvlIntegrity(*ppHead) ); */ + /* assert( amatchAvlIntegrity2(*ppHead) ); */ +} +/* +** End of the AVL Tree implementation +******************************************************************************/ + + +/* +** Various types. +** +** amatch_cost is the "cost" of an edit operation. +** +** amatch_len is the length of a matching string. +** +** amatch_langid is an ruleset identifier. +*/ +typedef int amatch_cost; +typedef signed char amatch_len; +typedef int amatch_langid; + +/* +** Limits +*/ +#define AMATCH_MX_LENGTH 50 /* Maximum length of a rule string */ +#define AMATCH_MX_LANGID 2147483647 /* Maximum rule ID */ +#define AMATCH_MX_COST 1000 /* Maximum single-rule cost */ + +/* +** A match or partial match +*/ +struct amatch_word { + amatch_word *pNext; /* Next on a list of all amatch_words */ + amatch_avl sCost; /* Linkage of this node into the cost tree */ + amatch_avl sWord; /* Linkage of this node into the word tree */ + amatch_cost rCost; /* Cost of the match so far */ + int iSeq; /* Sequence number */ + char zCost[10]; /* Cost key (text rendering of rCost) */ + short int nMatch; /* Input characters matched */ + char zWord[4]; /* Text of the word. Extra space appended as needed */ +}; + +/* +** Each transformation rule is stored as an instance of this object. +** All rules are kept on a linked list sorted by rCost. +*/ +struct amatch_rule { + amatch_rule *pNext; /* Next rule in order of increasing rCost */ + char *zFrom; /* Transform from (a string from user input) */ + amatch_cost rCost; /* Cost of this transformation */ + amatch_langid iLang; /* The langauge to which this rule belongs */ + amatch_len nFrom, nTo; /* Length of the zFrom and zTo strings */ + char zTo[4]; /* Tranform to V.W value (extra space appended) */ +}; + +/* +** A amatch virtual-table object +*/ +struct amatch_vtab { + sqlite3_vtab base; /* Base class - must be first */ + char *zClassName; /* Name of this class. Default: "amatch" */ + char *zDb; /* Name of database. (ex: "main") */ + char *zSelf; /* Name of this virtual table */ + char *zCostTab; /* Name of edit-cost-table */ + char *zVocabTab; /* Name of vocabulary table */ + char *zVocabWord; /* Name of vocabulary table word column */ + char *zVocabLang; /* Name of vocabulary table language column */ + amatch_rule *pRule; /* All active rules in this amatch */ + amatch_cost rIns; /* Generic insertion cost '' -> ? */ + amatch_cost rDel; /* Generic deletion cost ? -> '' */ + amatch_cost rSub; /* Generic substitution cost ? -> ? */ + sqlite3 *db; /* The database connection */ + sqlite3_stmt *pVCheck; /* Query to check zVocabTab */ + int nCursor; /* Number of active cursors */ +}; + +/* A amatch cursor object */ +struct amatch_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + sqlite3_int64 iRowid; /* The rowid of the current word */ + amatch_langid iLang; /* Use this language ID */ + amatch_cost rLimit; /* Maximum cost of any term */ + int nBuf; /* Space allocated for zBuf */ + int oomErr; /* True following an OOM error */ + int nWord; /* Number of amatch_word objects */ + char *zBuf; /* Temp-use buffer space */ + char *zInput; /* Input word to match against */ + amatch_vtab *pVtab; /* The virtual table this cursor belongs to */ + amatch_word *pAllWords; /* List of all amatch_word objects */ + amatch_word *pCurrent; /* Most recent solution */ + amatch_avl *pCost; /* amatch_word objects keyed by iCost */ + amatch_avl *pWord; /* amatch_word objects keyed by zWord */ +}; + +/* +** The two input rule lists are both sorted in order of increasing +** cost. Merge them together into a single list, sorted by cost, and +** return a pointer to the head of that list. +*/ +static amatch_rule *amatchMergeRules(amatch_rule *pA, amatch_rule *pB){ + amatch_rule head; + amatch_rule *pTail; + + pTail = &head; + while( pA && pB ){ + if( pA->rCost<=pB->rCost ){ + pTail->pNext = pA; + pTail = pA; + pA = pA->pNext; + }else{ + pTail->pNext = pB; + pTail = pB; + pB = pB->pNext; + } + } + if( pA==0 ){ + pTail->pNext = pB; + }else{ + pTail->pNext = pA; + } + return head.pNext; +} + +/* +** Statement pStmt currently points to a row in the amatch data table. This +** function allocates and populates a amatch_rule structure according to +** the content of the row. +** +** If successful, *ppRule is set to point to the new object and SQLITE_OK +** is returned. Otherwise, *ppRule is zeroed, *pzErr may be set to point +** to an error message and an SQLite error code returned. +*/ +static int amatchLoadOneRule( + amatch_vtab *p, /* Fuzzer virtual table handle */ + sqlite3_stmt *pStmt, /* Base rule on statements current row */ + amatch_rule **ppRule, /* OUT: New rule object */ + char **pzErr /* OUT: Error message */ +){ + sqlite3_int64 iLang = sqlite3_column_int64(pStmt, 0); + const char *zFrom = (const char *)sqlite3_column_text(pStmt, 1); + const char *zTo = (const char *)sqlite3_column_text(pStmt, 2); + amatch_cost rCost = sqlite3_column_int(pStmt, 3); + + int rc = SQLITE_OK; /* Return code */ + int nFrom; /* Size of string zFrom, in bytes */ + int nTo; /* Size of string zTo, in bytes */ + amatch_rule *pRule = 0; /* New rule object to return */ + + if( zFrom==0 ) zFrom = ""; + if( zTo==0 ) zTo = ""; + nFrom = (int)strlen(zFrom); + nTo = (int)strlen(zTo); + + /* Silently ignore null transformations */ + if( strcmp(zFrom, zTo)==0 ){ + if( zFrom[0]=='?' && zFrom[1]==0 ){ + if( p->rSub==0 || p->rSub>rCost ) p->rSub = rCost; + } + *ppRule = 0; + return SQLITE_OK; + } + + if( rCost<=0 || rCost>AMATCH_MX_COST ){ + *pzErr = sqlite3_mprintf("%s: cost must be between 1 and %d", + p->zClassName, AMATCH_MX_COST + ); + rc = SQLITE_ERROR; + }else + if( nFrom>AMATCH_MX_LENGTH || nTo>AMATCH_MX_LENGTH ){ + *pzErr = sqlite3_mprintf("%s: maximum string length is %d", + p->zClassName, AMATCH_MX_LENGTH + ); + rc = SQLITE_ERROR; + }else + if( iLang<0 || iLang>AMATCH_MX_LANGID ){ + *pzErr = sqlite3_mprintf("%s: iLang must be between 0 and %d", + p->zClassName, AMATCH_MX_LANGID + ); + rc = SQLITE_ERROR; + }else + if( strcmp(zFrom,"")==0 && strcmp(zTo,"?")==0 ){ + if( p->rIns==0 || p->rIns>rCost ) p->rIns = rCost; + }else + if( strcmp(zFrom,"?")==0 && strcmp(zTo,"")==0 ){ + if( p->rDel==0 || p->rDel>rCost ) p->rDel = rCost; + }else + { + pRule = sqlite3_malloc( sizeof(*pRule) + nFrom + nTo ); + if( pRule==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pRule, 0, sizeof(*pRule)); + pRule->zFrom = &pRule->zTo[nTo+1]; + pRule->nFrom = nFrom; + memcpy(pRule->zFrom, zFrom, nFrom+1); + memcpy(pRule->zTo, zTo, nTo+1); + pRule->nTo = nTo; + pRule->rCost = rCost; + pRule->iLang = (int)iLang; + } + } + + *ppRule = pRule; + return rc; +} + +/* +** Free all the content in the edit-cost-table +*/ +static void amatchFreeRules(amatch_vtab *p){ + while( p->pRule ){ + amatch_rule *pRule = p->pRule; + p->pRule = pRule->pNext; + sqlite3_free(pRule); + } + p->pRule = 0; +} + +/* +** Load the content of the amatch data table into memory. +*/ +static int amatchLoadRules( + sqlite3 *db, /* Database handle */ + amatch_vtab *p, /* Virtual amatch table to configure */ + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; /* Return code */ + char *zSql; /* SELECT used to read from rules table */ + amatch_rule *pHead = 0; + + zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", p->zDb, p->zCostTab); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + int rc2; /* finalize() return code */ + sqlite3_stmt *pStmt = 0; + rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); + if( rc!=SQLITE_OK ){ + *pzErr = sqlite3_mprintf("%s: %s", p->zClassName, sqlite3_errmsg(db)); + }else if( sqlite3_column_count(pStmt)!=4 ){ + *pzErr = sqlite3_mprintf("%s: %s has %d columns, expected 4", + p->zClassName, p->zCostTab, sqlite3_column_count(pStmt) + ); + rc = SQLITE_ERROR; + }else{ + while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ + amatch_rule *pRule = 0; + rc = amatchLoadOneRule(p, pStmt, &pRule, pzErr); + if( pRule ){ + pRule->pNext = pHead; + pHead = pRule; + } + } + } + rc2 = sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ) rc = rc2; + } + sqlite3_free(zSql); + + /* All rules are now in a singly linked list starting at pHead. This + ** block sorts them by cost and then sets amatch_vtab.pRule to point to + ** point to the head of the sorted list. + */ + if( rc==SQLITE_OK ){ + unsigned int i; + amatch_rule *pX; + amatch_rule *a[15]; + for(i=0; ipNext; + pX->pNext = 0; + for(i=0; a[i] && ipRule = amatchMergeRules(p->pRule, pX); + }else{ + /* An error has occurred. Setting p->pRule to point to the head of the + ** allocated list ensures that the list will be cleaned up in this case. + */ + assert( p->pRule==0 ); + p->pRule = pHead; + } + + return rc; +} + +/* +** This function converts an SQL quoted string into an unquoted string +** and returns a pointer to a buffer allocated using sqlite3_malloc() +** containing the result. The caller should eventually free this buffer +** using sqlite3_free. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static char *amatchDequote(const char *zIn){ + int nIn; /* Size of input string, in bytes */ + char *zOut; /* Output (dequoted) string */ + + nIn = (int)strlen(zIn); + zOut = sqlite3_malloc(nIn+1); + if( zOut ){ + char q = zIn[0]; /* Quote character (if any ) */ + + if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ + memcpy(zOut, zIn, nIn+1); + }else{ + int iOut = 0; /* Index of next byte to write to output */ + int iIn; /* Index of next byte to read from input */ + + if( q=='[' ) q = ']'; + for(iIn=1; iInpVCheck ){ + sqlite3_finalize(p->pVCheck); + p->pVCheck = 0; + } +} + +/* +** Deallocate an amatch_vtab object +*/ +static void amatchFree(amatch_vtab *p){ + if( p ){ + amatchFreeRules(p); + amatchVCheckClear(p); + sqlite3_free(p->zClassName); + sqlite3_free(p->zDb); + sqlite3_free(p->zCostTab); + sqlite3_free(p->zVocabTab); + sqlite3_free(p->zVocabWord); + sqlite3_free(p->zVocabLang); + memset(p, 0, sizeof(*p)); + sqlite3_free(p); + } +} + +/* +** xDisconnect/xDestroy method for the amatch module. +*/ +static int amatchDisconnect(sqlite3_vtab *pVtab){ + amatch_vtab *p = (amatch_vtab*)pVtab; + assert( p->nCursor==0 ); + amatchFree(p); + return SQLITE_OK; +} + +/* +** Check to see if the argument is of the form: +** +** KEY = VALUE +** +** If it is, return a pointer to the first character of VALUE. +** If not, return NULL. Spaces around the = are ignored. +*/ +static const char *amatchValueOfKey(const char *zKey, const char *zStr){ + int nKey = (int)strlen(zKey); + int nStr = (int)strlen(zStr); + int i; + if( nStr module name ("approximate_match") +** argv[1] -> database name +** argv[2] -> table name +** argv[3...] -> arguments +*/ +static int amatchConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + int rc = SQLITE_OK; /* Return code */ + amatch_vtab *pNew = 0; /* New virtual table */ + const char *zModule = argv[0]; + const char *zDb = argv[1]; + const char *zVal; + int i; + + (void)pAux; + *ppVtab = 0; + pNew = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + rc = SQLITE_NOMEM; + memset(pNew, 0, sizeof(*pNew)); + pNew->db = db; + pNew->zClassName = sqlite3_mprintf("%s", zModule); + if( pNew->zClassName==0 ) goto amatchConnectError; + pNew->zDb = sqlite3_mprintf("%s", zDb); + if( pNew->zDb==0 ) goto amatchConnectError; + pNew->zSelf = sqlite3_mprintf("%s", argv[2]); + if( pNew->zSelf==0 ) goto amatchConnectError; + for(i=3; izVocabTab); + pNew->zVocabTab = amatchDequote(zVal); + if( pNew->zVocabTab==0 ) goto amatchConnectError; + continue; + } + zVal = amatchValueOfKey("vocabulary_word", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zVocabWord); + pNew->zVocabWord = amatchDequote(zVal); + if( pNew->zVocabWord==0 ) goto amatchConnectError; + continue; + } + zVal = amatchValueOfKey("vocabulary_language", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zVocabLang); + pNew->zVocabLang = amatchDequote(zVal); + if( pNew->zVocabLang==0 ) goto amatchConnectError; + continue; + } + zVal = amatchValueOfKey("edit_distances", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zCostTab); + pNew->zCostTab = amatchDequote(zVal); + if( pNew->zCostTab==0 ) goto amatchConnectError; + continue; + } + *pzErr = sqlite3_mprintf("unrecognized argument: [%s]\n", argv[i]); + amatchFree(pNew); + *ppVtab = 0; + return SQLITE_ERROR; + } + rc = SQLITE_OK; + if( pNew->zCostTab==0 ){ + *pzErr = sqlite3_mprintf("no edit_distances table specified"); + rc = SQLITE_ERROR; + }else{ + rc = amatchLoadRules(db, pNew, pzErr); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_declare_vtab(db, + "CREATE TABLE x(word,distance,language," + "command HIDDEN,nword HIDDEN)" + ); +#define AMATCH_COL_WORD 0 +#define AMATCH_COL_DISTANCE 1 +#define AMATCH_COL_LANGUAGE 2 +#define AMATCH_COL_COMMAND 3 +#define AMATCH_COL_NWORD 4 + } + if( rc!=SQLITE_OK ){ + amatchFree(pNew); + } + *ppVtab = &pNew->base; + return rc; + +amatchConnectError: + amatchFree(pNew); + return rc; +} + +/* +** Open a new amatch cursor. +*/ +static int amatchOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + amatch_vtab *p = (amatch_vtab*)pVTab; + amatch_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVtab = p; + *ppCursor = &pCur->base; + p->nCursor++; + return SQLITE_OK; +} + +/* +** Free up all the memory allocated by a cursor. Set it rLimit to 0 +** to indicate that it is at EOF. +*/ +static void amatchClearCursor(amatch_cursor *pCur){ + amatch_word *pWord, *pNextWord; + for(pWord=pCur->pAllWords; pWord; pWord=pNextWord){ + pNextWord = pWord->pNext; + sqlite3_free(pWord); + } + pCur->pAllWords = 0; + sqlite3_free(pCur->zInput); + pCur->zInput = 0; + pCur->pCost = 0; + pCur->pWord = 0; + pCur->pCurrent = 0; + pCur->rLimit = 1000000; + pCur->iLang = 0; + pCur->nWord = 0; +} + +/* +** Close a amatch cursor. +*/ +static int amatchClose(sqlite3_vtab_cursor *cur){ + amatch_cursor *pCur = (amatch_cursor *)cur; + amatchClearCursor(pCur); + pCur->pVtab->nCursor--; + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Render a 24-bit unsigned integer as a 4-byte base-64 number. +*/ +static void amatchEncodeInt(int x, char *z){ + static const char a[] = + "0123456789" + "ABCDEFGHIJ" + "KLMNOPQRST" + "UVWXYZ^abc" + "defghijklm" + "nopqrstuvw" + "xyz~"; + z[0] = a[(x>>18)&0x3f]; + z[1] = a[(x>>12)&0x3f]; + z[2] = a[(x>>6)&0x3f]; + z[3] = a[x&0x3f]; +} + +/* +** Write the zCost[] field for a amatch_word object +*/ +static void amatchWriteCost(amatch_word *pWord){ + amatchEncodeInt(pWord->rCost, pWord->zCost); + amatchEncodeInt(pWord->iSeq, pWord->zCost+4); + pWord->zCost[8] = 0; +} + +/* +** Add a new amatch_word object to the queue. +** +** If a prior amatch_word object with the same zWord, and nMatch +** already exists, update its rCost (if the new rCost is less) but +** otherwise leave it unchanged. Do not add a duplicate. +** +** Do nothing if the cost exceeds threshold. +*/ +static void amatchAddWord( + amatch_cursor *pCur, + amatch_cost rCost, + int nMatch, + const char *zWordBase, + const char *zWordTail +){ + amatch_word *pWord; + amatch_avl *pNode; + amatch_avl *pOther; + int nBase, nTail; + char zBuf[4]; + + if( rCost>pCur->rLimit ){ + return; + } + nBase = (int)strlen(zWordBase); + nTail = (int)strlen(zWordTail); + if( nBase+nTail+3>pCur->nBuf ){ + pCur->nBuf = nBase+nTail+100; + pCur->zBuf = sqlite3_realloc(pCur->zBuf, pCur->nBuf); + if( pCur->zBuf==0 ){ + pCur->nBuf = 0; + return; + } + } + amatchEncodeInt(nMatch, zBuf); + memcpy(pCur->zBuf, zBuf+2, 2); + memcpy(pCur->zBuf+2, zWordBase, nBase); + memcpy(pCur->zBuf+2+nBase, zWordTail, nTail+1); + pNode = amatchAvlSearch(pCur->pWord, pCur->zBuf); + if( pNode ){ + pWord = pNode->pWord; + if( pWord->rCost>rCost ){ +#ifdef AMATCH_TRACE_1 + printf("UPDATE [%s][%.*s^%s] %d (\"%s\" \"%s\")\n", + pWord->zWord+2, pWord->nMatch, pCur->zInput, pCur->zInput, + pWord->rCost, pWord->zWord, pWord->zCost); +#endif + amatchAvlRemove(&pCur->pCost, &pWord->sCost); + pWord->rCost = rCost; + amatchWriteCost(pWord); +#ifdef AMATCH_TRACE_1 + printf(" ---> %d (\"%s\" \"%s\")\n", + pWord->rCost, pWord->zWord, pWord->zCost); +#endif + pOther = amatchAvlInsert(&pCur->pCost, &pWord->sCost); + assert( pOther==0 ); (void)pOther; + } + return; + } + pWord = sqlite3_malloc( sizeof(*pWord) + nBase + nTail - 1 ); + if( pWord==0 ) return; + memset(pWord, 0, sizeof(*pWord)); + pWord->rCost = rCost; + pWord->iSeq = pCur->nWord++; + amatchWriteCost(pWord); + pWord->nMatch = nMatch; + pWord->pNext = pCur->pAllWords; + pCur->pAllWords = pWord; + pWord->sCost.zKey = pWord->zCost; + pWord->sCost.pWord = pWord; + pOther = amatchAvlInsert(&pCur->pCost, &pWord->sCost); + assert( pOther==0 ); (void)pOther; + pWord->sWord.zKey = pWord->zWord; + pWord->sWord.pWord = pWord; + strcpy(pWord->zWord, pCur->zBuf); + pOther = amatchAvlInsert(&pCur->pWord, &pWord->sWord); + assert( pOther==0 ); (void)pOther; +#ifdef AMATCH_TRACE_1 + printf("INSERT [%s][%.*s^%s] %d (\"%s\" \"%s\")\n", pWord->zWord+2, + pWord->nMatch, pCur->zInput, pCur->zInput+pWord->nMatch, rCost, + pWord->zWord, pWord->zCost); +#endif +} + +/* +** Advance a cursor to its next row of output +*/ +static int amatchNext(sqlite3_vtab_cursor *cur){ + amatch_cursor *pCur = (amatch_cursor*)cur; + amatch_word *pWord = 0; + amatch_avl *pNode; + int isMatch = 0; + amatch_vtab *p = pCur->pVtab; + int nWord; + int rc; + int i; + const char *zW; + amatch_rule *pRule; + char *zBuf = 0; + char nBuf = 0; + char zNext[8]; + char zNextIn[8]; + int nNextIn; + + if( p->pVCheck==0 ){ + char *zSql; + if( p->zVocabLang && p->zVocabLang[0] ){ + zSql = sqlite3_mprintf( + "SELECT \"%s\" FROM \"%s\"", + " WHERE \"%w\">=?1 AND \"%w\"=?2" + " ORDER BY 1", + p->zVocabWord, p->zVocabTab, + p->zVocabWord, p->zVocabLang + ); + }else{ + zSql = sqlite3_mprintf( + "SELECT \"%s\" FROM \"%s\"" + " WHERE \"%w\">=?1" + " ORDER BY 1", + p->zVocabWord, p->zVocabTab, + p->zVocabWord + ); + } + rc = sqlite3_prepare_v2(p->db, zSql, -1, &p->pVCheck, 0); + sqlite3_free(zSql); + if( rc ) return rc; + } + sqlite3_bind_int(p->pVCheck, 2, pCur->iLang); + + do{ + pNode = amatchAvlFirst(pCur->pCost); + if( pNode==0 ){ + pWord = 0; + break; + } + pWord = pNode->pWord; + amatchAvlRemove(&pCur->pCost, &pWord->sCost); + +#ifdef AMATCH_TRACE_1 + printf("PROCESS [%s][%.*s^%s] %d (\"%s\" \"%s\")\n", + pWord->zWord+2, pWord->nMatch, pCur->zInput, pCur->zInput+pWord->nMatch, + pWord->rCost, pWord->zWord, pWord->zCost); +#endif + nWord = (int)strlen(pWord->zWord+2); + if( nWord+20>nBuf ){ + nBuf = nWord+100; + zBuf = sqlite3_realloc(zBuf, nBuf); + if( zBuf==0 ) return SQLITE_NOMEM; + } + strcpy(zBuf, pWord->zWord+2); + zNext[0] = 0; + zNextIn[0] = pCur->zInput[pWord->nMatch]; + if( zNextIn[0] ){ + for(i=1; i<=4 && (pCur->zInput[pWord->nMatch+i]&0xc0)==0x80; i++){ + zNextIn[i] = pCur->zInput[pWord->nMatch+i]; + } + zNextIn[i] = 0; + nNextIn = i; + }else{ + nNextIn = 0; + } + + if( zNextIn[0] && zNextIn[0]!='*' ){ + sqlite3_reset(p->pVCheck); + strcat(zBuf, zNextIn); + sqlite3_bind_text(p->pVCheck, 1, zBuf, nWord+nNextIn, SQLITE_STATIC); + rc = sqlite3_step(p->pVCheck); + if( rc==SQLITE_ROW ){ + zW = (const char*)sqlite3_column_text(p->pVCheck, 0); + if( strncmp(zBuf, zW, nWord+nNextIn)==0 ){ + amatchAddWord(pCur, pWord->rCost, pWord->nMatch+nNextIn, zBuf, ""); + } + } + zBuf[nWord] = 0; + } + + while( 1 ){ + strcpy(zBuf+nWord, zNext); + sqlite3_reset(p->pVCheck); + sqlite3_bind_text(p->pVCheck, 1, zBuf, -1, SQLITE_TRANSIENT); + rc = sqlite3_step(p->pVCheck); + if( rc!=SQLITE_ROW ) break; + zW = (const char*)sqlite3_column_text(p->pVCheck, 0); + strcpy(zBuf+nWord, zNext); + if( strncmp(zW, zBuf, nWord)!=0 ) break; + if( (zNextIn[0]=='*' && zNextIn[1]==0) + || (zNextIn[0]==0 && zW[nWord]==0) + ){ + isMatch = 1; + zNextIn[0] = 0; + nNextIn = 0; + break; + } + zNext[0] = zW[nWord]; + for(i=1; i<=4 && (zW[nWord+i]&0xc0)==0x80; i++){ + zNext[i] = zW[nWord+i]; + } + zNext[i] = 0; + zBuf[nWord] = 0; + if( p->rIns>0 ){ + amatchAddWord(pCur, pWord->rCost+p->rIns, pWord->nMatch, + zBuf, zNext); + } + if( p->rSub>0 ){ + amatchAddWord(pCur, pWord->rCost+p->rSub, pWord->nMatch+nNextIn, + zBuf, zNext); + } + if( p->rIns<0 && p->rSub<0 ) break; + zNext[i-1]++; /* FIX ME */ + } + sqlite3_reset(p->pVCheck); + + if( p->rDel>0 ){ + zBuf[nWord] = 0; + amatchAddWord(pCur, pWord->rCost+p->rDel, pWord->nMatch+nNextIn, + zBuf, ""); + } + + for(pRule=p->pRule; pRule; pRule=pRule->pNext){ + if( pRule->iLang!=pCur->iLang ) continue; + if( strncmp(pRule->zFrom, pCur->zInput+pWord->nMatch, pRule->nFrom)==0 ){ + amatchAddWord(pCur, pWord->rCost+pRule->rCost, + pWord->nMatch+pRule->nFrom, pWord->zWord+2, pRule->zTo); + } + } + }while( !isMatch ); + pCur->pCurrent = pWord; + sqlite3_free(zBuf); + return SQLITE_OK; +} + +/* +** Called to "rewind" a cursor back to the beginning so that +** it starts its output over again. Always called at least once +** prior to any amatchColumn, amatchRowid, or amatchEof call. +*/ +static int amatchFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + amatch_cursor *pCur = (amatch_cursor *)pVtabCursor; + const char *zWord = "*"; + int idx; + + amatchClearCursor(pCur); + idx = 0; + if( idxNum & 1 ){ + zWord = (const char*)sqlite3_value_text(argv[0]); + idx++; + } + if( idxNum & 2 ){ + pCur->rLimit = (amatch_cost)sqlite3_value_int(argv[idx]); + idx++; + } + if( idxNum & 4 ){ + pCur->iLang = (amatch_cost)sqlite3_value_int(argv[idx]); + idx++; + } + pCur->zInput = sqlite3_mprintf("%s", zWord); + if( pCur->zInput==0 ) return SQLITE_NOMEM; + amatchAddWord(pCur, 0, 0, "", ""); + amatchNext(pVtabCursor); + + return SQLITE_OK; +} + +/* +** Only the word and distance columns have values. All other columns +** return NULL +*/ +static int amatchColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ + amatch_cursor *pCur = (amatch_cursor*)cur; + switch( i ){ + case AMATCH_COL_WORD: { + sqlite3_result_text(ctx, pCur->pCurrent->zWord+2, -1, SQLITE_STATIC); + break; + } + case AMATCH_COL_DISTANCE: { + sqlite3_result_int(ctx, pCur->pCurrent->rCost); + break; + } + case AMATCH_COL_LANGUAGE: { + sqlite3_result_int(ctx, pCur->iLang); + break; + } + case AMATCH_COL_NWORD: { + sqlite3_result_int(ctx, pCur->nWord); + break; + } + default: { + sqlite3_result_null(ctx); + break; + } + } + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int amatchRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + amatch_cursor *pCur = (amatch_cursor*)cur; + *pRowid = pCur->iRowid; + return SQLITE_OK; +} + +/* +** EOF indicator +*/ +static int amatchEof(sqlite3_vtab_cursor *cur){ + amatch_cursor *pCur = (amatch_cursor*)cur; + return pCur->pCurrent==0; +} + +/* +** Search for terms of these forms: +** +** (A) word MATCH $str +** (B1) distance < $value +** (B2) distance <= $value +** (C) language == $language +** +** The distance< and distance<= are both treated as distance<=. +** The query plan number is a bit vector: +** +** bit 1: Term of the form (A) found +** bit 2: Term like (B1) or (B2) found +** bit 3: Term like (C) found +** +** If bit-1 is set, $str is always in filter.argv[0]. If bit-2 is set +** then $value is in filter.argv[0] if bit-1 is clear and is in +** filter.argv[1] if bit-1 is set. If bit-3 is set, then $ruleid is +** in filter.argv[0] if bit-1 and bit-2 are both zero, is in +** filter.argv[1] if exactly one of bit-1 and bit-2 are set, and is in +** filter.argv[2] if both bit-1 and bit-2 are set. +*/ +static int amatchBestIndex( + sqlite3_vtab *tab, + sqlite3_index_info *pIdxInfo +){ + int iPlan = 0; + int iDistTerm = -1; + int iLangTerm = -1; + int i; + const struct sqlite3_index_constraint *pConstraint; + + (void)tab; + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + if( (iPlan & 1)==0 + && pConstraint->iColumn==0 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 2)==0 + && pConstraint->iColumn==1 + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) + ){ + iPlan |= 2; + iDistTerm = i; + } + if( (iPlan & 4)==0 + && pConstraint->iColumn==2 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 4; + pIdxInfo->aConstraintUsage[i].omit = 1; + iLangTerm = i; + } + } + if( iPlan & 2 ){ + pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = 1+((iPlan&1)!=0); + } + if( iPlan & 4 ){ + int idx = 1; + if( iPlan & 1 ) idx++; + if( iPlan & 2 ) idx++; + pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx; + } + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==1 + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + pIdxInfo->estimatedCost = (double)10000; + + return SQLITE_OK; +} + +/* +** The xUpdate() method. +** +** This implementation disallows DELETE and UPDATE. The only thing +** allowed is INSERT into the "command" column. +*/ +static int amatchUpdate( + sqlite3_vtab *pVTab, + int argc, + sqlite3_value **argv, + sqlite_int64 *pRowid +){ + amatch_vtab *p = (amatch_vtab*)pVTab; + const unsigned char *zCmd; + (void)pRowid; + if( argc==1 ){ + pVTab->zErrMsg = sqlite3_mprintf("DELETE from %s is not allowed", + p->zSelf); + return SQLITE_ERROR; + } + if( sqlite3_value_type(argv[0])!=SQLITE_NULL ){ + pVTab->zErrMsg = sqlite3_mprintf("UPDATE of %s is not allowed", + p->zSelf); + return SQLITE_ERROR; + } + if( sqlite3_value_type(argv[2+AMATCH_COL_WORD])!=SQLITE_NULL + || sqlite3_value_type(argv[2+AMATCH_COL_DISTANCE])!=SQLITE_NULL + || sqlite3_value_type(argv[2+AMATCH_COL_LANGUAGE])!=SQLITE_NULL + ){ + pVTab->zErrMsg = sqlite3_mprintf( + "INSERT INTO %s allowed for column [command] only", p->zSelf); + return SQLITE_ERROR; + } + zCmd = sqlite3_value_text(argv[2+AMATCH_COL_COMMAND]); + if( zCmd==0 ) return SQLITE_OK; + + return SQLITE_OK; +} + +/* +** A virtual table module that implements the "approximate_match". +*/ +static sqlite3_module amatchModule = { + 0, /* iVersion */ + amatchConnect, /* xCreate */ + amatchConnect, /* xConnect */ + amatchBestIndex, /* xBestIndex */ + amatchDisconnect, /* xDisconnect */ + amatchDisconnect, /* xDestroy */ + amatchOpen, /* xOpen - open a cursor */ + amatchClose, /* xClose - close a cursor */ + amatchFilter, /* xFilter - configure scan constraints */ + amatchNext, /* xNext - advance a cursor */ + amatchEof, /* xEof - check for end of scan */ + amatchColumn, /* xColumn - read data */ + amatchRowid, /* xRowid - read data */ + amatchUpdate, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ + 0, /* xSavepoint */ + 0, /* xRelease */ + 0 /* xRollbackTo */ +}; + +/* +** Register the amatch virtual table +*/ +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_amatch_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Not used */ + rc = sqlite3_create_module(db, "approximate_match", &amatchModule, 0); + return rc; +} ADDED ext/misc/closure.c Index: ext/misc/closure.c ================================================================== --- ext/misc/closure.c +++ ext/misc/closure.c @@ -0,0 +1,942 @@ +/* +** 2013-04-16 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** This file contains code for a virtual table that finds the transitive +** closure of a parent/child relationship in a real table. The virtual +** table is called "transitive_closure". +** +** A transitive_closure virtual table is created like this: +** +** CREATE VIRTUAL TABLE x USING transitive_closure( +** tablename=, -- T +** idcolumn=, -- X +** parentcolumn= -- P +** ); +** +** When it is created, the new transitive_closure table may be supplied +** with default values for the name of a table T and columns T.X and T.P. +** The T.X and T.P columns must contain integers. The ideal case is for +** T.X to be the INTEGER PRIMARY KEY. The T.P column should reference +** the T.X column. The row referenced by T.P is the parent of the current row. +** +** The tablename, idcolumn, and parentcolumn supplied by the CREATE VIRTUAL +** TABLE statement may be overridden in individual queries by including +** terms like tablename='newtable', idcolumn='id2', or +** parentcolumn='parent3' in the WHERE clause of the query. +** +** For efficiency, it is essential that there be an index on the P column: +** +** CREATE Tidx1 ON T(P) +** +** Suppose a specific instance of the closure table is as follows: +** +** CREATE VIRTUAL TABLE ct1 USING transitive_closure( +** tablename='group', +** idcolumn='groupId', +** parentcolumn='parentId' +** ); +** +** Such an instance of the transitive_closure virtual table would be +** appropriate for walking a tree defined using a table like this, for example: +** +** CREATE TABLE group( +** groupId INTEGER PRIMARY KEY, +** parentId INTEGER REFERENCES group +** ); +** CREATE INDEX group_idx1 ON group(parentId); +** +** The group table above would presumably have other application-specific +** fields. The key point here is that rows of the group table form a +** tree. The purpose of the ct1 virtual table is to easily extract +** branches of that tree. +** +** Once it has been created, the ct1 virtual table can be queried +** as follows: +** +** SELECT * FROM element +** WHERE element.groupId IN (SELECT id FROM ct1 WHERE root=?1); +** +** The above query will return all elements that are part of group ?1 +** or children of group ?1 or grand-children of ?1 and so forth for all +** descendents of group ?1. The same query can be formulated as a join: +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupid=ct1.id +** AND ct1.root=?1; +** +** The depth of the transitive_closure (the number of generations of +** parent/child relations to follow) can be limited by setting "depth" +** column in the WHERE clause. So, for example, the following query +** finds only children and grandchildren but no further descendents: +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupid=ct1.id +** AND ct1.root=?1 +** AND ct1.depth<=2; +** +** The "ct1.depth<=2" term could be a strict equality "ct1.depth=2" in +** order to find only the grandchildren of ?1, not ?1 itself or the +** children of ?1. +** +** The root=?1 term must be supplied in WHERE clause or else the query +** of the ct1 virtual table will return an empty set. The tablename, +** idcolumn, and parentcolumn attributes can be overridden in the WHERE +** clause if desired. So, for example, the ct1 table could be repurposed +** to find ancestors rather than descendents by inverting the roles of +** the idcolumn and parentcolumn: +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupid=ct1.id +** AND ct1.root=?1 +** AND ct1.idcolumn='parentId' +** AND ct1.parentcolumn='groupId'; +** +** Multiple calls to ct1 could be combined. For example, the following +** query finds all elements that "cousins" of groupId ?1. That is to say +** elements where the groupId is a grandchild of the grandparent of ?1. +** (This definition of "cousins" also includes siblings and self.) +** +** SELECT element.* FROM element, ct1 +** WHERE element.groupId=ct1.id +** AND ct1.depth=2 +** AND ct1.root IN (SELECT id FROM ct1 +** WHERE root=?1 +** AND depth=2 +** AND idcolumn='parentId' +** AND parentcolumn='groupId'); +** +** In our example, the group.groupId column is unique and thus the +** subquery will return exactly one row. For that reason, the IN +** operator could be replaced by "=" to get the same result. But +** in the general case where the idcolumn is not unique, an IN operator +** would be required for this kind of query. +** +** Note that because the tablename, idcolumn, and parentcolumn can +** all be specified in the query, it is possible for an application +** to define a single transitive_closure virtual table for use on lots +** of different hierarchy tables. One might say: +** +** CREATE VIRTUAL TABLE temp.closure USING transitive_closure; +** +** As each database connection is being opened. Then the application +** would always have a "closure" virtual table handy to use for querying. +** +** SELECT element.* FROM element, closure +** WHERE element.groupid=ct1.id +** AND closure.root=?1 +** AND closure.tablename='group' +** AND closure.idname='groupId' +** AND closure.parentname='parentId'; +** +** See the documentation at http://www.sqlite.org/loadext.html for information +** on how to compile and use loadable extensions such as this one. +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include +#include +#include +#include +#include + +/* +** Forward declaration of objects used by this implementation +*/ +typedef struct closure_vtab closure_vtab; +typedef struct closure_cursor closure_cursor; +typedef struct closure_queue closure_queue; +typedef struct closure_avl closure_avl; + +/***************************************************************************** +** AVL Tree implementation +*/ +/* +** Objects that want to be members of the AVL tree should embedded an +** instance of this structure. +*/ +struct closure_avl { + sqlite3_int64 id; /* Id of this entry in the table */ + int iGeneration; /* Which generation is this entry part of */ + closure_avl *pList; /* A linked list of nodes */ + closure_avl *pBefore; /* Other elements less than id */ + closure_avl *pAfter; /* Other elements greater than id */ + closure_avl *pUp; /* Parent element */ + short int height; /* Height of this node. Leaf==1 */ + short int imbalance; /* Height difference between pBefore and pAfter */ +}; + +/* Recompute the closure_avl.height and closure_avl.imbalance fields for p. +** Assume that the children of p have correct heights. +*/ +static void closureAvlRecomputeHeight(closure_avl *p){ + short int hBefore = p->pBefore ? p->pBefore->height : 0; + short int hAfter = p->pAfter ? p->pAfter->height : 0; + p->imbalance = hBefore - hAfter; /* -: pAfter higher. +: pBefore higher */ + p->height = (hBefore>hAfter ? hBefore : hAfter)+1; +} + +/* +** P B +** / \ / \ +** B Z ==> X P +** / \ / \ +** X Y Y Z +** +*/ +static closure_avl *closureAvlRotateBefore(closure_avl *pP){ + closure_avl *pB = pP->pBefore; + closure_avl *pY = pB->pAfter; + pB->pUp = pP->pUp; + pB->pAfter = pP; + pP->pUp = pB; + pP->pBefore = pY; + if( pY ) pY->pUp = pP; + closureAvlRecomputeHeight(pP); + closureAvlRecomputeHeight(pB); + return pB; +} + +/* +** P A +** / \ / \ +** X A ==> P Z +** / \ / \ +** Y Z X Y +** +*/ +static closure_avl *closureAvlRotateAfter(closure_avl *pP){ + closure_avl *pA = pP->pAfter; + closure_avl *pY = pA->pBefore; + pA->pUp = pP->pUp; + pA->pBefore = pP; + pP->pUp = pA; + pP->pAfter = pY; + if( pY ) pY->pUp = pP; + closureAvlRecomputeHeight(pP); + closureAvlRecomputeHeight(pA); + return pA; +} + +/* +** Return a pointer to the pBefore or pAfter pointer in the parent +** of p that points to p. Or if p is the root node, return pp. +*/ +static closure_avl **closureAvlFromPtr(closure_avl *p, closure_avl **pp){ + closure_avl *pUp = p->pUp; + if( pUp==0 ) return pp; + if( pUp->pAfter==p ) return &pUp->pAfter; + return &pUp->pBefore; +} + +/* +** Rebalance all nodes starting with p and working up to the root. +** Return the new root. +*/ +static closure_avl *closureAvlBalance(closure_avl *p){ + closure_avl *pTop = p; + closure_avl **pp; + while( p ){ + closureAvlRecomputeHeight(p); + if( p->imbalance>=2 ){ + closure_avl *pB = p->pBefore; + if( pB->imbalance<0 ) p->pBefore = closureAvlRotateAfter(pB); + pp = closureAvlFromPtr(p,&p); + p = *pp = closureAvlRotateBefore(p); + }else if( p->imbalance<=(-2) ){ + closure_avl *pA = p->pAfter; + if( pA->imbalance>0 ) p->pAfter = closureAvlRotateBefore(pA); + pp = closureAvlFromPtr(p,&p); + p = *pp = closureAvlRotateAfter(p); + } + pTop = p; + p = p->pUp; + } + return pTop; +} + +/* Search the tree rooted at p for an entry with id. Return a pointer +** to the entry or return NULL. +*/ +static closure_avl *closureAvlSearch(closure_avl *p, sqlite3_int64 id){ + while( p && id!=p->id ){ + p = (idid) ? p->pBefore : p->pAfter; + } + return p; +} + +/* Find the first node (the one with the smallest key). +*/ +static closure_avl *closureAvlFirst(closure_avl *p){ + if( p ) while( p->pBefore ) p = p->pBefore; + return p; +} + +/* Return the node with the next larger key after p. +*/ +closure_avl *closureAvlNext(closure_avl *p){ + closure_avl *pPrev = 0; + while( p && p->pAfter==pPrev ){ + pPrev = p; + p = p->pUp; + } + if( p && pPrev==0 ){ + p = closureAvlFirst(p->pAfter); + } + return p; +} + +/* Insert a new node pNew. Return NULL on success. If the key is not +** unique, then do not perform the insert but instead leave pNew unchanged +** and return a pointer to an existing node with the same key. +*/ +static closure_avl *closureAvlInsert( + closure_avl **ppHead, /* Head of the tree */ + closure_avl *pNew /* New node to be inserted */ +){ + closure_avl *p = *ppHead; + if( p==0 ){ + p = pNew; + pNew->pUp = 0; + }else{ + while( p ){ + if( pNew->idid ){ + if( p->pBefore ){ + p = p->pBefore; + }else{ + p->pBefore = pNew; + pNew->pUp = p; + break; + } + }else if( pNew->id>p->id ){ + if( p->pAfter ){ + p = p->pAfter; + }else{ + p->pAfter = pNew; + pNew->pUp = p; + break; + } + }else{ + return p; + } + } + } + pNew->pBefore = 0; + pNew->pAfter = 0; + pNew->height = 1; + pNew->imbalance = 0; + *ppHead = closureAvlBalance(p); + return 0; +} + +/* Walk the tree can call xDestroy on each node +*/ +static void closureAvlDestroy(closure_avl *p, void (*xDestroy)(closure_avl*)){ + if( p ){ + closureAvlDestroy(p->pBefore, xDestroy); + closureAvlDestroy(p->pAfter, xDestroy); + xDestroy(p); + } +} +/* +** End of the AVL Tree implementation +******************************************************************************/ + +/* +** A closure virtual-table object +*/ +struct closure_vtab { + sqlite3_vtab base; /* Base class - must be first */ + char *zDb; /* Name of database. (ex: "main") */ + char *zSelf; /* Name of this virtual table */ + char *zTableName; /* Name of table holding parent/child relation */ + char *zIdColumn; /* Name of ID column of zTableName */ + char *zParentColumn; /* Name of PARENT column in zTableName */ + sqlite3 *db; /* The database connection */ + int nCursor; /* Number of pending cursors */ +}; + +/* A closure cursor object */ +struct closure_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + closure_vtab *pVtab; /* The virtual table this cursor belongs to */ + char *zTableName; /* Name of table holding parent/child relation */ + char *zIdColumn; /* Name of ID column of zTableName */ + char *zParentColumn; /* Name of PARENT column in zTableName */ + closure_avl *pCurrent; /* Current element of output */ + closure_avl *pClosure; /* The complete closure tree */ +}; + +/* A queue of AVL nodes */ +struct closure_queue { + closure_avl *pFirst; /* Oldest node on the queue */ + closure_avl *pLast; /* Youngest node on the queue */ +}; + +/* +** Add a node to the end of the queue +*/ +static void queuePush(closure_queue *pQueue, closure_avl *pNode){ + pNode->pList = 0; + if( pQueue->pLast ){ + pQueue->pLast->pList = pNode; + }else{ + pQueue->pFirst = pNode; + } + pQueue->pLast = pNode; +} + +/* +** Extract the oldest element (the front element) from the queue. +*/ +static closure_avl *queuePull(closure_queue *pQueue){ + closure_avl *p = pQueue->pFirst; + if( p ){ + pQueue->pFirst = p->pList; + if( pQueue->pFirst==0 ) pQueue->pLast = 0; + } + return p; +} + +/* +** This function converts an SQL quoted string into an unquoted string +** and returns a pointer to a buffer allocated using sqlite3_malloc() +** containing the result. The caller should eventually free this buffer +** using sqlite3_free. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static char *closureDequote(const char *zIn){ + int nIn; /* Size of input string, in bytes */ + char *zOut; /* Output (dequoted) string */ + + nIn = (int)strlen(zIn); + zOut = sqlite3_malloc(nIn+1); + if( zOut ){ + char q = zIn[0]; /* Quote character (if any ) */ + + if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ + memcpy(zOut, zIn, nIn+1); + }else{ + int iOut = 0; /* Index of next byte to write to output */ + int iIn; /* Index of next byte to read from input */ + + if( q=='[' ) q = ']'; + for(iIn=1; iInzDb); + sqlite3_free(p->zSelf); + sqlite3_free(p->zTableName); + sqlite3_free(p->zIdColumn); + sqlite3_free(p->zParentColumn); + memset(p, 0, sizeof(*p)); + sqlite3_free(p); + } +} + +/* +** xDisconnect/xDestroy method for the closure module. +*/ +static int closureDisconnect(sqlite3_vtab *pVtab){ + closure_vtab *p = (closure_vtab*)pVtab; + assert( p->nCursor==0 ); + closureFree(p); + return SQLITE_OK; +} + +/* +** Check to see if the argument is of the form: +** +** KEY = VALUE +** +** If it is, return a pointer to the first character of VALUE. +** If not, return NULL. Spaces around the = are ignored. +*/ +static const char *closureValueOfKey(const char *zKey, const char *zStr){ + int nKey = (int)strlen(zKey); + int nStr = (int)strlen(zStr); + int i; + if( nStr module name ("approximate_match") +** argv[1] -> database name +** argv[2] -> table name +** argv[3...] -> arguments +*/ +static int closureConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + int rc = SQLITE_OK; /* Return code */ + closure_vtab *pNew = 0; /* New virtual table */ + const char *zDb = argv[1]; + const char *zVal; + int i; + + (void)pAux; + *ppVtab = 0; + pNew = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + rc = SQLITE_NOMEM; + memset(pNew, 0, sizeof(*pNew)); + pNew->db = db; + pNew->zDb = sqlite3_mprintf("%s", zDb); + if( pNew->zDb==0 ) goto closureConnectError; + pNew->zSelf = sqlite3_mprintf("%s", argv[2]); + if( pNew->zSelf==0 ) goto closureConnectError; + for(i=3; izTableName); + pNew->zTableName = closureDequote(zVal); + if( pNew->zTableName==0 ) goto closureConnectError; + continue; + } + zVal = closureValueOfKey("idcolumn", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zIdColumn); + pNew->zIdColumn = closureDequote(zVal); + if( pNew->zIdColumn==0 ) goto closureConnectError; + continue; + } + zVal = closureValueOfKey("parentcolumn", argv[i]); + if( zVal ){ + sqlite3_free(pNew->zParentColumn); + pNew->zParentColumn = closureDequote(zVal); + if( pNew->zParentColumn==0 ) goto closureConnectError; + continue; + } + *pzErr = sqlite3_mprintf("unrecognized argument: [%s]\n", argv[i]); + closureFree(pNew); + *ppVtab = 0; + return SQLITE_ERROR; + } + rc = sqlite3_declare_vtab(db, + "CREATE TABLE x(id,depth,root HIDDEN,tablename HIDDEN," + "idcolumn HIDDEN,parentcolumn HIDDEN)" + ); +#define CLOSURE_COL_ID 0 +#define CLOSURE_COL_DEPTH 1 +#define CLOSURE_COL_ROOT 2 +#define CLOSURE_COL_TABLENAME 3 +#define CLOSURE_COL_IDCOLUMN 4 +#define CLOSURE_COL_PARENTCOLUMN 5 + if( rc!=SQLITE_OK ){ + closureFree(pNew); + } + *ppVtab = &pNew->base; + return rc; + +closureConnectError: + closureFree(pNew); + return rc; +} + +/* +** Open a new closure cursor. +*/ +static int closureOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + closure_vtab *p = (closure_vtab*)pVTab; + closure_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVtab = p; + *ppCursor = &pCur->base; + p->nCursor++; + return SQLITE_OK; +} + +/* +** Free up all the memory allocated by a cursor. Set it rLimit to 0 +** to indicate that it is at EOF. +*/ +static void closureClearCursor(closure_cursor *pCur){ + closureAvlDestroy(pCur->pClosure, (void(*)(closure_avl*))sqlite3_free); + sqlite3_free(pCur->zTableName); + sqlite3_free(pCur->zIdColumn); + sqlite3_free(pCur->zParentColumn); + pCur->zTableName = 0; + pCur->zIdColumn = 0; + pCur->zParentColumn = 0; + pCur->pCurrent = 0; + pCur->pClosure = 0; +} + +/* +** Close a closure cursor. +*/ +static int closureClose(sqlite3_vtab_cursor *cur){ + closure_cursor *pCur = (closure_cursor *)cur; + closureClearCursor(pCur); + pCur->pVtab->nCursor--; + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Advance a cursor to its next row of output +*/ +static int closureNext(sqlite3_vtab_cursor *cur){ + closure_cursor *pCur = (closure_cursor*)cur; + pCur->pCurrent = closureAvlNext(pCur->pCurrent); + return SQLITE_OK; +} + +/* +** Allocate and insert a node +*/ +static int closureInsertNode( + closure_queue *pQueue, /* Add new node to this queue */ + closure_cursor *pCur, /* The cursor into which to add the node */ + sqlite3_int64 id, /* The node ID */ + int iGeneration /* The generation number for this node */ +){ + closure_avl *pNew = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + memset(pNew, 0, sizeof(*pNew)); + pNew->id = id; + pNew->iGeneration = iGeneration; + closureAvlInsert(&pCur->pClosure, pNew); + queuePush(pQueue, pNew); + return SQLITE_OK; +} + +/* +** Called to "rewind" a cursor back to the beginning so that +** it starts its output over again. Always called at least once +** prior to any closureColumn, closureRowid, or closureEof call. +** +** This routine actually computes the closure. +** +** See the comment at the beginning of closureBestIndex() for a +** description of the meaning of idxNum. The idxStr parameter is +** not used. +*/ +static int closureFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + closure_cursor *pCur = (closure_cursor *)pVtabCursor; + closure_vtab *pVtab = pCur->pVtab; + sqlite3_int64 iRoot; + int mxGen = 999999999; + char *zSql; + sqlite3_stmt *pStmt; + closure_avl *pAvl; + int rc = SQLITE_OK; + const char *zTableName = pVtab->zTableName; + const char *zIdColumn = pVtab->zIdColumn; + const char *zParentColumn = pVtab->zParentColumn; + closure_queue sQueue; + + (void)idxStr; /* Unused parameter */ + (void)argc; /* Unused parameter */ + closureClearCursor(pCur); + memset(&sQueue, 0, sizeof(sQueue)); + if( (idxNum & 1)==0 ){ + /* No root=$root in the WHERE clause. Return an empty set */ + return SQLITE_OK; + } + iRoot = sqlite3_value_int64(argv[0]); + if( (idxNum & 0x000f0)!=0 ){ + mxGen = sqlite3_value_int(argv[(idxNum>>4)&0x0f]); + if( (idxNum & 0x00002)!=0 ) mxGen--; + } + if( (idxNum & 0x00f00)!=0 ){ + zTableName = (const char*)sqlite3_value_text(argv[(idxNum>>8)&0x0f]); + pCur->zTableName = sqlite3_mprintf("%s", zTableName); + } + if( (idxNum & 0x0f000)!=0 ){ + zIdColumn = (const char*)sqlite3_value_text(argv[(idxNum>>12)&0x0f]); + pCur->zIdColumn = sqlite3_mprintf("%s", zIdColumn); + } + if( (idxNum & 0x0f0000)!=0 ){ + zParentColumn = (const char*)sqlite3_value_text(argv[(idxNum>>16)&0x0f]); + pCur->zParentColumn = sqlite3_mprintf("%s", zParentColumn); + } + + zSql = sqlite3_mprintf( + "SELECT \"%w\".\"%w\" FROM \"%w\" WHERE \"%w\".\"%w\"=?1", + zTableName, zIdColumn, zTableName, zTableName, zParentColumn); + if( zSql==0 ){ + return SQLITE_NOMEM; + }else{ + rc = sqlite3_prepare_v2(pVtab->db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + if( rc ){ + sqlite3_free(pVtab->base.zErrMsg); + pVtab->base.zErrMsg = sqlite3_mprintf("%s", sqlite3_errmsg(pVtab->db)); + return rc; + } + } + if( rc==SQLITE_OK ){ + rc = closureInsertNode(&sQueue, pCur, iRoot, 0); + } + while( (pAvl = queuePull(&sQueue))!=0 ){ + if( pAvl->iGeneration>=mxGen ) continue; + sqlite3_bind_int64(pStmt, 1, pAvl->id); + while( rc==SQLITE_OK && sqlite3_step(pStmt)==SQLITE_ROW ){ + if( sqlite3_column_type(pStmt,0)==SQLITE_INTEGER ){ + sqlite3_int64 iNew = sqlite3_column_int64(pStmt, 0); + if( closureAvlSearch(pCur->pClosure, iNew)==0 ){ + rc = closureInsertNode(&sQueue, pCur, iNew, pAvl->iGeneration+1); + } + } + } + sqlite3_reset(pStmt); + } + sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ){ + pCur->pCurrent = closureAvlFirst(pCur->pClosure); + } + + return rc; +} + +/* +** Only the word and distance columns have values. All other columns +** return NULL +*/ +static int closureColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ + closure_cursor *pCur = (closure_cursor*)cur; + switch( i ){ + case CLOSURE_COL_ID: { + sqlite3_result_int64(ctx, pCur->pCurrent->id); + break; + } + case CLOSURE_COL_DEPTH: { + sqlite3_result_int(ctx, pCur->pCurrent->iGeneration); + break; + } + case CLOSURE_COL_ROOT: { + sqlite3_result_null(ctx); + break; + } + case CLOSURE_COL_TABLENAME: { + sqlite3_result_text(ctx, + pCur->zTableName ? pCur->zTableName : pCur->pVtab->zTableName, + -1, SQLITE_TRANSIENT); + break; + } + case CLOSURE_COL_IDCOLUMN: { + sqlite3_result_text(ctx, + pCur->zIdColumn ? pCur->zIdColumn : pCur->pVtab->zIdColumn, + -1, SQLITE_TRANSIENT); + break; + } + case CLOSURE_COL_PARENTCOLUMN: { + sqlite3_result_text(ctx, + pCur->zParentColumn ? pCur->zParentColumn : pCur->pVtab->zParentColumn, + -1, SQLITE_TRANSIENT); + break; + } + } + return SQLITE_OK; +} + +/* +** The rowid. For the closure table, this is the same as the "id" column. +*/ +static int closureRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + closure_cursor *pCur = (closure_cursor*)cur; + *pRowid = pCur->pCurrent->id; + return SQLITE_OK; +} + +/* +** EOF indicator +*/ +static int closureEof(sqlite3_vtab_cursor *cur){ + closure_cursor *pCur = (closure_cursor*)cur; + return pCur->pCurrent==0; +} + +/* +** Search for terms of these forms: +** +** (A) root = $root +** (B1) depth < $depth +** (B2) depth <= $depth +** (B3) depth = $depth +** (C) tablename = $tablename +** (D) idcolumn = $idcolumn +** (E) parentcolumn = $parentcolumn +** +** +** +** idxNum meaning +** ---------- ------------------------------------------------------ +** 0x00000001 Term of the form (A) found +** 0x00000002 The term of bit-2 is like (B1) +** 0x000000f0 Index in filter.argv[] of $depth. 0 if not used. +** 0x00000f00 Index in filter.argv[] of $tablename. 0 if not used. +** 0x0000f000 Index in filter.argv[] of $idcolumn. 0 if not used +** 0x000f0000 Index in filter.argv[] of $parentcolumn. 0 if not used. +** +** There must be a term of type (A). If there is not, then the index type +** is 0 and the query will return an empty set. +*/ +static int closureBestIndex( + sqlite3_vtab *pTab, /* The virtual table */ + sqlite3_index_info *pIdxInfo /* Information about the query */ +){ + int iPlan = 0; + int i; + int idx = 1; + const struct sqlite3_index_constraint *pConstraint; + closure_vtab *pVtab = (closure_vtab*)pTab; + + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + if( (iPlan & 1)==0 + && pConstraint->iColumn==CLOSURE_COL_ROOT + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 0x0000f0)==0 + && pConstraint->iColumn==CLOSURE_COL_DEPTH + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ) + ){ + iPlan |= idx<<4; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + if( pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ) iPlan |= 0x000002; + } + if( (iPlan & 0x000f00)==0 + && pConstraint->iColumn==CLOSURE_COL_TABLENAME + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= idx<<8; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 0x00f000)==0 + && pConstraint->iColumn==CLOSURE_COL_IDCOLUMN + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= idx<<12; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 0x0f0000)==0 + && pConstraint->iColumn==CLOSURE_COL_PARENTCOLUMN + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= idx<<16; + pIdxInfo->aConstraintUsage[i].argvIndex = ++idx; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + } + if( (pVtab->zTableName==0 && (iPlan & 0x000f00)==0) + || (pVtab->zIdColumn==0 && (iPlan & 0x00f000)==0) + || (pVtab->zParentColumn==0 && (iPlan & 0x0f0000)==0) + ){ + /* All of tablename, idcolumn, and parentcolumn must be specified + ** in either the CREATE VIRTUAL TABLE or in the WHERE clause constraints + ** or else the result is an empty set. */ + iPlan = 0; + } + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==CLOSURE_COL_ID + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + pIdxInfo->estimatedCost = (double)10000; + + return SQLITE_OK; +} + +/* +** A virtual table module that implements the "approximate_match". +*/ +static sqlite3_module closureModule = { + 0, /* iVersion */ + closureConnect, /* xCreate */ + closureConnect, /* xConnect */ + closureBestIndex, /* xBestIndex */ + closureDisconnect, /* xDisconnect */ + closureDisconnect, /* xDestroy */ + closureOpen, /* xOpen - open a cursor */ + closureClose, /* xClose - close a cursor */ + closureFilter, /* xFilter - configure scan constraints */ + closureNext, /* xNext - advance a cursor */ + closureEof, /* xEof - check for end of scan */ + closureColumn, /* xColumn - read data */ + closureRowid, /* xRowid - read data */ + 0, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ + 0, /* xSavepoint */ + 0, /* xRelease */ + 0 /* xRollbackTo */ +}; + +/* +** Register the closure virtual table +*/ +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_closure_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; + rc = sqlite3_create_module(db, "transitive_closure", &closureModule, 0); + return rc; +} ADDED ext/misc/fuzzer.c Index: ext/misc/fuzzer.c ================================================================== --- ext/misc/fuzzer.c +++ ext/misc/fuzzer.c @@ -0,0 +1,1171 @@ +/* +** 2011 March 24 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** Code for a demonstration virtual table that generates variations +** on an input word at increasing edit distances from the original. +** +** A fuzzer virtual table is created like this: +** +** CREATE VIRTUAL TABLE f USING fuzzer(); +** +** When it is created, the new fuzzer table must be supplied with the +** name of a "fuzzer data table", which must reside in the same database +** file as the new fuzzer table. The fuzzer data table contains the various +** transformations and their costs that the fuzzer logic uses to generate +** variations. +** +** The fuzzer data table must contain exactly four columns (more precisely, +** the statement "SELECT * FROM " must return records +** that consist of four columns). It does not matter what the columns are +** named. +** +** Each row in the fuzzer data table represents a single character +** transformation. The left most column of the row (column 0) contains an +** integer value - the identifier of the ruleset to which the transformation +** rule belongs (see "MULTIPLE RULE SETS" below). The second column of the +** row (column 0) contains the input character or characters. The third +** column contains the output character or characters. And the fourth column +** contains the integer cost of making the transformation. For example: +** +** CREATE TABLE f_data(ruleset, cFrom, cTo, Cost); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, '', 'a', 100); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'b', '', 87); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'o', 'oe', 38); +** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'oe', 'o', 40); +** +** The first row inserted into the fuzzer data table by the SQL script +** above indicates that the cost of inserting a letter 'a' is 100. (All +** costs are integers. We recommend that costs be scaled so that the +** average cost is around 100.) The second INSERT statement creates a rule +** saying that the cost of deleting a single letter 'b' is 87. The third +** and fourth INSERT statements mean that the cost of transforming a +** single letter "o" into the two-letter sequence "oe" is 38 and that the +** cost of transforming "oe" back into "o" is 40. +** +** The contents of the fuzzer data table are loaded into main memory when +** a fuzzer table is first created, and may be internally reloaded by the +** system at any subsequent time. Therefore, the fuzzer data table should be +** populated before the fuzzer table is created and not modified thereafter. +** If you do need to modify the contents of the fuzzer data table, it is +** recommended that the associated fuzzer table be dropped, the fuzzer data +** table edited, and the fuzzer table recreated within a single transaction. +** Alternatively, the fuzzer data table can be edited then the database +** connection can be closed and reopened. +** +** Once it has been created, the fuzzer table can be queried as follows: +** +** SELECT word, distance FROM f +** WHERE word MATCH 'abcdefg' +** AND distance<200; +** +** This first query outputs the string "abcdefg" and all strings that +** can be derived from that string by appling the specified transformations. +** The strings are output together with their total transformation cost +** (called "distance") and appear in order of increasing cost. No string +** is output more than once. If there are multiple ways to transform the +** target string into the output string then the lowest cost transform is +** the one that is returned. In the example, the search is limited to +** strings with a total distance of less than 200. +** +** The fuzzer is a read-only table. Any attempt to DELETE, INSERT, or +** UPDATE on a fuzzer table will throw an error. +** +** It is important to put some kind of a limit on the fuzzer output. This +** can be either in the form of a LIMIT clause at the end of the query, +** or better, a "distance +#include +#include +#include + +#ifndef SQLITE_OMIT_VIRTUALTABLE + +/* +** Forward declaration of objects used by this implementation +*/ +typedef struct fuzzer_vtab fuzzer_vtab; +typedef struct fuzzer_cursor fuzzer_cursor; +typedef struct fuzzer_rule fuzzer_rule; +typedef struct fuzzer_seen fuzzer_seen; +typedef struct fuzzer_stem fuzzer_stem; + +/* +** Various types. +** +** fuzzer_cost is the "cost" of an edit operation. +** +** fuzzer_len is the length of a matching string. +** +** fuzzer_ruleid is an ruleset identifier. +*/ +typedef int fuzzer_cost; +typedef signed char fuzzer_len; +typedef int fuzzer_ruleid; + +/* +** Limits +*/ +#define FUZZER_MX_LENGTH 50 /* Maximum length of a rule string */ +#define FUZZER_MX_RULEID 2147483647 /* Maximum rule ID */ +#define FUZZER_MX_COST 1000 /* Maximum single-rule cost */ +#define FUZZER_MX_OUTPUT_LENGTH 100 /* Maximum length of an output string */ + + +/* +** Each transformation rule is stored as an instance of this object. +** All rules are kept on a linked list sorted by rCost. +*/ +struct fuzzer_rule { + fuzzer_rule *pNext; /* Next rule in order of increasing rCost */ + char *zFrom; /* Transform from */ + fuzzer_cost rCost; /* Cost of this transformation */ + fuzzer_len nFrom, nTo; /* Length of the zFrom and zTo strings */ + fuzzer_ruleid iRuleset; /* The rule set to which this rule belongs */ + char zTo[4]; /* Transform to (extra space appended) */ +}; + +/* +** A stem object is used to generate variants. It is also used to record +** previously generated outputs. +** +** Every stem is added to a hash table as it is output. Generation of +** duplicate stems is suppressed. +** +** Active stems (those that might generate new outputs) are kepts on a linked +** list sorted by increasing cost. The cost is the sum of rBaseCost and +** pRule->rCost. +*/ +struct fuzzer_stem { + char *zBasis; /* Word being fuzzed */ + const fuzzer_rule *pRule; /* Current rule to apply */ + fuzzer_stem *pNext; /* Next stem in rCost order */ + fuzzer_stem *pHash; /* Next stem with same hash on zBasis */ + fuzzer_cost rBaseCost; /* Base cost of getting to zBasis */ + fuzzer_cost rCostX; /* Precomputed rBaseCost + pRule->rCost */ + fuzzer_len nBasis; /* Length of the zBasis string */ + fuzzer_len n; /* Apply pRule at this character offset */ +}; + +/* +** A fuzzer virtual-table object +*/ +struct fuzzer_vtab { + sqlite3_vtab base; /* Base class - must be first */ + char *zClassName; /* Name of this class. Default: "fuzzer" */ + fuzzer_rule *pRule; /* All active rules in this fuzzer */ + int nCursor; /* Number of active cursors */ +}; + +#define FUZZER_HASH 4001 /* Hash table size */ +#define FUZZER_NQUEUE 20 /* Number of slots on the stem queue */ + +/* A fuzzer cursor object */ +struct fuzzer_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + sqlite3_int64 iRowid; /* The rowid of the current word */ + fuzzer_vtab *pVtab; /* The virtual table this cursor belongs to */ + fuzzer_cost rLimit; /* Maximum cost of any term */ + fuzzer_stem *pStem; /* Stem with smallest rCostX */ + fuzzer_stem *pDone; /* Stems already processed to completion */ + fuzzer_stem *aQueue[FUZZER_NQUEUE]; /* Queue of stems with higher rCostX */ + int mxQueue; /* Largest used index in aQueue[] */ + char *zBuf; /* Temporary use buffer */ + int nBuf; /* Bytes allocated for zBuf */ + int nStem; /* Number of stems allocated */ + int iRuleset; /* Only process rules from this ruleset */ + fuzzer_rule nullRule; /* Null rule used first */ + fuzzer_stem *apHash[FUZZER_HASH]; /* Hash of previously generated terms */ +}; + +/* +** The two input rule lists are both sorted in order of increasing +** cost. Merge them together into a single list, sorted by cost, and +** return a pointer to the head of that list. +*/ +static fuzzer_rule *fuzzerMergeRules(fuzzer_rule *pA, fuzzer_rule *pB){ + fuzzer_rule head; + fuzzer_rule *pTail; + + pTail = &head; + while( pA && pB ){ + if( pA->rCost<=pB->rCost ){ + pTail->pNext = pA; + pTail = pA; + pA = pA->pNext; + }else{ + pTail->pNext = pB; + pTail = pB; + pB = pB->pNext; + } + } + if( pA==0 ){ + pTail->pNext = pB; + }else{ + pTail->pNext = pA; + } + return head.pNext; +} + +/* +** Statement pStmt currently points to a row in the fuzzer data table. This +** function allocates and populates a fuzzer_rule structure according to +** the content of the row. +** +** If successful, *ppRule is set to point to the new object and SQLITE_OK +** is returned. Otherwise, *ppRule is zeroed, *pzErr may be set to point +** to an error message and an SQLite error code returned. +*/ +static int fuzzerLoadOneRule( + fuzzer_vtab *p, /* Fuzzer virtual table handle */ + sqlite3_stmt *pStmt, /* Base rule on statements current row */ + fuzzer_rule **ppRule, /* OUT: New rule object */ + char **pzErr /* OUT: Error message */ +){ + sqlite3_int64 iRuleset = sqlite3_column_int64(pStmt, 0); + const char *zFrom = (const char *)sqlite3_column_text(pStmt, 1); + const char *zTo = (const char *)sqlite3_column_text(pStmt, 2); + int nCost = sqlite3_column_int(pStmt, 3); + + int rc = SQLITE_OK; /* Return code */ + int nFrom; /* Size of string zFrom, in bytes */ + int nTo; /* Size of string zTo, in bytes */ + fuzzer_rule *pRule = 0; /* New rule object to return */ + + if( zFrom==0 ) zFrom = ""; + if( zTo==0 ) zTo = ""; + nFrom = (int)strlen(zFrom); + nTo = (int)strlen(zTo); + + /* Silently ignore null transformations */ + if( strcmp(zFrom, zTo)==0 ){ + *ppRule = 0; + return SQLITE_OK; + } + + if( nCost<=0 || nCost>FUZZER_MX_COST ){ + *pzErr = sqlite3_mprintf("%s: cost must be between 1 and %d", + p->zClassName, FUZZER_MX_COST + ); + rc = SQLITE_ERROR; + }else + if( nFrom>FUZZER_MX_LENGTH || nTo>FUZZER_MX_LENGTH ){ + *pzErr = sqlite3_mprintf("%s: maximum string length is %d", + p->zClassName, FUZZER_MX_LENGTH + ); + rc = SQLITE_ERROR; + }else + if( iRuleset<0 || iRuleset>FUZZER_MX_RULEID ){ + *pzErr = sqlite3_mprintf("%s: ruleset must be between 0 and %d", + p->zClassName, FUZZER_MX_RULEID + ); + rc = SQLITE_ERROR; + }else{ + + pRule = sqlite3_malloc( sizeof(*pRule) + nFrom + nTo ); + if( pRule==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pRule, 0, sizeof(*pRule)); + pRule->zFrom = &pRule->zTo[nTo+1]; + pRule->nFrom = nFrom; + memcpy(pRule->zFrom, zFrom, nFrom+1); + memcpy(pRule->zTo, zTo, nTo+1); + pRule->nTo = nTo; + pRule->rCost = nCost; + pRule->iRuleset = (int)iRuleset; + } + } + + *ppRule = pRule; + return rc; +} + +/* +** Load the content of the fuzzer data table into memory. +*/ +static int fuzzerLoadRules( + sqlite3 *db, /* Database handle */ + fuzzer_vtab *p, /* Virtual fuzzer table to configure */ + const char *zDb, /* Database containing rules data */ + const char *zData, /* Table containing rules data */ + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; /* Return code */ + char *zSql; /* SELECT used to read from rules table */ + fuzzer_rule *pHead = 0; + + zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zData); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + int rc2; /* finalize() return code */ + sqlite3_stmt *pStmt = 0; + rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); + if( rc!=SQLITE_OK ){ + *pzErr = sqlite3_mprintf("%s: %s", p->zClassName, sqlite3_errmsg(db)); + }else if( sqlite3_column_count(pStmt)!=4 ){ + *pzErr = sqlite3_mprintf("%s: %s has %d columns, expected 4", + p->zClassName, zData, sqlite3_column_count(pStmt) + ); + rc = SQLITE_ERROR; + }else{ + while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ + fuzzer_rule *pRule = 0; + rc = fuzzerLoadOneRule(p, pStmt, &pRule, pzErr); + if( pRule ){ + pRule->pNext = pHead; + pHead = pRule; + } + } + } + rc2 = sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ) rc = rc2; + } + sqlite3_free(zSql); + + /* All rules are now in a singly linked list starting at pHead. This + ** block sorts them by cost and then sets fuzzer_vtab.pRule to point to + ** point to the head of the sorted list. + */ + if( rc==SQLITE_OK ){ + unsigned int i; + fuzzer_rule *pX; + fuzzer_rule *a[15]; + for(i=0; ipNext; + pX->pNext = 0; + for(i=0; a[i] && ipRule = fuzzerMergeRules(p->pRule, pX); + }else{ + /* An error has occurred. Setting p->pRule to point to the head of the + ** allocated list ensures that the list will be cleaned up in this case. + */ + assert( p->pRule==0 ); + p->pRule = pHead; + } + + return rc; +} + +/* +** This function converts an SQL quoted string into an unquoted string +** and returns a pointer to a buffer allocated using sqlite3_malloc() +** containing the result. The caller should eventually free this buffer +** using sqlite3_free. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static char *fuzzerDequote(const char *zIn){ + int nIn; /* Size of input string, in bytes */ + char *zOut; /* Output (dequoted) string */ + + nIn = (int)strlen(zIn); + zOut = sqlite3_malloc(nIn+1); + if( zOut ){ + char q = zIn[0]; /* Quote character (if any ) */ + + if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ + memcpy(zOut, zIn, nIn+1); + }else{ + int iOut = 0; /* Index of next byte to write to output */ + int iIn; /* Index of next byte to read from input */ + + if( q=='[' ) q = ']'; + for(iIn=1; iInnCursor==0 ); + while( p->pRule ){ + fuzzer_rule *pRule = p->pRule; + p->pRule = pRule->pNext; + sqlite3_free(pRule); + } + sqlite3_free(p); + return SQLITE_OK; +} + +/* +** xConnect/xCreate method for the fuzzer module. Arguments are: +** +** argv[0] -> module name ("fuzzer") +** argv[1] -> database name +** argv[2] -> table name +** argv[3] -> fuzzer rule table name +*/ +static int fuzzerConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + int rc = SQLITE_OK; /* Return code */ + fuzzer_vtab *pNew = 0; /* New virtual table */ + const char *zModule = argv[0]; + const char *zDb = argv[1]; + + if( argc!=4 ){ + *pzErr = sqlite3_mprintf( + "%s: wrong number of CREATE VIRTUAL TABLE arguments", zModule + ); + rc = SQLITE_ERROR; + }else{ + int nModule; /* Length of zModule, in bytes */ + + nModule = (int)strlen(zModule); + pNew = sqlite3_malloc( sizeof(*pNew) + nModule + 1); + if( pNew==0 ){ + rc = SQLITE_NOMEM; + }else{ + char *zTab; /* Dequoted name of fuzzer data table */ + + memset(pNew, 0, sizeof(*pNew)); + pNew->zClassName = (char*)&pNew[1]; + memcpy(pNew->zClassName, zModule, nModule+1); + + zTab = fuzzerDequote(argv[3]); + if( zTab==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = fuzzerLoadRules(db, pNew, zDb, zTab, pzErr); + sqlite3_free(zTab); + } + + if( rc==SQLITE_OK ){ + rc = sqlite3_declare_vtab(db, "CREATE TABLE x(word,distance,ruleset)"); + } + if( rc!=SQLITE_OK ){ + fuzzerDisconnect((sqlite3_vtab *)pNew); + pNew = 0; + } + } + } + + *ppVtab = (sqlite3_vtab *)pNew; + return rc; +} + +/* +** Open a new fuzzer cursor. +*/ +static int fuzzerOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + fuzzer_vtab *p = (fuzzer_vtab*)pVTab; + fuzzer_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVtab = p; + *ppCursor = &pCur->base; + p->nCursor++; + return SQLITE_OK; +} + +/* +** Free all stems in a list. +*/ +static void fuzzerClearStemList(fuzzer_stem *pStem){ + while( pStem ){ + fuzzer_stem *pNext = pStem->pNext; + sqlite3_free(pStem); + pStem = pNext; + } +} + +/* +** Free up all the memory allocated by a cursor. Set it rLimit to 0 +** to indicate that it is at EOF. +*/ +static void fuzzerClearCursor(fuzzer_cursor *pCur, int clearHash){ + int i; + fuzzerClearStemList(pCur->pStem); + fuzzerClearStemList(pCur->pDone); + for(i=0; iaQueue[i]); + pCur->rLimit = (fuzzer_cost)0; + if( clearHash && pCur->nStem ){ + pCur->mxQueue = 0; + pCur->pStem = 0; + pCur->pDone = 0; + memset(pCur->aQueue, 0, sizeof(pCur->aQueue)); + memset(pCur->apHash, 0, sizeof(pCur->apHash)); + } + pCur->nStem = 0; +} + +/* +** Close a fuzzer cursor. +*/ +static int fuzzerClose(sqlite3_vtab_cursor *cur){ + fuzzer_cursor *pCur = (fuzzer_cursor *)cur; + fuzzerClearCursor(pCur, 0); + sqlite3_free(pCur->zBuf); + pCur->pVtab->nCursor--; + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Compute the current output term for a fuzzer_stem. +*/ +static int fuzzerRender( + fuzzer_stem *pStem, /* The stem to be rendered */ + char **pzBuf, /* Write results into this buffer. realloc if needed */ + int *pnBuf /* Size of the buffer */ +){ + const fuzzer_rule *pRule = pStem->pRule; + int n; /* Size of output term without nul-term */ + char *z; /* Buffer to assemble output term in */ + + n = pStem->nBasis + pRule->nTo - pRule->nFrom; + if( (*pnBuf)n; + z = *pzBuf; + if( n<0 ){ + memcpy(z, pStem->zBasis, pStem->nBasis+1); + }else{ + memcpy(z, pStem->zBasis, n); + memcpy(&z[n], pRule->zTo, pRule->nTo); + memcpy(&z[n+pRule->nTo], &pStem->zBasis[n+pRule->nFrom], + pStem->nBasis-n-pRule->nFrom+1); + } + + assert( z[pStem->nBasis + pRule->nTo - pRule->nFrom]==0 ); + return SQLITE_OK; +} + +/* +** Compute a hash on zBasis. +*/ +static unsigned int fuzzerHash(const char *z){ + unsigned int h = 0; + while( *z ){ h = (h<<3) ^ (h>>29) ^ *(z++); } + return h % FUZZER_HASH; +} + +/* +** Current cost of a stem +*/ +static fuzzer_cost fuzzerCost(fuzzer_stem *pStem){ + return pStem->rCostX = pStem->rBaseCost + pStem->pRule->rCost; +} + +#if 0 +/* +** Print a description of a fuzzer_stem on stderr. +*/ +static void fuzzerStemPrint( + const char *zPrefix, + fuzzer_stem *pStem, + const char *zSuffix +){ + if( pStem->n<0 ){ + fprintf(stderr, "%s[%s](%d)-->self%s", + zPrefix, + pStem->zBasis, pStem->rBaseCost, + zSuffix + ); + }else{ + char *zBuf = 0; + int nBuf = 0; + if( fuzzerRender(pStem, &zBuf, &nBuf)!=SQLITE_OK ) return; + fprintf(stderr, "%s[%s](%d)-->{%s}(%d)%s", + zPrefix, + pStem->zBasis, pStem->rBaseCost, zBuf, pStem->, + zSuffix + ); + sqlite3_free(zBuf); + } +} +#endif + +/* +** Return 1 if the string to which the cursor is point has already +** been emitted. Return 0 if not. Return -1 on a memory allocation +** failures. +*/ +static int fuzzerSeen(fuzzer_cursor *pCur, fuzzer_stem *pStem){ + unsigned int h; + fuzzer_stem *pLookup; + + if( fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ + return -1; + } + h = fuzzerHash(pCur->zBuf); + pLookup = pCur->apHash[h]; + while( pLookup && strcmp(pLookup->zBasis, pCur->zBuf)!=0 ){ + pLookup = pLookup->pHash; + } + return pLookup!=0; +} + +/* +** If argument pRule is NULL, this function returns false. +** +** Otherwise, it returns true if rule pRule should be skipped. A rule +** should be skipped if it does not belong to rule-set iRuleset, or if +** applying it to stem pStem would create a string longer than +** FUZZER_MX_OUTPUT_LENGTH bytes. +*/ +static int fuzzerSkipRule( + const fuzzer_rule *pRule, /* Determine whether or not to skip this */ + fuzzer_stem *pStem, /* Stem rule may be applied to */ + int iRuleset /* Rule-set used by the current query */ +){ + return pRule && ( + (pRule->iRuleset!=iRuleset) + || (pStem->nBasis + pRule->nTo - pRule->nFrom)>FUZZER_MX_OUTPUT_LENGTH + ); +} + +/* +** Advance a fuzzer_stem to its next value. Return 0 if there are +** no more values that can be generated by this fuzzer_stem. Return +** -1 on a memory allocation failure. +*/ +static int fuzzerAdvance(fuzzer_cursor *pCur, fuzzer_stem *pStem){ + const fuzzer_rule *pRule; + while( (pRule = pStem->pRule)!=0 ){ + assert( pRule==&pCur->nullRule || pRule->iRuleset==pCur->iRuleset ); + while( pStem->n < pStem->nBasis - pRule->nFrom ){ + pStem->n++; + if( pRule->nFrom==0 + || memcmp(&pStem->zBasis[pStem->n], pRule->zFrom, pRule->nFrom)==0 + ){ + /* Found a rewrite case. Make sure it is not a duplicate */ + int rc = fuzzerSeen(pCur, pStem); + if( rc<0 ) return -1; + if( rc==0 ){ + fuzzerCost(pStem); + return 1; + } + } + } + pStem->n = -1; + do{ + pRule = pRule->pNext; + }while( fuzzerSkipRule(pRule, pStem, pCur->iRuleset) ); + pStem->pRule = pRule; + if( pRule && fuzzerCost(pStem)>pCur->rLimit ) pStem->pRule = 0; + } + return 0; +} + +/* +** The two input stem lists are both sorted in order of increasing +** rCostX. Merge them together into a single list, sorted by rCostX, and +** return a pointer to the head of that new list. +*/ +static fuzzer_stem *fuzzerMergeStems(fuzzer_stem *pA, fuzzer_stem *pB){ + fuzzer_stem head; + fuzzer_stem *pTail; + + pTail = &head; + while( pA && pB ){ + if( pA->rCostX<=pB->rCostX ){ + pTail->pNext = pA; + pTail = pA; + pA = pA->pNext; + }else{ + pTail->pNext = pB; + pTail = pB; + pB = pB->pNext; + } + } + if( pA==0 ){ + pTail->pNext = pB; + }else{ + pTail->pNext = pA; + } + return head.pNext; +} + +/* +** Load pCur->pStem with the lowest-cost stem. Return a pointer +** to the lowest-cost stem. +*/ +static fuzzer_stem *fuzzerLowestCostStem(fuzzer_cursor *pCur){ + fuzzer_stem *pBest, *pX; + int iBest; + int i; + + if( pCur->pStem==0 ){ + iBest = -1; + pBest = 0; + for(i=0; i<=pCur->mxQueue; i++){ + pX = pCur->aQueue[i]; + if( pX==0 ) continue; + if( pBest==0 || pBest->rCostX>pX->rCostX ){ + pBest = pX; + iBest = i; + } + } + if( pBest ){ + pCur->aQueue[iBest] = pBest->pNext; + pBest->pNext = 0; + pCur->pStem = pBest; + } + } + return pCur->pStem; +} + +/* +** Insert pNew into queue of pending stems. Then find the stem +** with the lowest rCostX and move it into pCur->pStem. +** list. The insert is done such the pNew is in the correct order +** according to fuzzer_stem.zBaseCost+fuzzer_stem.pRule->rCost. +*/ +static fuzzer_stem *fuzzerInsert(fuzzer_cursor *pCur, fuzzer_stem *pNew){ + fuzzer_stem *pX; + int i; + + /* If pCur->pStem exists and is greater than pNew, then make pNew + ** the new pCur->pStem and insert the old pCur->pStem instead. + */ + if( (pX = pCur->pStem)!=0 && pX->rCostX>pNew->rCostX ){ + pNew->pNext = 0; + pCur->pStem = pNew; + pNew = pX; + } + + /* Insert the new value */ + pNew->pNext = 0; + pX = pNew; + for(i=0; i<=pCur->mxQueue; i++){ + if( pCur->aQueue[i] ){ + pX = fuzzerMergeStems(pX, pCur->aQueue[i]); + pCur->aQueue[i] = 0; + }else{ + pCur->aQueue[i] = pX; + break; + } + } + if( i>pCur->mxQueue ){ + if( imxQueue = i; + pCur->aQueue[i] = pX; + }else{ + assert( pCur->mxQueue==FUZZER_NQUEUE-1 ); + pX = fuzzerMergeStems(pX, pCur->aQueue[FUZZER_NQUEUE-1]); + pCur->aQueue[FUZZER_NQUEUE-1] = pX; + } + } + + return fuzzerLowestCostStem(pCur); +} + +/* +** Allocate a new fuzzer_stem. Add it to the hash table but do not +** link it into either the pCur->pStem or pCur->pDone lists. +*/ +static fuzzer_stem *fuzzerNewStem( + fuzzer_cursor *pCur, + const char *zWord, + fuzzer_cost rBaseCost +){ + fuzzer_stem *pNew; + fuzzer_rule *pRule; + unsigned int h; + + pNew = sqlite3_malloc( sizeof(*pNew) + (int)strlen(zWord) + 1 ); + if( pNew==0 ) return 0; + memset(pNew, 0, sizeof(*pNew)); + pNew->zBasis = (char*)&pNew[1]; + pNew->nBasis = (int)strlen(zWord); + memcpy(pNew->zBasis, zWord, pNew->nBasis+1); + pRule = pCur->pVtab->pRule; + while( fuzzerSkipRule(pRule, pNew, pCur->iRuleset) ){ + pRule = pRule->pNext; + } + pNew->pRule = pRule; + pNew->n = -1; + pNew->rBaseCost = pNew->rCostX = rBaseCost; + h = fuzzerHash(pNew->zBasis); + pNew->pHash = pCur->apHash[h]; + pCur->apHash[h] = pNew; + pCur->nStem++; + return pNew; +} + + +/* +** Advance a cursor to its next row of output +*/ +static int fuzzerNext(sqlite3_vtab_cursor *cur){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + int rc; + fuzzer_stem *pStem, *pNew; + + pCur->iRowid++; + + /* Use the element the cursor is currently point to to create + ** a new stem and insert the new stem into the priority queue. + */ + pStem = pCur->pStem; + if( pStem->rCostX>0 ){ + rc = fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf); + if( rc==SQLITE_NOMEM ) return SQLITE_NOMEM; + pNew = fuzzerNewStem(pCur, pCur->zBuf, pStem->rCostX); + if( pNew ){ + if( fuzzerAdvance(pCur, pNew)==0 ){ + pNew->pNext = pCur->pDone; + pCur->pDone = pNew; + }else{ + if( fuzzerInsert(pCur, pNew)==pNew ){ + return SQLITE_OK; + } + } + }else{ + return SQLITE_NOMEM; + } + } + + /* Adjust the priority queue so that the first element of the + ** stem list is the next lowest cost word. + */ + while( (pStem = pCur->pStem)!=0 ){ + int res = fuzzerAdvance(pCur, pStem); + if( res<0 ){ + return SQLITE_NOMEM; + }else if( res>0 ){ + pCur->pStem = 0; + pStem = fuzzerInsert(pCur, pStem); + if( (rc = fuzzerSeen(pCur, pStem))!=0 ){ + if( rc<0 ) return SQLITE_NOMEM; + continue; + } + return SQLITE_OK; /* New word found */ + } + pCur->pStem = 0; + pStem->pNext = pCur->pDone; + pCur->pDone = pStem; + if( fuzzerLowestCostStem(pCur) ){ + rc = fuzzerSeen(pCur, pCur->pStem); + if( rc<0 ) return SQLITE_NOMEM; + if( rc==0 ){ + return SQLITE_OK; + } + } + } + + /* Reach this point only if queue has been exhausted and there is + ** nothing left to be output. */ + pCur->rLimit = (fuzzer_cost)0; + return SQLITE_OK; +} + +/* +** Called to "rewind" a cursor back to the beginning so that +** it starts its output over again. Always called at least once +** prior to any fuzzerColumn, fuzzerRowid, or fuzzerEof call. +*/ +static int fuzzerFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + fuzzer_cursor *pCur = (fuzzer_cursor *)pVtabCursor; + const char *zWord = ""; + fuzzer_stem *pStem; + int idx; + + fuzzerClearCursor(pCur, 1); + pCur->rLimit = 2147483647; + idx = 0; + if( idxNum & 1 ){ + zWord = (const char*)sqlite3_value_text(argv[0]); + idx++; + } + if( idxNum & 2 ){ + pCur->rLimit = (fuzzer_cost)sqlite3_value_int(argv[idx]); + idx++; + } + if( idxNum & 4 ){ + pCur->iRuleset = (fuzzer_cost)sqlite3_value_int(argv[idx]); + idx++; + } + pCur->nullRule.pNext = pCur->pVtab->pRule; + pCur->nullRule.rCost = 0; + pCur->nullRule.nFrom = 0; + pCur->nullRule.nTo = 0; + pCur->nullRule.zFrom = ""; + pCur->iRowid = 1; + assert( pCur->pStem==0 ); + + /* If the query term is longer than FUZZER_MX_OUTPUT_LENGTH bytes, this + ** query will return zero rows. */ + if( (int)strlen(zWord)pStem = pStem = fuzzerNewStem(pCur, zWord, (fuzzer_cost)0); + if( pStem==0 ) return SQLITE_NOMEM; + pStem->pRule = &pCur->nullRule; + pStem->n = pStem->nBasis; + }else{ + pCur->rLimit = 0; + } + + return SQLITE_OK; +} + +/* +** Only the word and distance columns have values. All other columns +** return NULL +*/ +static int fuzzerColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + if( i==0 ){ + /* the "word" column */ + if( fuzzerRender(pCur->pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ + return SQLITE_NOMEM; + } + sqlite3_result_text(ctx, pCur->zBuf, -1, SQLITE_TRANSIENT); + }else if( i==1 ){ + /* the "distance" column */ + sqlite3_result_int(ctx, pCur->pStem->rCostX); + }else{ + /* All other columns are NULL */ + sqlite3_result_null(ctx); + } + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int fuzzerRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + *pRowid = pCur->iRowid; + return SQLITE_OK; +} + +/* +** When the fuzzer_cursor.rLimit value is 0 or less, that is a signal +** that the cursor has nothing more to output. +*/ +static int fuzzerEof(sqlite3_vtab_cursor *cur){ + fuzzer_cursor *pCur = (fuzzer_cursor*)cur; + return pCur->rLimit<=(fuzzer_cost)0; +} + +/* +** Search for terms of these forms: +** +** (A) word MATCH $str +** (B1) distance < $value +** (B2) distance <= $value +** (C) ruleid == $ruleid +** +** The distance< and distance<= are both treated as distance<=. +** The query plan number is a bit vector: +** +** bit 1: Term of the form (A) found +** bit 2: Term like (B1) or (B2) found +** bit 3: Term like (C) found +** +** If bit-1 is set, $str is always in filter.argv[0]. If bit-2 is set +** then $value is in filter.argv[0] if bit-1 is clear and is in +** filter.argv[1] if bit-1 is set. If bit-3 is set, then $ruleid is +** in filter.argv[0] if bit-1 and bit-2 are both zero, is in +** filter.argv[1] if exactly one of bit-1 and bit-2 are set, and is in +** filter.argv[2] if both bit-1 and bit-2 are set. +*/ +static int fuzzerBestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ + int iPlan = 0; + int iDistTerm = -1; + int iRulesetTerm = -1; + int i; + const struct sqlite3_index_constraint *pConstraint; + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + if( (iPlan & 1)==0 + && pConstraint->iColumn==0 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + if( (iPlan & 2)==0 + && pConstraint->iColumn==1 + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) + ){ + iPlan |= 2; + iDistTerm = i; + } + if( (iPlan & 4)==0 + && pConstraint->iColumn==2 + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 4; + pIdxInfo->aConstraintUsage[i].omit = 1; + iRulesetTerm = i; + } + } + if( iPlan & 2 ){ + pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = 1+((iPlan&1)!=0); + } + if( iPlan & 4 ){ + int idx = 1; + if( iPlan & 1 ) idx++; + if( iPlan & 2 ) idx++; + pIdxInfo->aConstraintUsage[iRulesetTerm].argvIndex = idx; + } + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==1 + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + pIdxInfo->estimatedCost = (double)10000; + + return SQLITE_OK; +} + +/* +** A virtual table module that implements the "fuzzer". +*/ +static sqlite3_module fuzzerModule = { + 0, /* iVersion */ + fuzzerConnect, + fuzzerConnect, + fuzzerBestIndex, + fuzzerDisconnect, + fuzzerDisconnect, + fuzzerOpen, /* xOpen - open a cursor */ + fuzzerClose, /* xClose - close a cursor */ + fuzzerFilter, /* xFilter - configure scan constraints */ + fuzzerNext, /* xNext - advance a cursor */ + fuzzerEof, /* xEof - check for end of scan */ + fuzzerColumn, /* xColumn - read data */ + fuzzerRowid, /* xRowid - read data */ + 0, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ +}; + +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_fuzzer_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + rc = sqlite3_create_module(db, "fuzzer", &fuzzerModule, 0); + return rc; +} ADDED ext/misc/ieee754.c Index: ext/misc/ieee754.c ================================================================== --- ext/misc/ieee754.c +++ ext/misc/ieee754.c @@ -0,0 +1,131 @@ +/* +** 2013-04-17 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** This SQLite extension implements functions for the exact display +** and input of IEEE754 Binary64 floating-point numbers. +** +** ieee754(X) +** ieee754(Y,Z) +** +** In the first form, the value X should be a floating-point number. +** The function will return a string of the form 'ieee754(Y,Z)' where +** Y and Z are integers such that X==Y*pow(w.0,Z). +** +** In the second form, Y and Z are integers which are the mantissa and +** base-2 exponent of a new floating point number. The function returns +** a floating-point value equal to Y*pow(2.0,Z). +** +** Examples: +** +** ieee754(2.0) -> 'ieee754(2,0)' +** ieee754(45.25) -> 'ieee754(181,-2)' +** ieee754(2, 0) -> 2.0 +** ieee754(181, -2) -> 45.25 +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include +#include + +/* +** Implementation of the ieee754() function +*/ +static void ieee754func( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + if( argc==1 ){ + sqlite3_int64 m, a; + double r; + int e; + int isNeg; + char zResult[100]; + assert( sizeof(m)==sizeof(r) ); + if( sqlite3_value_type(argv[0])!=SQLITE_FLOAT ) return; + r = sqlite3_value_double(argv[0]); + if( r<0.0 ){ + isNeg = 1; + r = -r; + }else{ + isNeg = 0; + } + memcpy(&a,&r,sizeof(a)); + if( a==0 ){ + e = 0; + m = 0; + }else{ + e = a>>52; + m = a & ((((sqlite3_int64)1)<<52)-1); + m |= ((sqlite3_int64)1)<<52; + while( e<1075 && m>0 && (m&1)==0 ){ + m >>= 1; + e++; + } + if( isNeg ) m = -m; + } + sqlite3_snprintf(sizeof(zResult), zResult, "ieee754(%lld,%d)", + m, e-1075); + sqlite3_result_text(context, zResult, -1, SQLITE_TRANSIENT); + }else if( argc==2 ){ + sqlite3_int64 m, e, a; + double r; + int isNeg = 0; + m = sqlite3_value_int64(argv[0]); + e = sqlite3_value_int64(argv[1]); + if( m<0 ){ + isNeg = 1; + m = -m; + if( m<0 ) return; + }else if( m==0 && e>1000 && e<1000 ){ + sqlite3_result_double(context, 0.0); + return; + } + while( (m>>32)&0xffe00000 ){ + m >>= 1; + e++; + } + while( ((m>>32)&0xfff00000)==0 ){ + m <<= 1; + e--; + } + e += 1075; + if( e<0 ) e = m = 0; + if( e>0x7ff ) m = 0; + a = m & ((((sqlite3_int64)1)<<52)-1); + a |= e<<52; + if( isNeg ) a |= ((sqlite3_int64)1)<<63; + memcpy(&r, &a, sizeof(r)); + sqlite3_result_double(context, r); + } +} + + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_ieee_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + rc = sqlite3_create_function(db, "ieee754", 1, SQLITE_UTF8, 0, + ieee754func, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "ieee754", 2, SQLITE_UTF8, 0, + ieee754func, 0, 0); + } + return rc; +} ADDED ext/misc/nextchar.c Index: ext/misc/nextchar.c ================================================================== --- ext/misc/nextchar.c +++ ext/misc/nextchar.c @@ -0,0 +1,265 @@ +/* +** 2013-02-28 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** This file contains code to implement the next_char(A,T,F,W) SQL function. +** +** The next_char(A,T,F,H) function finds all valid "next" characters for +** string A given the vocabulary in T.F. The T.F field should be indexed. +** If the W value exists and is a non-empty string, then it is an SQL +** expression that limits the entries in T.F that will be considered. +** +** For example, suppose an application has a dictionary like this: +** +** CREATE TABLE dictionary(word TEXT UNIQUE); +** +** Further suppose that for user keypad entry, it is desired to disable +** (gray out) keys that are not valid as the next character. If the +** the user has previously entered (say) 'cha' then to find all allowed +** next characters (and thereby determine when keys should not be grayed +** out) run the following query: +** +** SELECT next_char('cha','dictionary','word'); +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 +#include + +/* +** A structure to hold context of the next_char() computation across +** nested function calls. +*/ +typedef struct nextCharContext nextCharContext; +struct nextCharContext { + sqlite3 *db; /* Database connection */ + sqlite3_stmt *pStmt; /* Prepared statement used to query */ + const unsigned char *zPrefix; /* Prefix to scan */ + int nPrefix; /* Size of zPrefix in bytes */ + int nAlloc; /* Space allocated to aResult */ + int nUsed; /* Space used in aResult */ + unsigned int *aResult; /* Array of next characters */ + int mallocFailed; /* True if malloc fails */ + int otherError; /* True for any other failure */ +}; + +/* +** Append a result character if the character is not already in the +** result. +*/ +static void nextCharAppend(nextCharContext *p, unsigned c){ + int i; + for(i=0; inUsed; i++){ + if( p->aResult[i]==c ) return; + } + if( p->nUsed+1 > p->nAlloc ){ + unsigned int *aNew; + int n = p->nAlloc*2 + 30; + aNew = sqlite3_realloc(p->aResult, n*sizeof(unsigned int)); + if( aNew==0 ){ + p->mallocFailed = 1; + return; + }else{ + p->aResult = aNew; + p->nAlloc = n; + } + } + p->aResult[p->nUsed++] = c; +} + +/* +** Write a character into z[] as UTF8. Return the number of bytes needed +** to hold the character +*/ +static int writeUtf8(unsigned char *z, unsigned c){ + if( c<0x00080 ){ + z[0] = (unsigned char)(c&0xff); + return 1; + } + if( c<0x00800 ){ + z[0] = 0xC0 + (unsigned char)((c>>6)&0x1F); + z[1] = 0x80 + (unsigned char)(c & 0x3F); + return 2; + } + if( c<0x10000 ){ + z[0] = 0xE0 + (unsigned char)((c>>12)&0x0F); + z[1] = 0x80 + (unsigned char)((c>>6) & 0x3F); + z[2] = 0x80 + (unsigned char)(c & 0x3F); + return 3; + } + z[0] = 0xF0 + (unsigned char)((c>>18) & 0x07); + z[1] = 0x80 + (unsigned char)((c>>12) & 0x3F); + z[2] = 0x80 + (unsigned char)((c>>6) & 0x3F); + z[3] = 0x80 + (unsigned char)(c & 0x3F); + return 4; +} + +/* +** Read a UTF8 character out of z[] and write it into *pOut. Return +** the number of bytes in z[] that were used to construct the character. +*/ +static int readUtf8(const unsigned char *z, unsigned *pOut){ + static const unsigned char validBits[] = { + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, + 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00, + }; + unsigned c = z[0]; + if( c<0xc0 ){ + *pOut = c; + return 1; + }else{ + int n = 1; + c = validBits[c-0xc0]; + while( (z[n] & 0xc0)==0x80 ){ + c = (c<<6) + (0x3f & z[n++]); + } + if( c<0x80 || (c&0xFFFFF800)==0xD800 || (c&0xFFFFFFFE)==0xFFFE ){ + c = 0xFFFD; + } + *pOut = c; + return n; + } +} + +/* +** The nextCharContext structure has been set up. Add all "next" characters +** to the result set. +*/ +static void findNextChars(nextCharContext *p){ + unsigned cPrev = 0; + unsigned char zPrev[8]; + int n, rc; + + for(;;){ + sqlite3_bind_text(p->pStmt, 1, (char*)p->zPrefix, p->nPrefix, + SQLITE_STATIC); + n = writeUtf8(zPrev, cPrev+1); + sqlite3_bind_text(p->pStmt, 2, (char*)zPrev, n, SQLITE_STATIC); + rc = sqlite3_step(p->pStmt); + if( rc==SQLITE_DONE ){ + sqlite3_reset(p->pStmt); + return; + }else if( rc!=SQLITE_ROW ){ + p->otherError = rc; + return; + }else{ + const unsigned char *zOut = sqlite3_column_text(p->pStmt, 0); + unsigned cNext; + n = readUtf8(zOut+p->nPrefix, &cNext); + sqlite3_reset(p->pStmt); + nextCharAppend(p, cNext); + cPrev = cNext; + if( p->mallocFailed ) return; + } + } +} + + +/* +** next_character(A,T,F,W) +** +** Return a string composted of all next possible characters after +** A for elements of T.F. If W is supplied, then it is an SQL expression +** that limits the elements in T.F that are considered. +*/ +static void nextCharFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + nextCharContext c; + const unsigned char *zTable = sqlite3_value_text(argv[1]); + const unsigned char *zField = sqlite3_value_text(argv[2]); + const unsigned char *zWhere; + char *zSql; + int rc; + + memset(&c, 0, sizeof(c)); + c.db = sqlite3_context_db_handle(context); + c.zPrefix = sqlite3_value_text(argv[0]); + c.nPrefix = sqlite3_value_bytes(argv[0]); + if( zTable==0 || zField==0 || c.zPrefix==0 ) return; + if( argc<4 + || (zWhere = sqlite3_value_text(argv[3]))==0 + || zWhere[0]==0 + ){ + zSql = sqlite3_mprintf( + "SELECT \"%w\" FROM \"%w\"" + " WHERE \"%w\">=(?1 || ?2)" + " AND \"%w\"<=(?1 || char(1114111))" /* 1114111 == 0x10ffff */ + " ORDER BY 1 ASC LIMIT 1", + zField, zTable, zField, zField); + }else{ + zSql = sqlite3_mprintf( + "SELECT \"%w\" FROM \"%w\"" + " WHERE \"%w\">=(?1 || ?2)" + " AND \"%w\"<=(?1 || char(1114111))" /* 1114111 == 0x10ffff */ + " AND (%s)" + " ORDER BY 1 ASC LIMIT 1", + zField, zTable, zField, zField, zWhere); + } + if( zSql==0 ){ + sqlite3_result_error_nomem(context); + return; + } + + rc = sqlite3_prepare_v2(c.db, zSql, -1, &c.pStmt, 0); + sqlite3_free(zSql); + if( rc ){ + sqlite3_result_error(context, sqlite3_errmsg(c.db), -1); + return; + } + findNextChars(&c); + if( c.mallocFailed ){ + sqlite3_result_error_nomem(context); + }else{ + unsigned char *pRes; + pRes = sqlite3_malloc( c.nUsed*4 + 1 ); + if( pRes==0 ){ + sqlite3_result_error_nomem(context); + }else{ + int i; + int n = 0; + for(i=0; i +#include +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 + +/* +** The following #defines change the names of some functions implemented in +** this file to prevent name collisions with C-library functions of the +** same name. +*/ +#define re_match sqlite3re_match +#define re_compile sqlite3re_compile +#define re_free sqlite3re_free + +/* The end-of-input character */ +#define RE_EOF 0 /* End of input */ + +/* The NFA is implemented as sequence of opcodes taken from the following +** set. Each opcode has a single integer argument. +*/ +#define RE_OP_MATCH 1 /* Match the one character in the argument */ +#define RE_OP_ANY 2 /* Match any one character. (Implements ".") */ +#define RE_OP_ANYSTAR 3 /* Special optimized version of .* */ +#define RE_OP_FORK 4 /* Continue to both next and opcode at iArg */ +#define RE_OP_GOTO 5 /* Jump to opcode at iArg */ +#define RE_OP_ACCEPT 6 /* Halt and indicate a successful match */ +#define RE_OP_CC_INC 7 /* Beginning of a [...] character class */ +#define RE_OP_CC_EXC 8 /* Beginning of a [^...] character class */ +#define RE_OP_CC_VALUE 9 /* Single value in a character class */ +#define RE_OP_CC_RANGE 10 /* Range of values in a character class */ +#define RE_OP_WORD 11 /* Perl word character [A-Za-z0-9_] */ +#define RE_OP_NOTWORD 12 /* Not a perl word character */ +#define RE_OP_DIGIT 13 /* digit: [0-9] */ +#define RE_OP_NOTDIGIT 14 /* Not a digit */ +#define RE_OP_SPACE 15 /* space: [ \t\n\r\v\f] */ +#define RE_OP_NOTSPACE 16 /* Not a digit */ +#define RE_OP_BOUNDARY 17 /* Boundary between word and non-word */ + +/* Each opcode is a "state" in the NFA */ +typedef unsigned short ReStateNumber; + +/* Because this is an NFA and not a DFA, multiple states can be active at +** once. An instance of the following object records all active states in +** the NFA. The implementation is optimized for the common case where the +** number of actives states is small. +*/ +typedef struct ReStateSet { + unsigned nState; /* Number of current states */ + ReStateNumber *aState; /* Current states */ +} ReStateSet; + +/* An input string read one character at a time. +*/ +typedef struct ReInput ReInput; +struct ReInput { + const unsigned char *z; /* All text */ + int i; /* Next byte to read */ + int mx; /* EOF when i>=mx */ +}; + +/* A compiled NFA (or an NFA that is in the process of being compiled) is +** an instance of the following object. +*/ +typedef struct ReCompiled ReCompiled; +struct ReCompiled { + ReInput sIn; /* Regular expression text */ + const char *zErr; /* Error message to return */ + char *aOp; /* Operators for the virtual machine */ + int *aArg; /* Arguments to each operator */ + unsigned (*xNextChar)(ReInput*); /* Next character function */ + unsigned char zInit[12]; /* Initial text to match */ + int nInit; /* Number of characters in zInit */ + unsigned nState; /* Number of entries in aOp[] and aArg[] */ + unsigned nAlloc; /* Slots allocated for aOp[] and aArg[] */ +}; + +/* Add a state to the given state set if it is not already there */ +static void re_add_state(ReStateSet *pSet, int newState){ + unsigned i; + for(i=0; inState; i++) if( pSet->aState[i]==newState ) return; + pSet->aState[pSet->nState++] = newState; +} + +/* Extract the next unicode character from *pzIn and return it. Advance +** *pzIn to the first byte past the end of the character returned. To +** be clear: this routine converts utf8 to unicode. This routine is +** optimized for the common case where the next character is a single byte. +*/ +static unsigned re_next_char(ReInput *p){ + unsigned c; + if( p->i>=p->mx ) return 0; + c = p->z[p->i++]; + if( c>=0x80 ){ + if( (c&0xe0)==0xc0 && p->imx && (p->z[p->i]&0xc0)==0x80 ){ + c = (c&0x1f)<<6 | (p->z[p->i++]&0x3f); + if( c<0x80 ) c = 0xfffd; + }else if( (c&0xf0)==0xe0 && p->i+1mx && (p->z[p->i]&0xc0)==0x80 + && (p->z[p->i+1]&0xc0)==0x80 ){ + c = (c&0x0f)<<12 | ((p->z[p->i]&0x3f)<<6) | (p->z[p->i+1]&0x3f); + p->i += 2; + if( c<=0x3ff || (c>=0xd800 && c<=0xdfff) ) c = 0xfffd; + }else if( (c&0xf8)==0xf0 && p->i+3mx && (p->z[p->i]&0xc0)==0x80 + && (p->z[p->i+1]&0xc0)==0x80 && (p->z[p->i+2]&0xc0)==0x80 ){ + c = (c&0x07)<<18 | ((p->z[p->i]&0x3f)<<12) | ((p->z[p->i+1]&0x3f)<<6) + | (p->z[p->i+2]&0x3f); + p->i += 3; + if( c<=0xffff || c>0x10ffff ) c = 0xfffd; + }else{ + c = 0xfffd; + } + } + return c; +} +static unsigned re_next_char_nocase(ReInput *p){ + unsigned c = re_next_char(p); + if( c>='A' && c<='Z' ) c += 'a' - 'A'; + return c; +} + +/* Return true if c is a perl "word" character: [A-Za-z0-9_] */ +static int re_word_char(int c){ + return (c>='0' && c<='9') || (c>='a' && c<='z') + || (c>='A' && c<='Z') || c=='_'; +} + +/* Return true if c is a "digit" character: [0-9] */ +static int re_digit_char(int c){ + return (c>='0' && c<='9'); +} + +/* Return true if c is a perl "space" character: [ \t\r\n\v\f] */ +static int re_space_char(int c){ + return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f'; +} + +/* Run a compiled regular expression on the zero-terminated input +** string zIn[]. Return true on a match and false if there is no match. +*/ +static int re_match(ReCompiled *pRe, const unsigned char *zIn, int nIn){ + ReStateSet aStateSet[2], *pThis, *pNext; + ReStateNumber aSpace[100]; + ReStateNumber *pToFree; + unsigned int i = 0; + unsigned int iSwap = 0; + int c = RE_EOF+1; + int cPrev = 0; + int rc = 0; + ReInput in; + + in.z = zIn; + in.i = 0; + in.mx = nIn>=0 ? nIn : (int)strlen((char const*)zIn); + + /* Look for the initial prefix match, if there is one. */ + if( pRe->nInit ){ + unsigned char x = pRe->zInit[0]; + while( in.i+pRe->nInit<=in.mx + && (zIn[in.i]!=x || + strncmp((const char*)zIn+in.i, (const char*)pRe->zInit, pRe->nInit)!=0) + ){ + in.i++; + } + if( in.i+pRe->nInit>in.mx ) return 0; + } + + if( pRe->nState<=(sizeof(aSpace)/(sizeof(aSpace[0])*2)) ){ + pToFree = 0; + aStateSet[0].aState = aSpace; + }else{ + pToFree = sqlite3_malloc( sizeof(ReStateNumber)*2*pRe->nState ); + if( pToFree==0 ) return -1; + aStateSet[0].aState = pToFree; + } + aStateSet[1].aState = &aStateSet[0].aState[pRe->nState]; + pNext = &aStateSet[1]; + pNext->nState = 0; + re_add_state(pNext, 0); + while( c!=RE_EOF && pNext->nState>0 ){ + cPrev = c; + c = pRe->xNextChar(&in); + pThis = pNext; + pNext = &aStateSet[iSwap]; + iSwap = 1 - iSwap; + pNext->nState = 0; + for(i=0; inState; i++){ + int x = pThis->aState[i]; + switch( pRe->aOp[x] ){ + case RE_OP_MATCH: { + if( pRe->aArg[x]==c ) re_add_state(pNext, x+1); + break; + } + case RE_OP_ANY: { + re_add_state(pNext, x+1); + break; + } + case RE_OP_WORD: { + if( re_word_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_NOTWORD: { + if( !re_word_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_DIGIT: { + if( re_digit_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_NOTDIGIT: { + if( !re_digit_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_SPACE: { + if( re_space_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_NOTSPACE: { + if( !re_space_char(c) ) re_add_state(pNext, x+1); + break; + } + case RE_OP_BOUNDARY: { + if( re_word_char(c)!=re_word_char(cPrev) ) re_add_state(pThis, x+1); + break; + } + case RE_OP_ANYSTAR: { + re_add_state(pNext, x); + re_add_state(pThis, x+1); + break; + } + case RE_OP_FORK: { + re_add_state(pThis, x+pRe->aArg[x]); + re_add_state(pThis, x+1); + break; + } + case RE_OP_GOTO: { + re_add_state(pThis, x+pRe->aArg[x]); + break; + } + case RE_OP_ACCEPT: { + rc = 1; + goto re_match_end; + } + case RE_OP_CC_INC: + case RE_OP_CC_EXC: { + int j = 1; + int n = pRe->aArg[x]; + int hit = 0; + for(j=1; j>0 && jaOp[x+j]==RE_OP_CC_VALUE ){ + if( pRe->aArg[x+j]==c ){ + hit = 1; + j = -1; + } + }else{ + if( pRe->aArg[x+j]<=c && pRe->aArg[x+j+1]>=c ){ + hit = 1; + j = -1; + }else{ + j++; + } + } + } + if( pRe->aOp[x]==RE_OP_CC_EXC ) hit = !hit; + if( hit ) re_add_state(pNext, x+n); + break; + } + } + } + } + for(i=0; inState; i++){ + if( pRe->aOp[pNext->aState[i]]==RE_OP_ACCEPT ){ rc = 1; break; } + } +re_match_end: + sqlite3_free(pToFree); + return rc; +} + +/* Resize the opcode and argument arrays for an RE under construction. +*/ +static int re_resize(ReCompiled *p, int N){ + char *aOp; + int *aArg; + aOp = sqlite3_realloc(p->aOp, N*sizeof(p->aOp[0])); + if( aOp==0 ) return 1; + p->aOp = aOp; + aArg = sqlite3_realloc(p->aArg, N*sizeof(p->aArg[0])); + if( aArg==0 ) return 1; + p->aArg = aArg; + p->nAlloc = N; + return 0; +} + +/* Insert a new opcode and argument into an RE under construction. The +** insertion point is just prior to existing opcode iBefore. +*/ +static int re_insert(ReCompiled *p, int iBefore, int op, int arg){ + int i; + if( p->nAlloc<=p->nState && re_resize(p, p->nAlloc*2) ) return 0; + for(i=p->nState; i>iBefore; i--){ + p->aOp[i] = p->aOp[i-1]; + p->aArg[i] = p->aArg[i-1]; + } + p->nState++; + p->aOp[iBefore] = op; + p->aArg[iBefore] = arg; + return iBefore; +} + +/* Append a new opcode and argument to the end of the RE under construction. +*/ +static int re_append(ReCompiled *p, int op, int arg){ + return re_insert(p, p->nState, op, arg); +} + +/* Make a copy of N opcodes starting at iStart onto the end of the RE +** under construction. +*/ +static void re_copy(ReCompiled *p, int iStart, int N){ + if( p->nState+N>=p->nAlloc && re_resize(p, p->nAlloc*2+N) ) return; + memcpy(&p->aOp[p->nState], &p->aOp[iStart], N*sizeof(p->aOp[0])); + memcpy(&p->aArg[p->nState], &p->aArg[iStart], N*sizeof(p->aArg[0])); + p->nState += N; +} + +/* Return true if c is a hexadecimal digit character: [0-9a-fA-F] +** If c is a hex digit, also set *pV = (*pV)*16 + valueof(c). If +** c is not a hex digit *pV is unchanged. +*/ +static int re_hex(int c, int *pV){ + if( c>='0' && c<='9' ){ + c -= '0'; + }else if( c>='a' && c<='f' ){ + c -= 'a' - 10; + }else if( c>='A' && c<='F' ){ + c -= 'A' - 10; + }else{ + return 0; + } + *pV = (*pV)*16 + (c & 0xff); + return 1; +} + +/* A backslash character has been seen, read the next character and +** return its interpretation. +*/ +static unsigned re_esc_char(ReCompiled *p){ + static const char zEsc[] = "afnrtv\\()*.+?[$^{|}]"; + static const char zTrans[] = "\a\f\n\r\t\v"; + int i, v = 0; + char c; + if( p->sIn.i>=p->sIn.mx ) return 0; + c = p->sIn.z[p->sIn.i]; + if( c=='u' && p->sIn.i+4sIn.mx ){ + const unsigned char *zIn = p->sIn.z + p->sIn.i; + if( re_hex(zIn[1],&v) + && re_hex(zIn[2],&v) + && re_hex(zIn[3],&v) + && re_hex(zIn[4],&v) + ){ + p->sIn.i += 5; + return v; + } + } + if( c=='x' && p->sIn.i+2sIn.mx ){ + const unsigned char *zIn = p->sIn.z + p->sIn.i; + if( re_hex(zIn[1],&v) + && re_hex(zIn[2],&v) + ){ + p->sIn.i += 3; + return v; + } + } + for(i=0; zEsc[i] && zEsc[i]!=c; i++){} + if( zEsc[i] ){ + if( i<6 ) c = zTrans[i]; + p->sIn.i++; + }else{ + p->zErr = "unknown \\ escape"; + } + return c; +} + +/* Forward declaration */ +static const char *re_subcompile_string(ReCompiled*); + +/* Peek at the next byte of input */ +static unsigned char rePeek(ReCompiled *p){ + return p->sIn.isIn.mx ? p->sIn.z[p->sIn.i] : 0; +} + +/* Compile RE text into a sequence of opcodes. Continue up to the +** first unmatched ")" character, then return. If an error is found, +** return a pointer to the error message string. +*/ +static const char *re_subcompile_re(ReCompiled *p){ + const char *zErr; + int iStart, iEnd, iGoto; + iStart = p->nState; + zErr = re_subcompile_string(p); + if( zErr ) return zErr; + while( rePeek(p)=='|' ){ + iEnd = p->nState; + re_insert(p, iStart, RE_OP_FORK, iEnd + 2 - iStart); + iGoto = re_append(p, RE_OP_GOTO, 0); + p->sIn.i++; + zErr = re_subcompile_string(p); + if( zErr ) return zErr; + p->aArg[iGoto] = p->nState - iGoto; + } + return 0; +} + +/* Compile an element of regular expression text (anything that can be +** an operand to the "|" operator). Return NULL on success or a pointer +** to the error message if there is a problem. +*/ +static const char *re_subcompile_string(ReCompiled *p){ + int iPrev = -1; + int iStart; + unsigned c; + const char *zErr; + while( (c = p->xNextChar(&p->sIn))!=0 ){ + iStart = p->nState; + switch( c ){ + case '|': + case '$': + case ')': { + p->sIn.i--; + return 0; + } + case '(': { + zErr = re_subcompile_re(p); + if( zErr ) return zErr; + if( rePeek(p)!=')' ) return "unmatched '('"; + p->sIn.i++; + break; + } + case '.': { + if( rePeek(p)=='*' ){ + re_append(p, RE_OP_ANYSTAR, 0); + p->sIn.i++; + }else{ + re_append(p, RE_OP_ANY, 0); + } + break; + } + case '*': { + if( iPrev<0 ) return "'*' without operand"; + re_insert(p, iPrev, RE_OP_GOTO, p->nState - iPrev + 1); + re_append(p, RE_OP_FORK, iPrev - p->nState + 1); + break; + } + case '+': { + if( iPrev<0 ) return "'+' without operand"; + re_append(p, RE_OP_FORK, iPrev - p->nState); + break; + } + case '?': { + if( iPrev<0 ) return "'?' without operand"; + re_insert(p, iPrev, RE_OP_FORK, p->nState - iPrev+1); + break; + } + case '{': { + int m = 0, n = 0; + int sz, j; + if( iPrev<0 ) return "'{m,n}' without operand"; + while( (c=rePeek(p))>='0' && c<='9' ){ m = m*10 + c - '0'; p->sIn.i++; } + n = m; + if( c==',' ){ + p->sIn.i++; + n = 0; + while( (c=rePeek(p))>='0' && c<='9' ){ n = n*10 + c-'0'; p->sIn.i++; } + } + if( c!='}' ) return "unmatched '{'"; + if( n>0 && nsIn.i++; + sz = p->nState - iPrev; + if( m==0 ){ + if( n==0 ) return "both m and n are zero in '{m,n}'"; + re_insert(p, iPrev, RE_OP_FORK, sz+1); + n--; + }else{ + for(j=1; j0 ){ + re_append(p, RE_OP_FORK, -sz); + } + break; + } + case '[': { + int iFirst = p->nState; + if( rePeek(p)=='^' ){ + re_append(p, RE_OP_CC_EXC, 0); + p->sIn.i++; + }else{ + re_append(p, RE_OP_CC_INC, 0); + } + while( (c = p->xNextChar(&p->sIn))!=0 ){ + if( c=='[' && rePeek(p)==':' ){ + return "POSIX character classes not supported"; + } + if( c=='\\' ) c = re_esc_char(p); + if( rePeek(p)=='-' ){ + re_append(p, RE_OP_CC_RANGE, c); + p->sIn.i++; + c = p->xNextChar(&p->sIn); + if( c=='\\' ) c = re_esc_char(p); + re_append(p, RE_OP_CC_RANGE, c); + }else{ + re_append(p, RE_OP_CC_VALUE, c); + } + if( rePeek(p)==']' ){ p->sIn.i++; break; } + } + if( c==0 ) return "unclosed '['"; + p->aArg[iFirst] = p->nState - iFirst; + break; + } + case '\\': { + int specialOp = 0; + switch( rePeek(p) ){ + case 'b': specialOp = RE_OP_BOUNDARY; break; + case 'd': specialOp = RE_OP_DIGIT; break; + case 'D': specialOp = RE_OP_NOTDIGIT; break; + case 's': specialOp = RE_OP_SPACE; break; + case 'S': specialOp = RE_OP_NOTSPACE; break; + case 'w': specialOp = RE_OP_WORD; break; + case 'W': specialOp = RE_OP_NOTWORD; break; + } + if( specialOp ){ + p->sIn.i++; + re_append(p, specialOp, 0); + }else{ + c = re_esc_char(p); + re_append(p, RE_OP_MATCH, c); + } + break; + } + default: { + re_append(p, RE_OP_MATCH, c); + break; + } + } + iPrev = iStart; + } + return 0; +} + +/* Free and reclaim all the memory used by a previously compiled +** regular expression. Applications should invoke this routine once +** for every call to re_compile() to avoid memory leaks. +*/ +void re_free(ReCompiled *pRe){ + if( pRe ){ + sqlite3_free(pRe->aOp); + sqlite3_free(pRe->aArg); + sqlite3_free(pRe); + } +} + +/* +** Compile a textual regular expression in zIn[] into a compiled regular +** expression suitable for us by re_match() and return a pointer to the +** compiled regular expression in *ppRe. Return NULL on success or an +** error message if something goes wrong. +*/ +const char *re_compile(ReCompiled **ppRe, const char *zIn, int noCase){ + ReCompiled *pRe; + const char *zErr; + int i, j; + + *ppRe = 0; + pRe = sqlite3_malloc( sizeof(*pRe) ); + if( pRe==0 ){ + return "out of memory"; + } + memset(pRe, 0, sizeof(*pRe)); + pRe->xNextChar = noCase ? re_next_char_nocase : re_next_char; + if( re_resize(pRe, 30) ){ + re_free(pRe); + return "out of memory"; + } + if( zIn[0]=='^' ){ + zIn++; + }else{ + re_append(pRe, RE_OP_ANYSTAR, 0); + } + pRe->sIn.z = (unsigned char*)zIn; + pRe->sIn.i = 0; + pRe->sIn.mx = (int)strlen(zIn); + zErr = re_subcompile_re(pRe); + if( zErr ){ + re_free(pRe); + return zErr; + } + if( rePeek(pRe)=='$' && pRe->sIn.i+1>=pRe->sIn.mx ){ + re_append(pRe, RE_OP_MATCH, RE_EOF); + re_append(pRe, RE_OP_ACCEPT, 0); + *ppRe = pRe; + }else if( pRe->sIn.i>=pRe->sIn.mx ){ + re_append(pRe, RE_OP_ACCEPT, 0); + *ppRe = pRe; + }else{ + re_free(pRe); + return "unrecognized character"; + } + + /* The following is a performance optimization. If the regex begins with + ** ".*" (if the input regex lacks an initial "^") and afterwards there are + ** one or more matching characters, enter those matching characters into + ** zInit[]. The re_match() routine can then search ahead in the input + ** string looking for the initial match without having to run the whole + ** regex engine over the string. Do not worry able trying to match + ** unicode characters beyond plane 0 - those are very rare and this is + ** just an optimization. */ + if( pRe->aOp[0]==RE_OP_ANYSTAR ){ + for(j=0, i=1; jzInit)-2 && pRe->aOp[i]==RE_OP_MATCH; i++){ + unsigned x = pRe->aArg[i]; + if( x<=127 ){ + pRe->zInit[j++] = x; + }else if( x<=0xfff ){ + pRe->zInit[j++] = 0xc0 | (x>>6); + pRe->zInit[j++] = 0x80 | (x&0x3f); + }else if( x<=0xffff ){ + pRe->zInit[j++] = 0xd0 | (x>>12); + pRe->zInit[j++] = 0x80 | ((x>>6)&0x3f); + pRe->zInit[j++] = 0x80 | (x&0x3f); + }else{ + break; + } + } + if( j>0 && pRe->zInit[j-1]==0 ) j--; + pRe->nInit = j; + } + return pRe->zErr; +} + +/* +** Implementation of the regexp() SQL function. This function implements +** the build-in REGEXP operator. The first argument to the function is the +** pattern and the second argument is the string. So, the SQL statements: +** +** A REGEXP B +** +** is implemented as regexp(B,A). +*/ +static void re_sql_func( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + ReCompiled *pRe; /* Compiled regular expression */ + const char *zPattern; /* The regular expression */ + const unsigned char *zStr;/* String being searched */ + const char *zErr; /* Compile error message */ + + pRe = sqlite3_get_auxdata(context, 0); + if( pRe==0 ){ + zPattern = (const char*)sqlite3_value_text(argv[0]); + if( zPattern==0 ) return; + zErr = re_compile(&pRe, zPattern, 0); + if( zErr ){ + re_free(pRe); + sqlite3_result_error(context, zErr, -1); + return; + } + if( pRe==0 ){ + sqlite3_result_error_nomem(context); + return; + } + sqlite3_set_auxdata(context, 0, pRe, (void(*)(void*))re_free); + } + zStr = (const unsigned char*)sqlite3_value_text(argv[1]); + if( zStr!=0 ){ + sqlite3_result_int(context, re_match(pRe, zStr, -1)); + } +} + +/* +** Invoke this routine to register the regexp() function with the +** SQLite database connection. +*/ +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_regexp_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); + rc = sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, 0, + re_sql_func, 0, 0); + return rc; +} ADDED ext/misc/spellfix.c Index: ext/misc/spellfix.c ================================================================== --- ext/misc/spellfix.c +++ ext/misc/spellfix.c @@ -0,0 +1,2837 @@ +/* +** 2012 April 10 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** This module implements the spellfix1 VIRTUAL TABLE that can be used +** to search a large vocabulary for close matches. See separate +** documentation (http://www.sqlite.org/spellfix1.html) for details. +*/ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 + +#ifndef SQLITE_AMALGAMATION +# include +# include +# include +# include +# define ALWAYS(X) 1 +# define NEVER(X) 0 + typedef unsigned char u8; + typedef unsigned short u16; +# include +#endif + +/* +** Character classes for ASCII characters: +** +** 0 '' Silent letters: H W +** 1 'A' Any vowel: A E I O U (Y) +** 2 'B' A bilabeal stop or fricative: B F P V W +** 3 'C' Other fricatives or back stops: C G J K Q S X Z +** 4 'D' Alveolar stops: D T +** 5 'H' Letter H at the beginning of a word +** 6 'L' Glide: L +** 7 'R' Semivowel: R +** 8 'M' Nasals: M N +** 9 'Y' Letter Y at the beginning of a word. +** 10 '9' Digits: 0 1 2 3 4 5 6 7 8 9 +** 11 ' ' White space +** 12 '?' Other. +*/ +#define CCLASS_SILENT 0 +#define CCLASS_VOWEL 1 +#define CCLASS_B 2 +#define CCLASS_C 3 +#define CCLASS_D 4 +#define CCLASS_H 5 +#define CCLASS_L 6 +#define CCLASS_R 7 +#define CCLASS_M 8 +#define CCLASS_Y 9 +#define CCLASS_DIGIT 10 +#define CCLASS_SPACE 11 +#define CCLASS_OTHER 12 + +/* +** The following table gives the character class for non-initial ASCII +** characters. +*/ +static const unsigned char midClass[] = { + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, + /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, + /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, + /* ' */ CCLASS_SILENT, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, + /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, + /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, + /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, + /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, + /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, + /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, + /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, + /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, + /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, + /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, + /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, + /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, + /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, + /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, + /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, + /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_VOWEL, + /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, + /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, + /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, + /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, + /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, + /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, + /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, + /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, + /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, + /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, + /* x */ CCLASS_C, /* y */ CCLASS_VOWEL, /* z */ CCLASS_C, + /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, + /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, +}; +/* +** This tables gives the character class for ASCII characters that form the +** initial character of a word. The only difference from midClass is with +** the letters H, W, and Y. +*/ +static const unsigned char initClass[] = { + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, + /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, + /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, + /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, + /* ' */ CCLASS_OTHER, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, + /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, + /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, + /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, + /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, + /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, + /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, + /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, + /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, + /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, + /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, + /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, + /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, + /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, + /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, + /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, + /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_Y, + /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, + /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, + /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, + /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, + /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, + /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, + /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, + /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, + /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, + /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, + /* x */ CCLASS_C, /* y */ CCLASS_Y, /* z */ CCLASS_C, + /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, + /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, +}; + +/* +** Mapping from the character class number (0-13) to a symbol for each +** character class. Note that initClass[] can be used to map the class +** symbol back into the class number. +*/ +static const unsigned char className[] = ".ABCDHLRMY9 ?"; + +/* +** Generate a "phonetic hash" from a string of ASCII characters +** in zIn[0..nIn-1]. +** +** * Map characters by character class as defined above. +** * Omit double-letters +** * Omit vowels beside R and L +** * Omit T when followed by CH +** * Omit W when followed by R +** * Omit D when followed by J or G +** * Omit K in KN or G in GN at the beginning of a word +** +** Space to hold the result is obtained from sqlite3_malloc() +** +** Return NULL if memory allocation fails. +*/ +static unsigned char *phoneticHash(const unsigned char *zIn, int nIn){ + unsigned char *zOut = sqlite3_malloc( nIn + 1 ); + int i; + int nOut = 0; + char cPrev = 0x77; + char cPrevX = 0x77; + const unsigned char *aClass = initClass; + + if( zOut==0 ) return 0; + if( nIn>2 ){ + switch( zIn[0] ){ + case 'g': + case 'k': { + if( zIn[1]=='n' ){ zIn++; nIn--; } + break; + } + } + } + for(i=0; i=0 ); + if( nOut==0 || c!=zOut[nOut-1] ) zOut[nOut++] = c; + } + zOut[nOut] = 0; + return zOut; +} + +/* +** This is an SQL function wrapper around phoneticHash(). See +** the description of phoneticHash() for additional information. +*/ +static void phoneticHashSqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn; + unsigned char *zOut; + + zIn = sqlite3_value_text(argv[0]); + if( zIn==0 ) return; + zOut = phoneticHash(zIn, sqlite3_value_bytes(argv[0])); + if( zOut==0 ){ + sqlite3_result_error_nomem(context); + }else{ + sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); + } +} + +/* +** Return the character class number for a character given its +** context. +*/ +static char characterClass(char cPrev, char c){ + return cPrev==0 ? initClass[c&0x7f] : midClass[c&0x7f]; +} + +/* +** Return the cost of inserting or deleting character c immediately +** following character cPrev. If cPrev==0, that means c is the first +** character of the word. +*/ +static int insertOrDeleteCost(char cPrev, char c, char cNext){ + char classC = characterClass(cPrev, c); + char classCprev; + + if( classC==CCLASS_SILENT ){ + /* Insert or delete "silent" characters such as H or W */ + return 1; + } + if( cPrev==c ){ + /* Repeated characters, or miss a repeat */ + return 10; + } + if( classC==CCLASS_VOWEL && (cPrev=='r' || cNext=='r') ){ + return 20; /* Insert a vowel before or after 'r' */ + } + classCprev = characterClass(cPrev, cPrev); + if( classC==classCprev ){ + if( classC==CCLASS_VOWEL ){ + /* Remove or add a new vowel to a vowel cluster */ + return 15; + }else{ + /* Remove or add a consonant not in the same class */ + return 50; + } + } + + /* any other character insertion or deletion */ + return 100; +} + +/* +** Divide the insertion cost by this factor when appending to the +** end of the word. +*/ +#define FINAL_INS_COST_DIV 4 + +/* +** Return the cost of substituting cTo in place of cFrom assuming +** the previous character is cPrev. If cPrev==0 then cTo is the first +** character of the word. +*/ +static int substituteCost(char cPrev, char cFrom, char cTo){ + char classFrom, classTo; + if( cFrom==cTo ){ + /* Exact match */ + return 0; + } + if( cFrom==(cTo^0x20) && ((cTo>='A' && cTo<='Z') || (cTo>='a' && cTo<='z')) ){ + /* differ only in case */ + return 0; + } + classFrom = characterClass(cPrev, cFrom); + classTo = characterClass(cPrev, cTo); + if( classFrom==classTo ){ + /* Same character class */ + return 40; + } + if( classFrom>=CCLASS_B && classFrom<=CCLASS_Y + && classTo>=CCLASS_B && classTo<=CCLASS_Y ){ + /* Convert from one consonant to another, but in a different class */ + return 75; + } + /* Any other subsitution */ + return 100; +} + +/* +** Given two strings zA and zB which are pure ASCII, return the cost +** of transforming zA into zB. If zA ends with '*' assume that it is +** a prefix of zB and give only minimal penalty for extra characters +** on the end of zB. +** +** Smaller numbers mean a closer match. +** +** Negative values indicate an error: +** -1 One of the inputs is NULL +** -2 Non-ASCII characters on input +** -3 Unable to allocate memory +** +** If pnMatch is not NULL, then *pnMatch is set to the number of bytes +** of zB that matched the pattern in zA. If zA does not end with a '*', +** then this value is always the number of bytes in zB (i.e. strlen(zB)). +** If zA does end in a '*', then it is the number of bytes in the prefix +** of zB that was deemed to match zA. +*/ +static int editdist1(const char *zA, const char *zB, int *pnMatch){ + int nA, nB; /* Number of characters in zA[] and zB[] */ + int xA, xB; /* Loop counters for zA[] and zB[] */ + char cA, cB; /* Current character of zA and zB */ + char cAprev, cBprev; /* Previous character of zA and zB */ + char cAnext, cBnext; /* Next character in zA and zB */ + int d; /* North-west cost value */ + int dc = 0; /* North-west character value */ + int res; /* Final result */ + int *m; /* The cost matrix */ + char *cx; /* Corresponding character values */ + int *toFree = 0; /* Malloced space */ + int mStack[60+15]; /* Stack space to use if not too much is needed */ + int nMatch = 0; + + /* Early out if either input is NULL */ + if( zA==0 || zB==0 ) return -1; + + /* Skip any common prefix */ + while( zA[0] && zA[0]==zB[0] ){ dc = zA[0]; zA++; zB++; nMatch++; } + if( pnMatch ) *pnMatch = nMatch; + if( zA[0]==0 && zB[0]==0 ) return 0; + +#if 0 + printf("A=\"%s\" B=\"%s\" dc=%c\n", zA, zB, dc?dc:' '); +#endif + + /* Verify input strings and measure their lengths */ + for(nA=0; zA[nA]; nA++){ + if( zA[nA]&0x80 ) return -2; + } + for(nB=0; zB[nB]; nB++){ + if( zB[nB]&0x80 ) return -2; + } + + /* Special processing if either string is empty */ + if( nA==0 ){ + cBprev = dc; + for(xB=res=0; (cB = zB[xB])!=0; xB++){ + res += insertOrDeleteCost(cBprev, cB, zB[xB+1])/FINAL_INS_COST_DIV; + cBprev = cB; + } + return res; + } + if( nB==0 ){ + cAprev = dc; + for(xA=res=0; (cA = zA[xA])!=0; xA++){ + res += insertOrDeleteCost(cAprev, cA, zA[xA+1]); + cAprev = cA; + } + return res; + } + + /* A is a prefix of B */ + if( zA[0]=='*' && zA[1]==0 ) return 0; + + /* Allocate and initialize the Wagner matrix */ + if( nB<(sizeof(mStack)*4)/(sizeof(mStack[0])*5) ){ + m = mStack; + }else{ + m = toFree = sqlite3_malloc( (nB+1)*5*sizeof(m[0])/4 ); + if( m==0 ) return -3; + } + cx = (char*)&m[nB+1]; + + /* Compute the Wagner edit distance */ + m[0] = 0; + cx[0] = dc; + cBprev = dc; + for(xB=1; xB<=nB; xB++){ + cBnext = zB[xB]; + cB = zB[xB-1]; + cx[xB] = cB; + m[xB] = m[xB-1] + insertOrDeleteCost(cBprev, cB, cBnext); + cBprev = cB; + } + cAprev = dc; + for(xA=1; xA<=nA; xA++){ + int lastA = (xA==nA); + cA = zA[xA-1]; + cAnext = zA[xA]; + if( cA=='*' && lastA ) break; + d = m[0]; + dc = cx[0]; + m[0] = d + insertOrDeleteCost(cAprev, cA, cAnext); + cBprev = 0; + for(xB=1; xB<=nB; xB++){ + int totalCost, insCost, delCost, subCost, ncx; + cB = zB[xB-1]; + cBnext = zB[xB]; + + /* Cost to insert cB */ + insCost = insertOrDeleteCost(cx[xB-1], cB, cBnext); + if( lastA ) insCost /= FINAL_INS_COST_DIV; + + /* Cost to delete cA */ + delCost = insertOrDeleteCost(cx[xB], cA, cBnext); + + /* Cost to substitute cA->cB */ + subCost = substituteCost(cx[xB-1], cA, cB); + + /* Best cost */ + totalCost = insCost + m[xB-1]; + ncx = cB; + if( (delCost + m[xB])nLang; i++){ + EditDist3Cost *pCost, *pNext; + pCost = p->a[i].pCost; + while( pCost ){ + pNext = pCost->pNext; + sqlite3_free(pCost); + pCost = pNext; + } + } + sqlite3_free(p->a); + memset(p, 0, sizeof(*p)); +} +static void editDist3ConfigDelete(void *pIn){ + EditDist3Config *p = (EditDist3Config*)pIn; + editDist3ConfigClear(p); + sqlite3_free(p); +} + +/* +** Load all edit-distance weights from a table. +*/ +static int editDist3ConfigLoad( + EditDist3Config *p, /* The edit distance configuration to load */ + sqlite3 *db, /* Load from this database */ + const char *zTable /* Name of the table from which to load */ +){ + sqlite3_stmt *pStmt; + int rc, rc2; + char *zSql; + int iLangPrev = -9999; + EditDist3Lang *pLang = 0; + + zSql = sqlite3_mprintf("SELECT iLang, cFrom, cTo, iCost" + " FROM \"%w\" WHERE iLang>=0 ORDER BY iLang", zTable); + if( zSql==0 ) return SQLITE_NOMEM; + rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + if( rc ) return rc; + editDist3ConfigClear(p); + while( sqlite3_step(pStmt)==SQLITE_ROW ){ + int iLang = sqlite3_column_int(pStmt, 0); + const char *zFrom = (const char*)sqlite3_column_text(pStmt, 1); + int nFrom = zFrom ? sqlite3_column_bytes(pStmt, 1) : 0; + const char *zTo = (const char*)sqlite3_column_text(pStmt, 2); + int nTo = zTo ? sqlite3_column_bytes(pStmt, 2) : 0; + int iCost = sqlite3_column_int(pStmt, 3); + + assert( zFrom!=0 || nFrom==0 ); + assert( zTo!=0 || nTo==0 ); + if( nFrom>100 || nTo>100 ) continue; + if( iCost<0 ) continue; + if( pLang==0 || iLang!=iLangPrev ){ + EditDist3Lang *pNew; + pNew = sqlite3_realloc(p->a, (p->nLang+1)*sizeof(p->a[0])); + if( pNew==0 ){ rc = SQLITE_NOMEM; break; } + p->a = pNew; + pLang = &p->a[p->nLang]; + p->nLang++; + pLang->iLang = iLang; + pLang->iInsCost = 100; + pLang->iDelCost = 100; + pLang->iSubCost = 150; + pLang->pCost = 0; + iLangPrev = iLang; + } + if( nFrom==1 && zFrom[0]=='?' && nTo==0 ){ + pLang->iDelCost = iCost; + }else if( nFrom==0 && nTo==1 && zTo[0]=='?' ){ + pLang->iInsCost = iCost; + }else if( nFrom==1 && nTo==1 && zFrom[0]=='?' && zTo[0]=='?' ){ + pLang->iSubCost = iCost; + }else{ + EditDist3Cost *pCost; + int nExtra = nFrom + nTo - 4; + if( nExtra<0 ) nExtra = 0; + pCost = sqlite3_malloc( sizeof(*pCost) + nExtra ); + if( pCost==0 ){ rc = SQLITE_NOMEM; break; } + pCost->nFrom = nFrom; + pCost->nTo = nTo; + pCost->iCost = iCost; + memcpy(pCost->a, zFrom, nFrom); + memcpy(pCost->a + nFrom, zTo, nTo); + pCost->pNext = pLang->pCost; + pLang->pCost = pCost; + } + } + rc2 = sqlite3_finalize(pStmt); + if( rc==SQLITE_OK ) rc = rc2; + return rc; +} + +/* +** Return the length (in bytes) of a utf-8 character. Or return a maximum +** of N. +*/ +static int utf8Len(unsigned char c, int N){ + int len = 1; + if( c>0x7f ){ + if( (c&0xe0)==0xc0 ){ + len = 2; + }else if( (c&0xf0)==0xe0 ){ + len = 3; + }else{ + len = 4; + } + } + if( len>N ) len = N; + return len; +} + +/* +** Return TRUE (non-zero) if the To side of the given cost matches +** the given string. +*/ +static int matchTo(EditDist3Cost *p, const char *z, int n){ + if( p->nTo>n ) return 0; + if( strncmp(p->a+p->nFrom, z, p->nTo)!=0 ) return 0; + return 1; +} + +/* +** Return TRUE (non-zero) if the From side of the given cost matches +** the given string. +*/ +static int matchFrom(EditDist3Cost *p, const char *z, int n){ + assert( p->nFrom<=n ); + if( strncmp(p->a, z, p->nFrom)!=0 ) return 0; + return 1; +} + +/* +** Return TRUE (non-zero) of the next FROM character and the next TO +** character are the same. +*/ +static int matchFromTo( + EditDist3FromString *pStr, /* Left hand string */ + int n1, /* Index of comparison character on the left */ + const char *z2, /* Right-handl comparison character */ + int n2 /* Bytes remaining in z2[] */ +){ + int b1 = pStr->a[n1].nByte; + if( b1>n2 ) return 0; + if( memcmp(pStr->z+n1, z2, b1)!=0 ) return 0; + return 1; +} + +/* +** Delete an EditDist3FromString objecct +*/ +static void editDist3FromStringDelete(EditDist3FromString *p){ + int i; + if( p ){ + for(i=0; in; i++){ + sqlite3_free(p->a[i].apDel); + sqlite3_free(p->a[i].apSubst); + } + sqlite3_free(p); + } +} + +/* +** Create a EditDist3FromString object. +*/ +static EditDist3FromString *editDist3FromStringNew( + const EditDist3Lang *pLang, + const char *z, + int n +){ + EditDist3FromString *pStr; + EditDist3Cost *p; + int i; + + if( z==0 ) return 0; + if( n<0 ) n = (int)strlen(z); + pStr = sqlite3_malloc( sizeof(*pStr) + sizeof(pStr->a[0])*n + n + 1 ); + if( pStr==0 ) return 0; + pStr->a = (EditDist3From*)&pStr[1]; + memset(pStr->a, 0, sizeof(pStr->a[0])*n); + pStr->n = n; + pStr->z = (char*)&pStr->a[n]; + memcpy(pStr->z, z, n+1); + if( n && z[n-1]=='*' ){ + pStr->isPrefix = 1; + n--; + pStr->n--; + pStr->z[n] = 0; + }else{ + pStr->isPrefix = 0; + } + + for(i=0; ia[i]; + memset(pFrom, 0, sizeof(*pFrom)); + pFrom->nByte = utf8Len((unsigned char)z[i], n-i); + for(p=pLang->pCost; p; p=p->pNext){ + EditDist3Cost **apNew; + if( i+p->nFrom>n ) continue; + if( matchFrom(p, z+i, n-i)==0 ) continue; + if( p->nTo==0 ){ + apNew = sqlite3_realloc(pFrom->apDel, + sizeof(*apNew)*(pFrom->nDel+1)); + if( apNew==0 ) break; + pFrom->apDel = apNew; + apNew[pFrom->nDel++] = p; + }else{ + apNew = sqlite3_realloc(pFrom->apSubst, + sizeof(*apNew)*(pFrom->nSubst+1)); + if( apNew==0 ) break; + pFrom->apSubst = apNew; + apNew[pFrom->nSubst++] = p; + } + } + if( p ){ + editDist3FromStringDelete(pStr); + pStr = 0; + break; + } + } + return pStr; +} + +/* +** Update entry m[i] such that it is the minimum of its current value +** and m[j]+iCost. +** +** If the iCost is 1,000,000 or greater, then consider the cost to be +** infinite and skip the update. +*/ +static void updateCost( + unsigned int *m, + int i, + int j, + int iCost +){ + assert( iCost>=0 ); + if( iCost<10000 ){ + unsigned int b = m[j] + iCost; + if( bpCost; p; p=p->pNext){ + EditDist3Cost **apNew; + if( p->nFrom>0 ) continue; + if( i2+p->nTo>n2 ) continue; + if( matchTo(p, z2+i2, n2-i2)==0 ) continue; + a2[i2].nIns++; + apNew = sqlite3_realloc(a2[i2].apIns, sizeof(*apNew)*a2[i2].nIns); + if( apNew==0 ){ + res = -1; /* Out of memory */ + goto editDist3Abort; + } + a2[i2].apIns = apNew; + a2[i2].apIns[a2[i2].nIns-1] = p; + } + } + + /* Prepare to compute the minimum edit distance */ + szRow = f.n+1; + memset(m, 0x01, (n2+1)*szRow*sizeof(m[0])); + m[0] = 0; + + /* First fill in the top-row of the matrix with FROM deletion costs */ + for(i1=0; i1iDelCost); + for(k=0; knFrom, i1, p->iCost); + } + } + + /* Fill in all subsequent rows, top-to-bottom, left-to-right */ + for(i2=0; i2iInsCost); + for(k=0; knTo), rxp, p->iCost); + } + for(i1=0; i1iDelCost); + for(k=0; knFrom, cxp, p->iCost); + } + updateCost(m, cx, cxu, pLang->iInsCost); + if( matchFromTo(&f, i1, z2+i2, n2-i2) ){ + updateCost(m, cx, cxd, 0); + } + updateCost(m, cx, cxd, pLang->iSubCost); + for(k=0; knFrom+szRow*p->nTo, cxd, p->iCost); + } + } + } + } + +#if 0 /* Enable for debugging */ + printf(" ^"); + for(i1=0; i19999 ) printf(" ****"); + else printf(" %4d", v); + } + printf("\n"); + for(i2=0; i29999 ) printf(" ****"); + else printf(" %4d", v); + } + printf("\n"); + } +#endif + + /* Free memory allocations and return the result */ + res = (int)m[szRow*(n2+1)-1]; + n = n2; + if( f.isPrefix ){ + for(i2=1; i2<=n2; i2++){ + int b = m[szRow*i2-1]; + if( b<=res ){ + res = b; + n = i2 - 1; + } + } + } + if( pnMatch ){ + int nExtra = 0; + for(k=0; knLang; i++){ + if( pConfig->a[i].iLang==iLang ) return &pConfig->a[i]; + } + return &editDist3Lang; +} + +/* +** Function: editdist3(A,B,iLang) +** editdist3(tablename) +** +** Return the cost of transforming string A into string B using edit +** weights for iLang. +** +** The second form loads edit weights into memory from a table. +*/ +static void editDist3SqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + EditDist3Config *pConfig = (EditDist3Config*)sqlite3_user_data(context); + sqlite3 *db = sqlite3_context_db_handle(context); + int rc; + if( argc==1 ){ + const char *zTable = (const char*)sqlite3_value_text(argv[0]); + rc = editDist3ConfigLoad(pConfig, db, zTable); + if( rc ) sqlite3_result_error_code(context, rc); + }else{ + const char *zA = (const char*)sqlite3_value_text(argv[0]); + const char *zB = (const char*)sqlite3_value_text(argv[1]); + int nA = sqlite3_value_bytes(argv[0]); + int nB = sqlite3_value_bytes(argv[1]); + int iLang = argc==3 ? sqlite3_value_int(argv[2]) : 0; + const EditDist3Lang *pLang = editDist3FindLang(pConfig, iLang); + EditDist3FromString *pFrom; + int dist; + + pFrom = editDist3FromStringNew(pLang, zA, nA); + if( pFrom==0 ){ + sqlite3_result_error_nomem(context); + return; + } + dist = editDist3Core(pFrom, zB, nB, pLang, 0); + editDist3FromStringDelete(pFrom); + if( dist==(-1) ){ + sqlite3_result_error_nomem(context); + }else{ + sqlite3_result_int(context, dist); + } + } +} + +/* +** Register the editDist3 function with SQLite +*/ +static int editDist3Install(sqlite3 *db){ + int rc; + EditDist3Config *pConfig = sqlite3_malloc( sizeof(*pConfig) ); + if( pConfig==0 ) return SQLITE_NOMEM; + memset(pConfig, 0, sizeof(*pConfig)); + rc = sqlite3_create_function_v2(db, "editdist3", + 2, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function_v2(db, "editdist3", + 3, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function_v2(db, "editdist3", + 1, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, + editDist3ConfigDelete); + }else{ + sqlite3_free(pConfig); + } + return rc; +} +/* End configurable cost unicode edit distance routines +****************************************************************************** +****************************************************************************** +** Begin transliterate unicode-to-ascii implementation +*/ + +#if !SQLITE_AMALGAMATION +/* +** This lookup table is used to help decode the first byte of +** a multi-byte UTF8 character. +*/ +static const unsigned char sqlite3Utf8Trans1[] = { + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, + 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00, +}; +#endif + +/* +** Return the value of the first UTF-8 character in the string. +*/ +static int utf8Read(const unsigned char *z, int n, int *pSize){ + int c, i; + + /* All callers to this routine (in the current implementation) + ** always have n>0. */ + if( NEVER(n==0) ){ + c = i = 0; + }else{ + c = z[0]; + i = 1; + if( c>=0xc0 ){ + c = sqlite3Utf8Trans1[c-0xc0]; + while( i0 ){ + c = utf8Read(zIn, nIn, &sz); + zIn += sz; + nIn -= sz; + if( c<=127 ){ + zOut[nOut++] = c; + }else{ + int xTop, xBtm, x; + xTop = sizeof(translit)/sizeof(translit[0]) - 1; + xBtm = 0; + while( xTop>=xBtm ){ + x = (xTop + xBtm)/2; + if( translit[x].cFrom==c ){ + zOut[nOut++] = translit[x].cTo0; + if( translit[x].cTo1 ){ + zOut[nOut++] = translit[x].cTo1; + /* Add an extra "ch" after the "sh" for Щ and щ */ + if( c==0x0429 || c== 0x0449 ){ + zOut[nOut++] = 'c'; + zOut[nOut++] = 'h'; + } + } + c = 0; + break; + }else if( translit[x].cFrom>c ){ + xTop = x-1; + }else{ + xBtm = x+1; + } + } + if( c ) zOut[nOut++] = '?'; + } + } + zOut[nOut] = 0; + return zOut; +} + +/* +** Return the number of characters in the shortest prefix of the input +** string that transliterates to an ASCII string nTrans bytes or longer. +** Or, if the transliteration of the input string is less than nTrans +** bytes in size, return the number of characters in the input string. +*/ +static int translen_to_charlen(const char *zIn, int nIn, int nTrans){ + int i, c, sz, nOut; + int nChar; + + i = nOut = 0; + for(nChar=0; i=128 ){ + int xTop, xBtm, x; + xTop = sizeof(translit)/sizeof(translit[0]) - 1; + xBtm = 0; + while( xTop>=xBtm ){ + x = (xTop + xBtm)/2; + if( translit[x].cFrom==c ){ + if( translit[x].cTo1 ) nOut++; + if( c==0x0429 || c== 0x0449 ) nOut += 2; + break; + }else if( translit[x].cFrom>c ){ + xTop = x-1; + }else{ + xBtm = x+1; + } + } + } + } + + return nChar; +} + + +/* +** spellfix1_translit(X) +** +** Convert a string that contains non-ASCII Roman characters into +** pure ASCII. +*/ +static void transliterateSqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn = sqlite3_value_text(argv[0]); + int nIn = sqlite3_value_bytes(argv[0]); + unsigned char *zOut = transliterate(zIn, nIn); + if( zOut==0 ){ + sqlite3_result_error_nomem(context); + }else{ + sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); + } +} + +/* +** spellfix1_scriptcode(X) +** +** Try to determine the dominant script used by the word X and return +** its ISO 15924 numeric code. +** +** The current implementation only understands the following scripts: +** +** 215 (Latin) +** 220 (Cyrillic) +** 200 (Greek) +** +** This routine will return 998 if the input X contains characters from +** two or more of the above scripts or 999 if X contains no characters +** from any of the above scripts. +*/ +static void scriptCodeSqlFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + const unsigned char *zIn = sqlite3_value_text(argv[0]); + int nIn = sqlite3_value_bytes(argv[0]); + int c, sz; + int scriptMask = 0; + int res; +# define SCRIPT_LATIN 0x0001 +# define SCRIPT_CYRILLIC 0x0002 +# define SCRIPT_GREEK 0x0004 + + while( nIn>0 ){ + c = utf8Read(zIn, nIn, &sz); + zIn += sz; + nIn -= sz; + if( c<0x02af ){ + scriptMask |= SCRIPT_LATIN; + }else if( c>=0x0400 && c<=0x04ff ){ + scriptMask |= SCRIPT_CYRILLIC; + }else if( c>=0x0386 && c<=0x03ce ){ + scriptMask |= SCRIPT_GREEK; + } + } + switch( scriptMask ){ + case 0: res = 999; break; + case SCRIPT_LATIN: res = 215; break; + case SCRIPT_CYRILLIC: res = 220; break; + case SCRIPT_GREEK: res = 200; break; + default: res = 998; break; + } + sqlite3_result_int(context, res); +} + +/* End transliterate +****************************************************************************** +****************************************************************************** +** Begin spellfix1 virtual table. +*/ + +/* Maximum length of a phonehash used for querying the shadow table */ +#define SPELLFIX_MX_HASH 8 + +/* Maximum number of hash strings to examine per query */ +#define SPELLFIX_MX_RUN 1 + +typedef struct spellfix1_vtab spellfix1_vtab; +typedef struct spellfix1_cursor spellfix1_cursor; + +/* Fuzzy-search virtual table object */ +struct spellfix1_vtab { + sqlite3_vtab base; /* Base class - must be first */ + sqlite3 *db; /* Database connection */ + char *zDbName; /* Name of database holding this table */ + char *zTableName; /* Name of the virtual table */ + char *zCostTable; /* Table holding edit-distance cost numbers */ + EditDist3Config *pConfig3; /* Parsed edit distance costs */ +}; + +/* Fuzzy-search cursor object */ +struct spellfix1_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + spellfix1_vtab *pVTab; /* The table to which this cursor belongs */ + char *zPattern; /* rhs of MATCH clause */ + int nRow; /* Number of rows of content */ + int nAlloc; /* Number of allocated rows */ + int iRow; /* Current row of content */ + int iLang; /* Value of the langid= constraint */ + int iTop; /* Value of the top= constraint */ + int iScope; /* Value of the scope= constraint */ + int nSearch; /* Number of vocabulary items checked */ + sqlite3_stmt *pFullScan; /* Shadow query for a full table scan */ + struct spellfix1_row { /* For each row of content */ + sqlite3_int64 iRowid; /* Rowid for this row */ + char *zWord; /* Text for this row */ + int iRank; /* Rank for this row */ + int iDistance; /* Distance from pattern for this row */ + int iScore; /* Score for sorting */ + int iMatchlen; /* Value of matchlen column (or -1) */ + char zHash[SPELLFIX_MX_HASH]; /* the phonehash used for this match */ + } *a; +}; + +/* +** Construct one or more SQL statements from the format string given +** and then evaluate those statements. The success code is written +** into *pRc. +** +** If *pRc is initially non-zero then this routine is a no-op. +*/ +static void spellfix1DbExec( + int *pRc, /* Success code */ + sqlite3 *db, /* Database in which to run SQL */ + const char *zFormat, /* Format string for SQL */ + ... /* Arguments to the format string */ +){ + va_list ap; + char *zSql; + if( *pRc ) return; + va_start(ap, zFormat); + zSql = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + if( zSql==0 ){ + *pRc = SQLITE_NOMEM; + }else{ + *pRc = sqlite3_exec(db, zSql, 0, 0, 0); + sqlite3_free(zSql); + } +} + +/* +** xDisconnect/xDestroy method for the fuzzy-search module. +*/ +static int spellfix1Uninit(int isDestroy, sqlite3_vtab *pVTab){ + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + int rc = SQLITE_OK; + if( isDestroy ){ + sqlite3 *db = p->db; + spellfix1DbExec(&rc, db, "DROP TABLE IF EXISTS \"%w\".\"%w_vocab\"", + p->zDbName, p->zTableName); + } + if( rc==SQLITE_OK ){ + sqlite3_free(p->zTableName); + editDist3ConfigDelete(p->pConfig3); + sqlite3_free(p->zCostTable); + sqlite3_free(p); + } + return rc; +} +static int spellfix1Disconnect(sqlite3_vtab *pVTab){ + return spellfix1Uninit(0, pVTab); +} +static int spellfix1Destroy(sqlite3_vtab *pVTab){ + return spellfix1Uninit(1, pVTab); +} + +/* +** Make a copy of a string. Remove leading and trailing whitespace +** and dequote it. +*/ +static char *spellfix1Dequote(const char *zIn){ + char *zOut; + int i, j; + char c; + while( isspace(zIn[0]) ) zIn++; + zOut = sqlite3_mprintf("%s", zIn); + if( zOut==0 ) return 0; + i = (int)strlen(zOut); +#if 0 /* The parser will never leave spaces at the end */ + while( i>0 && isspace(zOut[i-1]) ){ i--; } +#endif + zOut[i] = 0; + c = zOut[0]; + if( c=='\'' || c=='"' ){ + for(i=1, j=0; ALWAYS(zOut[i]); i++){ + zOut[j++] = zOut[i]; + if( zOut[i]==c ){ + if( zOut[i+1]==c ){ + i++; + }else{ + zOut[j-1] = 0; + break; + } + } + } + } + return zOut; +} + + +/* +** xConnect/xCreate method for the spellfix1 module. Arguments are: +** +** argv[0] -> module name ("spellfix1") +** argv[1] -> database name +** argv[2] -> table name +** argv[3].. -> optional arguments (i.e. "edit_cost_table" parameter) +*/ +static int spellfix1Init( + int isCreate, + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVTab, + char **pzErr +){ + spellfix1_vtab *pNew = 0; + const char *zModule = argv[0]; + const char *zDbName = argv[1]; + const char *zTableName = argv[2]; + int nDbName; + int rc = SQLITE_OK; + int i; + + nDbName = (int)strlen(zDbName); + pNew = sqlite3_malloc( sizeof(*pNew) + nDbName + 1); + if( pNew==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pNew, 0, sizeof(*pNew)); + pNew->zDbName = (char*)&pNew[1]; + memcpy(pNew->zDbName, zDbName, nDbName+1); + pNew->zTableName = sqlite3_mprintf("%s", zTableName); + pNew->db = db; + if( pNew->zTableName==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = sqlite3_declare_vtab(db, + "CREATE TABLE x(word,rank,distance,langid, " + "score, matchlen, phonehash HIDDEN, " + "top HIDDEN, scope HIDDEN, srchcnt HIDDEN, " + "soundslike HIDDEN, command HIDDEN)" + ); +#define SPELLFIX_COL_WORD 0 +#define SPELLFIX_COL_RANK 1 +#define SPELLFIX_COL_DISTANCE 2 +#define SPELLFIX_COL_LANGID 3 +#define SPELLFIX_COL_SCORE 4 +#define SPELLFIX_COL_MATCHLEN 5 +#define SPELLFIX_COL_PHONEHASH 6 +#define SPELLFIX_COL_TOP 7 +#define SPELLFIX_COL_SCOPE 8 +#define SPELLFIX_COL_SRCHCNT 9 +#define SPELLFIX_COL_SOUNDSLIKE 10 +#define SPELLFIX_COL_COMMAND 11 + } + if( rc==SQLITE_OK && isCreate ){ + sqlite3_uint64 r; + spellfix1DbExec(&rc, db, + "CREATE TABLE IF NOT EXISTS \"%w\".\"%w_vocab\"(\n" + " id INTEGER PRIMARY KEY,\n" + " rank INT,\n" + " langid INT,\n" + " word TEXT,\n" + " k1 TEXT,\n" + " k2 TEXT\n" + ");\n", + zDbName, zTableName + ); + sqlite3_randomness(sizeof(r), &r); + spellfix1DbExec(&rc, db, + "CREATE INDEX IF NOT EXISTS \"%w\".\"%w_index_%llx\" " + "ON \"%w_vocab\"(langid,k2);", + zDbName, zModule, r, zTableName + ); + } + for(i=3; rc==SQLITE_OK && ibase); + }else{ + *ppVTab = (sqlite3_vtab *)pNew; + } + return rc; +} + +/* +** The xConnect and xCreate methods +*/ +static int spellfix1Connect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVTab, + char **pzErr +){ + return spellfix1Init(0, db, pAux, argc, argv, ppVTab, pzErr); +} +static int spellfix1Create( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVTab, + char **pzErr +){ + return spellfix1Init(1, db, pAux, argc, argv, ppVTab, pzErr); +} + +/* +** Clear all of the content from a cursor. +*/ +static void spellfix1ResetCursor(spellfix1_cursor *pCur){ + int i; + for(i=0; inRow; i++){ + sqlite3_free(pCur->a[i].zWord); + } + pCur->nRow = 0; + pCur->iRow = 0; + pCur->nSearch = 0; + if( pCur->pFullScan ){ + sqlite3_finalize(pCur->pFullScan); + pCur->pFullScan = 0; + } +} + +/* +** Resize the cursor to hold up to N rows of content +*/ +static void spellfix1ResizeCursor(spellfix1_cursor *pCur, int N){ + struct spellfix1_row *aNew; + assert( N>=pCur->nRow ); + aNew = sqlite3_realloc(pCur->a, sizeof(pCur->a[0])*N); + if( aNew==0 && N>0 ){ + spellfix1ResetCursor(pCur); + sqlite3_free(pCur->a); + pCur->nAlloc = 0; + pCur->a = 0; + }else{ + pCur->nAlloc = N; + pCur->a = aNew; + } +} + + +/* +** Close a fuzzy-search cursor. +*/ +static int spellfix1Close(sqlite3_vtab_cursor *cur){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + spellfix1ResetCursor(pCur); + spellfix1ResizeCursor(pCur, 0); + sqlite3_free(pCur->zPattern); + sqlite3_free(pCur); + return SQLITE_OK; +} + +/* +** Search for terms of these forms: +** +** (A) word MATCH $str +** (B) langid == $langid +** (C) top = $top +** (D) scope = $scope +** (E) distance < $distance +** (F) distance <= $distance +** +** The plan number is a bit mask formed with these bits: +** +** 0x01 (A) is found +** 0x02 (B) is found +** 0x04 (C) is found +** 0x08 (D) is found +** 0x10 (E) is found +** 0x20 (F) is found +** +** filter.argv[*] values contains $str, $langid, $top, and $scope, +** if specified and in that order. +*/ +static int spellfix1BestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ + int iPlan = 0; + int iLangTerm = -1; + int iTopTerm = -1; + int iScopeTerm = -1; + int iDistTerm = -1; + int i; + const struct sqlite3_index_constraint *pConstraint; + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + + /* Terms of the form: word MATCH $str */ + if( (iPlan & 1)==0 + && pConstraint->iColumn==SPELLFIX_COL_WORD + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH + ){ + iPlan |= 1; + pIdxInfo->aConstraintUsage[i].argvIndex = 1; + pIdxInfo->aConstraintUsage[i].omit = 1; + } + + /* Terms of the form: langid = $langid */ + if( (iPlan & 2)==0 + && pConstraint->iColumn==SPELLFIX_COL_LANGID + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 2; + iLangTerm = i; + } + + /* Terms of the form: top = $top */ + if( (iPlan & 4)==0 + && pConstraint->iColumn==SPELLFIX_COL_TOP + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 4; + iTopTerm = i; + } + + /* Terms of the form: scope = $scope */ + if( (iPlan & 8)==0 + && pConstraint->iColumn==SPELLFIX_COL_SCOPE + && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ + ){ + iPlan |= 8; + iScopeTerm = i; + } + + /* Terms of the form: distance < $dist or distance <= $dist */ + if( (iPlan & (16|32))==0 + && pConstraint->iColumn==SPELLFIX_COL_DISTANCE + && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT + || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) + ){ + iPlan |= pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ? 16 : 32; + iDistTerm = i; + } + } + if( iPlan&1 ){ + int idx = 2; + pIdxInfo->idxNum = iPlan; + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].iColumn==SPELLFIX_COL_SCORE + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; /* Default order by iScore */ + } + if( iPlan&2 ){ + pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iLangTerm].omit = 1; + } + if( iPlan&4 ){ + pIdxInfo->aConstraintUsage[iTopTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iTopTerm].omit = 1; + } + if( iPlan&8 ){ + pIdxInfo->aConstraintUsage[iScopeTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iScopeTerm].omit = 1; + } + if( iPlan&(16|32) ){ + pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = idx++; + pIdxInfo->aConstraintUsage[iDistTerm].omit = 1; + } + pIdxInfo->estimatedCost = (double)10000; + }else{ + pIdxInfo->idxNum = 0; + pIdxInfo->estimatedCost = (double)10000000; + } + return SQLITE_OK; +} + +/* +** Open a new fuzzy-search cursor. +*/ +static int spellfix1Open(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + spellfix1_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + pCur->pVTab = p; + *ppCursor = &pCur->base; + return SQLITE_OK; +} + +/* +** Adjust a distance measurement by the words rank in order to show +** preference to common words. +*/ +static int spellfix1Score(int iDistance, int iRank){ + int iLog2; + for(iLog2=0; iRank>0; iLog2++, iRank>>=1){} + return iDistance + 32 - iLog2; +} + +/* +** Compare two spellfix1_row objects for sorting purposes in qsort() such +** that they sort in order of increasing distance. +*/ +static int spellfix1RowCompare(const void *A, const void *B){ + const struct spellfix1_row *a = (const struct spellfix1_row*)A; + const struct spellfix1_row *b = (const struct spellfix1_row*)B; + return a->iScore - b->iScore; +} + +/* +** A structure used to pass information from spellfix1FilterForMatch() +** into spellfix1RunQuery(). +*/ +typedef struct MatchQuery { + spellfix1_cursor *pCur; /* The cursor being queried */ + sqlite3_stmt *pStmt; /* shadow table query statment */ + char zHash[SPELLFIX_MX_HASH]; /* The current phonehash for zPattern */ + const char *zPattern; /* Transliterated input string */ + int nPattern; /* Length of zPattern */ + EditDist3FromString *pMatchStr3; /* Original unicode string */ + EditDist3Config *pConfig3; /* Edit-distance cost coefficients */ + const EditDist3Lang *pLang; /* The selected language coefficients */ + int iLang; /* The language id */ + int iScope; /* Default scope */ + int iMaxDist; /* Maximum allowed edit distance, or -1 */ + int rc; /* Error code */ + int nRun; /* Number of prior runs for the same zPattern */ + char azPrior[SPELLFIX_MX_RUN][SPELLFIX_MX_HASH]; /* Prior hashes */ +} MatchQuery; + +/* +** Run a query looking for the best matches against zPattern using +** zHash as the character class seed hash. +*/ +static void spellfix1RunQuery(MatchQuery *p, const char *zQuery, int nQuery){ + const char *zK1; + const char *zWord; + int iDist; + int iRank; + int iScore; + int iWorst = 0; + int idx; + int idxWorst = -1; + int i; + int iScope = p->iScope; + spellfix1_cursor *pCur = p->pCur; + sqlite3_stmt *pStmt = p->pStmt; + char zHash1[SPELLFIX_MX_HASH]; + char zHash2[SPELLFIX_MX_HASH]; + char *zClass; + int nClass; + int rc; + + if( pCur->a==0 || p->rc ) return; /* Prior memory allocation failure */ + zClass = (char*)phoneticHash((unsigned char*)zQuery, nQuery); + if( zClass==0 ){ + p->rc = SQLITE_NOMEM; + return; + } + nClass = (int)strlen(zClass); + if( nClass>SPELLFIX_MX_HASH-2 ){ + nClass = SPELLFIX_MX_HASH-2; + zClass[nClass] = 0; + } + if( nClass<=iScope ){ + if( nClass>2 ){ + iScope = nClass-1; + }else{ + iScope = nClass; + } + } + memcpy(zHash1, zClass, iScope); + sqlite3_free(zClass); + zHash1[iScope] = 0; + memcpy(zHash2, zHash1, iScope); + zHash2[iScope] = 'Z'; + zHash2[iScope+1] = 0; +#if SPELLFIX_MX_RUN>1 + for(i=0; inRun; i++){ + if( strcmp(p->azPrior[i], zHash1)==0 ) return; + } +#endif + assert( p->nRunazPrior[p->nRun++], zHash1, iScope+1); + if( sqlite3_bind_text(pStmt, 1, zHash1, -1, SQLITE_STATIC)==SQLITE_NOMEM + || sqlite3_bind_text(pStmt, 2, zHash2, -1, SQLITE_STATIC)==SQLITE_NOMEM + ){ + p->rc = SQLITE_NOMEM; + return; + } +#if SPELLFIX_MX_RUN>1 + for(i=0; inRow; i++){ + if( pCur->a[i].iScore>iWorst ){ + iWorst = pCur->a[i].iScore; + idxWorst = i; + } + } +#endif + while( sqlite3_step(pStmt)==SQLITE_ROW ){ + int iMatchlen = -1; + iRank = sqlite3_column_int(pStmt, 2); + if( p->pMatchStr3 ){ + int nWord = sqlite3_column_bytes(pStmt, 1); + zWord = (const char*)sqlite3_column_text(pStmt, 1); + iDist = editDist3Core(p->pMatchStr3, zWord, nWord, p->pLang, &iMatchlen); + }else{ + zK1 = (const char*)sqlite3_column_text(pStmt, 3); + if( zK1==0 ) continue; + iDist = editdist1(p->zPattern, zK1, 0); + } + if( iDist<0 ){ + p->rc = SQLITE_NOMEM; + break; + } + pCur->nSearch++; + iScore = spellfix1Score(iDist,iRank); + if( p->iMaxDist>=0 ){ + if( iDist>p->iMaxDist ) continue; + if( pCur->nRow>=pCur->nAlloc-1 ){ + spellfix1ResizeCursor(pCur, pCur->nAlloc*2 + 10); + if( pCur->a==0 ) break; + } + idx = pCur->nRow; + }else if( pCur->nRownAlloc ){ + idx = pCur->nRow; + }else if( iScorea[idx].zWord); + }else{ + continue; + } + pCur->a[idx].zWord = sqlite3_mprintf("%s", sqlite3_column_text(pStmt, 1)); + if( pCur->a[idx].zWord==0 ){ + p->rc = SQLITE_NOMEM; + break; + } + pCur->a[idx].iRowid = sqlite3_column_int64(pStmt, 0); + pCur->a[idx].iRank = iRank; + pCur->a[idx].iDistance = iDist; + pCur->a[idx].iScore = iScore; + pCur->a[idx].iMatchlen = iMatchlen; + memcpy(pCur->a[idx].zHash, zHash1, iScope+1); + if( pCur->nRownAlloc ) pCur->nRow++; + if( pCur->nRow==pCur->nAlloc ){ + iWorst = pCur->a[0].iScore; + idxWorst = 0; + for(i=1; inRow; i++){ + iScore = pCur->a[i].iScore; + if( iWorstrc = rc; +} + +/* +** This version of the xFilter method work if the MATCH term is present +** and we are doing a scan. +*/ +static int spellfix1FilterForMatch( + spellfix1_cursor *pCur, + int idxNum, + int argc, + sqlite3_value **argv +){ + const unsigned char *zMatchThis; /* RHS of the MATCH operator */ + EditDist3FromString *pMatchStr3 = 0; /* zMatchThis as an editdist string */ + char *zPattern; /* Transliteration of zMatchThis */ + int nPattern; /* Length of zPattern */ + int iLimit = 20; /* Max number of rows of output */ + int iScope = 3; /* Use this many characters of zClass */ + int iLang = 0; /* Language code */ + char *zSql; /* SQL of shadow table query */ + sqlite3_stmt *pStmt = 0; /* Shadow table query */ + int rc; /* Result code */ + int idx = 1; /* Next available filter parameter */ + spellfix1_vtab *p = pCur->pVTab; /* The virtual table that owns pCur */ + MatchQuery x; /* For passing info to RunQuery() */ + + /* Load the cost table if we have not already done so */ + if( p->zCostTable!=0 && p->pConfig3==0 ){ + p->pConfig3 = sqlite3_malloc( sizeof(p->pConfig3[0]) ); + if( p->pConfig3==0 ) return SQLITE_NOMEM; + memset(p->pConfig3, 0, sizeof(p->pConfig3[0])); + rc = editDist3ConfigLoad(p->pConfig3, p->db, p->zCostTable); + if( rc ) return rc; + } + memset(&x, 0, sizeof(x)); + x.iScope = 3; /* Default scope if none specified by "WHERE scope=N" */ + x.iMaxDist = -1; /* Maximum allowed edit distance */ + + if( idxNum&2 ){ + iLang = sqlite3_value_int(argv[idx++]); + } + if( idxNum&4 ){ + iLimit = sqlite3_value_int(argv[idx++]); + if( iLimit<1 ) iLimit = 1; + } + if( idxNum&8 ){ + x.iScope = sqlite3_value_int(argv[idx++]); + if( x.iScope<1 ) x.iScope = 1; + if( x.iScope>SPELLFIX_MX_HASH-2 ) x.iScope = SPELLFIX_MX_HASH-2; + } + if( idxNum&(16|32) ){ + x.iMaxDist = sqlite3_value_int(argv[idx++]); + if( idxNum&16 ) x.iMaxDist--; + if( x.iMaxDist<0 ) x.iMaxDist = 0; + } + spellfix1ResetCursor(pCur); + spellfix1ResizeCursor(pCur, iLimit); + zMatchThis = sqlite3_value_text(argv[0]); + if( zMatchThis==0 ) return SQLITE_OK; + if( p->pConfig3 ){ + x.pLang = editDist3FindLang(p->pConfig3, iLang); + pMatchStr3 = editDist3FromStringNew(x.pLang, (const char*)zMatchThis, -1); + if( pMatchStr3==0 ){ + x.rc = SQLITE_NOMEM; + goto filter_exit; + } + }else{ + x.pLang = 0; + } + zPattern = (char*)transliterate(zMatchThis, sqlite3_value_bytes(argv[0])); + sqlite3_free(pCur->zPattern); + pCur->zPattern = zPattern; + if( zPattern==0 ){ + x.rc = SQLITE_NOMEM; + goto filter_exit; + } + nPattern = (int)strlen(zPattern); + if( zPattern[nPattern-1]=='*' ) nPattern--; + zSql = sqlite3_mprintf( + "SELECT id, word, rank, k1" + " FROM \"%w\".\"%w_vocab\"" + " WHERE langid=%d AND k2>=?1 AND k2zDbName, p->zTableName, iLang + ); + if( zSql==0 ){ + x.rc = SQLITE_NOMEM; + pStmt = 0; + goto filter_exit; + } + rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + pCur->iLang = iLang; + x.pCur = pCur; + x.pStmt = pStmt; + x.zPattern = zPattern; + x.nPattern = nPattern; + x.pMatchStr3 = pMatchStr3; + x.iLang = iLang; + x.rc = rc; + x.pConfig3 = p->pConfig3; + if( x.rc==SQLITE_OK ){ + spellfix1RunQuery(&x, zPattern, nPattern); + } + + if( pCur->a ){ + qsort(pCur->a, pCur->nRow, sizeof(pCur->a[0]), spellfix1RowCompare); + pCur->iTop = iLimit; + pCur->iScope = iScope; + }else{ + x.rc = SQLITE_NOMEM; + } + +filter_exit: + sqlite3_finalize(pStmt); + editDist3FromStringDelete(pMatchStr3); + return x.rc; +} + +/* +** This version of xFilter handles a full-table scan case +*/ +static int spellfix1FilterForFullScan( + spellfix1_cursor *pCur, + int idxNum, + int argc, + sqlite3_value **argv +){ + int rc; + char *zSql; + spellfix1_vtab *pVTab = pCur->pVTab; + spellfix1ResetCursor(pCur); + zSql = sqlite3_mprintf( + "SELECT word, rank, NULL, langid, id FROM \"%w\".\"%w_vocab\"", + pVTab->zDbName, pVTab->zTableName); + if( zSql==0 ) return SQLITE_NOMEM; + rc = sqlite3_prepare_v2(pVTab->db, zSql, -1, &pCur->pFullScan, 0); + sqlite3_free(zSql); + pCur->nRow = pCur->iRow = 0; + if( rc==SQLITE_OK ){ + rc = sqlite3_step(pCur->pFullScan); + if( rc==SQLITE_ROW ){ pCur->iRow = -1; rc = SQLITE_OK; } + if( rc==SQLITE_DONE ){ rc = SQLITE_OK; } + }else{ + pCur->iRow = 0; + } + return rc; +} + + +/* +** Called to "rewind" a cursor back to the beginning so that +** it starts its output over again. Always called at least once +** prior to any spellfix1Column, spellfix1Rowid, or spellfix1Eof call. +*/ +static int spellfix1Filter( + sqlite3_vtab_cursor *cur, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + int rc; + if( idxNum & 1 ){ + rc = spellfix1FilterForMatch(pCur, idxNum, argc, argv); + }else{ + rc = spellfix1FilterForFullScan(pCur, idxNum, argc, argv); + } + return rc; +} + + +/* +** Advance a cursor to its next row of output +*/ +static int spellfix1Next(sqlite3_vtab_cursor *cur){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + int rc = SQLITE_OK; + if( pCur->iRow < pCur->nRow ){ + if( pCur->pFullScan ){ + rc = sqlite3_step(pCur->pFullScan); + if( rc!=SQLITE_ROW ) pCur->iRow = pCur->nRow; + if( rc==SQLITE_ROW || rc==SQLITE_DONE ) rc = SQLITE_OK; + }else{ + pCur->iRow++; + } + } + return rc; +} + +/* +** Return TRUE if we are at the end-of-file +*/ +static int spellfix1Eof(sqlite3_vtab_cursor *cur){ + spellfix1_cursor *pCur = (spellfix1_cursor *)cur; + return pCur->iRow>=pCur->nRow; +} + +/* +** Return columns from the current row. +*/ +static int spellfix1Column( + sqlite3_vtab_cursor *cur, + sqlite3_context *ctx, + int i +){ + spellfix1_cursor *pCur = (spellfix1_cursor*)cur; + if( pCur->pFullScan ){ + if( i<=SPELLFIX_COL_LANGID ){ + sqlite3_result_value(ctx, sqlite3_column_value(pCur->pFullScan, i)); + }else{ + sqlite3_result_null(ctx); + } + return SQLITE_OK; + } + switch( i ){ + case SPELLFIX_COL_WORD: { + sqlite3_result_text(ctx, pCur->a[pCur->iRow].zWord, -1, SQLITE_STATIC); + break; + } + case SPELLFIX_COL_RANK: { + sqlite3_result_int(ctx, pCur->a[pCur->iRow].iRank); + break; + } + case SPELLFIX_COL_DISTANCE: { + sqlite3_result_int(ctx, pCur->a[pCur->iRow].iDistance); + break; + } + case SPELLFIX_COL_LANGID: { + sqlite3_result_int(ctx, pCur->iLang); + break; + } + case SPELLFIX_COL_SCORE: { + sqlite3_result_int(ctx, pCur->a[pCur->iRow].iScore); + break; + } + case SPELLFIX_COL_MATCHLEN: { + int iMatchlen = pCur->a[pCur->iRow].iMatchlen; + if( iMatchlen<0 ){ + int nPattern = (int)strlen(pCur->zPattern); + char *zWord = pCur->a[pCur->iRow].zWord; + int nWord = (int)strlen(zWord); + + if( nPattern>0 && pCur->zPattern[nPattern-1]=='*' ){ + char *zTranslit; + int res; + zTranslit = (char *)transliterate((unsigned char *)zWord, nWord); + if( !zTranslit ) return SQLITE_NOMEM; + res = editdist1(pCur->zPattern, zTranslit, &iMatchlen); + sqlite3_free(zTranslit); + if( res<0 ) return SQLITE_NOMEM; + iMatchlen = translen_to_charlen(zWord, nWord, iMatchlen); + }else{ + iMatchlen = utf8Charlen(zWord, nWord); + } + } + + sqlite3_result_int(ctx, iMatchlen); + break; + } + case SPELLFIX_COL_PHONEHASH: { + sqlite3_result_text(ctx, pCur->a[pCur->iRow].zHash, -1, SQLITE_STATIC); + break; + } + case SPELLFIX_COL_TOP: { + sqlite3_result_int(ctx, pCur->iTop); + break; + } + case SPELLFIX_COL_SCOPE: { + sqlite3_result_int(ctx, pCur->iScope); + break; + } + case SPELLFIX_COL_SRCHCNT: { + sqlite3_result_int(ctx, pCur->nSearch); + break; + } + default: { + sqlite3_result_null(ctx); + break; + } + } + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int spellfix1Rowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + spellfix1_cursor *pCur = (spellfix1_cursor*)cur; + if( pCur->pFullScan ){ + *pRowid = sqlite3_column_int64(pCur->pFullScan, 4); + }else{ + *pRowid = pCur->a[pCur->iRow].iRowid; + } + return SQLITE_OK; +} + +/* +** The xUpdate() method. +*/ +static int spellfix1Update( + sqlite3_vtab *pVTab, + int argc, + sqlite3_value **argv, + sqlite_int64 *pRowid +){ + int rc = SQLITE_OK; + sqlite3_int64 rowid, newRowid; + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + sqlite3 *db = p->db; + + if( argc==1 ){ + /* A delete operation on the rowid given by argv[0] */ + rowid = *pRowid = sqlite3_value_int64(argv[0]); + spellfix1DbExec(&rc, db, "DELETE FROM \"%w\".\"%w_vocab\" " + " WHERE id=%lld", + p->zDbName, p->zTableName, rowid); + }else{ + const unsigned char *zWord = sqlite3_value_text(argv[SPELLFIX_COL_WORD+2]); + int nWord = sqlite3_value_bytes(argv[SPELLFIX_COL_WORD+2]); + int iLang = sqlite3_value_int(argv[SPELLFIX_COL_LANGID+2]); + int iRank = sqlite3_value_int(argv[SPELLFIX_COL_RANK+2]); + const unsigned char *zSoundslike = + sqlite3_value_text(argv[SPELLFIX_COL_SOUNDSLIKE+2]); + int nSoundslike = sqlite3_value_bytes(argv[SPELLFIX_COL_SOUNDSLIKE+2]); + char *zK1, *zK2; + int i; + char c; + + if( zWord==0 ){ + /* Inserts of the form: INSERT INTO table(command) VALUES('xyzzy'); + ** cause zWord to be NULL, so we look at the "command" column to see + ** what special actions to take */ + const char *zCmd = + (const char*)sqlite3_value_text(argv[SPELLFIX_COL_COMMAND+2]); + if( zCmd==0 ){ + pVTab->zErrMsg = sqlite3_mprintf("%s.word may not be NULL", + p->zTableName); + return SQLITE_CONSTRAINT_NOTNULL; + } + if( strcmp(zCmd,"reset")==0 ){ + /* Reset the edit cost table (if there is one). */ + editDist3ConfigDelete(p->pConfig3); + p->pConfig3 = 0; + return SQLITE_OK; + } + if( strncmp(zCmd,"edit_cost_table=",16)==0 ){ + editDist3ConfigDelete(p->pConfig3); + p->pConfig3 = 0; + sqlite3_free(p->zCostTable); + p->zCostTable = spellfix1Dequote(zCmd+16); + if( p->zCostTable==0 ) return SQLITE_NOMEM; + if( p->zCostTable[0]==0 || sqlite3_stricmp(p->zCostTable,"null")==0 ){ + sqlite3_free(p->zCostTable); + p->zCostTable = 0; + } + return SQLITE_OK; + } + pVTab->zErrMsg = sqlite3_mprintf("unknown value for %s.command: \"%w\"", + p->zTableName, zCmd); + return SQLITE_ERROR; + } + if( iRank<1 ) iRank = 1; + if( zSoundslike ){ + zK1 = (char*)transliterate(zSoundslike, nSoundslike); + }else{ + zK1 = (char*)transliterate(zWord, nWord); + } + if( zK1==0 ) return SQLITE_NOMEM; + for(i=0; (c = zK1[i])!=0; i++){ + if( c>='A' && c<='Z' ) zK1[i] += 'a' - 'A'; + } + zK2 = (char*)phoneticHash((const unsigned char*)zK1, i); + if( zK2==0 ){ + sqlite3_free(zK1); + return SQLITE_NOMEM; + } + if( sqlite3_value_type(argv[0])==SQLITE_NULL ){ + spellfix1DbExec(&rc, db, + "INSERT INTO \"%w\".\"%w_vocab\"(rank,langid,word,k1,k2) " + "VALUES(%d,%d,%Q,%Q,%Q)", + p->zDbName, p->zTableName, + iRank, iLang, zWord, zK1, zK2 + ); + *pRowid = sqlite3_last_insert_rowid(db); + }else{ + rowid = sqlite3_value_int64(argv[0]); + newRowid = *pRowid = sqlite3_value_int64(argv[1]); + spellfix1DbExec(&rc, db, + "UPDATE \"%w\".\"%w_vocab\" SET id=%lld, rank=%d, langid=%d," + " word=%Q, k1=%Q, k2=%Q WHERE id=%lld", + p->zDbName, p->zTableName, newRowid, iRank, iLang, + zWord, zK1, zK2, rowid + ); + } + sqlite3_free(zK1); + sqlite3_free(zK2); + } + return rc; +} + +/* +** Rename the spellfix1 table. +*/ +static int spellfix1Rename(sqlite3_vtab *pVTab, const char *zNew){ + spellfix1_vtab *p = (spellfix1_vtab*)pVTab; + sqlite3 *db = p->db; + int rc = SQLITE_OK; + char *zNewName = sqlite3_mprintf("%s", zNew); + if( zNewName==0 ){ + return SQLITE_NOMEM; + } + spellfix1DbExec(&rc, db, + "ALTER TABLE \"%w\".\"%w_vocab\" RENAME TO \"%w_vocab\"", + p->zDbName, p->zTableName, zNewName + ); + if( rc==SQLITE_OK ){ + sqlite3_free(p->zTableName); + p->zTableName = zNewName; + }else{ + sqlite3_free(zNewName); + } + return rc; +} + + +/* +** A virtual table module that provides fuzzy search. +*/ +static sqlite3_module spellfix1Module = { + 0, /* iVersion */ + spellfix1Create, /* xCreate - handle CREATE VIRTUAL TABLE */ + spellfix1Connect, /* xConnect - reconnected to an existing table */ + spellfix1BestIndex, /* xBestIndex - figure out how to do a query */ + spellfix1Disconnect, /* xDisconnect - close a connection */ + spellfix1Destroy, /* xDestroy - handle DROP TABLE */ + spellfix1Open, /* xOpen - open a cursor */ + spellfix1Close, /* xClose - close a cursor */ + spellfix1Filter, /* xFilter - configure scan constraints */ + spellfix1Next, /* xNext - advance a cursor */ + spellfix1Eof, /* xEof - check for end of scan */ + spellfix1Column, /* xColumn - read data */ + spellfix1Rowid, /* xRowid - read data */ + spellfix1Update, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + spellfix1Rename, /* xRename */ +}; + +/* +** Register the various functions and the virtual table. +*/ +static int spellfix1Register(sqlite3 *db){ + int rc = SQLITE_OK; + int i; + rc = sqlite3_create_function(db, "spellfix1_translit", 1, SQLITE_UTF8, 0, + transliterateSqlFunc, 0, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "spellfix1_editdist", 2, SQLITE_UTF8, 0, + editdistSqlFunc, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "spellfix1_phonehash", 1, SQLITE_UTF8, 0, + phoneticHashSqlFunc, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function(db, "spellfix1_scriptcode", 1, SQLITE_UTF8, 0, + scriptCodeSqlFunc, 0, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_module(db, "spellfix1", &spellfix1Module, 0); + } + if( rc==SQLITE_OK ){ + rc = editDist3Install(db); + } + + /* Verify sanity of the translit[] table */ + for(i=0; i +#include + +#ifndef SQLITE_OMIT_VIRTUALTABLE + + +/* A wholenumber cursor object */ +typedef struct wholenumber_cursor wholenumber_cursor; +struct wholenumber_cursor { + sqlite3_vtab_cursor base; /* Base class - must be first */ + sqlite3_int64 iValue; /* Current value */ + sqlite3_int64 mxValue; /* Maximum value */ +}; + +/* Methods for the wholenumber module */ +static int wholenumberConnect( + sqlite3 *db, + void *pAux, + int argc, const char *const*argv, + sqlite3_vtab **ppVtab, + char **pzErr +){ + sqlite3_vtab *pNew; + pNew = *ppVtab = sqlite3_malloc( sizeof(*pNew) ); + if( pNew==0 ) return SQLITE_NOMEM; + sqlite3_declare_vtab(db, "CREATE TABLE x(value)"); + memset(pNew, 0, sizeof(*pNew)); + return SQLITE_OK; +} +/* Note that for this virtual table, the xCreate and xConnect +** methods are identical. */ + +static int wholenumberDisconnect(sqlite3_vtab *pVtab){ + sqlite3_free(pVtab); + return SQLITE_OK; +} +/* The xDisconnect and xDestroy methods are also the same */ + + +/* +** Open a new wholenumber cursor. +*/ +static int wholenumberOpen(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ + wholenumber_cursor *pCur; + pCur = sqlite3_malloc( sizeof(*pCur) ); + if( pCur==0 ) return SQLITE_NOMEM; + memset(pCur, 0, sizeof(*pCur)); + *ppCursor = &pCur->base; + return SQLITE_OK; +} + +/* +** Close a wholenumber cursor. +*/ +static int wholenumberClose(sqlite3_vtab_cursor *cur){ + sqlite3_free(cur); + return SQLITE_OK; +} + + +/* +** Advance a cursor to its next row of output +*/ +static int wholenumberNext(sqlite3_vtab_cursor *cur){ + wholenumber_cursor *pCur = (wholenumber_cursor*)cur; + pCur->iValue++; + return SQLITE_OK; +} + +/* +** Return the value associated with a wholenumber. +*/ +static int wholenumberColumn( + sqlite3_vtab_cursor *cur, + sqlite3_context *ctx, + int i +){ + wholenumber_cursor *pCur = (wholenumber_cursor*)cur; + sqlite3_result_int64(ctx, pCur->iValue); + return SQLITE_OK; +} + +/* +** The rowid. +*/ +static int wholenumberRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ + wholenumber_cursor *pCur = (wholenumber_cursor*)cur; + *pRowid = pCur->iValue; + return SQLITE_OK; +} + +/* +** When the wholenumber_cursor.rLimit value is 0 or less, that is a signal +** that the cursor has nothing more to output. +*/ +static int wholenumberEof(sqlite3_vtab_cursor *cur){ + wholenumber_cursor *pCur = (wholenumber_cursor*)cur; + return pCur->iValue>pCur->mxValue || pCur->iValue==0; +} + +/* +** Called to "rewind" a cursor back to the beginning so that +** it starts its output over again. Always called at least once +** prior to any wholenumberColumn, wholenumberRowid, or wholenumberEof call. +** +** idxNum Constraints +** ------ --------------------- +** 0 (none) +** 1 value > $argv0 +** 2 value >= $argv0 +** 4 value < $argv0 +** 8 value <= $argv0 +** +** 5 value > $argv0 AND value < $argv1 +** 6 value >= $argv0 AND value < $argv1 +** 9 value > $argv0 AND value <= $argv1 +** 10 value >= $argv0 AND value <= $argv1 +*/ +static int wholenumberFilter( + sqlite3_vtab_cursor *pVtabCursor, + int idxNum, const char *idxStr, + int argc, sqlite3_value **argv +){ + wholenumber_cursor *pCur = (wholenumber_cursor *)pVtabCursor; + sqlite3_int64 v; + int i = 0; + pCur->iValue = 1; + pCur->mxValue = 0xffffffff; /* 4294967295 */ + if( idxNum & 3 ){ + v = sqlite3_value_int64(argv[0]) + (idxNum&1); + if( v>pCur->iValue && v<=pCur->mxValue ) pCur->iValue = v; + i++; + } + if( idxNum & 12 ){ + v = sqlite3_value_int64(argv[i]) - ((idxNum>>2)&1); + if( v>=pCur->iValue && vmxValue ) pCur->mxValue = v; + } + return SQLITE_OK; +} + +/* +** Search for terms of these forms: +** +** (1) value > $value +** (2) value >= $value +** (4) value < $value +** (8) value <= $value +** +** idxNum is an ORed combination of 1 or 2 with 4 or 8. +*/ +static int wholenumberBestIndex( + sqlite3_vtab *tab, + sqlite3_index_info *pIdxInfo +){ + int i; + int idxNum = 0; + int argvIdx = 1; + int ltIdx = -1; + int gtIdx = -1; + const struct sqlite3_index_constraint *pConstraint; + pConstraint = pIdxInfo->aConstraint; + for(i=0; inConstraint; i++, pConstraint++){ + if( pConstraint->usable==0 ) continue; + if( (idxNum & 3)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_GT ){ + idxNum |= 1; + ltIdx = i; + } + if( (idxNum & 3)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_GE ){ + idxNum |= 2; + ltIdx = i; + } + if( (idxNum & 12)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ){ + idxNum |= 4; + gtIdx = i; + } + if( (idxNum & 12)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE ){ + idxNum |= 8; + gtIdx = i; + } + } + pIdxInfo->idxNum = idxNum; + if( ltIdx>=0 ){ + pIdxInfo->aConstraintUsage[ltIdx].argvIndex = argvIdx++; + pIdxInfo->aConstraintUsage[ltIdx].omit = 1; + } + if( gtIdx>=0 ){ + pIdxInfo->aConstraintUsage[gtIdx].argvIndex = argvIdx; + pIdxInfo->aConstraintUsage[gtIdx].omit = 1; + } + if( pIdxInfo->nOrderBy==1 + && pIdxInfo->aOrderBy[0].desc==0 + ){ + pIdxInfo->orderByConsumed = 1; + } + if( (idxNum & 12)==0 ){ + pIdxInfo->estimatedCost = (double)100000000; + }else if( (idxNum & 3)==0 ){ + pIdxInfo->estimatedCost = (double)5; + }else{ + pIdxInfo->estimatedCost = (double)1; + } + return SQLITE_OK; +} + +/* +** A virtual table module that provides read-only access to a +** Tcl global variable namespace. +*/ +static sqlite3_module wholenumberModule = { + 0, /* iVersion */ + wholenumberConnect, + wholenumberConnect, + wholenumberBestIndex, + wholenumberDisconnect, + wholenumberDisconnect, + wholenumberOpen, /* xOpen - open a cursor */ + wholenumberClose, /* xClose - close a cursor */ + wholenumberFilter, /* xFilter - configure scan constraints */ + wholenumberNext, /* xNext - advance a cursor */ + wholenumberEof, /* xEof - check for end of scan */ + wholenumberColumn, /* xColumn - read data */ + wholenumberRowid, /* xRowid - read data */ + 0, /* xUpdate */ + 0, /* xBegin */ + 0, /* xSync */ + 0, /* xCommit */ + 0, /* xRollback */ + 0, /* xFindMethod */ + 0, /* xRename */ +}; + +#endif /* SQLITE_OMIT_VIRTUALTABLE */ + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_wholenumber_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + int rc = SQLITE_OK; + SQLITE_EXTENSION_INIT2(pApi); +#ifndef SQLITE_OMIT_VIRTUALTABLE + rc = sqlite3_create_module(db, "wholenumber", &wholenumberModule, 0); +#endif + return rc; +} Index: ext/rtree/rtree.c ================================================================== --- ext/rtree/rtree.c +++ ext/rtree/rtree.c @@ -2663,11 +2663,11 @@ RtreeNode *pLeaf = 0; /* Leaf node containing record iDelete */ int iCell; /* Index of iDelete cell in pLeaf */ RtreeNode *pRoot; /* Root node of rtree structure */ - /* Obtain a reference to the root node to initialise Rtree.iDepth */ + /* Obtain a reference to the root node to initialize Rtree.iDepth */ rc = nodeAcquire(pRtree, 1, 0, &pRoot); /* Obtain a reference to the leaf node that contains the entry ** about to be deleted. */ Index: ext/rtree/rtree1.test ================================================================== --- ext/rtree/rtree1.test +++ ext/rtree/rtree1.test @@ -15,10 +15,11 @@ if {![info exists testdir]} { set testdir [file join [file dirname [info script]] .. .. test] } source [file join [file dirname [info script]] rtree_util.tcl] source $testdir/tester.tcl +set testprefix rtree1 # Test plan: # # rtree-1.*: Creating/destroying r-tree tables. # rtree-2.*: Test the implicit constraints - unique rowid and Index: ext/rtree/rtree5.test ================================================================== --- ext/rtree/rtree5.test +++ ext/rtree/rtree5.test @@ -59,11 +59,11 @@ } {1} do_test rtree5-1.10 { execsql { SELECT (1<<31)-5, (1<<31)-1, -1*(1<<31), -1*(1<<31)+5 } } {2147483643 2147483647 -2147483648 -2147483643} -do_test rtree5-1.10 { +do_test rtree5-1.11 { execsql { INSERT INTO t1 VALUES(2, (1<<31)-5, (1<<31)-1, -1*(1<<31), -1*(1<<31)+5) } } {} do_test rtree5-1.12 { ADDED magic.txt Index: magic.txt ================================================================== --- magic.txt +++ magic.txt @@ -0,0 +1,28 @@ +# This file contains suggested magic(5) text for the unix file(1) +# utility for recognizing SQLite3 databases. +# +# When SQLite is used as an application file format, it is desirable to +# have file(1) recognize the database file as being with the specific +# application. You can set the application_id for a database file +# using: +# +# PRAGMA application_id = INTEGER; +# +# INTEGER can be any signed 32-bit integer. That integer is written as +# a 4-byte big-endian integer into offset 68 of the database header. +# +# The Monotone application used "PRAGMA user_version=1598903374;" to set +# its identifier long before "PRAGMA application_id" became available. +# The user_version is very similar to application_id except that it is +# stored at offset 68 instead of offset 60. The application_id pragma +# is preferred. The rule using offset 60 for Monotone is for historical +# compatibility only. +# +0 string =SQLite\ format\ 3 +>68 belong =0x0f055111 Fossil repository - +>68 belong =0x0f055112 Fossil checkout - +>68 belong =0x0f055113 Fossil global configuration - +>68 belong =0x42654462 Bentley Systems BeSQLite Database - +>68 belong =0x42654c6e Bentley Systems Localization File - +>60 belong =0x5f4d544e Monotone source repository - +>0 string =SQLite SQLite3 database Index: main.mk ================================================================== --- main.mk +++ main.mk @@ -53,10 +53,11 @@ LIBOBJ+= alter.o analyze.o attach.o auth.o \ backup.o bitvec.o btmutex.o btree.o build.o \ callback.o complete.o ctime.o date.o delete.o expr.o fault.o fkey.o \ fts3.o fts3_aux.o fts3_expr.o fts3_hash.o fts3_icu.o fts3_porter.o \ fts3_snippet.o fts3_tokenizer.o fts3_tokenizer1.o \ + fts3_tokenize_vtab.o \ fts3_unicode.o fts3_unicode2.o \ fts3_write.o func.o global.o hash.o \ icu.o insert.o journal.o legacy.o loadext.o \ main.o malloc.o mem0.o mem1.o mem2.o mem3.o mem5.o \ memjournal.o \ @@ -195,10 +196,11 @@ $(TOP)/ext/fts3/fts3_porter.c \ $(TOP)/ext/fts3/fts3_snippet.c \ $(TOP)/ext/fts3/fts3_tokenizer.h \ $(TOP)/ext/fts3/fts3_tokenizer.c \ $(TOP)/ext/fts3/fts3_tokenizer1.c \ + $(TOP)/ext/fts3/fts3_tokenize_vtab.c \ $(TOP)/ext/fts3/fts3_unicode.c \ $(TOP)/ext/fts3/fts3_unicode2.c \ $(TOP)/ext/fts3/fts3_write.c SRC += \ $(TOP)/ext/icu/sqliteicu.h \ @@ -240,11 +242,10 @@ $(TOP)/src/test_config.c \ $(TOP)/src/test_demovfs.c \ $(TOP)/src/test_devsym.c \ $(TOP)/src/test_fs.c \ $(TOP)/src/test_func.c \ - $(TOP)/src/test_fuzzer.c \ $(TOP)/src/test_hexio.c \ $(TOP)/src/test_init.c \ $(TOP)/src/test_intarray.c \ $(TOP)/src/test_journal.c \ $(TOP)/src/test_malloc.c \ @@ -252,11 +253,10 @@ $(TOP)/src/test_mutex.c \ $(TOP)/src/test_onefile.c \ $(TOP)/src/test_osinst.c \ $(TOP)/src/test_pcache.c \ $(TOP)/src/test_quota.c \ - $(TOP)/src/test_regexp.c \ $(TOP)/src/test_rtree.c \ $(TOP)/src/test_schema.c \ $(TOP)/src/test_server.c \ $(TOP)/src/test_stat.c \ $(TOP)/src/test_sqllog.c \ @@ -263,12 +263,24 @@ $(TOP)/src/test_superlock.c \ $(TOP)/src/test_syscall.c \ $(TOP)/src/test_tclvar.c \ $(TOP)/src/test_thread.c \ $(TOP)/src/test_vfs.c \ - $(TOP)/src/test_wholenumber.c \ $(TOP)/src/test_wsd.c + +# Extensions to be statically loaded. +# +TESTSRC += \ + $(TOP)/ext/misc/amatch.c \ + $(TOP)/ext/misc/closure.c \ + $(TOP)/ext/misc/fuzzer.c \ + $(TOP)/ext/misc/ieee754.c \ + $(TOP)/ext/misc/nextchar.c \ + $(TOP)/ext/misc/regexp.c \ + $(TOP)/ext/misc/spellfix.c \ + $(TOP)/ext/misc/wholenumber.c + #TESTSRC += $(TOP)/ext/fts2/fts2_tokenizer.c #TESTSRC += $(TOP)/ext/fts3/fts3_tokenizer.c TESTSRC2 = \ @@ -363,10 +375,14 @@ sqlite3$(EXE): $(TOP)/src/shell.c libsqlite3.a sqlite3.h $(TCCX) $(READLINE_FLAGS) -o sqlite3$(EXE) \ $(TOP)/src/shell.c \ libsqlite3.a $(LIBREADLINE) $(TLIBS) $(THREADLIB) +mptester$(EXE): sqlite3.c $(TOP)/mptest/mptest.c + $(TCCX) -o $@ -I. $(TOP)/mptest/mptest.c sqlite3.c \ + $(TLIBS) $(THREADLIB) + sqlite3.o: sqlite3.c $(TCCX) -c sqlite3.c # This target creates a directory named "tsrc" and fills it with # copies of all of the C source code and header files needed to @@ -383,10 +399,11 @@ mv vdbe.new tsrc/vdbe.c touch target_source sqlite3.c: target_source $(TOP)/tool/mksqlite3c.tcl tclsh $(TOP)/tool/mksqlite3c.tcl + cp tsrc/shell.c tsrc/sqlite3ext.h . echo '#ifndef USE_SYSTEM_SQLITE' >tclsqlite3.c cat sqlite3.c >>tclsqlite3.c echo '#endif /* USE_SYSTEM_SQLITE */' >>tclsqlite3.c cat $(TOP)/src/tclsqlite.c >>tclsqlite3.c @@ -508,10 +525,13 @@ fts3_tokenizer.o: $(TOP)/ext/fts3/fts3_tokenizer.c $(HDR) $(EXTHDR) $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer.c fts3_tokenizer1.o: $(TOP)/ext/fts3/fts3_tokenizer1.c $(HDR) $(EXTHDR) $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer1.c + +fts3_tokenize_vtab.o: $(TOP)/ext/fts3/fts3_tokenize_vtab.c $(HDR) $(EXTHDR) + $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenize_vtab.c fts3_unicode.o: $(TOP)/ext/fts3/fts3_unicode.c $(HDR) $(EXTHDR) $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_unicode.c fts3_unicode2.o: $(TOP)/ext/fts3/fts3_unicode2.c $(HDR) $(EXTHDR) @@ -619,7 +639,10 @@ rm -f amalgamation-testfixture amalgamation-testfixture.exe rm -f fts3-testfixture fts3-testfixture.exe rm -f testfixture testfixture.exe rm -f threadtest3 threadtest3.exe rm -f sqlite3.c fts?amal.c tclsqlite3.c + rm -f sqlite3rc.h + rm -f shell.c sqlite3ext.h rm -f sqlite3_analyzer sqlite3_analyzer.exe sqlite3_analyzer.c rm -f sqlite-*-output.vsix + rm -f mptester mptester.exe ADDED mptest/config01.test Index: mptest/config01.test ================================================================== --- mptest/config01.test +++ mptest/config01.test @@ -0,0 +1,46 @@ +/* +** Configure five tasks in different ways, then run tests. +*/ +--if vfsname() GLOB 'unix' +PRAGMA page_size=8192; +--task 1 + PRAGMA journal_mode=PERSIST; + PRAGMA mmap_size=0; +--end +--task 2 + PRAGMA journal_mode=TRUNCATE; + PRAGMA mmap_size=28672; +--end +--task 3 + PRAGMA journal_mode=MEMORY; +--end +--task 4 + PRAGMA journal_mode=OFF; +--end +--task 4 + PRAGMA mmap_size(268435456); +--end +--source multiwrite01.test +--wait all +PRAGMA page_size=16384; +VACUUM; +CREATE TABLE pgsz(taskid, sz INTEGER); +--task 1 + INSERT INTO pgsz VALUES(1, eval('PRAGMA page_size')); +--end +--task 2 + INSERT INTO pgsz VALUES(2, eval('PRAGMA page_size')); +--end +--task 3 + INSERT INTO pgsz VALUES(3, eval('PRAGMA page_size')); +--end +--task 4 + INSERT INTO pgsz VALUES(4, eval('PRAGMA page_size')); +--end +--task 5 + INSERT INTO pgsz VALUES(5, eval('PRAGMA page_size')); +--end +--source multiwrite01.test +--wait all +SELECT sz FROM pgsz; +--match 16384 16384 16384 16384 16384 ADDED mptest/config02.test Index: mptest/config02.test ================================================================== --- mptest/config02.test +++ mptest/config02.test @@ -0,0 +1,123 @@ +/* +** Configure five tasks in different ways, then run tests. +*/ +PRAGMA page_size=512; +--task 1 + PRAGMA mmap_size=0; +--end +--task 2 + PRAGMA mmap_size=28672; +--end +--task 3 + PRAGMA mmap_size=8192; +--end +--task 4 + PRAGMA mmap_size=65536; +--end +--task 5 + PRAGMA mmap_size=268435456; +--end +--source multiwrite01.test +--source crash02.subtest +PRAGMA page_size=1024; +VACUUM; +CREATE TABLE pgsz(taskid, sz INTEGER); +--task 1 + INSERT INTO pgsz VALUES(1, eval('PRAGMA page_size')); +--end +--task 2 + INSERT INTO pgsz VALUES(2, eval('PRAGMA page_size')); +--end +--task 3 + INSERT INTO pgsz VALUES(3, eval('PRAGMA page_size')); +--end +--task 4 + INSERT INTO pgsz VALUES(4, eval('PRAGMA page_size')); +--end +--task 5 + INSERT INTO pgsz VALUES(5, eval('PRAGMA page_size')); +--end +--source multiwrite01.test +--source crash02.subtest +--wait all +SELECT sz FROM pgsz; +--match 1024 1024 1024 1024 1024 +PRAGMA page_size=2048; +VACUUM; +DELETE FROM pgsz; +--task 1 + INSERT INTO pgsz VALUES(1, eval('PRAGMA page_size')); +--end +--task 2 + INSERT INTO pgsz VALUES(2, eval('PRAGMA page_size')); +--end +--task 3 + INSERT INTO pgsz VALUES(3, eval('PRAGMA page_size')); +--end +--task 4 + INSERT INTO pgsz VALUES(4, eval('PRAGMA page_size')); +--end +--task 5 + INSERT INTO pgsz VALUES(5, eval('PRAGMA page_size')); +--end +--source multiwrite01.test +--source crash02.subtest +--wait all +SELECT sz FROM pgsz; +--match 2048 2048 2048 2048 2048 +PRAGMA page_size=8192; +VACUUM; +DELETE FROM pgsz; +--task 1 + INSERT INTO pgsz VALUES(1, eval('PRAGMA page_size')); +--end +--task 2 + INSERT INTO pgsz VALUES(2, eval('PRAGMA page_size')); +--end +--task 3 + INSERT INTO pgsz VALUES(3, eval('PRAGMA page_size')); +--end +--task 4 + INSERT INTO pgsz VALUES(4, eval('PRAGMA page_size')); +--end +--task 5 + INSERT INTO pgsz VALUES(5, eval('PRAGMA page_size')); +--end +--source multiwrite01.test +--source crash02.subtest +--wait all +SELECT sz FROM pgsz; +--match 8192 8192 8192 8192 8192 +PRAGMA page_size=16384; +VACUUM; +DELETE FROM pgsz; +--task 1 + INSERT INTO pgsz VALUES(1, eval('PRAGMA page_size')); +--end +--task 2 + INSERT INTO pgsz VALUES(2, eval('PRAGMA page_size')); +--end +--task 3 + INSERT INTO pgsz VALUES(3, eval('PRAGMA page_size')); +--end +--task 4 + INSERT INTO pgsz VALUES(4, eval('PRAGMA page_size')); +--end +--task 5 + INSERT INTO pgsz VALUES(5, eval('PRAGMA page_size')); +--end +--source multiwrite01.test +--source crash02.subtest +--wait all +SELECT sz FROM pgsz; +--match 16384 16384 16384 16384 16384 +PRAGMA auto_vacuum=FULL; +VACUUM; +--source multiwrite01.test +--source crash02.subtest +--wait all +PRAGMA auto_vacuum=FULL; +PRAGMA page_size=512; +VACUUM; +--source multiwrite01.test +--source crash02.subtest ADDED mptest/crash01.test Index: mptest/crash01.test ================================================================== --- mptest/crash01.test +++ mptest/crash01.test @@ -0,0 +1,102 @@ +/* Test cases involving incomplete transactions that must be rolled back. +*/ +--task 1 + DROP TABLE IF EXISTS t1; + CREATE TABLE t1(a INTEGER PRIMARY KEY, b); + --sleep 1 + INSERT INTO t1 VALUES(1, randomblob(2000)); + INSERT INTO t1 VALUES(2, randomblob(1000)); + --sleep 1 + INSERT INTO t1 SELECT a+2, randomblob(1500) FROM t1; + INSERT INTO t1 SELECT a+4, randomblob(1500) FROM t1; + INSERT INTO t1 SELECT a+8, randomblob(1500) FROM t1; + --sleep 1 + INSERT INTO t1 SELECT a+16, randomblob(1500) FROM t1; + --sleep 1 + INSERT INTO t1 SELECT a+32, randomblob(1500) FROM t1; + SELECT count(*) FROM t1; + --match 64 + SELECT avg(length(b)) FROM t1; + --match 1500.0 + --sleep 2 + UPDATE t1 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t1; + --match 247 + SELECT a FROM t1 WHERE b='x17y'; + --match 17 + CREATE INDEX t1b ON t1(b); + SELECT a FROM t1 WHERE b='x17y'; + --match 17 + SELECT a FROM t1 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end +--wait 1 +--task 2 + CREATE TABLE t2(a INTEGER PRIMARY KEY, b); + INSERT INTO t2 SELECT a, b FROM t1; + UPDATE t1 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t2; + --match 247 + SELECT a FROM t2 WHERE b='x17y'; + --match 17 + CREATE INDEX t2b ON t2(b); + SELECT a FROM t2 WHERE b='x17y'; + --match 17 + SELECT a FROM t2 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end +--task 3 + CREATE TABLE t3(a INTEGER PRIMARY KEY, b); + INSERT INTO t3 SELECT a, b FROM t1; + UPDATE t1 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t3; + --match 247 + SELECT a FROM t3 WHERE b='x17y'; + --match 17 + CREATE INDEX t3b ON t3(b); + SELECT a FROM t3 WHERE b='x17y'; + --match 17 + SELECT a FROM t3 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end +--task 4 + CREATE TABLE t4(a INTEGER PRIMARY KEY, b); + INSERT INTO t4 SELECT a, b FROM t1; + UPDATE t1 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t4; + --match 247 + SELECT a FROM t4 WHERE b='x17y'; + --match 17 + CREATE INDEX t4b ON t4(b); + SELECT a FROM t4 WHERE b='x17y'; + --match 17 + SELECT a FROM t4 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end +--task 5 + CREATE TABLE t5(a INTEGER PRIMARY KEY, b); + INSERT INTO t5 SELECT a, b FROM t1; + UPDATE t1 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t5; + --match 247 + SELECT a FROM t5 WHERE b='x17y'; + --match 17 + CREATE INDEX t5b ON t5(b); + SELECT a FROM t5 WHERE b='x17y'; + --match 17 + SELECT a FROM t5 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end + +--wait all +/* After the database file has been set up, run the crash2 subscript +** multiple times. */ +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest +--source crash02.subtest ADDED mptest/crash02.subtest Index: mptest/crash02.subtest ================================================================== --- mptest/crash02.subtest +++ mptest/crash02.subtest @@ -0,0 +1,53 @@ +/* +** This script is called from crash01.test and config02.test and perhaps other +** script. After the database file has been set up, make a big rollback +** journal in client 1, then crash client 1. +** Then in the other clients, do an integrity check. +*/ +--task 1 leave-hot-journal + --sleep 5 + --finish + PRAGMA cache_size=10; + BEGIN; + UPDATE t1 SET b=randomblob(20000); + UPDATE t2 SET b=randomblob(20000); + UPDATE t3 SET b=randomblob(20000); + UPDATE t4 SET b=randomblob(20000); + UPDATE t5 SET b=randomblob(20000); + UPDATE t1 SET b=NULL; + UPDATE t2 SET b=NULL; + UPDATE t3 SET b=NULL; + UPDATE t4 SET b=NULL; + UPDATE t5 SET b=NULL; + --print Task one crashing an incomplete transaction + --exit 1 +--end +--task 2 integrity_check-2 + SELECT count(*) FROM t1; + --match 64 + --sleep 100 + PRAGMA integrity_check(10); + --match ok +--end +--task 3 integrity_check-3 + SELECT count(*) FROM t1; + --match 64 + --sleep 100 + PRAGMA integrity_check(10); + --match ok +--end +--task 4 integrity_check-4 + SELECT count(*) FROM t1; + --match 64 + --sleep 100 + PRAGMA integrity_check(10); + --match ok +--end +--task 5 integrity_check-5 + SELECT count(*) FROM t1; + --match 64 + --sleep 100 + PRAGMA integrity_check(10); + --match ok +--end +--wait all ADDED mptest/mptest.c Index: mptest/mptest.c ================================================================== --- mptest/mptest.c +++ mptest/mptest.c @@ -0,0 +1,1401 @@ +/* +** 2013-04-05 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** +** This is a program used for testing SQLite, and specifically for testing +** the ability of independent processes to access the same SQLite database +** concurrently. +** +** Compile this program as follows: +** +** gcc -g -c -Wall sqlite3.c $(OPTS) +** gcc -g -o mptest mptest.c sqlite3.o $(LIBS) +** +** Recommended options: +** +** -DHAVE_USLEEP +** -DSQLITE_NO_SYNC +** -DSQLITE_THREADSAFE=0 +** -DSQLITE_OMIT_LOAD_EXTENSION +** +** Run like this: +** +** ./mptest $database $script +** +** where $database is the database to use for testing and $script is a +** test script. +*/ +#include "sqlite3.h" +#include +#if defined(_WIN32) +# define WIN32_LEAN_AND_MEAN +# include +#else +# include +#endif +#include +#include +#include +#include + +/* The suffix to append to the child command lines, if any */ +#if defined(_WIN32) +# define GETPID (int)GetCurrentProcessId +#else +# define GETPID getpid +#endif + +/* Mark a parameter as unused to suppress compiler warnings */ +#define UNUSED_PARAMETER(x) (void)x + +/* Global data +*/ +static struct Global { + char *argv0; /* Name of the executable */ + const char *zVfs; /* Name of VFS to use. Often NULL meaning "default" */ + char *zDbFile; /* Name of the database */ + sqlite3 *db; /* Open connection to database */ + char *zErrLog; /* Filename for error log */ + FILE *pErrLog; /* Where to write errors */ + char *zLog; /* Name of output log file */ + FILE *pLog; /* Where to write log messages */ + char zName[32]; /* Symbolic name of this process */ + int taskId; /* Task ID. 0 means supervisor. */ + int iTrace; /* Tracing level */ + int bSqlTrace; /* True to trace SQL commands */ + int bIgnoreSqlErrors; /* Ignore errors in SQL statements */ + int nError; /* Number of errors */ + int nTest; /* Number of --match operators */ + int iTimeout; /* Milliseconds until a busy timeout */ + int bSync; /* Call fsync() */ +} g; + +/* Default timeout */ +#define DEFAULT_TIMEOUT 10000 + +/* +** Print a message adding zPrefix[] to the beginning of every line. +*/ +static void printWithPrefix(FILE *pOut, const char *zPrefix, const char *zMsg){ + while( zMsg && zMsg[0] ){ + int i; + for(i=0; zMsg[i] && zMsg[i]!='\n' && zMsg[i]!='\r'; i++){} + fprintf(pOut, "%s%.*s\n", zPrefix, i, zMsg); + zMsg += i; + while( zMsg[0]=='\n' || zMsg[0]=='\r' ) zMsg++; + } +} + +/* +** Compare two pointers to strings, where the pointers might be NULL. +*/ +static int safe_strcmp(const char *a, const char *b){ + if( a==b ) return 0; + if( a==0 ) return -1; + if( b==0 ) return 1; + return strcmp(a,b); +} + +/* +** Return TRUE if string z[] matches glob pattern zGlob[]. +** Return FALSE if the pattern does not match. +** +** Globbing rules: +** +** '*' Matches any sequence of zero or more characters. +** +** '?' Matches exactly one character. +** +** [...] Matches one character from the enclosed list of +** characters. +** +** [^...] Matches one character not in the enclosed list. +** +** '#' Matches any sequence of one or more digits with an +** optional + or - sign in front +*/ +int strglob(const char *zGlob, const char *z){ + int c, c2; + int invert; + int seen; + + while( (c = (*(zGlob++)))!=0 ){ + if( c=='*' ){ + while( (c=(*(zGlob++))) == '*' || c=='?' ){ + if( c=='?' && (*(z++))==0 ) return 0; + } + if( c==0 ){ + return 1; + }else if( c=='[' ){ + while( *z && strglob(zGlob-1,z) ){ + z++; + } + return (*z)!=0; + } + while( (c2 = (*(z++)))!=0 ){ + while( c2!=c ){ + c2 = *(z++); + if( c2==0 ) return 0; + } + if( strglob(zGlob,z) ) return 1; + } + return 0; + }else if( c=='?' ){ + if( (*(z++))==0 ) return 0; + }else if( c=='[' ){ + int prior_c = 0; + seen = 0; + invert = 0; + c = *(z++); + if( c==0 ) return 0; + c2 = *(zGlob++); + if( c2=='^' ){ + invert = 1; + c2 = *(zGlob++); + } + if( c2==']' ){ + if( c==']' ) seen = 1; + c2 = *(zGlob++); + } + while( c2 && c2!=']' ){ + if( c2=='-' && zGlob[0]!=']' && zGlob[0]!=0 && prior_c>0 ){ + c2 = *(zGlob++); + if( c>=prior_c && c<=c2 ) seen = 1; + prior_c = 0; + }else{ + if( c==c2 ){ + seen = 1; + } + prior_c = c2; + } + c2 = *(zGlob++); + } + if( c2==0 || (seen ^ invert)==0 ) return 0; + }else if( c=='#' ){ + if( (z[0]=='-' || z[0]=='+') && isdigit(z[1]) ) z++; + if( !isdigit(z[0]) ) return 0; + z++; + while( isdigit(z[0]) ){ z++; } + }else{ + if( c!=(*(z++)) ) return 0; + } + } + return *z==0; +} + +/* +** Close output stream pOut if it is not stdout or stderr +*/ +static void maybeClose(FILE *pOut){ + if( pOut!=stdout && pOut!=stderr ) fclose(pOut); +} + +/* +** Print an error message +*/ +static void errorMessage(const char *zFormat, ...){ + va_list ap; + char *zMsg; + char zPrefix[30]; + va_start(ap, zFormat); + zMsg = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + sqlite3_snprintf(sizeof(zPrefix), zPrefix, "%s:ERROR: ", g.zName); + if( g.pLog ){ + printWithPrefix(g.pLog, zPrefix, zMsg); + fflush(g.pLog); + } + if( g.pErrLog && safe_strcmp(g.zErrLog,g.zLog) ){ + printWithPrefix(g.pErrLog, zPrefix, zMsg); + fflush(g.pErrLog); + } + sqlite3_free(zMsg); + g.nError++; +} + +/* Forward declaration */ +static int trySql(const char*, ...); + +/* +** Print an error message and then quit. +*/ +static void fatalError(const char *zFormat, ...){ + va_list ap; + char *zMsg; + char zPrefix[30]; + va_start(ap, zFormat); + zMsg = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + sqlite3_snprintf(sizeof(zPrefix), zPrefix, "%s:FATAL: ", g.zName); + if( g.pLog ){ + printWithPrefix(g.pLog, zPrefix, zMsg); + fflush(g.pLog); + maybeClose(g.pLog); + } + if( g.pErrLog && safe_strcmp(g.zErrLog,g.zLog) ){ + printWithPrefix(g.pErrLog, zPrefix, zMsg); + fflush(g.pErrLog); + maybeClose(g.pErrLog); + } + sqlite3_free(zMsg); + if( g.db ){ + int nTry = 0; + g.iTimeout = 0; + while( trySql("UPDATE client SET wantHalt=1;")==SQLITE_BUSY + && (nTry++)<100 ){ + sqlite3_sleep(10); + } + } + sqlite3_close(g.db); + exit(1); +} + + +/* +** Print a log message +*/ +static void logMessage(const char *zFormat, ...){ + va_list ap; + char *zMsg; + char zPrefix[30]; + va_start(ap, zFormat); + zMsg = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + sqlite3_snprintf(sizeof(zPrefix), zPrefix, "%s: ", g.zName); + if( g.pLog ){ + printWithPrefix(g.pLog, zPrefix, zMsg); + fflush(g.pLog); + } + sqlite3_free(zMsg); +} + +/* +** Return the length of a string omitting trailing whitespace +*/ +static int clipLength(const char *z){ + int n = (int)strlen(z); + while( n>0 && isspace(z[n-1]) ){ n--; } + return n; +} + +/* +** Auxiliary SQL function to return the name of the VFS +*/ +static void vfsNameFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + sqlite3 *db = sqlite3_context_db_handle(context); + char *zVfs = 0; + UNUSED_PARAMETER(argc); + UNUSED_PARAMETER(argv); + sqlite3_file_control(db, "main", SQLITE_FCNTL_VFSNAME, &zVfs); + if( zVfs ){ + sqlite3_result_text(context, zVfs, -1, sqlite3_free); + } +} + +/* +** Busy handler with a g.iTimeout-millisecond timeout +*/ +static int busyHandler(void *pCD, int count){ + UNUSED_PARAMETER(pCD); + if( count*10>g.iTimeout ){ + if( g.iTimeout>0 ) errorMessage("timeout after %dms", g.iTimeout); + return 0; + } + sqlite3_sleep(10); + return 1; +} + +/* +** SQL Trace callback +*/ +static void sqlTraceCallback(void *NotUsed1, const char *zSql){ + UNUSED_PARAMETER(NotUsed1); + logMessage("[%.*s]", clipLength(zSql), zSql); +} + +/* +** SQL error log callback +*/ +static void sqlErrorCallback(void *pArg, int iErrCode, const char *zMsg){ + UNUSED_PARAMETER(pArg); + if( iErrCode==SQLITE_ERROR && g.bIgnoreSqlErrors ) return; + if( (iErrCode&0xff)==SQLITE_SCHEMA && g.iTrace<3 ) return; + if( g.iTimeout==0 && (iErrCode&0xff)==SQLITE_BUSY && g.iTrace<3 ) return; + if( (iErrCode&0xff)==SQLITE_NOTICE ){ + logMessage("(info) %s", zMsg); + }else{ + errorMessage("(errcode=%d) %s", iErrCode, zMsg); + } +} + +/* +** Prepare an SQL statement. Issue a fatal error if unable. +*/ +static sqlite3_stmt *prepareSql(const char *zFormat, ...){ + va_list ap; + char *zSql; + int rc; + sqlite3_stmt *pStmt = 0; + va_start(ap, zFormat); + zSql = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + rc = sqlite3_prepare_v2(g.db, zSql, -1, &pStmt, 0); + if( rc!=SQLITE_OK ){ + sqlite3_finalize(pStmt); + fatalError("%s\n%s\n", sqlite3_errmsg(g.db), zSql); + } + sqlite3_free(zSql); + return pStmt; +} + +/* +** Run arbitrary SQL. Issue a fatal error on failure. +*/ +static void runSql(const char *zFormat, ...){ + va_list ap; + char *zSql; + int rc; + va_start(ap, zFormat); + zSql = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + rc = sqlite3_exec(g.db, zSql, 0, 0, 0); + if( rc!=SQLITE_OK ){ + fatalError("%s\n%s\n", sqlite3_errmsg(g.db), zSql); + } + sqlite3_free(zSql); +} + +/* +** Try to run arbitrary SQL. Return success code. +*/ +static int trySql(const char *zFormat, ...){ + va_list ap; + char *zSql; + int rc; + va_start(ap, zFormat); + zSql = sqlite3_vmprintf(zFormat, ap); + va_end(ap); + rc = sqlite3_exec(g.db, zSql, 0, 0, 0); + sqlite3_free(zSql); + return rc; +} + +/* Structure for holding an arbitrary length string +*/ +typedef struct String String; +struct String { + char *z; /* the string */ + int n; /* Slots of z[] used */ + int nAlloc; /* Slots of z[] allocated */ +}; + +/* Free a string */ +static void stringFree(String *p){ + if( p->z ) sqlite3_free(p->z); + memset(p, 0, sizeof(*p)); +} + +/* Append n bytes of text to a string. If n<0 append the entire string. */ +static void stringAppend(String *p, const char *z, int n){ + if( n<0 ) n = (int)strlen(z); + if( p->n+n>=p->nAlloc ){ + int nAlloc = p->nAlloc*2 + n + 100; + char *z = sqlite3_realloc(p->z, nAlloc); + if( z==0 ) fatalError("out of memory"); + p->z = z; + p->nAlloc = nAlloc; + } + memcpy(p->z+p->n, z, n); + p->n += n; + p->z[p->n] = 0; +} + +/* Reset a string to an empty string */ +static void stringReset(String *p){ + if( p->z==0 ) stringAppend(p, " ", 1); + p->n = 0; + p->z[0] = 0; +} + +/* Append a new token onto the end of the string */ +static void stringAppendTerm(String *p, const char *z){ + int i; + if( p->n ) stringAppend(p, " ", 1); + if( z==0 ){ + stringAppend(p, "nil", 3); + return; + } + for(i=0; z[i] && !isspace(z[i]); i++){} + if( i>0 && z[i]==0 ){ + stringAppend(p, z, i); + return; + } + stringAppend(p, "'", 1); + while( z[0] ){ + for(i=0; z[i] && z[i]!='\''; i++){} + if( z[i] ){ + stringAppend(p, z, i+1); + stringAppend(p, "'", 1); + z += i+1; + }else{ + stringAppend(p, z, i); + break; + } + } + stringAppend(p, "'", 1); +} + +/* +** Callback function for evalSql() +*/ +static int evalCallback(void *pCData, int argc, char **argv, char **azCol){ + String *p = (String*)pCData; + int i; + UNUSED_PARAMETER(azCol); + for(i=0; i0 ); + rc = sqlite3_exec(g.db, zSql, evalCallback, p, &zErrMsg); + sqlite3_free(zSql); + if( rc ){ + char zErr[30]; + sqlite3_snprintf(sizeof(zErr), zErr, "error(%d)", rc); + stringAppendTerm(p, zErr); + if( zErrMsg ){ + stringAppendTerm(p, zErrMsg); + sqlite3_free(zErrMsg); + } + } + return rc; +} + +/* +** Auxiliary SQL function to recursively evaluate SQL. +*/ +static void evalFunc( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + sqlite3 *db = sqlite3_context_db_handle(context); + const char *zSql = (const char*)sqlite3_value_text(argv[0]); + String res; + char *zErrMsg = 0; + int rc; + UNUSED_PARAMETER(argc); + memset(&res, 0, sizeof(res)); + rc = sqlite3_exec(db, zSql, evalCallback, &res, &zErrMsg); + if( zErrMsg ){ + sqlite3_result_error(context, zErrMsg, -1); + sqlite3_free(zErrMsg); + }else if( rc ){ + sqlite3_result_error_code(context, rc); + }else{ + sqlite3_result_text(context, res.z, -1, SQLITE_TRANSIENT); + } + stringFree(&res); +} + +/* +** Look up the next task for client iClient in the database. +** Return the task script and the task number and mark that +** task as being under way. +*/ +static int startScript( + int iClient, /* The client number */ + char **pzScript, /* Write task script here */ + int *pTaskId, /* Write task number here */ + char **pzTaskName /* Name of the task */ +){ + sqlite3_stmt *pStmt = 0; + int taskId; + int rc; + int totalTime = 0; + + *pzScript = 0; + g.iTimeout = 0; + while(1){ + rc = trySql("BEGIN IMMEDIATE"); + if( rc==SQLITE_BUSY ){ + sqlite3_sleep(10); + totalTime += 10; + continue; + } + if( rc!=SQLITE_OK ){ + fatalError("in startScript: %s", sqlite3_errmsg(g.db)); + } + if( g.nError || g.nTest ){ + runSql("UPDATE counters SET nError=nError+%d, nTest=nTest+%d", + g.nError, g.nTest); + g.nError = 0; + g.nTest = 0; + } + pStmt = prepareSql("SELECT 1 FROM client WHERE id=%d AND wantHalt",iClient); + rc = sqlite3_step(pStmt); + sqlite3_finalize(pStmt); + if( rc==SQLITE_ROW ){ + runSql("DELETE FROM client WHERE id=%d", iClient); + g.iTimeout = DEFAULT_TIMEOUT; + runSql("COMMIT TRANSACTION;"); + return SQLITE_DONE; + } + pStmt = prepareSql( + "SELECT script, id, name FROM task" + " WHERE client=%d AND starttime IS NULL" + " ORDER BY id LIMIT 1", iClient); + rc = sqlite3_step(pStmt); + if( rc==SQLITE_ROW ){ + int n = sqlite3_column_bytes(pStmt, 0); + *pzScript = sqlite3_malloc(n+1); + strcpy(*pzScript, (const char*)sqlite3_column_text(pStmt, 0)); + *pTaskId = taskId = sqlite3_column_int(pStmt, 1); + *pzTaskName = sqlite3_mprintf("%s", sqlite3_column_text(pStmt, 2)); + sqlite3_finalize(pStmt); + runSql("UPDATE task" + " SET starttime=strftime('%%Y-%%m-%%d %%H:%%M:%%f','now')" + " WHERE id=%d;", taskId); + g.iTimeout = DEFAULT_TIMEOUT; + runSql("COMMIT TRANSACTION;"); + return SQLITE_OK; + } + sqlite3_finalize(pStmt); + if( rc==SQLITE_DONE ){ + if( totalTime>30000 ){ + errorMessage("Waited over 30 seconds with no work. Giving up."); + runSql("DELETE FROM client WHERE id=%d; COMMIT;", iClient); + sqlite3_close(g.db); + exit(1); + } + while( trySql("COMMIT")==SQLITE_BUSY ){ + sqlite3_sleep(10); + totalTime += 10; + } + sqlite3_sleep(100); + totalTime += 100; + continue; + } + fatalError("%s", sqlite3_errmsg(g.db)); + } + g.iTimeout = DEFAULT_TIMEOUT; +} + +/* +** Mark a script as having finished. Remove the CLIENT table entry +** if bShutdown is true. +*/ +static int finishScript(int iClient, int taskId, int bShutdown){ + runSql("UPDATE task" + " SET endtime=strftime('%%Y-%%m-%%d %%H:%%M:%%f','now')" + " WHERE id=%d;", taskId); + if( bShutdown ){ + runSql("DELETE FROM client WHERE id=%d", iClient); + } + return SQLITE_OK; +} + +/* +** Start up a client process for iClient, if it is not already +** running. If the client is already running, then this routine +** is a no-op. +*/ +static void startClient(int iClient){ + runSql("INSERT OR IGNORE INTO client VALUES(%d,0)", iClient); + if( sqlite3_changes(g.db) ){ + char *zSys; + int rc; + zSys = sqlite3_mprintf("%s \"%s\" --client %d --trace %d", + g.argv0, g.zDbFile, iClient, g.iTrace); + if( g.bSqlTrace ){ + zSys = sqlite3_mprintf("%z --sqltrace", zSys); + } + if( g.bSync ){ + zSys = sqlite3_mprintf("%z --sync", zSys); + } + if( g.zVfs ){ + zSys = sqlite3_mprintf("%z --vfs \"%s\"", zSys, g.zVfs); + } + if( g.iTrace>=2 ) logMessage("system('%q')", zSys); +#if !defined(_WIN32) + zSys = sqlite3_mprintf("%z &", zSys); + rc = system(zSys); + if( rc ) errorMessage("system() fails with error code %d", rc); +#else + { + STARTUPINFOA startupInfo; + PROCESS_INFORMATION processInfo; + memset(&startupInfo, 0, sizeof(startupInfo)); + startupInfo.cb = sizeof(startupInfo); + memset(&processInfo, 0, sizeof(processInfo)); + rc = CreateProcessA(NULL, zSys, NULL, NULL, FALSE, 0, NULL, NULL, + &startupInfo, &processInfo); + if( rc ){ + CloseHandle(processInfo.hThread); + CloseHandle(processInfo.hProcess); + }else{ + errorMessage("CreateProcessA() fails with error code %lu", + GetLastError()); + } + } +#endif + sqlite3_free(zSys); + } +} + +/* +** Read the entire content of a file into memory +*/ +static char *readFile(const char *zFilename){ + FILE *in = fopen(zFilename, "rb"); + long sz; + char *z; + if( in==0 ){ + fatalError("cannot open \"%s\" for reading", zFilename); + } + fseek(in, 0, SEEK_END); + sz = ftell(in); + rewind(in); + z = sqlite3_malloc( sz+1 ); + sz = (long)fread(z, 1, sz, in); + z[sz] = 0; + fclose(in); + return z; +} + +/* +** Return the length of the next token. +*/ +static int tokenLength(const char *z, int *pnLine){ + int n = 0; + if( isspace(z[0]) || (z[0]=='/' && z[1]=='*') ){ + int inC = 0; + int c; + if( z[0]=='/' ){ + inC = 1; + n = 2; + } + while( (c = z[n++])!=0 ){ + if( c=='\n' ) (*pnLine)++; + if( isspace(c) ) continue; + if( inC && c=='*' && z[n]=='/' ){ + n++; + inC = 0; + }else if( !inC && c=='/' && z[n]=='*' ){ + n++; + inC = 1; + }else if( !inC ){ + break; + } + } + n--; + }else if( z[0]=='-' && z[1]=='-' ){ + for(n=2; z[n] && z[n]!='\n'; n++){} + if( z[n] ){ (*pnLine)++; n++; } + }else if( z[0]=='"' || z[0]=='\'' ){ + int delim = z[0]; + for(n=1; z[n]; n++){ + if( z[n]=='\n' ) (*pnLine)++; + if( z[n]==delim ){ + n++; + if( z[n+1]!=delim ) break; + } + } + }else{ + int c; + for(n=1; (c = z[n])!=0 && !isspace(c) && c!='"' && c!='\'' && c!=';'; n++){} + } + return n; +} + +/* +** Copy a single token into a string buffer. +*/ +static int extractToken(const char *zIn, int nIn, char *zOut, int nOut){ + int i; + if( nIn<=0 ){ + zOut[0] = 0; + return 0; + } + for(i=0; i0 ){ + pStmt = prepareSql( + "SELECT 1 FROM task" + " WHERE client=%d" + " AND client IN (SELECT id FROM client)" + " AND endtime IS NULL", + iClient); + }else{ + pStmt = prepareSql( + "SELECT 1 FROM task" + " WHERE client IN (SELECT id FROM client)" + " AND endtime IS NULL"); + } + g.iTimeout = 0; + while( ((rc = sqlite3_step(pStmt))==SQLITE_BUSY || rc==SQLITE_ROW) + && iTimeout>0 + ){ + sqlite3_reset(pStmt); + sqlite3_sleep(50); + iTimeout -= 50; + } + sqlite3_finalize(pStmt); + g.iTimeout = DEFAULT_TIMEOUT; + if( rc!=SQLITE_DONE ){ + if( zErrPrefix==0 ) zErrPrefix = ""; + if( iClient>0 ){ + errorMessage("%stimeout waiting for client %d", zErrPrefix, iClient); + }else{ + errorMessage("%stimeout waiting for all clients", zErrPrefix); + } + } +} + +/* Return a pointer to the tail of a filename +*/ +static char *filenameTail(char *z){ + int i, j; + for(i=j=0; z[i]; i++) if( z[i]=='/' ) j = i+1; + return z+j; +} + +/* +** Interpret zArg as a boolean value. Return either 0 or 1. +*/ +static int booleanValue(char *zArg){ + int i; + if( zArg==0 ) return 0; + for(i=0; zArg[i]>='0' && zArg[i]<='9'; i++){} + if( i>0 && zArg[i]==0 ) return atoi(zArg); + if( sqlite3_stricmp(zArg, "on")==0 || sqlite3_stricmp(zArg,"yes")==0 ){ + return 1; + } + if( sqlite3_stricmp(zArg, "off")==0 || sqlite3_stricmp(zArg,"no")==0 ){ + return 0; + } + errorMessage("unknown boolean: [%s]", zArg); + return 0; +} + + +/* This routine exists as a convenient place to set a debugger +** breakpoint. +*/ +static void test_breakpoint(void){ static volatile int cnt = 0; cnt++; } + +/* Maximum number of arguments to a --command */ +#define MX_ARG 2 + +/* +** Run a script. +*/ +static void runScript( + int iClient, /* The client number, or 0 for the master */ + int taskId, /* The task ID for clients. 0 for master */ + char *zScript, /* Text of the script */ + char *zFilename /* File from which script was read. */ +){ + int lineno = 1; + int prevLine = 1; + int ii = 0; + int iBegin = 0; + int n, c, j; + int len; + int nArg; + String sResult; + char zCmd[30]; + char zError[1000]; + char azArg[MX_ARG][100]; + + memset(&sResult, 0, sizeof(sResult)); + stringReset(&sResult); + while( (c = zScript[ii])!=0 ){ + prevLine = lineno; + len = tokenLength(zScript+ii, &lineno); + if( isspace(c) || (c=='/' && zScript[ii+1]=='*') ){ + ii += len; + continue; + } + if( c!='-' || zScript[ii+1]!='-' || !isalpha(zScript[ii+2]) ){ + ii += len; + continue; + } + + /* Run any prior SQL before processing the new --command */ + if( ii>iBegin ){ + char *zSql = sqlite3_mprintf("%.*s", ii-iBegin, zScript+iBegin); + evalSql(&sResult, zSql); + sqlite3_free(zSql); + iBegin = ii + len; + } + + /* Parse the --command */ + if( g.iTrace>=2 ) logMessage("%.*s", len, zScript+ii); + n = extractToken(zScript+ii+2, len-2, zCmd, sizeof(zCmd)); + for(nArg=0; n=len-2 ) break; + n += extractToken(zScript+ii+2+n, len-2-n, + azArg[nArg], sizeof(azArg[nArg])); + } + for(j=nArg; j0 then exit without shutting down + ** SQLite. (In other words, simulate a crash.) + */ + if( strcmp(zCmd, "exit")==0 ){ + int rc = atoi(azArg[0]); + finishScript(iClient, taskId, 1); + if( rc==0 ) sqlite3_close(g.db); + exit(rc); + }else + + /* + ** --testcase NAME + ** + ** Begin a new test case. Announce in the log that the test case + ** has begun. + */ + if( strcmp(zCmd, "testcase")==0 ){ + if( g.iTrace==1 ) logMessage("%.*s", len - 1, zScript+ii); + stringReset(&sResult); + }else + + /* + ** --finish + ** + ** Mark the current task as having finished, even if it is not. + ** This can be used in conjunction with --exit to simulate a crash. + */ + if( strcmp(zCmd, "finish")==0 && iClient>0 ){ + finishScript(iClient, taskId, 1); + }else + + /* + ** --reset + ** + ** Reset accumulated results back to an empty string + */ + if( strcmp(zCmd, "reset")==0 ){ + stringReset(&sResult); + }else + + /* + ** --match ANSWER... + ** + ** Check to see if output matches ANSWER. Report an error if not. + */ + if( strcmp(zCmd, "match")==0 ){ + int jj; + char *zAns = zScript+ii; + for(jj=7; jj=0 && zFilename[k]!='/'; k--){} + if( k>0 ){ + zNewFile = zToDel = sqlite3_mprintf("%.*s/%s", k,zFilename,zNewFile); + } + } + zNewScript = readFile(zNewFile); + if( g.iTrace ) logMessage("begin script [%s]\n", zNewFile); + runScript(0, 0, zNewScript, zNewFile); + sqlite3_free(zNewScript); + if( g.iTrace ) logMessage("end script [%s]\n", zNewFile); + sqlite3_free(zToDel); + }else + + /* + ** --print MESSAGE.... + ** + ** Output the remainder of the line to the log file + */ + if( strcmp(zCmd, "print")==0 ){ + int jj; + for(jj=7; jj0 ){ + startClient(iNewClient); + } + }else + + /* + ** --wait CLIENT TIMEOUT + ** + ** Wait until all tasks complete for the given client. If CLIENT is + ** "all" then wait for all clients to complete. Wait no longer than + ** TIMEOUT milliseconds (default 10,000) + */ + if( strcmp(zCmd, "wait")==0 && iClient==0 ){ + int iTimeout = nArg>=2 ? atoi(azArg[1]) : 10000; + sqlite3_snprintf(sizeof(zError),zError,"line %d of %s\n", + prevLine, zFilename); + waitForClient(atoi(azArg[0]), iTimeout, zError); + }else + + /* + ** --task CLIENT + ** + ** --end + ** + ** Assign work to a client. Start the client if it is not running + ** already. + */ + if( strcmp(zCmd, "task")==0 && iClient==0 ){ + int iTarget = atoi(azArg[0]); + int iEnd; + char *zTask; + char *zTName; + iEnd = findEnd(zScript+ii+len, &lineno); + if( iTarget<0 ){ + errorMessage("line %d of %s: bad client number: %d", + prevLine, zFilename, iTarget); + }else{ + zTask = sqlite3_mprintf("%.*s", iEnd, zScript+ii+len); + if( nArg>1 ){ + zTName = sqlite3_mprintf("%s", azArg[1]); + }else{ + zTName = sqlite3_mprintf("%s:%d", filenameTail(zFilename), prevLine); + } + startClient(iTarget); + runSql("INSERT INTO task(client,script,name)" + " VALUES(%d,'%q',%Q)", iTarget, zTask, zTName); + sqlite3_free(zTask); + sqlite3_free(zTName); + } + iEnd += tokenLength(zScript+ii+len+iEnd, &lineno); + len += iEnd; + iBegin = ii+len; + }else + + /* + ** --breakpoint + ** + ** This command calls "test_breakpoint()" which is a routine provided + ** as a convenient place to set a debugger breakpoint. + */ + if( strcmp(zCmd, "breakpoint")==0 ){ + test_breakpoint(); + }else + + /* + ** --show-sql-errors BOOLEAN + ** + ** Turn display of SQL errors on and off. + */ + if( strcmp(zCmd, "show-sql-errors")==0 ){ + g.bIgnoreSqlErrors = nArg>=1 ? !booleanValue(azArg[0]) : 1; + }else + + + /* error */{ + errorMessage("line %d of %s: unknown command --%s", + prevLine, zFilename, zCmd); + } + ii += len; + } + if( iBegin= nArg ) break; + z = azArg[i]; + if( z[0]!='-' ) continue; + z++; + if( z[0]=='-' ){ + if( z[1]==0 ) break; + z++; + } + if( strcmp(z,zOption)==0 ){ + if( hasArg && i==nArg-1 ){ + fatalError("command-line option \"--%s\" requires an argument", z); + } + if( hasArg ){ + zReturn = azArg[i+1]; + }else{ + zReturn = azArg[i]; + } + j = i+1+(hasArg!=0); + while( j0 ){ + printf("With SQLite " SQLITE_VERSION " " SQLITE_SOURCE_ID "\n" ); + for(i=0; (zCOption = sqlite3_compileoption_get(i))!=0; i++){ + printf("-DSQLITE_%s\n", zCOption); + } + fflush(stdout); + } + iClient = 0; + unlink(g.zDbFile); + openFlags |= SQLITE_OPEN_CREATE; + } + rc = sqlite3_open_v2(g.zDbFile, &g.db, openFlags, g.zVfs); + if( rc ) fatalError("cannot open [%s]", g.zDbFile); + sqlite3_enable_load_extension(g.db, 1); + sqlite3_busy_handler(g.db, busyHandler, 0); + sqlite3_create_function(g.db, "vfsname", 0, SQLITE_UTF8, 0, + vfsNameFunc, 0, 0); + sqlite3_create_function(g.db, "eval", 1, SQLITE_UTF8, 0, + evalFunc, 0, 0); + g.iTimeout = DEFAULT_TIMEOUT; + if( g.bSqlTrace ) sqlite3_trace(g.db, sqlTraceCallback, 0); + if( !g.bSync ) trySql("PRAGMA synchronous=OFF"); + if( iClient>0 ){ + if( n>0 ) unrecognizedArguments(argv[0], n, argv+2); + if( g.iTrace ) logMessage("start-client"); + while(1){ + char *zTaskName = 0; + rc = startScript(iClient, &zScript, &taskId, &zTaskName); + if( rc==SQLITE_DONE ) break; + if( g.iTrace ) logMessage("begin %s (%d)", zTaskName, taskId); + runScript(iClient, taskId, zScript, zTaskName); + if( g.iTrace ) logMessage("end %s (%d)", zTaskName, taskId); + finishScript(iClient, taskId, 0); + sqlite3_free(zTaskName); + sqlite3_sleep(10); + } + if( g.iTrace ) logMessage("end-client"); + }else{ + sqlite3_stmt *pStmt; + int iTimeout; + if( n==0 ){ + fatalError("missing script filename"); + } + if( n>1 ) unrecognizedArguments(argv[0], n, argv+2); + runSql( + "CREATE TABLE task(\n" + " id INTEGER PRIMARY KEY,\n" + " name TEXT,\n" + " client INTEGER,\n" + " starttime DATE,\n" + " endtime DATE,\n" + " script TEXT\n" + ");" + "CREATE INDEX task_i1 ON task(client, starttime);\n" + "CREATE INDEX task_i2 ON task(client, endtime);\n" + "CREATE TABLE counters(nError,nTest);\n" + "INSERT INTO counters VALUES(0,0);\n" + "CREATE TABLE client(id INTEGER PRIMARY KEY, wantHalt);\n" + ); + zScript = readFile(argv[2]); + if( g.iTrace ) logMessage("begin script [%s]\n", argv[2]); + runScript(0, 0, zScript, argv[2]); + sqlite3_free(zScript); + if( g.iTrace ) logMessage("end script [%s]\n", argv[2]); + waitForClient(0, 2000, "during shutdown...\n"); + trySql("UPDATE client SET wantHalt=1"); + sqlite3_sleep(10); + g.iTimeout = 0; + iTimeout = 1000; + while( ((rc = trySql("SELECT 1 FROM client"))==SQLITE_BUSY + || rc==SQLITE_ROW) && iTimeout>0 ){ + sqlite3_sleep(10); + iTimeout -= 10; + } + sqlite3_sleep(100); + pStmt = prepareSql("SELECT nError, nTest FROM counters"); + iTimeout = 1000; + while( (rc = sqlite3_step(pStmt))==SQLITE_BUSY && iTimeout>0 ){ + sqlite3_sleep(10); + iTimeout -= 10; + } + if( rc==SQLITE_ROW ){ + g.nError += sqlite3_column_int(pStmt, 0); + g.nTest += sqlite3_column_int(pStmt, 1); + } + sqlite3_finalize(pStmt); + } + sqlite3_close(g.db); + maybeClose(g.pLog); + maybeClose(g.pErrLog); + if( iClient==0 ){ + printf("Summary: %d errors in %d tests\n", g.nError, g.nTest); + } + return g.nError>0; +} ADDED mptest/multiwrite01.test Index: mptest/multiwrite01.test ================================================================== --- mptest/multiwrite01.test +++ mptest/multiwrite01.test @@ -0,0 +1,405 @@ +/* +** This script sets up five different tasks all writing and updating +** the database at the same time, but each in its own table. +*/ +--task 1 build-t1 + DROP TABLE IF EXISTS t1; + CREATE TABLE t1(a INTEGER PRIMARY KEY, b); + --sleep 1 + INSERT INTO t1 VALUES(1, randomblob(2000)); + INSERT INTO t1 VALUES(2, randomblob(1000)); + --sleep 1 + INSERT INTO t1 SELECT a+2, randomblob(1500) FROM t1; + INSERT INTO t1 SELECT a+4, randomblob(1500) FROM t1; + INSERT INTO t1 SELECT a+8, randomblob(1500) FROM t1; + --sleep 1 + INSERT INTO t1 SELECT a+16, randomblob(1500) FROM t1; + --sleep 1 + INSERT INTO t1 SELECT a+32, randomblob(1500) FROM t1; + SELECT count(*) FROM t1; + --match 64 + SELECT avg(length(b)) FROM t1; + --match 1500.0 + --sleep 2 + UPDATE t1 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t1; + --match 247 + SELECT a FROM t1 WHERE b='x17y'; + --match 17 + CREATE INDEX t1b ON t1(b); + SELECT a FROM t1 WHERE b='x17y'; + --match 17 + SELECT a FROM t1 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end + + +--task 2 build-t2 + DROP TABLE IF EXISTS t2; + CREATE TABLE t2(a INTEGER PRIMARY KEY, b); + --sleep 1 + INSERT INTO t2 VALUES(1, randomblob(2000)); + INSERT INTO t2 VALUES(2, randomblob(1000)); + --sleep 1 + INSERT INTO t2 SELECT a+2, randomblob(1500) FROM t2; + INSERT INTO t2 SELECT a+4, randomblob(1500) FROM t2; + INSERT INTO t2 SELECT a+8, randomblob(1500) FROM t2; + --sleep 1 + INSERT INTO t2 SELECT a+16, randomblob(1500) FROM t2; + --sleep 1 + INSERT INTO t2 SELECT a+32, randomblob(1500) FROM t2; + SELECT count(*) FROM t2; + --match 64 + SELECT avg(length(b)) FROM t2; + --match 1500.0 + --sleep 2 + UPDATE t2 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t2; + --match 247 + SELECT a FROM t2 WHERE b='x17y'; + --match 17 + CREATE INDEX t2b ON t2(b); + SELECT a FROM t2 WHERE b='x17y'; + --match 17 + SELECT a FROM t2 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end + +--task 3 build-t3 + DROP TABLE IF EXISTS t3; + CREATE TABLE t3(a INTEGER PRIMARY KEY, b); + --sleep 1 + INSERT INTO t3 VALUES(1, randomblob(2000)); + INSERT INTO t3 VALUES(2, randomblob(1000)); + --sleep 1 + INSERT INTO t3 SELECT a+2, randomblob(1500) FROM t3; + INSERT INTO t3 SELECT a+4, randomblob(1500) FROM t3; + INSERT INTO t3 SELECT a+8, randomblob(1500) FROM t3; + --sleep 1 + INSERT INTO t3 SELECT a+16, randomblob(1500) FROM t3; + --sleep 1 + INSERT INTO t3 SELECT a+32, randomblob(1500) FROM t3; + SELECT count(*) FROM t3; + --match 64 + SELECT avg(length(b)) FROM t3; + --match 1500.0 + --sleep 2 + UPDATE t3 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t3; + --match 247 + SELECT a FROM t3 WHERE b='x17y'; + --match 17 + CREATE INDEX t3b ON t3(b); + SELECT a FROM t3 WHERE b='x17y'; + --match 17 + SELECT a FROM t3 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end + +--task 4 build-t4 + DROP TABLE IF EXISTS t4; + CREATE TABLE t4(a INTEGER PRIMARY KEY, b); + --sleep 1 + INSERT INTO t4 VALUES(1, randomblob(2000)); + INSERT INTO t4 VALUES(2, randomblob(1000)); + --sleep 1 + INSERT INTO t4 SELECT a+2, randomblob(1500) FROM t4; + INSERT INTO t4 SELECT a+4, randomblob(1500) FROM t4; + INSERT INTO t4 SELECT a+8, randomblob(1500) FROM t4; + --sleep 1 + INSERT INTO t4 SELECT a+16, randomblob(1500) FROM t4; + --sleep 1 + INSERT INTO t4 SELECT a+32, randomblob(1500) FROM t4; + SELECT count(*) FROM t4; + --match 64 + SELECT avg(length(b)) FROM t4; + --match 1500.0 + --sleep 2 + UPDATE t4 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t4; + --match 247 + SELECT a FROM t4 WHERE b='x17y'; + --match 17 + CREATE INDEX t4b ON t4(b); + SELECT a FROM t4 WHERE b='x17y'; + --match 17 + SELECT a FROM t4 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end + +--task 5 build-t5 + DROP TABLE IF EXISTS t5; + CREATE TABLE t5(a INTEGER PRIMARY KEY, b); + --sleep 1 + INSERT INTO t5 VALUES(1, randomblob(2000)); + INSERT INTO t5 VALUES(2, randomblob(1000)); + --sleep 1 + INSERT INTO t5 SELECT a+2, randomblob(1500) FROM t5; + INSERT INTO t5 SELECT a+4, randomblob(1500) FROM t5; + INSERT INTO t5 SELECT a+8, randomblob(1500) FROM t5; + --sleep 1 + INSERT INTO t5 SELECT a+16, randomblob(1500) FROM t5; + --sleep 1 + INSERT INTO t5 SELECT a+32, randomblob(1500) FROM t5; + SELECT count(*) FROM t5; + --match 64 + SELECT avg(length(b)) FROM t5; + --match 1500.0 + --sleep 2 + UPDATE t5 SET b='x'||a||'y'; + SELECT total(length(b)) FROM t5; + --match 247 + SELECT a FROM t5 WHERE b='x17y'; + --match 17 + CREATE INDEX t5b ON t5(b); + SELECT a FROM t5 WHERE b='x17y'; + --match 17 + SELECT a FROM t5 WHERE b GLOB 'x2?y' ORDER BY b DESC LIMIT 5; + --match 29 28 27 26 25 +--end + +--wait all +SELECT count(*), total(length(b)) FROM t1; +--match 64 247 +SELECT count(*), total(length(b)) FROM t2; +--match 64 247 +SELECT count(*), total(length(b)) FROM t3; +--match 64 247 +SELECT count(*), total(length(b)) FROM t4; +--match 64 247 +SELECT count(*), total(length(b)) FROM t5; +--match 64 247 + +--task 1 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 5 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 3 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 2 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 4 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--wait all + +--task 5 + DROP INDEX t5b; + --sleep 5 + PRAGMA integrity_check(10); + --match ok + CREATE INDEX t5b ON t5(b DESC); +--end +--task 3 + DROP INDEX t3b; + --sleep 5 + PRAGMA integrity_check(10); + --match ok + CREATE INDEX t3b ON t3(b DESC); +--end +--task 1 + DROP INDEX t1b; + --sleep 5 + PRAGMA integrity_check(10); + --match ok + CREATE INDEX t1b ON t1(b DESC); +--end +--task 2 + DROP INDEX t2b; + --sleep 5 + PRAGMA integrity_check(10); + --match ok + CREATE INDEX t2b ON t2(b DESC); +--end +--task 4 + DROP INDEX t4b; + --sleep 5 + PRAGMA integrity_check(10); + --match ok + CREATE INDEX t4b ON t4(b DESC); +--end +--wait all + +--task 1 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 5 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 3 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 2 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 4 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--wait all + +VACUUM; +PRAGMA integrity_check(10); +--match ok + +--task 1 + UPDATE t1 SET b=randomblob(20000); + --sleep 5 + UPDATE t1 SET b='x'||a||'y'; + SELECT a FROM t1 WHERE b='x63y'; + --match 63 +--end +--task 2 + UPDATE t2 SET b=randomblob(20000); + --sleep 5 + UPDATE t2 SET b='x'||a||'y'; + SELECT a FROM t2 WHERE b='x63y'; + --match 63 +--end +--task 3 + UPDATE t3 SET b=randomblob(20000); + --sleep 5 + UPDATE t3 SET b='x'||a||'y'; + SELECT a FROM t3 WHERE b='x63y'; + --match 63 +--end +--task 4 + UPDATE t4 SET b=randomblob(20000); + --sleep 5 + UPDATE t4 SET b='x'||a||'y'; + SELECT a FROM t4 WHERE b='x63y'; + --match 63 +--end +--task 5 + UPDATE t5 SET b=randomblob(20000); + --sleep 5 + UPDATE t5 SET b='x'||a||'y'; + SELECT a FROM t5 WHERE b='x63y'; + --match 63 +--end +--wait all + +--task 1 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 5 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 3 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 2 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--task 4 + SELECT t1.a FROM t1, t2 + WHERE t2.b GLOB 'x3?y' AND t1.b=('x'||(t2.a+3)||'y') + ORDER BY t1.a LIMIT 4 + --match 33 34 35 36 + SELECT t3.a FROM t3, t4 + WHERE t4.b GLOB 'x4?y' AND t3.b=('x'||(t4.a+5)||'y') + ORDER BY t3.a LIMIT 7 + --match 45 46 47 48 49 50 51 +--end +--wait all Index: src/attach.c ================================================================== --- src/attach.c +++ src/attach.c @@ -107,11 +107,11 @@ zErrDyn = sqlite3MPrintf(db, "database %s is already in use", zName); goto attach_error; } } - /* Allocate the new entry in the db->aDb[] array and initialise the schema + /* Allocate the new entry in the db->aDb[] array and initialize the schema ** hash tables. */ if( db->aDb==db->aDbStatic ){ aNew = sqlite3DbMallocRaw(db, sizeof(db->aDb[0])*3 ); if( aNew==0 ) return; @@ -124,11 +124,11 @@ aNew = &db->aDb[db->nDb]; memset(aNew, 0, sizeof(*aNew)); /* Open the database file. If the btree is successfully opened, use ** it to obtain the database schema. At this point the schema may - ** or may not be initialised. + ** or may not be initialized. */ flags = db->openFlags; rc = sqlite3ParseUri(db->pVfs->zName, zFile, &flags, &pVfs, &zPath, &zErr); if( rc!=SQLITE_OK ){ if( rc==SQLITE_NOMEM ) db->mallocFailed = 1; Index: src/backup.c ================================================================== --- src/backup.c +++ src/backup.c @@ -395,11 +395,12 @@ assert( nSrcPage>=0 ); for(ii=0; (nPage<0 || iiiNext<=(Pgno)nSrcPage && !rc; ii++){ const Pgno iSrcPg = p->iNext; /* Source page number */ if( iSrcPg!=PENDING_BYTE_PAGE(p->pSrc->pBt) ){ DbPage *pSrcPg; /* Source page object */ - rc = sqlite3PagerGet(pSrcPager, iSrcPg, &pSrcPg); + rc = sqlite3PagerAcquire(pSrcPager, iSrcPg, &pSrcPg, + PAGER_ACQUIRE_READONLY); if( rc==SQLITE_OK ){ rc = backupOnePage(p, iSrcPg, sqlite3PagerGetData(pSrcPg), 0); sqlite3PagerUnref(pSrcPg); } } Index: src/bitvec.c ================================================================== --- src/bitvec.c +++ src/bitvec.c @@ -70,11 +70,11 @@ /* ** A bitmap is an instance of the following structure. ** -** This bitmap records the existance of zero or more bits +** This bitmap records the existence of zero or more bits ** with values between 1 and iSize, inclusive. ** ** There are three possible representations of the bitmap. ** If iSize<=BITVEC_NBIT, then Bitvec.u.aBitmap[] is a straight ** bitmap. The least significant bit is bit 1. Index: src/btree.c ================================================================== --- src/btree.c +++ src/btree.c @@ -572,10 +572,23 @@ */ static void btreeClearHasContent(BtShared *pBt){ sqlite3BitvecDestroy(pBt->pHasContent); pBt->pHasContent = 0; } + +/* +** Release all of the apPage[] pages for a cursor. +*/ +static void btreeReleaseAllCursorPages(BtCursor *pCur){ + int i; + for(i=0; i<=pCur->iPage; i++){ + releasePage(pCur->apPage[i]); + pCur->apPage[i] = 0; + } + pCur->iPage = -1; +} + /* ** Save the current cursor position in the variables BtCursor.nKey ** and BtCursor.pKey. The cursor's state is set to CURSOR_REQUIRESEEK. ** @@ -612,16 +625,11 @@ } } assert( !pCur->apPage[0]->intKey || !pCur->pKey ); if( rc==SQLITE_OK ){ - int i; - for(i=0; i<=pCur->iPage; i++){ - releasePage(pCur->apPage[i]); - pCur->apPage[i] = 0; - } - pCur->iPage = -1; + btreeReleaseAllCursorPages(pCur); pCur->eState = CURSOR_REQUIRESEEK; } invalidateOverflowCache(pCur); return rc; @@ -635,15 +643,19 @@ static int saveAllCursors(BtShared *pBt, Pgno iRoot, BtCursor *pExcept){ BtCursor *p; assert( sqlite3_mutex_held(pBt->mutex) ); assert( pExcept==0 || pExcept->pBt==pBt ); for(p=pBt->pCursor; p; p=p->pNext){ - if( p!=pExcept && (0==iRoot || p->pgnoRoot==iRoot) && - p->eState==CURSOR_VALID ){ - int rc = saveCursorPosition(p); - if( SQLITE_OK!=rc ){ - return rc; + if( p!=pExcept && (0==iRoot || p->pgnoRoot==iRoot) ){ + if( p->eState==CURSOR_VALID ){ + int rc = saveCursorPosition(p); + if( SQLITE_OK!=rc ){ + return rc; + } + }else{ + testcase( p->iPage>0 ); + btreeReleaseAllCursorPages(p); } } } return SQLITE_OK; } @@ -1567,17 +1579,21 @@ */ static int btreeGetPage( BtShared *pBt, /* The btree */ Pgno pgno, /* Number of the page to fetch */ MemPage **ppPage, /* Return the page in this parameter */ - int noContent /* Do not load page content if true */ + int noContent, /* Do not load page content if true */ + int bReadonly /* True if a read-only (mmap) page is ok */ ){ int rc; DbPage *pDbPage; + int flags = (noContent ? PAGER_ACQUIRE_NOCONTENT : 0) + | (bReadonly ? PAGER_ACQUIRE_READONLY : 0); + assert( noContent==0 || bReadonly==0 ); assert( sqlite3_mutex_held(pBt->mutex) ); - rc = sqlite3PagerAcquire(pBt->pPager, pgno, (DbPage**)&pDbPage, noContent); + rc = sqlite3PagerAcquire(pBt->pPager, pgno, (DbPage**)&pDbPage, flags); if( rc ) return rc; *ppPage = btreePageFromDbPage(pDbPage, pgno, pBt); return SQLITE_OK; } @@ -1616,21 +1632,22 @@ ** ** If an error occurs, then the value *ppPage is set to is undefined. It ** may remain unchanged, or it may be set to an invalid value. */ static int getAndInitPage( - BtShared *pBt, /* The database file */ - Pgno pgno, /* Number of the page to get */ - MemPage **ppPage /* Write the page pointer here */ + BtShared *pBt, /* The database file */ + Pgno pgno, /* Number of the page to get */ + MemPage **ppPage, /* Write the page pointer here */ + int bReadonly /* True if a read-only (mmap) page is ok */ ){ int rc; assert( sqlite3_mutex_held(pBt->mutex) ); if( pgno>btreePagecount(pBt) ){ rc = SQLITE_CORRUPT_BKPT; }else{ - rc = btreeGetPage(pBt, pgno, ppPage, 0); + rc = btreeGetPage(pBt, pgno, ppPage, 0, bReadonly); if( rc==SQLITE_OK ){ rc = btreeInitPage(*ppPage); if( rc!=SQLITE_OK ){ releasePage(*ppPage); } @@ -1857,10 +1874,11 @@ goto btree_open_out; } rc = sqlite3PagerOpen(pVfs, &pBt->pPager, zFilename, EXTRA_SIZE, flags, vfsFlags, pageReinit); if( rc==SQLITE_OK ){ + sqlite3PagerSetMmapLimit(pBt->pPager, db->szMmap); rc = sqlite3PagerReadFileheader(pBt->pPager,sizeof(zDbHeader),zDbHeader); } if( rc!=SQLITE_OK ){ goto btree_open_out; } @@ -2122,10 +2140,23 @@ sqlite3BtreeEnter(p); sqlite3PagerSetCachesize(pBt->pPager, mxPage); sqlite3BtreeLeave(p); return SQLITE_OK; } + +/* +** Change the limit on the amount of the database file that may be +** memory mapped. +*/ +int sqlite3BtreeSetMmapLimit(Btree *p, sqlite3_int64 szMmap){ + BtShared *pBt = p->pBt; + assert( sqlite3_mutex_held(p->db->mutex) ); + sqlite3BtreeEnter(p); + sqlite3PagerSetMmapLimit(pBt->pPager, szMmap); + sqlite3BtreeLeave(p); + return SQLITE_OK; +} /* ** Change the way data is synced to disk in order to increase or decrease ** how well the database resists damage due to OS crashes and power ** failures. Level 1 is the same as asynchronous (no syncs() occur and @@ -2227,11 +2258,11 @@ ** ** This is useful in one special case in the backup API code where it is ** known that the shared b-tree mutex is held, but the mutex on the ** database handle that owns *p is not. In this case if sqlite3BtreeEnter() ** were to be called, it might collide with some other operation on the -** database handle that owns *p, causing undefined behaviour. +** database handle that owns *p, causing undefined behavior. */ int sqlite3BtreeGetReserveNoMutex(Btree *p){ assert( sqlite3_mutex_held(p->pBt->mutex) ); return p->pBt->pageSize - p->pBt->usableSize; } @@ -2348,11 +2379,11 @@ assert( sqlite3_mutex_held(pBt->mutex) ); assert( pBt->pPage1==0 ); rc = sqlite3PagerSharedLock(pBt->pPager); if( rc!=SQLITE_OK ) return rc; - rc = btreeGetPage(pBt, 1, &pPage1, 0); + rc = btreeGetPage(pBt, 1, &pPage1, 0, 0); if( rc!=SQLITE_OK ) return rc; /* Do some checking to help insure the file we opened really is ** a valid database file. */ @@ -2907,11 +2938,11 @@ /* Fix the database pointer on page iPtrPage that pointed at iDbPage so ** that it points at iFreePage. Also fix the pointer map entry for ** iPtrPage. */ if( eType!=PTRMAP_ROOTPAGE ){ - rc = btreeGetPage(pBt, iPtrPage, &pPtrPage, 0); + rc = btreeGetPage(pBt, iPtrPage, &pPtrPage, 0, 0); if( rc!=SQLITE_OK ){ return rc; } rc = sqlite3PagerWrite(pPtrPage->pDbPage); if( rc!=SQLITE_OK ){ @@ -2991,11 +3022,11 @@ Pgno iFreePg; /* Index of free page to move pLastPg to */ MemPage *pLastPg; u8 eMode = BTALLOC_ANY; /* Mode parameter for allocateBtreePage() */ Pgno iNear = 0; /* nearby parameter for allocateBtreePage() */ - rc = btreeGetPage(pBt, iLastPg, &pLastPg, 0); + rc = btreeGetPage(pBt, iLastPg, &pLastPg, 0, 0); if( rc!=SQLITE_OK ){ return rc; } /* If bCommit is zero, this loop runs exactly once and page pLastPg @@ -3083,12 +3114,15 @@ Pgno nFin = finalDbSize(pBt, nOrig, nFree); if( nOrig0 ){ - invalidateAllOverflowCache(pBt); - rc = incrVacuumStep(pBt, nFin, nOrig, 0); + rc = saveAllCursors(pBt, 0, 0); + if( rc==SQLITE_OK ){ + invalidateAllOverflowCache(pBt); + rc = incrVacuumStep(pBt, nFin, nOrig, 0); + } if( rc==SQLITE_OK ){ rc = sqlite3PagerWrite(pBt->pPage1->pDbPage); put4byte(&pBt->pPage1->aData[28], pBt->nPage); } }else{ @@ -3132,11 +3166,13 @@ } nFree = get4byte(&pBt->pPage1->aData[36]); nFin = finalDbSize(pBt, nOrig, nFree); if( nFin>nOrig ) return SQLITE_CORRUPT_BKPT; - + if( nFinnFin && rc==SQLITE_OK; iFree--){ rc = incrVacuumStep(pBt, nFin, iFree, 1); } if( (rc==SQLITE_DONE || rc==SQLITE_OK) && nFree>0 ){ rc = sqlite3PagerWrite(pBt->pPage1->pDbPage); @@ -3149,11 +3185,11 @@ if( rc!=SQLITE_OK ){ sqlite3PagerRollback(pPager); } } - assert( nRef==sqlite3PagerRefcount(pPager) ); + assert( nRef>=sqlite3PagerRefcount(pPager) ); return rc; } #else /* ifndef SQLITE_OMIT_AUTOVACUUM */ # define setChildPtrmaps(x) SQLITE_OK @@ -3405,11 +3441,11 @@ } /* The rollback may have destroyed the pPage1->aData value. So ** call btreeGetPage() on page 1 again to make ** sure pPage1->aData is set correctly. */ - if( btreeGetPage(pBt, 1, &pPage1, 0)==SQLITE_OK ){ + if( btreeGetPage(pBt, 1, &pPage1, 0, 0)==SQLITE_OK ){ int nPage = get4byte(28+(u8*)pPage1->aData); testcase( nPage==0 ); if( nPage==0 ) sqlite3PagerPagecount(pBt->pPager, &nPage); testcase( pBt->nPage!=nPage ); pBt->nPage = nPage; @@ -3839,11 +3875,11 @@ } #endif assert( next==0 || rc==SQLITE_DONE ); if( rc==SQLITE_OK ){ - rc = btreeGetPage(pBt, ovfl, &pPage, 0); + rc = btreeGetPage(pBt, ovfl, &pPage, 0, (ppPage==0)); assert( rc==SQLITE_OK || pPage==0 ); if( rc==SQLITE_OK ){ next = get4byte(pPage->aData); } } @@ -4060,11 +4096,13 @@ }else #endif { DbPage *pDbPage; - rc = sqlite3PagerGet(pBt->pPager, nextPage, &pDbPage); + rc = sqlite3PagerAcquire(pBt->pPager, nextPage, &pDbPage, + (eOp==0 ? PAGER_ACQUIRE_READONLY : 0) + ); if( rc==SQLITE_OK ){ aPayload = sqlite3PagerGetData(pDbPage); nextPage = get4byte(aPayload); rc = copyPayload(&aPayload[offset+4], pBuf, a, eOp, pDbPage); sqlite3PagerUnref(pDbPage); @@ -4239,14 +4277,15 @@ BtShared *pBt = pCur->pBt; assert( cursorHoldsMutex(pCur) ); assert( pCur->eState==CURSOR_VALID ); assert( pCur->iPageiPage>=0 ); if( pCur->iPage>=(BTCURSOR_MAX_DEPTH-1) ){ return SQLITE_CORRUPT_BKPT; } - rc = getAndInitPage(pBt, newPgno, &pNewPage); + rc = getAndInitPage(pBt, newPgno, &pNewPage, (pCur->wrFlag==0)); if( rc ) return rc; pCur->apPage[i+1] = pNewPage; pCur->aiIdx[i+1] = 0; pCur->iPage++; @@ -4359,11 +4398,11 @@ pCur->iPage = 0; }else if( pCur->pgnoRoot==0 ){ pCur->eState = CURSOR_INVALID; return SQLITE_OK; }else{ - rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0]); + rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0], pCur->wrFlag==0); if( rc!=SQLITE_OK ){ pCur->eState = CURSOR_INVALID; return rc; } pCur->iPage = 0; @@ -4973,11 +5012,11 @@ } testcase( iTrunk==mxPage ); if( iTrunk>mxPage ){ rc = SQLITE_CORRUPT_BKPT; }else{ - rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0); + rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0, 0); } if( rc ){ pTrunk = 0; goto end_allocate_page; } @@ -5037,11 +5076,11 @@ if( iNewTrunk>mxPage ){ rc = SQLITE_CORRUPT_BKPT; goto end_allocate_page; } testcase( iNewTrunk==mxPage ); - rc = btreeGetPage(pBt, iNewTrunk, &pNewTrunk, 0); + rc = btreeGetPage(pBt, iNewTrunk, &pNewTrunk, 0, 0); if( rc!=SQLITE_OK ){ goto end_allocate_page; } rc = sqlite3PagerWrite(pNewTrunk->pDbPage); if( rc!=SQLITE_OK ){ @@ -5117,11 +5156,11 @@ if( closestpDbPage); if( rc!=SQLITE_OK ){ releasePage(*ppPage); } @@ -5165,11 +5204,11 @@ ** becomes a new pointer-map page, the second is used by the caller. */ MemPage *pPg = 0; TRACE(("ALLOCATE: %d from end of file (pointer-map page)\n", pBt->nPage)); assert( pBt->nPage!=PENDING_BYTE_PAGE(pBt) ); - rc = btreeGetPage(pBt, pBt->nPage, &pPg, bNoContent); + rc = btreeGetPage(pBt, pBt->nPage, &pPg, bNoContent, 0); if( rc==SQLITE_OK ){ rc = sqlite3PagerWrite(pPg->pDbPage); releasePage(pPg); } if( rc ) return rc; @@ -5179,11 +5218,11 @@ #endif put4byte(28 + (u8*)pBt->pPage1->aData, pBt->nPage); *pPgno = pBt->nPage; assert( *pPgno!=PENDING_BYTE_PAGE(pBt) ); - rc = btreeGetPage(pBt, *pPgno, ppPage, bNoContent); + rc = btreeGetPage(pBt, *pPgno, ppPage, bNoContent, 0); if( rc ) return rc; rc = sqlite3PagerWrite((*ppPage)->pDbPage); if( rc!=SQLITE_OK ){ releasePage(*ppPage); } @@ -5247,11 +5286,11 @@ if( pBt->btsFlags & BTS_SECURE_DELETE ){ /* If the secure_delete option is enabled, then ** always fully overwrite deleted information with zeros. */ - if( (!pPage && ((rc = btreeGetPage(pBt, iPage, &pPage, 0))!=0) ) + if( (!pPage && ((rc = btreeGetPage(pBt, iPage, &pPage, 0, 0))!=0) ) || ((rc = sqlite3PagerWrite(pPage->pDbPage))!=0) ){ goto freepage_out; } memset(pPage->aData, 0, pPage->pBt->pageSize); @@ -5274,11 +5313,11 @@ */ if( nFree!=0 ){ u32 nLeaf; /* Initial number of leaf cells on trunk page */ iTrunk = get4byte(&pPage1->aData[32]); - rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0); + rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0, 0); if( rc!=SQLITE_OK ){ goto freepage_out; } nLeaf = get4byte(&pTrunk->aData[4]); @@ -5320,11 +5359,11 @@ ** the page being freed as a leaf page of the first trunk in the free-list. ** Possibly because the free-list is empty, or possibly because the ** first trunk in the free-list is full. Either way, the page being freed ** will become the new first trunk page in the free-list. */ - if( pPage==0 && SQLITE_OK!=(rc = btreeGetPage(pBt, iPage, &pPage, 0)) ){ + if( pPage==0 && SQLITE_OK!=(rc = btreeGetPage(pBt, iPage, &pPage, 0, 0)) ){ goto freepage_out; } rc = sqlite3PagerWrite(pPage->pDbPage); if( rc!=SQLITE_OK ){ goto freepage_out; @@ -5507,11 +5546,11 @@ ** for that page now. ** ** If this is the first overflow page, then write a partial entry ** to the pointer-map. If we write nothing to this pointer-map slot, ** then the optimistic overflow chain processing in clearCell() - ** may misinterpret the uninitialised values and delete the + ** may misinterpret the uninitialized values and delete the ** wrong pages from the database. */ if( pBt->autoVacuum && rc==SQLITE_OK ){ u8 eType = (pgnoPtrmap?PTRMAP_OVERFLOW2:PTRMAP_OVERFLOW1); ptrmapPut(pBt, pgnoOvfl, eType, pgnoPtrmap, &rc); @@ -6121,11 +6160,11 @@ }else{ pRight = findCell(pParent, i+nxDiv-pParent->nOverflow); } pgno = get4byte(pRight); while( 1 ){ - rc = getAndInitPage(pBt, pgno, &apOld[i]); + rc = getAndInitPage(pBt, pgno, &apOld[i], 0); if( rc ){ memset(apOld, 0, (i+1)*sizeof(MemPage*)); goto balance_cleanup; } nMaxCells += 1+apOld[i]->nCell+apOld[i]->nOverflow; @@ -6980,11 +7019,11 @@ assert( pPage->leaf ); } insertCell(pPage, idx, newCell, szNew, 0, 0, &rc); assert( rc!=SQLITE_OK || pPage->nCell>0 || pPage->nOverflow>0 ); - /* If no error has occured and pPage has an overflow cell, call balance() + /* If no error has occurred and pPage has an overflow cell, call balance() ** to redistribute the cells within the tree. Since balance() may move ** the cursor, zero the BtCursor.info.nSize and BtCursor.validNKey ** variables. ** ** Previous versions of SQLite called moveToRoot() to move the cursor @@ -7209,14 +7248,21 @@ ** is already journaled. */ u8 eType = 0; Pgno iPtrPage = 0; + /* Save the positions of any open cursors. This is required in + ** case they are holding a reference to an xFetch reference + ** corresponding to page pgnoRoot. */ + rc = saveAllCursors(pBt, 0, 0); releasePage(pPageMove); + if( rc!=SQLITE_OK ){ + return rc; + } /* Move the page currently at pgnoRoot to pgnoMove. */ - rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0); + rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0, 0); if( rc!=SQLITE_OK ){ return rc; } rc = ptrmapGet(pBt, pgnoRoot, &eType, &iPtrPage); if( eType==PTRMAP_ROOTPAGE || eType==PTRMAP_FREEPAGE ){ @@ -7233,11 +7279,11 @@ /* Obtain the page at pgnoRoot */ if( rc!=SQLITE_OK ){ return rc; } - rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0); + rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0, 0); if( rc!=SQLITE_OK ){ return rc; } rc = sqlite3PagerWrite(pRoot->pDbPage); if( rc!=SQLITE_OK ){ @@ -7309,11 +7355,11 @@ assert( sqlite3_mutex_held(pBt->mutex) ); if( pgno>btreePagecount(pBt) ){ return SQLITE_CORRUPT_BKPT; } - rc = getAndInitPage(pBt, pgno, &pPage); + rc = getAndInitPage(pBt, pgno, &pPage, 0); if( rc ) return rc; for(i=0; inCell; i++){ pCell = findCell(pPage, i); if( !pPage->leaf ){ rc = clearDatabasePage(pBt, get4byte(pCell), 1, pnChange); @@ -7411,11 +7457,11 @@ if( NEVER(pBt->pCursor) ){ sqlite3ConnectionBlocked(p->db, pBt->pCursor->pBtree->db); return SQLITE_LOCKED_SHAREDCACHE; } - rc = btreeGetPage(pBt, (Pgno)iTable, &pPage, 0); + rc = btreeGetPage(pBt, (Pgno)iTable, &pPage, 0, 0); if( rc ) return rc; rc = sqlite3BtreeClearTable(p, iTable, 0); if( rc ){ releasePage(pPage); return rc; @@ -7446,21 +7492,21 @@ ** number in the database. So move the page that does into the ** gap left by the deleted root-page. */ MemPage *pMove; releasePage(pPage); - rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0); + rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0, 0); if( rc!=SQLITE_OK ){ return rc; } rc = relocatePage(pBt, pMove, PTRMAP_ROOTPAGE, 0, iTable, 0); releasePage(pMove); if( rc!=SQLITE_OK ){ return rc; } pMove = 0; - rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0); + rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0, 0); freePage(pMove, &rc); releasePage(pMove); if( rc!=SQLITE_OK ){ return rc; } @@ -7868,11 +7914,11 @@ */ pBt = pCheck->pBt; usableSize = pBt->usableSize; if( iPage==0 ) return 0; if( checkRef(pCheck, iPage, zParentContext) ) return 0; - if( (rc = btreeGetPage(pBt, (Pgno)iPage, &pPage, 0))!=0 ){ + if( (rc = btreeGetPage(pBt, (Pgno)iPage, &pPage, 0, 0))!=0 ){ checkAppendMsg(pCheck, zContext, "unable to get the page. error code=%d", rc); return 0; } @@ -8339,10 +8385,21 @@ } assert( pCsr->eState!=CURSOR_REQUIRESEEK ); if( pCsr->eState!=CURSOR_VALID ){ return SQLITE_ABORT; } + + /* Save the positions of all other cursors open on this table. This is + ** required in case any of them are holding references to an xFetch + ** version of the b-tree page modified by the accessPayload call below. + ** + ** Note that pCsr must be open on a BTREE_INTKEY table and saveCursorPosition() + ** and hence saveAllCursors() cannot fail on a BTREE_INTKEY table, hence + ** saveAllCursors can only return SQLITE_OK. + */ + VVA_ONLY(rc =) saveAllCursors(pCsr->pBt, pCsr->pgnoRoot, pCsr); + assert( rc==SQLITE_OK ); /* Check some assumptions: ** (a) the cursor is open for writing, ** (b) there is a read/write transaction open, ** (c) the connection holds a write-lock on the table (if required), Index: src/btree.h ================================================================== --- src/btree.h +++ src/btree.h @@ -61,10 +61,11 @@ #define BTREE_SINGLE 4 /* The file contains at most 1 b-tree */ #define BTREE_UNORDERED 8 /* Use of a hash implementation is OK */ int sqlite3BtreeClose(Btree*); int sqlite3BtreeSetCacheSize(Btree*,int); +int sqlite3BtreeSetMmapLimit(Btree*,sqlite3_int64); int sqlite3BtreeSetSafetyLevel(Btree*,int,int,int); int sqlite3BtreeSyncDisabled(Btree*); int sqlite3BtreeSetPageSize(Btree *p, int nPagesize, int nReserve, int eFix); int sqlite3BtreeGetPageSize(Btree*); int sqlite3BtreeMaxPageCount(Btree*,int); @@ -137,10 +138,11 @@ #define BTREE_DEFAULT_CACHE_SIZE 3 #define BTREE_LARGEST_ROOT_PAGE 4 #define BTREE_TEXT_ENCODING 5 #define BTREE_USER_VERSION 6 #define BTREE_INCR_VACUUM 7 +#define BTREE_APPLICATION_ID 8 /* ** Values that may be OR'd together to form the second argument of an ** sqlite3BtreeCursorHints() call. */ Index: src/build.c ================================================================== --- src/build.c +++ src/build.c @@ -2098,11 +2098,11 @@ #endif /* Drop all SQLITE_MASTER table and index entries that refer to the ** table. The program name loops through the master table and deletes ** every row that refers to a table of the same name as the one being - ** dropped. Triggers are handled seperately because a trigger can be + ** dropped. Triggers are handled separately because a trigger can be ** created in the temp database that refers to a table in another ** database. */ sqlite3NestedParse(pParse, "DELETE FROM %Q.%s WHERE tbl_name=%Q and type!='trigger'", @@ -2390,13 +2390,10 @@ int addr1; /* Address of top of loop */ int addr2; /* Address to jump to for next iteration */ int tnum; /* Root page of index */ Vdbe *v; /* Generate code into this virtual machine */ KeyInfo *pKey; /* KeyInfo for index */ -#ifdef SQLITE_OMIT_MERGE_SORT - int regIdxKey; /* Registers containing the index key */ -#endif int regRecord; /* Register holding assemblied index record */ sqlite3 *db = pParse->db; /* The database connection */ int iDb = sqlite3SchemaToIndex(db, pIndex->pSchema); #ifndef SQLITE_OMIT_AUTHORIZATION @@ -2420,25 +2417,20 @@ pKey = sqlite3IndexKeyinfo(pParse, pIndex); sqlite3VdbeAddOp4(v, OP_OpenWrite, iIdx, tnum, iDb, (char *)pKey, P4_KEYINFO_HANDOFF); sqlite3VdbeChangeP5(v, OPFLAG_BULKCSR|((memRootPage>=0)?OPFLAG_P2ISREG:0)); -#ifndef SQLITE_OMIT_MERGE_SORT /* Open the sorter cursor if we are to use one. */ iSorter = pParse->nTab++; sqlite3VdbeAddOp4(v, OP_SorterOpen, iSorter, 0, 0, (char*)pKey, P4_KEYINFO); -#else - iSorter = iTab; -#endif /* Open the table. Loop through all rows of the table, inserting index ** records into the sorter. */ sqlite3OpenTable(pParse, iTab, iDb, pTab, OP_OpenRead); addr1 = sqlite3VdbeAddOp2(v, OP_Rewind, iTab, 0); regRecord = sqlite3GetTempReg(pParse); -#ifndef SQLITE_OMIT_MERGE_SORT sqlite3GenerateIndexKey(pParse, pIndex, iTab, regRecord, 1); sqlite3VdbeAddOp2(v, OP_SorterInsert, iSorter, regRecord); sqlite3VdbeAddOp2(v, OP_Next, iTab, addr1+1); sqlite3VdbeJumpHere(v, addr1); addr1 = sqlite3VdbeAddOp2(v, OP_SorterSort, iSorter, 0); @@ -2454,34 +2446,10 @@ addr2 = sqlite3VdbeCurrentAddr(v); } sqlite3VdbeAddOp2(v, OP_SorterData, iSorter, regRecord); sqlite3VdbeAddOp3(v, OP_IdxInsert, iIdx, regRecord, 1); sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT); -#else - regIdxKey = sqlite3GenerateIndexKey(pParse, pIndex, iTab, regRecord, 1); - addr2 = addr1 + 1; - if( pIndex->onError!=OE_None ){ - const int regRowid = regIdxKey + pIndex->nColumn; - const int j2 = sqlite3VdbeCurrentAddr(v) + 2; - void * const pRegKey = SQLITE_INT_TO_PTR(regIdxKey); - - /* The registers accessed by the OP_IsUnique opcode were allocated - ** using sqlite3GetTempRange() inside of the sqlite3GenerateIndexKey() - ** call above. Just before that function was freed they were released - ** (made available to the compiler for reuse) using - ** sqlite3ReleaseTempRange(). So in some ways having the OP_IsUnique - ** opcode use the values stored within seems dangerous. However, since - ** we can be sure that no other temp registers have been allocated - ** since sqlite3ReleaseTempRange() was called, it is safe to do so. - */ - sqlite3VdbeAddOp4(v, OP_IsUnique, iIdx, j2, regRowid, pRegKey, P4_INT32); - sqlite3HaltConstraint(pParse, SQLITE_CONSTRAINT_UNIQUE, - "indexed columns are not unique", P4_STATIC); - } - sqlite3VdbeAddOp3(v, OP_IdxInsert, iIdx, regRecord, 0); - sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT); -#endif sqlite3ReleaseTempReg(pParse, regRecord); sqlite3VdbeAddOp2(v, OP_SorterNext, iSorter, addr2); sqlite3VdbeJumpHere(v, addr1); sqlite3VdbeAddOp1(v, OP_Close, iTab); @@ -2837,11 +2805,11 @@ /* This constraint creates the same index as a previous ** constraint specified somewhere in the CREATE TABLE statement. ** However the ON CONFLICT clauses are different. If both this ** constraint and the previous equivalent constraint have explicit ** ON CONFLICT clauses this is an error. Otherwise, use the - ** explicitly specified behaviour for the index. + ** explicitly specified behavior for the index. */ if( !(pIdx->onError==OE_Default || pIndex->onError==OE_Default) ){ sqlite3ErrorMsg(pParse, "conflicting ON CONFLICT clauses specified", 0); } Index: src/ctime.c ================================================================== --- src/ctime.c +++ src/ctime.c @@ -46,18 +46,18 @@ "CHECK_PAGES", #endif #ifdef SQLITE_COVERAGE_TEST "COVERAGE_TEST", #endif -#ifdef SQLITE_CURDIR - "CURDIR", -#endif #ifdef SQLITE_DEBUG "DEBUG", #endif #ifdef SQLITE_DEFAULT_LOCKING_MODE "DEFAULT_LOCKING_MODE=" CTIMEOPT_VAL(SQLITE_DEFAULT_LOCKING_MODE), +#endif +#if defined(SQLITE_DEFAULT_MMAP_SIZE) && !defined(SQLITE_DEFAULT_MMAP_SIZE_xc) + "DEFAULT_MMAP_SIZE=" CTIMEOPT_VAL(SQLITE_DEFAULT_MMAP_SIZE), #endif #ifdef SQLITE_DISABLE_DIRSYNC "DISABLE_DIRSYNC", #endif #ifdef SQLITE_DISABLE_LFS @@ -145,10 +145,13 @@ "INT64_TYPE", #endif #ifdef SQLITE_LOCK_TRACE "LOCK_TRACE", #endif +#if defined(SQLITE_MAX_MMAP_SIZE) && !defined(SQLITE_MAX_MMAP_SIZE_xc) + "MAX_MMAP_SIZE=" CTIMEOPT_VAL(SQLITE_MAX_MMAP_SIZE), +#endif #ifdef SQLITE_MAX_SCHEMA_RETRY "MAX_SCHEMA_RETRY=" CTIMEOPT_VAL(SQLITE_MAX_SCHEMA_RETRY), #endif #ifdef SQLITE_MEMDEBUG "MEMDEBUG", @@ -202,15 +205,10 @@ "OMIT_CAST", #endif #ifdef SQLITE_OMIT_CHECK "OMIT_CHECK", #endif -/* // redundant -** #ifdef SQLITE_OMIT_COMPILEOPTION_DIAGS -** "OMIT_COMPILEOPTION_DIAGS", -** #endif -*/ #ifdef SQLITE_OMIT_COMPLETE "OMIT_COMPLETE", #endif #ifdef SQLITE_OMIT_COMPOUND_SELECT "OMIT_COMPOUND_SELECT", @@ -261,13 +259,10 @@ "OMIT_LOOKASIDE", #endif #ifdef SQLITE_OMIT_MEMORYDB "OMIT_MEMORYDB", #endif -#ifdef SQLITE_OMIT_MERGE_SORT - "OMIT_MERGE_SORT", -#endif #ifdef SQLITE_OMIT_OR_OPTIMIZATION "OMIT_OR_OPTIMIZATION", #endif #ifdef SQLITE_OMIT_PAGER_PRAGMAS "OMIT_PAGER_PRAGMAS", @@ -351,17 +346,17 @@ "SOUNDEX", #endif #ifdef SQLITE_TCL "TCL", #endif -#ifdef SQLITE_TEMP_STORE +#if defined(SQLITE_TEMP_STORE) && !defined(SQLITE_TEMP_STORE_xc) "TEMP_STORE=" CTIMEOPT_VAL(SQLITE_TEMP_STORE), #endif #ifdef SQLITE_TEST "TEST", #endif -#ifdef SQLITE_THREADSAFE +#if defined(SQLITE_THREADSAFE) "THREADSAFE=" CTIMEOPT_VAL(SQLITE_THREADSAFE), #endif #ifdef SQLITE_USE_ALLOCA "USE_ALLOCA", #endif @@ -383,12 +378,15 @@ n = sqlite3Strlen30(zOptName); /* Since ArraySize(azCompileOpt) is normally in single digits, a ** linear search is adequate. No need for a binary search. */ for(i=0; iu.i = 0; return WRC_Abort; } static int exprIsConst(Expr *p, int initFlag){ Walker w; + memset(&w, 0, sizeof(w)); w.u.i = initFlag; w.xExprCallback = exprNodeIsConstant; w.xSelectCallback = selectNodeIsConstant; sqlite3WalkExpr(&w, p); return w.u.i; @@ -3426,12 +3427,12 @@ */ void sqlite3ExprCodeConstants(Parse *pParse, Expr *pExpr){ Walker w; if( pParse->cookieGoto ) return; if( OptimizationDisabled(pParse->db, SQLITE_FactorOutConst) ) return; + memset(&w, 0, sizeof(w)); w.xExprCallback = evalConstExpr; - w.xSelectCallback = 0; w.pParse = pParse; sqlite3WalkExpr(&w, pExpr); } @@ -3540,11 +3541,11 @@ int regFree1 = 0; int regFree2 = 0; int r1, r2; assert( jumpIfNull==SQLITE_JUMPIFNULL || jumpIfNull==0 ); - if( NEVER(v==0) ) return; /* Existance of VDBE checked by caller */ + if( NEVER(v==0) ) return; /* Existence of VDBE checked by caller */ if( NEVER(pExpr==0) ) return; /* No way this can happen */ op = pExpr->op; switch( op ){ case TK_AND: { int d2 = sqlite3VdbeMakeLabel(v); @@ -3660,11 +3661,11 @@ int regFree1 = 0; int regFree2 = 0; int r1, r2; assert( jumpIfNull==SQLITE_JUMPIFNULL || jumpIfNull==0 ); - if( NEVER(v==0) ) return; /* Existance of VDBE checked by caller */ + if( NEVER(v==0) ) return; /* Existence of VDBE checked by caller */ if( pExpr==0 ) return; /* The value of pExpr->op and op are related as follows: ** ** pExpr->op op Index: src/func.c ================================================================== --- src/func.c +++ src/func.c @@ -693,10 +693,17 @@ prevEscape = 0; } } return *zString==0; } + +/* +** The sqlite3_strglob() interface. +*/ +int sqlite3_strglob(const char *zGlobPattern, const char *zString){ + return patternCompare((u8*)zGlobPattern, (u8*)zString, &globInfo, 0)==0; +} /* ** Count the number of times that the LIKE operator (or GLOB which is ** just a variation of LIKE) gets called. This is used for testing ** only. Index: src/global.c ================================================================== --- src/global.c +++ src/global.c @@ -154,10 +154,12 @@ {0,0,0,0,0,0,0,0,0}, /* mutex */ {0,0,0,0,0,0,0,0,0,0,0,0,0},/* pcache2 */ (void*)0, /* pHeap */ 0, /* nHeap */ 0, 0, /* mnHeap, mxHeap */ + SQLITE_DEFAULT_MMAP_SIZE, /* szMmap */ + SQLITE_MAX_MMAP_SIZE, /* mxMmap */ (void*)0, /* pScratch */ 0, /* szScratch */ 0, /* nScratch */ (void*)0, /* pPage */ 0, /* szPage */ Index: src/hash.h ================================================================== --- src/hash.h +++ src/hash.h @@ -7,11 +7,11 @@ ** May you do good and not evil. ** May you find forgiveness for yourself and forgive others. ** May you share freely, never taking more than you give. ** ************************************************************************* -** This is the header file for the generic hash-table implemenation +** This is the header file for the generic hash-table implementation ** used in SQLite. */ #ifndef _SQLITE_HASH_H_ #define _SQLITE_HASH_H_ Index: src/legacy.c ================================================================== --- src/legacy.c +++ src/legacy.c @@ -36,24 +36,23 @@ ){ int rc = SQLITE_OK; /* Return code */ const char *zLeftover; /* Tail of unprocessed SQL */ sqlite3_stmt *pStmt = 0; /* The current SQL statement */ char **azCols = 0; /* Names of result columns */ - int nRetry = 0; /* Number of retry attempts */ int callbackIsInit; /* True if callback data is initialized */ if( !sqlite3SafetyCheckOk(db) ) return SQLITE_MISUSE_BKPT; if( zSql==0 ) zSql = ""; sqlite3_mutex_enter(db->mutex); sqlite3Error(db, SQLITE_OK, 0); - while( (rc==SQLITE_OK || (rc==SQLITE_SCHEMA && (++nRetry)<2)) && zSql[0] ){ + while( rc==SQLITE_OK && zSql[0] ){ int nCol; char **azVals = 0; pStmt = 0; - rc = sqlite3_prepare(db, zSql, -1, &pStmt, &zLeftover); + rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, &zLeftover); assert( rc==SQLITE_OK || pStmt==0 ); if( rc!=SQLITE_OK ){ continue; } if( !pStmt ){ @@ -106,15 +105,12 @@ } if( rc!=SQLITE_ROW ){ rc = sqlite3VdbeFinalize((Vdbe *)pStmt); pStmt = 0; - if( rc!=SQLITE_SCHEMA ){ - nRetry = 0; - zSql = zLeftover; - while( sqlite3Isspace(zSql[0]) ) zSql++; - } + zSql = zLeftover; + while( sqlite3Isspace(zSql[0]) ) zSql++; break; } } sqlite3DbFree(db, azCols); Index: src/loadext.c ================================================================== --- src/loadext.c +++ src/loadext.c @@ -413,12 +413,27 @@ ){ sqlite3_vfs *pVfs = db->pVfs; void *handle; int (*xInit)(sqlite3*,char**,const sqlite3_api_routines*); char *zErrmsg = 0; + const char *zEntry; + char *zAltEntry = 0; void **aHandle; int nMsg = 300 + sqlite3Strlen30(zFile); + int ii; + + /* Shared library endings to try if zFile cannot be loaded as written */ + static const char *azEndings[] = { +#if SQLITE_OS_WIN + "dll" +#elif defined(__APPLE__) + "dylib" +#else + "so" +#endif + }; + if( pzErrMsg ) *pzErrMsg = 0; /* Ticket #1863. To avoid a creating security problems for older ** applications that relink against newer versions of SQLite, the @@ -431,15 +446,21 @@ *pzErrMsg = sqlite3_mprintf("not authorized"); } return SQLITE_ERROR; } - if( zProc==0 ){ - zProc = "sqlite3_extension_init"; - } + zEntry = zProc ? zProc : "sqlite3_extension_init"; handle = sqlite3OsDlOpen(pVfs, zFile); +#if SQLITE_OS_UNIX || SQLITE_OS_WIN + for(ii=0; ii sqlite3_example_init + ** C:/lib/mathfuncs.dll ==> sqlite3_mathfuncs_init + */ + if( xInit==0 && zProc==0 ){ + int iFile, iEntry, c; + int ncFile = sqlite3Strlen30(zFile); + zAltEntry = sqlite3_malloc(ncFile+30); + if( zAltEntry==0 ){ + sqlite3OsDlClose(pVfs, handle); + return SQLITE_NOMEM; + } + memcpy(zAltEntry, "sqlite3_", 8); + for(iFile=ncFile-1; iFile>=0 && zFile[iFile]!='/'; iFile--){} + iFile++; + if( sqlite3_strnicmp(zFile+iFile, "lib", 3)==0 ) iFile += 3; + for(iEntry=8; (c = zFile[iFile])!=0 && c!='.'; iFile++){ + if( sqlite3Isalpha(c) ){ + zAltEntry[iEntry++] = (char)sqlite3UpperToLower[(unsigned)c]; + } + } + memcpy(zAltEntry+iEntry, "_init", 6); + zEntry = zAltEntry; + xInit = (int(*)(sqlite3*,char**,const sqlite3_api_routines*)) + sqlite3OsDlSym(pVfs, handle, zEntry); + } if( xInit==0 ){ if( pzErrMsg ){ - nMsg += sqlite3Strlen30(zProc); + nMsg += sqlite3Strlen30(zEntry); *pzErrMsg = zErrmsg = sqlite3_malloc(nMsg); if( zErrmsg ){ sqlite3_snprintf(nMsg, zErrmsg, - "no entry point [%s] in shared library [%s]", zProc,zFile); + "no entry point [%s] in shared library [%s]", zEntry, zFile); sqlite3OsDlError(pVfs, nMsg-1, zErrmsg); } - sqlite3OsDlClose(pVfs, handle); } + sqlite3OsDlClose(pVfs, handle); + sqlite3_free(zAltEntry); return SQLITE_ERROR; - }else if( xInit(db, &zErrmsg, &sqlite3Apis) ){ + } + sqlite3_free(zAltEntry); + if( xInit(db, &zErrmsg, &sqlite3Apis) ){ if( pzErrMsg ){ *pzErrMsg = sqlite3_mprintf("error during initialization: %s", zErrmsg); } sqlite3_free(zErrmsg); sqlite3OsDlClose(pVfs, handle); Index: src/main.c ================================================================== --- src/main.c +++ src/main.c @@ -493,10 +493,23 @@ sqlite3GlobalConfig.xSqllog = va_arg(ap, SQLLOGFUNC_t); sqlite3GlobalConfig.pSqllogArg = va_arg(ap, void *); break; } #endif + + case SQLITE_CONFIG_MMAP_SIZE: { + sqlite3_int64 szMmap = va_arg(ap, sqlite3_int64); + sqlite3_int64 mxMmap = va_arg(ap, sqlite3_int64); + if( mxMmap<0 || mxMmap>SQLITE_MAX_MMAP_SIZE ){ + mxMmap = SQLITE_MAX_MMAP_SIZE; + } + sqlite3GlobalConfig.mxMmap = mxMmap; + if( szMmap<0 ) szMmap = SQLITE_DEFAULT_MMAP_SIZE; + if( szMmap>mxMmap) szMmap = mxMmap; + sqlite3GlobalConfig.szMmap = szMmap; + break; + } default: { rc = SQLITE_ERROR; break; } @@ -883,11 +896,11 @@ return; } /* If we reach this point, it means that the database connection has ** closed all sqlite3_stmt and sqlite3_backup objects and has been - ** pased to sqlite3_close (meaning that it is a zombie). Therefore, + ** passed to sqlite3_close (meaning that it is a zombie). Therefore, ** go ahead and free all resources. */ /* Free any outstanding Savepoint structures. */ sqlite3CloseSavepoints(db); @@ -1014,10 +1027,114 @@ if( db->xRollbackCallback && (inTrans || !db->autoCommit) ){ db->xRollbackCallback(db->pRollbackArg); } } +/* +** Return a static string containing the name corresponding to the error code +** specified in the argument. +*/ +#if defined(SQLITE_DEBUG) || defined(SQLITE_TEST) || \ + defined(SQLITE_DEBUG_OS_TRACE) +const char *sqlite3ErrName(int rc){ + const char *zName = 0; + int i, origRc = rc; + for(i=0; i<2 && zName==0; i++, rc &= 0xff){ + switch( rc ){ + case SQLITE_OK: zName = "SQLITE_OK"; break; + case SQLITE_ERROR: zName = "SQLITE_ERROR"; break; + case SQLITE_INTERNAL: zName = "SQLITE_INTERNAL"; break; + case SQLITE_PERM: zName = "SQLITE_PERM"; break; + case SQLITE_ABORT: zName = "SQLITE_ABORT"; break; + case SQLITE_ABORT_ROLLBACK: zName = "SQLITE_ABORT_ROLLBACK"; break; + case SQLITE_BUSY: zName = "SQLITE_BUSY"; break; + case SQLITE_BUSY_RECOVERY: zName = "SQLITE_BUSY_RECOVERY"; break; + case SQLITE_LOCKED: zName = "SQLITE_LOCKED"; break; + case SQLITE_LOCKED_SHAREDCACHE: zName = "SQLITE_LOCKED_SHAREDCACHE";break; + case SQLITE_NOMEM: zName = "SQLITE_NOMEM"; break; + case SQLITE_READONLY: zName = "SQLITE_READONLY"; break; + case SQLITE_READONLY_RECOVERY: zName = "SQLITE_READONLY_RECOVERY"; break; + case SQLITE_READONLY_CANTLOCK: zName = "SQLITE_READONLY_CANTLOCK"; break; + case SQLITE_READONLY_ROLLBACK: zName = "SQLITE_READONLY_ROLLBACK"; break; + case SQLITE_INTERRUPT: zName = "SQLITE_INTERRUPT"; break; + case SQLITE_IOERR: zName = "SQLITE_IOERR"; break; + case SQLITE_IOERR_READ: zName = "SQLITE_IOERR_READ"; break; + case SQLITE_IOERR_SHORT_READ: zName = "SQLITE_IOERR_SHORT_READ"; break; + case SQLITE_IOERR_WRITE: zName = "SQLITE_IOERR_WRITE"; break; + case SQLITE_IOERR_FSYNC: zName = "SQLITE_IOERR_FSYNC"; break; + case SQLITE_IOERR_DIR_FSYNC: zName = "SQLITE_IOERR_DIR_FSYNC"; break; + case SQLITE_IOERR_TRUNCATE: zName = "SQLITE_IOERR_TRUNCATE"; break; + case SQLITE_IOERR_FSTAT: zName = "SQLITE_IOERR_FSTAT"; break; + case SQLITE_IOERR_UNLOCK: zName = "SQLITE_IOERR_UNLOCK"; break; + case SQLITE_IOERR_RDLOCK: zName = "SQLITE_IOERR_RDLOCK"; break; + case SQLITE_IOERR_DELETE: zName = "SQLITE_IOERR_DELETE"; break; + case SQLITE_IOERR_BLOCKED: zName = "SQLITE_IOERR_BLOCKED"; break; + case SQLITE_IOERR_NOMEM: zName = "SQLITE_IOERR_NOMEM"; break; + case SQLITE_IOERR_ACCESS: zName = "SQLITE_IOERR_ACCESS"; break; + case SQLITE_IOERR_CHECKRESERVEDLOCK: + zName = "SQLITE_IOERR_CHECKRESERVEDLOCK"; break; + case SQLITE_IOERR_LOCK: zName = "SQLITE_IOERR_LOCK"; break; + case SQLITE_IOERR_CLOSE: zName = "SQLITE_IOERR_CLOSE"; break; + case SQLITE_IOERR_DIR_CLOSE: zName = "SQLITE_IOERR_DIR_CLOSE"; break; + case SQLITE_IOERR_SHMOPEN: zName = "SQLITE_IOERR_SHMOPEN"; break; + case SQLITE_IOERR_SHMSIZE: zName = "SQLITE_IOERR_SHMSIZE"; break; + case SQLITE_IOERR_SHMLOCK: zName = "SQLITE_IOERR_SHMLOCK"; break; + case SQLITE_IOERR_SHMMAP: zName = "SQLITE_IOERR_SHMMAP"; break; + case SQLITE_IOERR_SEEK: zName = "SQLITE_IOERR_SEEK"; break; + case SQLITE_IOERR_DELETE_NOENT: zName = "SQLITE_IOERR_DELETE_NOENT";break; + case SQLITE_IOERR_MMAP: zName = "SQLITE_IOERR_MMAP"; break; + case SQLITE_CORRUPT: zName = "SQLITE_CORRUPT"; break; + case SQLITE_CORRUPT_VTAB: zName = "SQLITE_CORRUPT_VTAB"; break; + case SQLITE_NOTFOUND: zName = "SQLITE_NOTFOUND"; break; + case SQLITE_FULL: zName = "SQLITE_FULL"; break; + case SQLITE_CANTOPEN: zName = "SQLITE_CANTOPEN"; break; + case SQLITE_CANTOPEN_NOTEMPDIR: zName = "SQLITE_CANTOPEN_NOTEMPDIR";break; + case SQLITE_CANTOPEN_ISDIR: zName = "SQLITE_CANTOPEN_ISDIR"; break; + case SQLITE_CANTOPEN_FULLPATH: zName = "SQLITE_CANTOPEN_FULLPATH"; break; + case SQLITE_PROTOCOL: zName = "SQLITE_PROTOCOL"; break; + case SQLITE_EMPTY: zName = "SQLITE_EMPTY"; break; + case SQLITE_SCHEMA: zName = "SQLITE_SCHEMA"; break; + case SQLITE_TOOBIG: zName = "SQLITE_TOOBIG"; break; + case SQLITE_CONSTRAINT: zName = "SQLITE_CONSTRAINT"; break; + case SQLITE_CONSTRAINT_UNIQUE: zName = "SQLITE_CONSTRAINT_UNIQUE"; break; + case SQLITE_CONSTRAINT_TRIGGER: zName = "SQLITE_CONSTRAINT_TRIGGER";break; + case SQLITE_CONSTRAINT_FOREIGNKEY: + zName = "SQLITE_CONSTRAINT_FOREIGNKEY"; break; + case SQLITE_CONSTRAINT_CHECK: zName = "SQLITE_CONSTRAINT_CHECK"; break; + case SQLITE_CONSTRAINT_PRIMARYKEY: + zName = "SQLITE_CONSTRAINT_PRIMARYKEY"; break; + case SQLITE_CONSTRAINT_NOTNULL: zName = "SQLITE_CONSTRAINT_NOTNULL";break; + case SQLITE_CONSTRAINT_COMMITHOOK: + zName = "SQLITE_CONSTRAINT_COMMITHOOK"; break; + case SQLITE_CONSTRAINT_VTAB: zName = "SQLITE_CONSTRAINT_VTAB"; break; + case SQLITE_CONSTRAINT_FUNCTION: + zName = "SQLITE_CONSTRAINT_FUNCTION"; break; + case SQLITE_MISMATCH: zName = "SQLITE_MISMATCH"; break; + case SQLITE_MISUSE: zName = "SQLITE_MISUSE"; break; + case SQLITE_NOLFS: zName = "SQLITE_NOLFS"; break; + case SQLITE_AUTH: zName = "SQLITE_AUTH"; break; + case SQLITE_FORMAT: zName = "SQLITE_FORMAT"; break; + case SQLITE_RANGE: zName = "SQLITE_RANGE"; break; + case SQLITE_NOTADB: zName = "SQLITE_NOTADB"; break; + case SQLITE_ROW: zName = "SQLITE_ROW"; break; + case SQLITE_NOTICE: zName = "SQLITE_NOTICE"; break; + case SQLITE_NOTICE_RECOVER_WAL: zName = "SQLITE_NOTICE_RECOVER_WAL";break; + case SQLITE_NOTICE_RECOVER_ROLLBACK: + zName = "SQLITE_NOTICE_RECOVER_ROLLBACK"; break; + case SQLITE_WARNING: zName = "SQLITE_WARNING"; break; + case SQLITE_DONE: zName = "SQLITE_DONE"; break; + } + } + if( zName==0 ){ + static char zBuf[50]; + sqlite3_snprintf(sizeof(zBuf), zBuf, "SQLITE_UNKNOWN(%d)", origRc); + zName = zBuf; + } + return zName; +} +#endif + /* ** Return a static string that describes the kind of error specified in the ** argument. */ const char *sqlite3ErrStr(int rc){ @@ -2314,10 +2431,11 @@ assert( sizeof(db->aLimit)==sizeof(aHardLimit) ); memcpy(db->aLimit, aHardLimit, sizeof(db->aLimit)); db->autoCommit = 1; db->nextAutovac = -1; + db->szMmap = sqlite3GlobalConfig.szMmap; db->nextPagesize = 0; db->flags |= SQLITE_ShortColNames | SQLITE_AutoIndex | SQLITE_EnableTrigger #if SQLITE_DEFAULT_FILE_FORMAT<4 | SQLITE_LegacyFileFmt #endif Index: src/memjournal.c ================================================================== --- src/memjournal.c +++ src/memjournal.c @@ -228,11 +228,13 @@ 0, /* xSectorSize */ 0, /* xDeviceCharacteristics */ 0, /* xShmMap */ 0, /* xShmLock */ 0, /* xShmBarrier */ - 0 /* xShmUnlock */ + 0, /* xShmUnmap */ + 0, /* xFetch */ + 0 /* xUnfetch */ }; /* ** Open a journal file. */ Index: src/os.c ================================================================== --- src/os.c +++ src/os.c @@ -138,10 +138,30 @@ void volatile **pp /* OUT: Pointer to mapping */ ){ DO_OS_MALLOC_TEST(id); return id->pMethods->xShmMap(id, iPage, pgsz, bExtend, pp); } + +#if SQLITE_MAX_MMAP_SIZE>0 +/* The real implementation of xFetch and xUnfetch */ +int sqlite3OsFetch(sqlite3_file *id, i64 iOff, int iAmt, void **pp){ + DO_OS_MALLOC_TEST(id); + return id->pMethods->xFetch(id, iOff, iAmt, pp); +} +int sqlite3OsUnfetch(sqlite3_file *id, i64 iOff, void *p){ + return id->pMethods->xUnfetch(id, iOff, p); +} +#else +/* No-op stubs to use when memory-mapped I/O is disabled */ +int sqlite3OsFetch(sqlite3_file *id, i64 iOff, int iAmt, void **pp){ + *pp = 0; + return SQLITE_OK; +} +int sqlite3OsUnfetch(sqlite3_file *id, i64 iOff, void *p){ + return SQLITE_OK; +} +#endif /* ** The next group of routines are convenience wrappers around the ** VFS methods. */ Index: src/os.h ================================================================== --- src/os.h +++ src/os.h @@ -97,18 +97,10 @@ */ #if !defined(SQLITE_OS_WINRT) # define SQLITE_OS_WINRT 0 #endif -/* -** When compiled for WinCE or WinRT, there is no concept of the current -** directory. - */ -#if !SQLITE_OS_WINCE && !SQLITE_OS_WINRT -# define SQLITE_CURDIR 1 -#endif - /* If the SET_FULLSYNC macro is not defined above, then make it ** a no-op */ #ifndef SET_FULLSYNC # define SET_FULLSYNC(x,y) @@ -257,10 +249,12 @@ int sqlite3OsDeviceCharacteristics(sqlite3_file *id); int sqlite3OsShmMap(sqlite3_file *,int,int,int,void volatile **); int sqlite3OsShmLock(sqlite3_file *id, int, int, int); void sqlite3OsShmBarrier(sqlite3_file *id); int sqlite3OsShmUnmap(sqlite3_file *id, int); +int sqlite3OsFetch(sqlite3_file *id, i64, int, void **); +int sqlite3OsUnfetch(sqlite3_file *, i64, void *); /* ** Functions for accessing sqlite3_vfs methods */ Index: src/os_unix.c ================================================================== --- src/os_unix.c +++ src/os_unix.c @@ -124,11 +124,11 @@ #include #include #include #include #include -#ifndef SQLITE_OMIT_WAL +#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0 #include #endif #if SQLITE_ENABLE_LOCKING_STYLE @@ -223,10 +223,15 @@ void *lockingContext; /* Locking style specific state */ UnixUnusedFd *pUnused; /* Pre-allocated UnixUnusedFd */ const char *zPath; /* Name of the file */ unixShm *pShm; /* Shared memory segment information */ int szChunk; /* Configured by FCNTL_CHUNK_SIZE */ + int nFetchOut; /* Number of outstanding xFetch refs */ + sqlite3_int64 mmapSize; /* Usable size of mapping at pMapRegion */ + sqlite3_int64 mmapSizeActual; /* Actual size of mapping at pMapRegion */ + sqlite3_int64 mmapSizeMax; /* Configured FCNTL_MMAP_SIZE value */ + void *pMapRegion; /* Memory mapped region */ #ifdef __QNXNTO__ int sectorSize; /* Device sector size */ int deviceCharacteristics; /* Precomputed device characteristics */ #endif #if SQLITE_ENABLE_LOCKING_STYLE @@ -247,11 +252,13 @@ ** one described by ticket #3584. */ unsigned char transCntrChng; /* True if the transaction counter changed */ unsigned char dbUpdate; /* True if any part of database file changed */ unsigned char inNormalWrite; /* True if in a normal write operation */ + #endif + #ifdef SQLITE_TEST /* In test mode, increase the size of this structure a bit so that ** it is larger than the struct CrashFile defined in test6.c. */ char aPadding[32]; @@ -271,10 +278,11 @@ #endif #define UNIXFILE_PSOW 0x10 /* SQLITE_IOCAP_POWERSAFE_OVERWRITE */ #define UNIXFILE_DELETE 0x20 /* Delete on close */ #define UNIXFILE_URI 0x40 /* Filename might have query parameters */ #define UNIXFILE_NOLOCK 0x80 /* Do no file locking */ +#define UNIXFILE_WARNED 0x0100 /* verifyDbFile() warnings have been issued */ /* ** Include code that is common to all os_*.c files */ #include "os_common.h" @@ -304,10 +312,21 @@ #define threadid pthread_self() #else #define threadid 0 #endif +/* +** HAVE_MREMAP defaults to true on Linux and false everywhere else. +*/ +#if !defined(HAVE_MREMAP) +# if defined(__linux__) && defined(_GNU_SOURCE) +# define HAVE_MREMAP 1 +# else +# define HAVE_MREMAP 0 +# endif +#endif + /* ** Different Unix systems declare open() in different ways. Same use ** open(const char*,int,mode_t). Others use open(const char*,int,...). ** The difference is important when using a pointer to the function. ** @@ -335,11 +354,11 @@ ** they may be overridden at runtime to facilitate fault injection during ** testing and sandboxing. The following array holds the names and pointers ** to all overrideable system calls. */ static struct unix_syscall { - const char *zName; /* Name of the sytem call */ + const char *zName; /* Name of the system call */ sqlite3_syscall_ptr pCurrent; /* Current value of the system call */ sqlite3_syscall_ptr pDefault; /* Default value */ } aSyscall[] = { { "open", (sqlite3_syscall_ptr)posixOpen, 0 }, #define osOpen ((int(*)(const char*,int,int))aSyscall[0].pCurrent) @@ -435,10 +454,23 @@ #define osRmdir ((int(*)(const char*))aSyscall[19].pCurrent) { "fchown", (sqlite3_syscall_ptr)posixFchown, 0 }, #define osFchown ((int(*)(int,uid_t,gid_t))aSyscall[20].pCurrent) + { "mmap", (sqlite3_syscall_ptr)mmap, 0 }, +#define osMmap ((void*(*)(void*,size_t,int,int,int,off_t))aSyscall[21].pCurrent) + + { "munmap", (sqlite3_syscall_ptr)munmap, 0 }, +#define osMunmap ((void*(*)(void*,size_t))aSyscall[22].pCurrent) + +#if HAVE_MREMAP + { "mremap", (sqlite3_syscall_ptr)mremap, 0 }, +#else + { "mremap", (sqlite3_syscall_ptr)0, 0 }, +#endif +#define osMremap ((void*(*)(void*,size_t,size_t,int,...))aSyscall[23].pCurrent) + }; /* End of the overrideable system calls */ /* ** This is the xSetSystemCall() method of sqlite3_vfs for all of the ** "unix" VFSes. Return SQLITE_OK opon successfully updating the @@ -764,11 +796,10 @@ default: return sqliteIOErr; } } - /****************************************************************************** ****************** Begin Unique File ID Utility Used By VxWorks *************** ** @@ -1102,11 +1133,10 @@ #else /* Non-threadsafe build, use strerror(). */ zErr = strerror(iErrno); #endif - assert( errcode!=SQLITE_OK ); if( zPath==0 ) zPath = ""; sqlite3_log(errcode, "os_unix.c:%d: (%d) %s(%s) - %s", iLine, iErrno, zFunc, zPath, zErr ); @@ -1267,10 +1297,54 @@ } *ppInode = pInode; return SQLITE_OK; } + +/* +** Check a unixFile that is a database. Verify the following: +** +** (1) There is exactly one hard link on the file +** (2) The file is not a symbolic link +** (3) The file has not been renamed or unlinked +** +** Issue sqlite3_log(SQLITE_WARNING,...) messages if anything is not right. +*/ +static void verifyDbFile(unixFile *pFile){ + struct stat buf; + int rc; + if( pFile->ctrlFlags & UNIXFILE_WARNED ){ + /* One or more of the following warnings have already been issued. Do not + ** repeat them so as not to clutter the error log */ + return; + } + rc = osFstat(pFile->h, &buf); + if( rc!=0 ){ + sqlite3_log(SQLITE_WARNING, "cannot fstat db file %s", pFile->zPath); + pFile->ctrlFlags |= UNIXFILE_WARNED; + return; + } + if( buf.st_nlink==0 && (pFile->ctrlFlags & UNIXFILE_DELETE)==0 ){ + sqlite3_log(SQLITE_WARNING, "file unlinked while open: %s", pFile->zPath); + pFile->ctrlFlags |= UNIXFILE_WARNED; + return; + } + if( buf.st_nlink>1 ){ + sqlite3_log(SQLITE_WARNING, "multiple links to file: %s", pFile->zPath); + pFile->ctrlFlags |= UNIXFILE_WARNED; + return; + } + if( pFile->pInode!=0 + && ((rc = osStat(pFile->zPath, &buf))!=0 + || buf.st_ino!=pFile->pInode->fileId.ino) + ){ + sqlite3_log(SQLITE_WARNING, "file renamed while open: %s", pFile->zPath); + pFile->ctrlFlags |= UNIXFILE_WARNED; + return; + } +} + /* ** This routine checks if there is a RESERVED lock held on the specified ** file by this or any other process. If such a lock is held, set *pResOut ** to a non-zero value otherwise *pResOut is set to zero. The return value @@ -1798,12 +1872,16 @@ ** ** If the locking level of the file descriptor is already at or below ** the requested locking level, this routine is a no-op. */ static int unixUnlock(sqlite3_file *id, int eFileLock){ + assert( eFileLock==SHARED_LOCK || ((unixFile *)id)->nFetchOut==0 ); return posixUnlock(id, eFileLock, 0); } + +static int unixMapfile(unixFile *pFd, i64 nByte); +static void unixUnmapfile(unixFile *pFd); /* ** This function performs the parts of the "close file" operation ** common to all locking schemes. It closes the directory and file ** handles, if they are valid, and sets all fields of the unixFile @@ -1813,10 +1891,11 @@ ** even on VxWorks. A mutex will be acquired on VxWorks by the ** vxworksReleaseFileId() routine. */ static int closeUnixFile(sqlite3_file *id){ unixFile *pFile = (unixFile*)id; + unixUnmapfile(pFile); if( pFile->h>=0 ){ robust_close(pFile, pFile->h, __LINE__); pFile->h = -1; } #if OS_VXWORKS @@ -1839,10 +1918,11 @@ ** Close a file. */ static int unixClose(sqlite3_file *id){ int rc = SQLITE_OK; unixFile *pFile = (unixFile *)id; + verifyDbFile(pFile); unixUnlock(id, NO_LOCK); unixEnterMutex(); /* unixFile.pInode is always valid here. Otherwise, a different close ** routine (e.g. nolockClose()) would be called instead. @@ -1907,11 +1987,11 @@ ******************************************************************************/ /****************************************************************************** ************************* Begin dot-file Locking ****************************** ** -** The dotfile locking implementation uses the existance of separate lock +** The dotfile locking implementation uses the existence of separate lock ** files (really a directory) to control access to the database. This works ** on just about every filesystem imaginable. But there are serious downsides: ** ** (1) There is zero concurrency. A single reader blocks all other ** connections from reading or writing the database. @@ -1922,11 +2002,11 @@ ** Nevertheless, a dotlock is an appropriate locking mode for use if no ** other locking strategy is available. ** ** Dotfile locking works by creating a subdirectory in the same directory as ** the database and with the same name but with a ".lock" extension added. -** The existance of a lock directory implies an EXCLUSIVE lock. All other +** The existence of a lock directory implies an EXCLUSIVE lock. All other ** lock types (SHARED, RESERVED, PENDING) are mapped into EXCLUSIVE. */ /* ** The file suffix added to the data base filename in order to create the @@ -3070,19 +3150,38 @@ sqlite3_int64 offset ){ unixFile *pFile = (unixFile *)id; int got; assert( id ); + assert( offset>=0 ); + assert( amt>0 ); /* If this is a database file (not a journal, master-journal or temp ** file), the bytes in the locking range should never be read or written. */ #if 0 assert( pFile->pUnused==0 || offset>=PENDING_BYTE+512 || offset+amt<=PENDING_BYTE ); #endif + +#if SQLITE_MAX_MMAP_SIZE>0 + /* Deal with as much of this read request as possible by transfering + ** data from the memory mapping using memcpy(). */ + if( offsetmmapSize ){ + if( offset+amt <= pFile->mmapSize ){ + memcpy(pBuf, &((u8 *)(pFile->pMapRegion))[offset], amt); + return SQLITE_OK; + }else{ + int nCopy = pFile->mmapSize - offset; + memcpy(pBuf, &((u8 *)(pFile->pMapRegion))[offset], nCopy); + pBuf = &((u8 *)pBuf)[nCopy]; + amt -= nCopy; + offset += nCopy; + } + } +#endif got = seekAndRead(pFile, offset, pBuf, amt); if( got==amt ){ return SQLITE_OK; }else if( got<0 ){ @@ -3093,52 +3192,65 @@ /* Unread parts of the buffer must be zero-filled */ memset(&((char*)pBuf)[got], 0, amt-got); return SQLITE_IOERR_SHORT_READ; } } + +/* +** Attempt to seek the file-descriptor passed as the first argument to +** absolute offset iOff, then attempt to write nBuf bytes of data from +** pBuf to it. If an error occurs, return -1 and set *piErrno. Otherwise, +** return the actual number of bytes written (which may be less than +** nBuf). +*/ +static int seekAndWriteFd( + int fd, /* File descriptor to write to */ + i64 iOff, /* File offset to begin writing at */ + const void *pBuf, /* Copy data from this buffer to the file */ + int nBuf, /* Size of buffer pBuf in bytes */ + int *piErrno /* OUT: Error number if error occurs */ +){ + int rc = 0; /* Value returned by system call */ + + assert( nBuf==(nBuf&0x1ffff) ); + nBuf &= 0x1ffff; + TIMER_START; + +#if defined(USE_PREAD) + do{ rc = osPwrite(fd, pBuf, nBuf, iOff); }while( rc<0 && errno==EINTR ); +#elif defined(USE_PREAD64) + do{ rc = osPwrite64(fd, pBuf, nBuf, iOff);}while( rc<0 && errno==EINTR); +#else + do{ + i64 iSeek = lseek(fd, iOff, SEEK_SET); + SimulateIOError( iSeek-- ); + + if( iSeek!=iOff ){ + if( piErrno ) *piErrno = (iSeek==-1 ? errno : 0); + return -1; + } + rc = osWrite(fd, pBuf, nBuf); + }while( rc<0 && errno==EINTR ); +#endif + + TIMER_END; + OSTRACE(("WRITE %-3d %5d %7lld %llu\n", fd, rc, iOff, TIMER_ELAPSED)); + + if( rc<0 && piErrno ) *piErrno = errno; + return rc; +} + /* ** Seek to the offset in id->offset then read cnt bytes into pBuf. ** Return the number of bytes actually read. Update the offset. ** ** To avoid stomping the errno value on a failed write the lastErrno value ** is set before returning. */ static int seekAndWrite(unixFile *id, i64 offset, const void *pBuf, int cnt){ - int got; -#if (!defined(USE_PREAD) && !defined(USE_PREAD64)) - i64 newOffset; -#endif - assert( cnt==(cnt&0x1ffff) ); - cnt &= 0x1ffff; - TIMER_START; -#if defined(USE_PREAD) - do{ got = osPwrite(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR ); -#elif defined(USE_PREAD64) - do{ got = osPwrite64(id->h, pBuf, cnt, offset);}while( got<0 && errno==EINTR); -#else - do{ - newOffset = lseek(id->h, offset, SEEK_SET); - SimulateIOError( newOffset-- ); - if( newOffset!=offset ){ - if( newOffset == -1 ){ - ((unixFile*)id)->lastErrno = errno; - }else{ - ((unixFile*)id)->lastErrno = 0; - } - return -1; - } - got = osWrite(id->h, pBuf, cnt); - }while( got<0 && errno==EINTR ); -#endif - TIMER_END; - if( got<0 ){ - ((unixFile*)id)->lastErrno = errno; - } - - OSTRACE(("WRITE %-3d %5d %7lld %llu\n", id->h, got, offset, TIMER_ELAPSED)); - return got; + return seekAndWriteFd(id->h, offset, pBuf, cnt, &id->lastErrno); } /* ** Write data from a buffer into a file. Return SQLITE_OK on success @@ -3181,10 +3293,27 @@ SimulateIOErrorBenign(0); if( rc!=4 || memcmp(oldCntr, &((char*)pBuf)[24-offset], 4)!=0 ){ pFile->transCntrChng = 1; /* The transaction counter has changed */ } } + } +#endif + +#if SQLITE_MAX_MMAP_SIZE>0 + /* Deal with as much of this write request as possible by transfering + ** data from the memory mapping using memcpy(). */ + if( offsetmmapSize ){ + if( offset+amt <= pFile->mmapSize ){ + memcpy(&((u8 *)(pFile->pMapRegion))[offset], pBuf, amt); + return SQLITE_OK; + }else{ + int nCopy = pFile->mmapSize - offset; + memcpy(&((u8 *)(pFile->pMapRegion))[offset], pBuf, nCopy); + pBuf = &((u8 *)pBuf)[nCopy]; + amt -= nCopy; + offset += nCopy; + } } #endif while( amt>0 && (wrote = seekAndWrite(pFile, offset, pBuf, amt))>0 ){ amt -= wrote; @@ -3411,11 +3540,11 @@ pFile->lastErrno = errno; return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath); } /* Also fsync the directory containing the file if the DIRSYNC flag - ** is set. This is a one-time occurrance. Many systems (examples: AIX) + ** is set. This is a one-time occurrence. Many systems (examples: AIX) ** are unable to fsync a directory, so ignore errors on the fsync. */ if( pFile->ctrlFlags & UNIXFILE_DIRSYNC ){ int dirfd; OSTRACE(("DIRSYNC %s (have_fullfsync=%d fullsync=%d)\n", pFile->zPath, @@ -3465,10 +3594,18 @@ */ if( pFile->inNormalWrite && nByte==0 ){ pFile->transCntrChng = 1; } #endif + + /* If the file was just truncated to a size smaller than the currently + ** mapped region, reduce the effective mapping size as well. SQLite will + ** use read() and write() to access data beyond this point from now on. + */ + if( nBytemmapSize ){ + pFile->mmapSize = nByte; + } return SQLITE_OK; } } @@ -3553,10 +3690,23 @@ iWrite += nBlk; } #endif } } + + if( pFile->mmapSizeMax>0 && nByte>pFile->mmapSize ){ + int rc; + if( pFile->szChunk<=0 ){ + if( robust_ftruncate(pFile->h, nByte) ){ + pFile->lastErrno = errno; + return unixLogError(SQLITE_IOERR_TRUNCATE, "ftruncate", pFile->zPath); + } + } + + rc = unixMapfile(pFile, nByte); + return rc; + } return SQLITE_OK; } /* @@ -3620,10 +3770,22 @@ if( zTFile ){ unixGetTempname(pFile->pVfs->mxPathname, zTFile); *(char**)pArg = zTFile; } return SQLITE_OK; + } + case SQLITE_FCNTL_MMAP_SIZE: { + i64 newLimit = *(i64*)pArg; + if( newLimit>sqlite3GlobalConfig.mxMmap ){ + newLimit = sqlite3GlobalConfig.mxMmap; + } + *(i64*)pArg = pFile->mmapSizeMax; + if( newLimit>=0 ){ + pFile->mmapSizeMax = newLimit; + if( newLimitmmapSize ) pFile->mmapSize = newLimit; + } + return SQLITE_OK; } #ifdef SQLITE_DEBUG /* The pager calls this method to signal that it has done ** a rollback and that the database is therefore unchanged and ** it hence it is OK for the transaction change counter to be @@ -3933,11 +4095,11 @@ int i; assert( p->pInode==pFd->pInode ); sqlite3_mutex_free(p->mutex); for(i=0; inRegion; i++){ if( p->h>=0 ){ - munmap(p->apRegion[i], p->szRegion); + osMunmap(p->apRegion[i], p->szRegion); }else{ sqlite3_free(p->apRegion[i]); } } sqlite3_free(p->apRegion); @@ -4173,28 +4335,36 @@ } if( sStat.st_sizeh, sStat.st_size, nByte)!=0 ){ - rc = unixLogError(SQLITE_IOERR_SHMSIZE, "fallocate", - pShmNode->zFilename); + if( !bExtend ){ goto shmpage_out; } -#else - if( robust_ftruncate(pShmNode->h, nByte) ){ - rc = unixLogError(SQLITE_IOERR_SHMSIZE, "ftruncate", - pShmNode->zFilename); - goto shmpage_out; + + /* Alternatively, if bExtend is true, extend the file. Do this by + ** writing a single byte to the end of each (OS) page being + ** allocated or extended. Technically, we need only write to the + ** last page in order to extend the file. But writing to all new + ** pages forces the OS to allocate them immediately, which reduces + ** the chances of SIGBUS while accessing the mapped region later on. + */ + else{ + static const int pgsz = 4096; + int iPg; + + /* Write to the last byte of each newly allocated or extended page */ + assert( (nByte % pgsz)==0 ); + for(iPg=(sStat.st_size/pgsz); iPg<(nByte/pgsz); iPg++){ + if( seekAndWriteFd(pShmNode->h, iPg*pgsz + pgsz-1, "", 1, 0)!=1 ){ + const char *zFile = pShmNode->zFilename; + rc = unixLogError(SQLITE_IOERR_SHMSIZE, "write", zFile); + goto shmpage_out; + } + } } -#endif } } /* Map the requested memory region into this processes address space. */ apNew = (char **)sqlite3_realloc( @@ -4206,11 +4376,11 @@ } pShmNode->apRegion = apNew; while(pShmNode->nRegion<=iRegion){ void *pMem; if( pShmNode->h>=0 ){ - pMem = mmap(0, szRegion, + pMem = osMmap(0, szRegion, pShmNode->isReadonly ? PROT_READ : PROT_READ|PROT_WRITE, MAP_SHARED, pShmNode->h, szRegion*(i64)pShmNode->nRegion ); if( pMem==MAP_FAILED ){ rc = unixLogError(SQLITE_IOERR_SHMMAP, "mmap", pShmNode->zFilename); @@ -4422,10 +4592,240 @@ # define unixShmMap 0 # define unixShmLock 0 # define unixShmBarrier 0 # define unixShmUnmap 0 #endif /* #ifndef SQLITE_OMIT_WAL */ + +/* +** If it is currently memory mapped, unmap file pFd. +*/ +static void unixUnmapfile(unixFile *pFd){ + assert( pFd->nFetchOut==0 ); +#if SQLITE_MAX_MMAP_SIZE>0 + if( pFd->pMapRegion ){ + osMunmap(pFd->pMapRegion, pFd->mmapSizeActual); + pFd->pMapRegion = 0; + pFd->mmapSize = 0; + pFd->mmapSizeActual = 0; + } +#endif +} + +#if SQLITE_MAX_MMAP_SIZE>0 +/* +** Return the system page size. +*/ +static int unixGetPagesize(void){ +#if HAVE_MREMAP + return 512; +#elif defined(_BSD_SOURCE) + return getpagesize(); +#else + return (int)sysconf(_SC_PAGESIZE); +#endif +} +#endif /* SQLITE_MAX_MMAP_SIZE>0 */ + +#if SQLITE_MAX_MMAP_SIZE>0 +/* +** Attempt to set the size of the memory mapping maintained by file +** descriptor pFd to nNew bytes. Any existing mapping is discarded. +** +** If successful, this function sets the following variables: +** +** unixFile.pMapRegion +** unixFile.mmapSize +** unixFile.mmapSizeActual +** +** If unsuccessful, an error message is logged via sqlite3_log() and +** the three variables above are zeroed. In this case SQLite should +** continue accessing the database using the xRead() and xWrite() +** methods. +*/ +static void unixRemapfile( + unixFile *pFd, /* File descriptor object */ + i64 nNew /* Required mapping size */ +){ + const char *zErr = "mmap"; + int h = pFd->h; /* File descriptor open on db file */ + u8 *pOrig = (u8 *)pFd->pMapRegion; /* Pointer to current file mapping */ + i64 nOrig = pFd->mmapSizeActual; /* Size of pOrig region in bytes */ + u8 *pNew = 0; /* Location of new mapping */ + int flags = PROT_READ; /* Flags to pass to mmap() */ + + assert( pFd->nFetchOut==0 ); + assert( nNew>pFd->mmapSize ); + assert( nNew<=pFd->mmapSizeMax ); + assert( nNew>0 ); + assert( pFd->mmapSizeActual>=pFd->mmapSize ); + assert( MAP_FAILED!=0 ); + + if( (pFd->ctrlFlags & UNIXFILE_RDONLY)==0 ) flags |= PROT_WRITE; + + if( pOrig ){ + const int szSyspage = unixGetPagesize(); + i64 nReuse = (pFd->mmapSize & ~(szSyspage-1)); + u8 *pReq = &pOrig[nReuse]; + + /* Unmap any pages of the existing mapping that cannot be reused. */ + if( nReuse!=nOrig ){ + osMunmap(pReq, nOrig-nReuse); + } + +#if HAVE_MREMAP + pNew = osMremap(pOrig, nReuse, nNew, MREMAP_MAYMOVE); + zErr = "mremap"; +#else + pNew = osMmap(pReq, nNew-nReuse, flags, MAP_SHARED, h, nReuse); + if( pNew!=MAP_FAILED ){ + if( pNew!=pReq ){ + osMunmap(pNew, nNew - nReuse); + pNew = 0; + }else{ + pNew = pOrig; + } + } +#endif + + /* The attempt to extend the existing mapping failed. Free it. */ + if( pNew==MAP_FAILED || pNew==0 ){ + osMunmap(pOrig, nReuse); + } + } + + /* If pNew is still NULL, try to create an entirely new mapping. */ + if( pNew==0 ){ + pNew = osMmap(0, nNew, flags, MAP_SHARED, h, 0); + } + + if( pNew==MAP_FAILED ){ + pNew = 0; + nNew = 0; + unixLogError(SQLITE_OK, zErr, pFd->zPath); + + /* If the mmap() above failed, assume that all subsequent mmap() calls + ** will probably fail too. Fall back to using xRead/xWrite exclusively + ** in this case. */ + pFd->mmapSizeMax = 0; + } + pFd->pMapRegion = (void *)pNew; + pFd->mmapSize = pFd->mmapSizeActual = nNew; +} +#endif + +/* +** Memory map or remap the file opened by file-descriptor pFd (if the file +** is already mapped, the existing mapping is replaced by the new). Or, if +** there already exists a mapping for this file, and there are still +** outstanding xFetch() references to it, this function is a no-op. +** +** If parameter nByte is non-negative, then it is the requested size of +** the mapping to create. Otherwise, if nByte is less than zero, then the +** requested size is the size of the file on disk. The actual size of the +** created mapping is either the requested size or the value configured +** using SQLITE_FCNTL_MMAP_LIMIT, whichever is smaller. +** +** SQLITE_OK is returned if no error occurs (even if the mapping is not +** recreated as a result of outstanding references) or an SQLite error +** code otherwise. +*/ +static int unixMapfile(unixFile *pFd, i64 nByte){ +#if SQLITE_MAX_MMAP_SIZE>0 + i64 nMap = nByte; + int rc; + + assert( nMap>=0 || pFd->nFetchOut==0 ); + if( pFd->nFetchOut>0 ) return SQLITE_OK; + + if( nMap<0 ){ + struct stat statbuf; /* Low-level file information */ + rc = osFstat(pFd->h, &statbuf); + if( rc!=SQLITE_OK ){ + return SQLITE_IOERR_FSTAT; + } + nMap = statbuf.st_size; + } + if( nMap>pFd->mmapSizeMax ){ + nMap = pFd->mmapSizeMax; + } + + if( nMap!=pFd->mmapSize ){ + if( nMap>0 ){ + unixRemapfile(pFd, nMap); + }else{ + unixUnmapfile(pFd); + } + } +#endif + + return SQLITE_OK; +} + +/* +** If possible, return a pointer to a mapping of file fd starting at offset +** iOff. The mapping must be valid for at least nAmt bytes. +** +** If such a pointer can be obtained, store it in *pp and return SQLITE_OK. +** Or, if one cannot but no error occurs, set *pp to 0 and return SQLITE_OK. +** Finally, if an error does occur, return an SQLite error code. The final +** value of *pp is undefined in this case. +** +** If this function does return a pointer, the caller must eventually +** release the reference by calling unixUnfetch(). +*/ +static int unixFetch(sqlite3_file *fd, i64 iOff, int nAmt, void **pp){ +#if SQLITE_MAX_MMAP_SIZE>0 + unixFile *pFd = (unixFile *)fd; /* The underlying database file */ +#endif + *pp = 0; + +#if SQLITE_MAX_MMAP_SIZE>0 + if( pFd->mmapSizeMax>0 ){ + if( pFd->pMapRegion==0 ){ + int rc = unixMapfile(pFd, -1); + if( rc!=SQLITE_OK ) return rc; + } + if( pFd->mmapSize >= iOff+nAmt ){ + *pp = &((u8 *)pFd->pMapRegion)[iOff]; + pFd->nFetchOut++; + } + } +#endif + return SQLITE_OK; +} + +/* +** If the third argument is non-NULL, then this function releases a +** reference obtained by an earlier call to unixFetch(). The second +** argument passed to this function must be the same as the corresponding +** argument that was passed to the unixFetch() invocation. +** +** Or, if the third argument is NULL, then this function is being called +** to inform the VFS layer that, according to POSIX, any existing mapping +** may now be invalid and should be unmapped. +*/ +static int unixUnfetch(sqlite3_file *fd, i64 iOff, void *p){ + unixFile *pFd = (unixFile *)fd; /* The underlying database file */ + UNUSED_PARAMETER(iOff); + + /* If p==0 (unmap the entire file) then there must be no outstanding + ** xFetch references. Or, if p!=0 (meaning it is an xFetch reference), + ** then there must be at least one outstanding. */ + assert( (p==0)==(pFd->nFetchOut==0) ); + + /* If p!=0, it must match the iOff value. */ + assert( p==0 || p==&((u8 *)pFd->pMapRegion)[iOff] ); + + if( p ){ + pFd->nFetchOut--; + }else{ + unixUnmapfile(pFd); + } + + assert( pFd->nFetchOut>=0 ); + return SQLITE_OK; +} /* ** Here ends the implementation of all sqlite3_file methods. ** ********************** End sqlite3_file Methods ******************************* @@ -4481,11 +4881,13 @@ unixSectorSize, /* xSectorSize */ \ unixDeviceCharacteristics, /* xDeviceCapabilities */ \ unixShmMap, /* xShmMap */ \ unixShmLock, /* xShmLock */ \ unixShmBarrier, /* xShmBarrier */ \ - unixShmUnmap /* xShmUnmap */ \ + unixShmUnmap, /* xShmUnmap */ \ + unixFetch, /* xFetch */ \ + unixUnfetch, /* xUnfetch */ \ }; \ static const sqlite3_io_methods *FINDER##Impl(const char *z, unixFile *p){ \ UNUSED_PARAMETER(z); UNUSED_PARAMETER(p); \ return &METHOD; \ } \ @@ -4498,11 +4900,11 @@ ** are also created. */ IOMETHODS( posixIoFinder, /* Finder function name */ posixIoMethods, /* sqlite3_io_methods object name */ - 2, /* shared memory is enabled */ + 3, /* shared memory and mmap are enabled */ unixClose, /* xClose method */ unixLock, /* xLock method */ unixUnlock, /* xUnlock method */ unixCheckReservedLock /* xCheckReservedLock method */ ) @@ -4749,10 +5151,11 @@ OSTRACE(("OPEN %-3d %s\n", h, zFilename)); pNew->h = h; pNew->pVfs = pVfs; pNew->zPath = zFilename; pNew->ctrlFlags = (u8)ctrlFlags; + pNew->mmapSizeMax = sqlite3GlobalConfig.mxMmap; if( sqlite3_uri_boolean(((ctrlFlags & UNIXFILE_URI) ? zFilename : 0), "psow", SQLITE_POWERSAFE_OVERWRITE) ){ pNew->ctrlFlags |= UNIXFILE_PSOW; } if( strcmp(pVfs->zName,"unix-excl")==0 ){ @@ -4785,11 +5188,11 @@ #endif ){ unixEnterMutex(); rc = findInodeInfo(pNew, &pNew->pInode); if( rc!=SQLITE_OK ){ - /* If an error occured in findInodeInfo(), close the file descriptor + /* If an error occurred in findInodeInfo(), close the file descriptor ** immediately, before releasing the mutex. findInodeInfo() may fail ** in two scenarios: ** ** (a) A call to fstat() failed. ** (b) A malloc failed. @@ -4884,19 +5287,19 @@ #if OS_VXWORKS if( rc!=SQLITE_OK ){ if( h>=0 ) robust_close(pNew, h, __LINE__); h = -1; osUnlink(zFilename); - isDelete = 0; + pNew->ctrlFlags |= UNIXFILE_DELETE; } - if( isDelete ) pNew->ctrlFlags |= UNIXFILE_DELETE; #endif if( rc!=SQLITE_OK ){ if( h>=0 ) robust_close(pNew, h, __LINE__); }else{ pNew->pMethod = pLockingStyle; OpenCounter(+1); + verifyDbFile(pNew); } return rc; } /* @@ -5423,11 +5826,11 @@ #endif return rc; } /* -** Test the existance of or access permissions of file zPath. The +** Test the existence of or access permissions of file zPath. The ** test performed depends on the value of flags: ** ** SQLITE_ACCESS_EXISTS: Return 1 if the file exists ** SQLITE_ACCESS_READWRITE: Return 1 if the file is read and writable. ** SQLITE_ACCESS_READONLY: Return 1 if the file is readable. @@ -6986,11 +7389,11 @@ }; unsigned int i; /* Loop counter */ /* Double-check that the aSyscall[] array has been constructed ** correctly. See ticket [bb3a86e890c8e96ab] */ - assert( ArraySize(aSyscall)==21 ); + assert( ArraySize(aSyscall)==24 ); /* Register all VFSes defined in the aVfs[] array */ for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){ sqlite3_vfs_register(&aVfs[i], i==0); } Index: src/os_win.c ================================================================== --- src/os_win.c +++ src/os_win.c @@ -148,15 +148,24 @@ HANDLE hMutex; /* Mutex used to control access to shared lock */ HANDLE hShared; /* Shared memory segment used for locking */ winceLock local; /* Locks obtained by this instance of winFile */ winceLock *shared; /* Global shared lock memory for the file */ #endif +#if SQLITE_MAX_MMAP_SIZE>0 + int nFetchOut; /* Number of outstanding xFetch references */ + HANDLE hMap; /* Handle for accessing memory mapping */ + void *pMapRegion; /* Area memory mapped */ + sqlite3_int64 mmapSize; /* Usable size of mapped region */ + sqlite3_int64 mmapSizeActual; /* Actual size of mapped region */ + sqlite3_int64 mmapSizeMax; /* Configured FCNTL_MMAP_SIZE value */ +#endif }; /* ** Allowed values for winFile.ctrlFlags */ +#define WINFILE_RDONLY 0x02 /* Connection is read only */ #define WINFILE_PERSIST_WAL 0x04 /* Persistent WAL mode */ #define WINFILE_PSOW 0x10 /* SQLITE_IOCAP_POWERSAFE_OVERWRITE */ /* * The size of the buffer used by sqlite3_win32_write_debug(). @@ -306,11 +315,11 @@ ** they may be overridden at runtime to facilitate fault injection during ** testing and sandboxing. The following array holds the names and pointers ** to all overrideable system calls. */ static struct win_syscall { - const char *zName; /* Name of the sytem call */ + const char *zName; /* Name of the system call */ sqlite3_syscall_ptr pCurrent; /* Current value of the system call */ sqlite3_syscall_ptr pDefault; /* Default value */ } aSyscall[] = { #if !SQLITE_OS_WINCE && !SQLITE_OS_WINRT { "AreFileApisANSI", (SYSCALL)AreFileApisANSI, 0 }, @@ -1512,11 +1521,11 @@ osLocalFree(zTemp); } } #endif if( 0 == dwLen ){ - sqlite3_snprintf(nBuf, zBuf, "OsError 0x%x (%u)", lastErrno, lastErrno); + sqlite3_snprintf(nBuf, zBuf, "OsError 0x%lx (%lu)", lastErrno, lastErrno); }else{ /* copy a maximum of nBuf chars to output buffer */ sqlite3_snprintf(nBuf, zBuf, "%s", zOut); /* free the UTF8 buffer */ sqlite3_free(zOut); @@ -1555,11 +1564,11 @@ assert( errcode!=SQLITE_OK ); if( zPath==0 ) zPath = ""; for(i=0; zMsg[i] && zMsg[i]!='\r' && zMsg[i]!='\n'; i++){} zMsg[i] = 0; sqlite3_log(errcode, - "os_win.c:%d: (%d) %s(%s) - %s", + "os_win.c:%d: (%lu) %s(%s) - %s", iLine, lastErrno, zFunc, zPath, zMsg ); return errcode; } @@ -2016,30 +2025,34 @@ LONG upperBits; /* Most sig. 32 bits of new offset */ LONG lowerBits; /* Least sig. 32 bits of new offset */ DWORD dwRet; /* Value returned by SetFilePointer() */ DWORD lastErrno; /* Value returned by GetLastError() */ + OSTRACE(("SEEK file=%p, offset=%lld\n", pFile->h, iOffset)); + upperBits = (LONG)((iOffset>>32) & 0x7fffffff); lowerBits = (LONG)(iOffset & 0xffffffff); /* API oddity: If successful, SetFilePointer() returns a dword ** containing the lower 32-bits of the new file-offset. Or, if it fails, ** it returns INVALID_SET_FILE_POINTER. However according to MSDN, ** INVALID_SET_FILE_POINTER may also be a valid new offset. So to determine - ** whether an error has actually occured, it is also necessary to call + ** whether an error has actually occurred, it is also necessary to call ** GetLastError(). */ dwRet = osSetFilePointer(pFile->h, lowerBits, &upperBits, FILE_BEGIN); if( (dwRet==INVALID_SET_FILE_POINTER && ((lastErrno = osGetLastError())!=NO_ERROR)) ){ pFile->lastErrno = lastErrno; winLogError(SQLITE_IOERR_SEEK, pFile->lastErrno, "seekWinFile", pFile->zPath); + OSTRACE(("SEEK file=%p, rc=SQLITE_IOERR_SEEK\n", pFile->h)); return 1; } + OSTRACE(("SEEK file=%p, rc=SQLITE_OK\n", pFile->h)); return 0; #else /* ** Same as above, except that this implementation works for WinRT. */ @@ -2052,16 +2065,23 @@ if(!bRet){ pFile->lastErrno = osGetLastError(); winLogError(SQLITE_IOERR_SEEK, pFile->lastErrno, "seekWinFile", pFile->zPath); + OSTRACE(("SEEK file=%p, rc=SQLITE_IOERR_SEEK\n", pFile->h)); return 1; } + OSTRACE(("SEEK file=%p, rc=SQLITE_OK\n", pFile->h)); return 0; #endif } + +#if SQLITE_MAX_MMAP_SIZE>0 +/* Forward references to VFS methods */ +static int winUnmapfile(winFile*); +#endif /* ** Close a file. ** ** It is reported that an attempt to close a handle might sometimes @@ -2078,12 +2098,18 @@ assert( id!=0 ); #ifndef SQLITE_OMIT_WAL assert( pFile->pShm==0 ); #endif - OSTRACE(("CLOSE %d\n", pFile->h)); assert( pFile->h!=NULL && pFile->h!=INVALID_HANDLE_VALUE ); + OSTRACE(("CLOSE file=%p\n", pFile->h)); + +#if SQLITE_MAX_MMAP_SIZE>0 + rc = winUnmapfile(pFile); + if( rc!=SQLITE_OK ) return rc; +#endif + do{ rc = osCloseHandle(pFile->h); /* SimulateIOError( rc=0; cnt=MX_CLOSE_ATTEMPT; ); */ }while( rc==0 && ++cnt < MX_CLOSE_ATTEMPT && (sqlite3_win32_sleep(100), 1) ); #if SQLITE_OS_WINCE @@ -2099,15 +2125,15 @@ sqlite3_win32_sleep(100); /* Wait a little before trying again */ } sqlite3_free(pFile->zDeleteOnClose); } #endif - OSTRACE(("CLOSE %d %s\n", pFile->h, rc ? "ok" : "failed")); if( rc ){ pFile->h = NULL; } OpenCounter(-1); + OSTRACE(("CLOSE file=%p, rc=%s\n", pFile->h, rc ? "ok" : "failed")); return rc ? SQLITE_OK : winLogError(SQLITE_IOERR_CLOSE, osGetLastError(), "winClose", pFile->zPath); } @@ -2128,15 +2154,37 @@ winFile *pFile = (winFile*)id; /* file handle */ DWORD nRead; /* Number of bytes actually read from file */ int nRetry = 0; /* Number of retrys */ assert( id!=0 ); + assert( amt>0 ); + assert( offset>=0 ); SimulateIOError(return SQLITE_IOERR_READ); - OSTRACE(("READ %d lock=%d\n", pFile->h, pFile->locktype)); + OSTRACE(("READ file=%p, buffer=%p, amount=%d, offset=%lld, lock=%d\n", + pFile->h, pBuf, amt, offset, pFile->locktype)); + +#if SQLITE_MAX_MMAP_SIZE>0 + /* Deal with as much of this read request as possible by transfering + ** data from the memory mapping using memcpy(). */ + if( offsetmmapSize ){ + if( offset+amt <= pFile->mmapSize ){ + memcpy(pBuf, &((u8 *)(pFile->pMapRegion))[offset], amt); + OSTRACE(("READ-MMAP file=%p, rc=SQLITE_OK\n", pFile->h)); + return SQLITE_OK; + }else{ + int nCopy = (int)(pFile->mmapSize - offset); + memcpy(pBuf, &((u8 *)(pFile->pMapRegion))[offset], nCopy); + pBuf = &((u8 *)pBuf)[nCopy]; + amt -= nCopy; + offset += nCopy; + } + } +#endif #if SQLITE_OS_WINCE if( seekWinFile(pFile, offset) ){ + OSTRACE(("READ file=%p, rc=SQLITE_FULL\n", pFile->h)); return SQLITE_FULL; } while( !osReadFile(pFile->h, pBuf, amt, &nRead, 0) ){ #else memset(&overlapped, 0, sizeof(OVERLAPPED)); @@ -2146,20 +2194,23 @@ osGetLastError()!=ERROR_HANDLE_EOF ){ #endif DWORD lastErrno; if( retryIoerr(&nRetry, &lastErrno) ) continue; pFile->lastErrno = lastErrno; + OSTRACE(("READ file=%p, rc=SQLITE_IOERR_READ\n", pFile->h)); return winLogError(SQLITE_IOERR_READ, pFile->lastErrno, "winRead", pFile->zPath); } logIoerr(nRetry); if( nRead<(DWORD)amt ){ /* Unread parts of the buffer must be zero-filled */ memset(&((char*)pBuf)[nRead], 0, amt-nRead); + OSTRACE(("READ file=%p, rc=SQLITE_IOERR_SHORT_READ\n", pFile->h)); return SQLITE_IOERR_SHORT_READ; } + OSTRACE(("READ file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } /* ** Write data from a buffer into a file. Return SQLITE_OK on success @@ -2169,20 +2220,39 @@ sqlite3_file *id, /* File to write into */ const void *pBuf, /* The bytes to be written */ int amt, /* Number of bytes to write */ sqlite3_int64 offset /* Offset into the file to begin writing at */ ){ - int rc = 0; /* True if error has occured, else false */ + int rc = 0; /* True if error has occurred, else false */ winFile *pFile = (winFile*)id; /* File handle */ int nRetry = 0; /* Number of retries */ assert( amt>0 ); assert( pFile ); SimulateIOError(return SQLITE_IOERR_WRITE); SimulateDiskfullError(return SQLITE_FULL); - OSTRACE(("WRITE %d lock=%d\n", pFile->h, pFile->locktype)); + OSTRACE(("WRITE file=%p, buffer=%p, amount=%d, offset=%lld, lock=%d\n", + pFile->h, pBuf, amt, offset, pFile->locktype)); + +#if SQLITE_MAX_MMAP_SIZE>0 + /* Deal with as much of this write request as possible by transfering + ** data from the memory mapping using memcpy(). */ + if( offsetmmapSize ){ + if( offset+amt <= pFile->mmapSize ){ + memcpy(&((u8 *)(pFile->pMapRegion))[offset], pBuf, amt); + OSTRACE(("WRITE-MMAP file=%p, rc=SQLITE_OK\n", pFile->h)); + return SQLITE_OK; + }else{ + int nCopy = (int)(pFile->mmapSize - offset); + memcpy(&((u8 *)(pFile->pMapRegion))[offset], pBuf, nCopy); + pBuf = &((u8 *)pBuf)[nCopy]; + amt -= nCopy; + offset += nCopy; + } + } +#endif #if SQLITE_OS_WINCE rc = seekWinFile(pFile, offset); if( rc==0 ){ #else @@ -2231,31 +2301,35 @@ } if( rc ){ if( ( pFile->lastErrno==ERROR_HANDLE_DISK_FULL ) || ( pFile->lastErrno==ERROR_DISK_FULL )){ + OSTRACE(("WRITE file=%p, rc=SQLITE_FULL\n", pFile->h)); return SQLITE_FULL; } + OSTRACE(("WRITE file=%p, rc=SQLITE_IOERR_WRITE\n", pFile->h)); return winLogError(SQLITE_IOERR_WRITE, pFile->lastErrno, "winWrite", pFile->zPath); }else{ logIoerr(nRetry); } + OSTRACE(("WRITE file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } /* ** Truncate an open file to a specified size */ static int winTruncate(sqlite3_file *id, sqlite3_int64 nByte){ winFile *pFile = (winFile*)id; /* File handle object */ int rc = SQLITE_OK; /* Return code for this function */ + DWORD lastErrno; assert( pFile ); - - OSTRACE(("TRUNCATE %d %lld\n", pFile->h, nByte)); SimulateIOError(return SQLITE_IOERR_TRUNCATE); + OSTRACE(("TRUNCATE file=%p, size=%lld, lock=%d\n", + pFile->h, nByte, pFile->locktype)); /* If the user has configured a chunk-size for this file, truncate the ** file so that it consists of an integer number of chunks (i.e. the ** actual file size after the operation may be larger than the requested ** size). @@ -2265,18 +2339,29 @@ } /* SetEndOfFile() returns non-zero when successful, or zero when it fails. */ if( seekWinFile(pFile, nByte) ){ rc = winLogError(SQLITE_IOERR_TRUNCATE, pFile->lastErrno, - "winTruncate1", pFile->zPath); - }else if( 0==osSetEndOfFile(pFile->h) ){ - pFile->lastErrno = osGetLastError(); + "winTruncate1", pFile->zPath); + }else if( 0==osSetEndOfFile(pFile->h) && + ((lastErrno = osGetLastError())!=ERROR_USER_MAPPED_FILE) ){ + pFile->lastErrno = lastErrno; rc = winLogError(SQLITE_IOERR_TRUNCATE, pFile->lastErrno, - "winTruncate2", pFile->zPath); + "winTruncate2", pFile->zPath); } - OSTRACE(("TRUNCATE %d %lld %s\n", pFile->h, nByte, rc ? "failed" : "ok")); +#if SQLITE_MAX_MMAP_SIZE>0 + /* If the file was truncated to a size smaller than the currently + ** mapped region, reduce the effective mapping size as well. SQLite will + ** use read() and write() to access data beyond this point from now on. + */ + if( pFile->pMapRegion && nBytemmapSize ){ + pFile->mmapSize = nByte; + } +#endif + + OSTRACE(("TRUNCATE file=%p, rc=%s\n", pFile->h, sqlite3ErrName(rc))); return rc; } #ifdef SQLITE_TEST /* @@ -2312,16 +2397,17 @@ /* Check that one of SQLITE_SYNC_NORMAL or FULL was passed */ assert((flags&0x0F)==SQLITE_SYNC_NORMAL || (flags&0x0F)==SQLITE_SYNC_FULL ); - OSTRACE(("SYNC %d lock=%d\n", pFile->h, pFile->locktype)); - /* Unix cannot, but some systems may return SQLITE_FULL from here. This ** line is to test that doing so does not cause any problems. */ SimulateDiskfullError( return SQLITE_FULL ); + + OSTRACE(("SYNC file=%p, flags=%x, lock=%d\n", + pFile->h, flags, pFile->locktype)); #ifndef SQLITE_TEST UNUSED_PARAMETER(flags); #else if( (flags&0x0F)==SQLITE_SYNC_FULL ){ @@ -2337,13 +2423,15 @@ return SQLITE_OK; #else rc = osFlushFileBuffers(pFile->h); SimulateIOError( rc=FALSE ); if( rc ){ + OSTRACE(("SYNC file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; }else{ pFile->lastErrno = osGetLastError(); + OSTRACE(("SYNC file=%p, rc=SQLITE_IOERR_FSYNC\n", pFile->h)); return winLogError(SQLITE_IOERR_FSYNC, pFile->lastErrno, "winSync", pFile->zPath); } #endif } @@ -2354,11 +2442,14 @@ static int winFileSize(sqlite3_file *id, sqlite3_int64 *pSize){ winFile *pFile = (winFile*)id; int rc = SQLITE_OK; assert( id!=0 ); + assert( pSize!=0 ); SimulateIOError(return SQLITE_IOERR_FSTAT); + OSTRACE(("SIZE file=%p, pSize=%p\n", pFile->h, pSize)); + #if SQLITE_OS_WINRT { FILE_STANDARD_INFO info; if( osGetFileInformationByHandleEx(pFile->h, FileStandardInfo, &info, sizeof(info)) ){ @@ -2383,10 +2474,12 @@ rc = winLogError(SQLITE_IOERR_FSTAT, pFile->lastErrno, "winFileSize", pFile->zPath); } } #endif + OSTRACE(("SIZE file=%p, pSize=%p, *pSize=%lld, rc=%s\n", + pFile->h, pSize, *pSize, sqlite3ErrName(rc))); return rc; } /* ** LOCKFILE_FAIL_IMMEDIATELY is undefined on some Windows systems. @@ -2424,10 +2517,11 @@ ** Different API routines are called depending on whether or not this ** is Win9x or WinNT. */ static int getReadLock(winFile *pFile){ int res; + OSTRACE(("READ-LOCK file=%p, lock=%d\n", pFile->h, pFile->locktype)); if( isNT() ){ #if SQLITE_OS_WINCE /* ** NOTE: Windows CE is handled differently here due its lack of the Win32 ** API LockFileEx. @@ -2449,19 +2543,21 @@ #endif if( res == 0 ){ pFile->lastErrno = osGetLastError(); /* No need to log a failure to lock */ } + OSTRACE(("READ-LOCK file=%p, rc=%s\n", pFile->h, sqlite3ErrName(res))); return res; } /* ** Undo a readlock */ static int unlockReadLock(winFile *pFile){ int res; DWORD lastErrno; + OSTRACE(("READ-UNLOCK file=%p, lock=%d\n", pFile->h, pFile->locktype)); if( isNT() ){ res = winUnlockFile(&pFile->h, SHARED_FIRST, 0, SHARED_SIZE, 0); } #ifdef SQLITE_WIN32_HAS_ANSI else{ @@ -2471,10 +2567,11 @@ if( res==0 && ((lastErrno = osGetLastError())!=ERROR_NOT_LOCKED) ){ pFile->lastErrno = lastErrno; winLogError(SQLITE_IOERR_UNLOCK, pFile->lastErrno, "unlockReadLock", pFile->zPath); } + OSTRACE(("READ-UNLOCK file=%p, rc=%s\n", pFile->h, sqlite3ErrName(res))); return res; } /* ** Lock the file with the lock specified by parameter locktype - one @@ -2509,18 +2606,19 @@ int gotPendingLock = 0;/* True if we acquired a PENDING lock this time */ winFile *pFile = (winFile*)id; DWORD lastErrno = NO_ERROR; assert( id!=0 ); - OSTRACE(("LOCK %d %d was %d(%d)\n", - pFile->h, locktype, pFile->locktype, pFile->sharedLockByte)); + OSTRACE(("LOCK file=%p, oldLock=%d(%d), newLock=%d\n", + pFile->h, pFile->locktype, pFile->sharedLockByte, locktype)); /* If there is already a lock of this type or more restrictive on the ** OsFile, do nothing. Don't use the end_lock: exit path, as ** sqlite3OsEnterMutex() hasn't been called yet. */ if( pFile->locktype>=locktype ){ + OSTRACE(("LOCK-HELD file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } /* Make sure the locking sequence is correct */ @@ -2544,11 +2642,12 @@ ** around problems caused by indexing and/or anti-virus software on ** Windows systems. ** If you are using this code as a model for alternative VFSes, do not ** copy this retry logic. It is a hack intended for Windows only. */ - OSTRACE(("could not get a PENDING lock. cnt=%d\n", cnt)); + OSTRACE(("LOCK-PENDING-FAIL file=%p, count=%d, rc=%s\n", + pFile->h, cnt, sqlite3ErrName(res))); if( cnt ) sqlite3_win32_sleep(1); } gotPendingLock = res; if( !res ){ lastErrno = osGetLastError(); @@ -2589,18 +2688,16 @@ /* Acquire an EXCLUSIVE lock */ if( locktype==EXCLUSIVE_LOCK && res ){ assert( pFile->locktype>=SHARED_LOCK ); res = unlockReadLock(pFile); - OSTRACE(("unreadlock = %d\n", res)); res = winLockFile(&pFile->h, SQLITE_LOCKFILE_FLAGS, SHARED_FIRST, 0, SHARED_SIZE, 0); if( res ){ newLocktype = EXCLUSIVE_LOCK; }else{ lastErrno = osGetLastError(); - OSTRACE(("error-code = %d\n", lastErrno)); getReadLock(pFile); } } /* If we are holding a PENDING lock that ought to be released, then @@ -2614,16 +2711,18 @@ ** return the appropriate result code. */ if( res ){ rc = SQLITE_OK; }else{ - OSTRACE(("LOCK FAILED %d trying for %d but got %d\n", pFile->h, - locktype, newLocktype)); + OSTRACE(("LOCK-FAIL file=%p, wanted=%d, got=%d\n", + pFile->h, locktype, newLocktype)); pFile->lastErrno = lastErrno; rc = SQLITE_BUSY; } pFile->locktype = (u8)newLocktype; + OSTRACE(("LOCK file=%p, lock=%d, rc=%s\n", + pFile->h, pFile->locktype, sqlite3ErrName(rc))); return rc; } /* ** This routine checks if there is a RESERVED lock held on the specified @@ -2633,24 +2732,27 @@ static int winCheckReservedLock(sqlite3_file *id, int *pResOut){ int rc; winFile *pFile = (winFile*)id; SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; ); + OSTRACE(("TEST-WR-LOCK file=%p, pResOut=%p\n", pFile->h, pResOut)); assert( id!=0 ); if( pFile->locktype>=RESERVED_LOCK ){ rc = 1; - OSTRACE(("TEST WR-LOCK %d %d (local)\n", pFile->h, rc)); + OSTRACE(("TEST-WR-LOCK file=%p, rc=%d (local)\n", pFile->h, rc)); }else{ - rc = winLockFile(&pFile->h, SQLITE_LOCKFILE_FLAGS, RESERVED_BYTE, 0, 1, 0); + rc = winLockFile(&pFile->h, SQLITE_LOCKFILEEX_FLAGS,RESERVED_BYTE, 0, 1, 0); if( rc ){ winUnlockFile(&pFile->h, RESERVED_BYTE, 0, 1, 0); } rc = !rc; - OSTRACE(("TEST WR-LOCK %d %d (remote)\n", pFile->h, rc)); + OSTRACE(("TEST-WR-LOCK file=%p, rc=%d (remote)\n", pFile->h, rc)); } *pResOut = rc; + OSTRACE(("TEST-WR-LOCK file=%p, pResOut=%p, *pResOut=%d, rc=SQLITE_OK\n", + pFile->h, pResOut, *pResOut)); return SQLITE_OK; } /* ** Lower the locking level on file descriptor id to locktype. locktype @@ -2667,12 +2769,12 @@ int type; winFile *pFile = (winFile*)id; int rc = SQLITE_OK; assert( pFile!=0 ); assert( locktype<=SHARED_LOCK ); - OSTRACE(("UNLOCK %d to %d was %d(%d)\n", pFile->h, locktype, - pFile->locktype, pFile->sharedLockByte)); + OSTRACE(("UNLOCK file=%p, oldLock=%d(%d), newLock=%d\n", + pFile->h, pFile->locktype, pFile->sharedLockByte, locktype)); type = pFile->locktype; if( type>=EXCLUSIVE_LOCK ){ winUnlockFile(&pFile->h, SHARED_FIRST, 0, SHARED_SIZE, 0); if( locktype==SHARED_LOCK && !getReadLock(pFile) ){ /* This should never happen. We should always be able to @@ -2689,10 +2791,12 @@ } if( type>=PENDING_LOCK ){ winUnlockFile(&pFile->h, PENDING_BYTE, 0, 1, 0); } pFile->locktype = (u8)locktype; + OSTRACE(("UNLOCK file=%p, lock=%d, rc=%s\n", + pFile->h, pFile->locktype, sqlite3ErrName(rc))); return rc; } /* ** If *pArg is inititially negative then this is a query. Set *pArg to @@ -2716,21 +2820,25 @@ /* ** Control and query of the open file handle. */ static int winFileControl(sqlite3_file *id, int op, void *pArg){ winFile *pFile = (winFile*)id; + OSTRACE(("FCNTL file=%p, op=%d, pArg=%p\n", pFile->h, op, pArg)); switch( op ){ case SQLITE_FCNTL_LOCKSTATE: { *(int*)pArg = pFile->locktype; + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_LAST_ERRNO: { *(int*)pArg = (int)pFile->lastErrno; + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_CHUNK_SIZE: { pFile->szChunk = *(int *)pArg; + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_SIZE_HINT: { if( pFile->szChunk>0 ){ sqlite3_int64 oldSz; @@ -2741,24 +2849,29 @@ SimulateIOErrorBenign(1); rc = winTruncate(id, newSz); SimulateIOErrorBenign(0); } } + OSTRACE(("FCNTL file=%p, rc=%s\n", pFile->h, sqlite3ErrName(rc))); return rc; } + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_PERSIST_WAL: { winModeBit(pFile, WINFILE_PERSIST_WAL, (int*)pArg); + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_POWERSAFE_OVERWRITE: { winModeBit(pFile, WINFILE_PSOW, (int*)pArg); + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_VFSNAME: { *(char**)pArg = sqlite3_mprintf("win32"); + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_WIN32_AV_RETRY: { int *a = (int*)pArg; if( a[0]>0 ){ @@ -2769,21 +2882,36 @@ if( a[1]>0 ){ win32IoerrRetryDelay = a[1]; }else{ a[1] = win32IoerrRetryDelay; } + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } case SQLITE_FCNTL_TEMPFILENAME: { char *zTFile = sqlite3MallocZero( pFile->pVfs->mxPathname ); if( zTFile ){ getTempname(pFile->pVfs->mxPathname, zTFile); *(char**)pArg = zTFile; } + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); + return SQLITE_OK; + } +#if SQLITE_MAX_MMAP_SIZE>0 + case SQLITE_FCNTL_MMAP_SIZE: { + i64 newLimit = *(i64*)pArg; + if( newLimit>sqlite3GlobalConfig.mxMmap ){ + newLimit = sqlite3GlobalConfig.mxMmap; + } + *(i64*)pArg = pFile->mmapSizeMax; + if( newLimit>=0 ) pFile->mmapSizeMax = newLimit; + OSTRACE(("FCNTL file=%p, rc=SQLITE_OK\n", pFile->h)); return SQLITE_OK; } +#endif } + OSTRACE(("FCNTL file=%p, rc=SQLITE_NOTFOUND\n", pFile->h)); return SQLITE_NOTFOUND; } /* ** Return the sector size in bytes of the underlying block device for @@ -2940,10 +3068,13 @@ int rc = 0; /* Result code form Lock/UnlockFileEx() */ /* Access to the winShmNode object is serialized by the caller */ assert( sqlite3_mutex_held(pFile->mutex) || pFile->nRef==0 ); + OSTRACE(("SHM-LOCK file=%p, lock=%d, offset=%d, size=%d\n", + pFile->hFile.h, lockType, ofst, nByte)); + /* Release/Acquire the system-level lock */ if( lockType==_SHM_UNLCK ){ rc = winUnlockFile(&pFile->hFile.h, ofst, 0, nByte, 0); }else{ /* Initialize the locking parameters */ @@ -2957,15 +3088,13 @@ }else{ pFile->lastErrno = osGetLastError(); rc = SQLITE_BUSY; } - OSTRACE(("SHM-LOCK %d %s %s 0x%08lx\n", - pFile->hFile.h, - rc==SQLITE_OK ? "ok" : "failed", - lockType==_SHM_UNLCK ? "UnlockFileEx" : "LockFileEx", - pFile->lastErrno)); + OSTRACE(("SHM-LOCK file=%p, func=%s, errno=%lu, rc=%s\n", + pFile->hFile.h, (lockType == _SHM_UNLCK) ? "winUnlockFile" : + "winLockFile", pFile->lastErrno, sqlite3ErrName(rc))); return rc; } /* Forward references to VFS methods */ @@ -2981,24 +3110,24 @@ static void winShmPurge(sqlite3_vfs *pVfs, int deleteFlag){ winShmNode **pp; winShmNode *p; BOOL bRc; assert( winShmMutexHeld() ); + OSTRACE(("SHM-PURGE pid=%lu, deleteFlag=%d\n", + osGetCurrentProcessId(), deleteFlag)); pp = &winShmNodeList; while( (p = *pp)!=0 ){ if( p->nRef==0 ){ int i; if( p->mutex ) sqlite3_mutex_free(p->mutex); for(i=0; inRegion; i++){ bRc = osUnmapViewOfFile(p->aRegion[i].pMap); - OSTRACE(("SHM-PURGE pid-%d unmap region=%d %s\n", - (int)osGetCurrentProcessId(), i, - bRc ? "ok" : "failed")); + OSTRACE(("SHM-PURGE-UNMAP pid=%lu, region=%d, rc=%s\n", + osGetCurrentProcessId(), i, bRc ? "ok" : "failed")); bRc = osCloseHandle(p->aRegion[i].hMap); - OSTRACE(("SHM-PURGE pid-%d close region=%d %s\n", - (int)osGetCurrentProcessId(), i, - bRc ? "ok" : "failed")); + OSTRACE(("SHM-PURGE-CLOSE pid=%lu, region=%d, rc=%s\n", + osGetCurrentProcessId(), i, bRc ? "ok" : "failed")); } if( p->hFile.h!=NULL && p->hFile.h!=INVALID_HANDLE_VALUE ){ SimulateIOErrorBenign(1); winClose((sqlite3_file *)&p->hFile); SimulateIOErrorBenign(0); @@ -3273,13 +3402,13 @@ p->exclMask |= mask; } } } sqlite3_mutex_leave(pShmNode->mutex); - OSTRACE(("SHM-LOCK shmid-%d, pid-%d got %03x,%03x %s\n", - p->id, (int)osGetCurrentProcessId(), p->sharedMask, p->exclMask, - rc ? "failed" : "ok")); + OSTRACE(("SHM-LOCK pid=%lu, id=%d, sharedMask=%03x, exclMask=%03x, rc=%s\n", + osGetCurrentProcessId(), p->id, p->sharedMask, p->exclMask, + sqlite3ErrName(rc))); return rc; } /* ** Implement a memory barrier or memory fence on shared memory. @@ -3396,12 +3525,12 @@ #elif defined(SQLITE_WIN32_HAS_ANSI) hMap = osCreateFileMappingA(pShmNode->hFile.h, NULL, PAGE_READWRITE, 0, nByte, NULL ); #endif - OSTRACE(("SHM-MAP pid-%d create region=%d nbyte=%d %s\n", - (int)osGetCurrentProcessId(), pShmNode->nRegion, nByte, + OSTRACE(("SHM-MAP-CREATE pid=%lu, region=%d, size=%d, rc=%s\n", + osGetCurrentProcessId(), pShmNode->nRegion, nByte, hMap ? "ok" : "failed")); if( hMap ){ int iOffset = pShmNode->nRegion*szRegion; int iOffsetShift = iOffset % winSysInfo.dwAllocationGranularity; #if SQLITE_OS_WINRT @@ -3411,12 +3540,12 @@ #else pMap = osMapViewOfFile(hMap, FILE_MAP_WRITE | FILE_MAP_READ, 0, iOffset - iOffsetShift, szRegion + iOffsetShift ); #endif - OSTRACE(("SHM-MAP pid-%d map region=%d offset=%d size=%d %s\n", - (int)osGetCurrentProcessId(), pShmNode->nRegion, iOffset, + OSTRACE(("SHM-MAP-MAP pid=%lu, region=%d, offset=%d, size=%d, rc=%s\n", + osGetCurrentProcessId(), pShmNode->nRegion, iOffset, szRegion, pMap ? "ok" : "failed")); } if( !pMap ){ pShmNode->lastErrno = osGetLastError(); rc = winLogError(SQLITE_IOERR_SHMMAP, pShmNode->lastErrno, @@ -3448,10 +3577,234 @@ # define winShmMap 0 # define winShmLock 0 # define winShmBarrier 0 # define winShmUnmap 0 #endif /* #ifndef SQLITE_OMIT_WAL */ + +/* +** Cleans up the mapped region of the specified file, if any. +*/ +#if SQLITE_MAX_MMAP_SIZE>0 +static int winUnmapfile(winFile *pFile){ + assert( pFile!=0 ); + OSTRACE(("UNMAP-FILE pid=%lu, pFile=%p, hMap=%p, pMapRegion=%p, " + "mmapSize=%lld, mmapSizeActual=%lld, mmapSizeMax=%lld\n", + osGetCurrentProcessId(), pFile, pFile->hMap, pFile->pMapRegion, + pFile->mmapSize, pFile->mmapSizeActual, pFile->mmapSizeMax)); + if( pFile->pMapRegion ){ + if( !osUnmapViewOfFile(pFile->pMapRegion) ){ + pFile->lastErrno = osGetLastError(); + OSTRACE(("UNMAP-FILE pid=%lu, pFile=%p, pMapRegion=%p, " + "rc=SQLITE_IOERR_MMAP\n", osGetCurrentProcessId(), pFile, + pFile->pMapRegion)); + return winLogError(SQLITE_IOERR_MMAP, pFile->lastErrno, + "winUnmap1", pFile->zPath); + } + pFile->pMapRegion = 0; + pFile->mmapSize = 0; + pFile->mmapSizeActual = 0; + } + if( pFile->hMap!=NULL ){ + if( !osCloseHandle(pFile->hMap) ){ + pFile->lastErrno = osGetLastError(); + OSTRACE(("UNMAP-FILE pid=%lu, pFile=%p, hMap=%p, rc=SQLITE_IOERR_MMAP\n", + osGetCurrentProcessId(), pFile, pFile->hMap)); + return winLogError(SQLITE_IOERR_MMAP, pFile->lastErrno, + "winUnmap2", pFile->zPath); + } + pFile->hMap = NULL; + } + OSTRACE(("UNMAP-FILE pid=%lu, pFile=%p, rc=SQLITE_OK\n", + osGetCurrentProcessId(), pFile)); + return SQLITE_OK; +} + +/* +** Memory map or remap the file opened by file-descriptor pFd (if the file +** is already mapped, the existing mapping is replaced by the new). Or, if +** there already exists a mapping for this file, and there are still +** outstanding xFetch() references to it, this function is a no-op. +** +** If parameter nByte is non-negative, then it is the requested size of +** the mapping to create. Otherwise, if nByte is less than zero, then the +** requested size is the size of the file on disk. The actual size of the +** created mapping is either the requested size or the value configured +** using SQLITE_FCNTL_MMAP_SIZE, whichever is smaller. +** +** SQLITE_OK is returned if no error occurs (even if the mapping is not +** recreated as a result of outstanding references) or an SQLite error +** code otherwise. +*/ +static int winMapfile(winFile *pFd, sqlite3_int64 nByte){ + sqlite3_int64 nMap = nByte; + int rc; + + assert( nMap>=0 || pFd->nFetchOut==0 ); + OSTRACE(("MAP-FILE pid=%lu, pFile=%p, size=%lld\n", + osGetCurrentProcessId(), pFd, nByte)); + + if( pFd->nFetchOut>0 ) return SQLITE_OK; + + if( nMap<0 ){ + rc = winFileSize((sqlite3_file*)pFd, &nMap); + if( rc ){ + OSTRACE(("MAP-FILE pid=%lu, pFile=%p, rc=SQLITE_IOERR_FSTAT\n", + osGetCurrentProcessId(), pFd)); + return SQLITE_IOERR_FSTAT; + } + } + if( nMap>pFd->mmapSizeMax ){ + nMap = pFd->mmapSizeMax; + } + nMap &= ~(sqlite3_int64)(winSysInfo.dwPageSize - 1); + + if( nMap==0 && pFd->mmapSize>0 ){ + winUnmapfile(pFd); + } + if( nMap!=pFd->mmapSize ){ + void *pNew = 0; + DWORD protect = PAGE_READONLY; + DWORD flags = FILE_MAP_READ; + + winUnmapfile(pFd); + if( (pFd->ctrlFlags & WINFILE_RDONLY)==0 ){ + protect = PAGE_READWRITE; + flags |= FILE_MAP_WRITE; + } +#if SQLITE_OS_WINRT + pFd->hMap = osCreateFileMappingFromApp(pFd->h, NULL, protect, nMap, NULL); +#elif defined(SQLITE_WIN32_HAS_WIDE) + pFd->hMap = osCreateFileMappingW(pFd->h, NULL, protect, + (DWORD)((nMap>>32) & 0xffffffff), + (DWORD)(nMap & 0xffffffff), NULL); +#elif defined(SQLITE_WIN32_HAS_ANSI) + pFd->hMap = osCreateFileMappingA(pFd->h, NULL, protect, + (DWORD)((nMap>>32) & 0xffffffff), + (DWORD)(nMap & 0xffffffff), NULL); +#endif + if( pFd->hMap==NULL ){ + pFd->lastErrno = osGetLastError(); + rc = winLogError(SQLITE_IOERR_MMAP, pFd->lastErrno, + "winMapfile", pFd->zPath); + /* Log the error, but continue normal operation using xRead/xWrite */ + OSTRACE(("MAP-FILE-CREATE pid=%lu, pFile=%p, rc=SQLITE_IOERR_MMAP\n", + osGetCurrentProcessId(), pFd)); + return SQLITE_OK; + } + assert( (nMap % winSysInfo.dwPageSize)==0 ); +#if SQLITE_OS_WINRT + pNew = osMapViewOfFileFromApp(pFd->hMap, flags, 0, nMap); +#else + assert( sizeof(SIZE_T)==sizeof(sqlite3_int64) || nMap<=0xffffffff ); + pNew = osMapViewOfFile(pFd->hMap, flags, 0, 0, (SIZE_T)nMap); +#endif + if( pNew==NULL ){ + osCloseHandle(pFd->hMap); + pFd->hMap = NULL; + pFd->lastErrno = osGetLastError(); + winLogError(SQLITE_IOERR_MMAP, pFd->lastErrno, + "winMapfile", pFd->zPath); + OSTRACE(("MAP-FILE-MAP pid=%lu, pFile=%p, rc=SQLITE_IOERR_MMAP\n", + osGetCurrentProcessId(), pFd)); + return SQLITE_OK; + } + pFd->pMapRegion = pNew; + pFd->mmapSize = nMap; + pFd->mmapSizeActual = nMap; + } + + OSTRACE(("MAP-FILE pid=%lu, pFile=%p, rc=SQLITE_OK\n", + osGetCurrentProcessId(), pFd)); + return SQLITE_OK; +} +#endif /* SQLITE_MAX_MMAP_SIZE>0 */ + +/* +** If possible, return a pointer to a mapping of file fd starting at offset +** iOff. The mapping must be valid for at least nAmt bytes. +** +** If such a pointer can be obtained, store it in *pp and return SQLITE_OK. +** Or, if one cannot but no error occurs, set *pp to 0 and return SQLITE_OK. +** Finally, if an error does occur, return an SQLite error code. The final +** value of *pp is undefined in this case. +** +** If this function does return a pointer, the caller must eventually +** release the reference by calling winUnfetch(). +*/ +static int winFetch(sqlite3_file *fd, i64 iOff, int nAmt, void **pp){ +#if SQLITE_MAX_MMAP_SIZE>0 + winFile *pFd = (winFile*)fd; /* The underlying database file */ +#endif + *pp = 0; + + OSTRACE(("FETCH pid=%lu, pFile=%p, offset=%lld, amount=%d, pp=%p\n", + osGetCurrentProcessId(), fd, iOff, nAmt, pp)); + +#if SQLITE_MAX_MMAP_SIZE>0 + if( pFd->mmapSizeMax>0 ){ + if( pFd->pMapRegion==0 ){ + int rc = winMapfile(pFd, -1); + if( rc!=SQLITE_OK ){ + OSTRACE(("FETCH pid=%lu, pFile=%p, rc=%s\n", + osGetCurrentProcessId(), pFd, sqlite3ErrName(rc))); + return rc; + } + } + if( pFd->mmapSize >= iOff+nAmt ){ + *pp = &((u8 *)pFd->pMapRegion)[iOff]; + pFd->nFetchOut++; + } + } +#endif + + OSTRACE(("FETCH pid=%lu, pFile=%p, pp=%p, *pp=%p, rc=SQLITE_OK\n", + osGetCurrentProcessId(), fd, pp, *pp)); + return SQLITE_OK; +} + +/* +** If the third argument is non-NULL, then this function releases a +** reference obtained by an earlier call to winFetch(). The second +** argument passed to this function must be the same as the corresponding +** argument that was passed to the winFetch() invocation. +** +** Or, if the third argument is NULL, then this function is being called +** to inform the VFS layer that, according to POSIX, any existing mapping +** may now be invalid and should be unmapped. +*/ +static int winUnfetch(sqlite3_file *fd, i64 iOff, void *p){ +#if SQLITE_MAX_MMAP_SIZE>0 + winFile *pFd = (winFile*)fd; /* The underlying database file */ + + /* If p==0 (unmap the entire file) then there must be no outstanding + ** xFetch references. Or, if p!=0 (meaning it is an xFetch reference), + ** then there must be at least one outstanding. */ + assert( (p==0)==(pFd->nFetchOut==0) ); + + /* If p!=0, it must match the iOff value. */ + assert( p==0 || p==&((u8 *)pFd->pMapRegion)[iOff] ); + + OSTRACE(("UNFETCH pid=%lu, pFile=%p, offset=%lld, p=%p\n", + osGetCurrentProcessId(), pFd, iOff, p)); + + if( p ){ + pFd->nFetchOut--; + }else{ + /* FIXME: If Windows truly always prevents truncating or deleting a + ** file while a mapping is held, then the following winUnmapfile() call + ** is unnecessary can can be omitted - potentially improving + ** performance. */ + winUnmapfile(pFd); + } + + assert( pFd->nFetchOut>=0 ); +#endif + + OSTRACE(("UNFETCH pid=%lu, pFile=%p, rc=SQLITE_OK\n", + osGetCurrentProcessId(), fd)); + return SQLITE_OK; +} /* ** Here ends the implementation of all sqlite3_file methods. ** ********************** End sqlite3_file Methods ******************************* @@ -3460,11 +3813,11 @@ /* ** This vector defines all the methods that can operate on an ** sqlite3_file for win32. */ static const sqlite3_io_methods winIoMethod = { - 2, /* iVersion */ + 3, /* iVersion */ winClose, /* xClose */ winRead, /* xRead */ winWrite, /* xWrite */ winTruncate, /* xTruncate */ winSync, /* xSync */ @@ -3476,11 +3829,13 @@ winSectorSize, /* xSectorSize */ winDeviceCharacteristics, /* xDeviceCharacteristics */ winShmMap, /* xShmMap */ winShmLock, /* xShmLock */ winShmBarrier, /* xShmBarrier */ - winShmUnmap /* xShmUnmap */ + winShmUnmap, /* xShmUnmap */ + winFetch, /* xFetch */ + winUnfetch /* xUnfetch */ }; /**************************************************************************** **************************** sqlite3_vfs methods **************************** ** @@ -3540,10 +3895,11 @@ zMulti = unicodeToUtf8(zWidePath); if( zMulti ){ sqlite3_snprintf(MAX_PATH-30, zTempPath, "%s", zMulti); sqlite3_free(zMulti); }else{ + OSTRACE(("TEMP-FILENAME rc=SQLITE_IOERR_NOMEM\n")); return SQLITE_IOERR_NOMEM; } } #ifdef SQLITE_WIN32_HAS_ANSI else{ @@ -3553,10 +3909,11 @@ zUtf8 = sqlite3_win32_mbcs_to_utf8(zMbcsPath); if( zUtf8 ){ sqlite3_snprintf(MAX_PATH-30, zTempPath, "%s", zUtf8); sqlite3_free(zUtf8); }else{ + OSTRACE(("TEMP-FILENAME rc=SQLITE_IOERR_NOMEM\n")); return SQLITE_IOERR_NOMEM; } } #endif #endif @@ -3565,10 +3922,11 @@ ** name. If it is not, return SQLITE_ERROR. */ nTempPath = sqlite3Strlen30(zTempPath); if( (nTempPath + sqlite3Strlen30(SQLITE_TEMP_FILE_PREFIX) + 18) >= nBuf ){ + OSTRACE(("TEMP-FILENAME rc=SQLITE_ERROR\n")); return SQLITE_ERROR; } for(i=nTempPath; i>0 && zTempPath[i-1]=='\\'; i--){} zTempPath[i] = 0; @@ -3582,12 +3940,12 @@ zBuf[j] = (char)zChars[ ((unsigned char)zBuf[j])%(sizeof(zChars)-1) ]; } zBuf[j] = 0; zBuf[j+1] = 0; - OSTRACE(("TEMP FILENAME: %s\n", zBuf)); - return SQLITE_OK; + OSTRACE(("TEMP-FILENAME name=%s, rc=SQLITE_OK\n", zBuf)); + return SQLITE_OK; } /* ** Return TRUE if the named file is really a directory. Return false if ** it is something other than a directory, or if there is any kind of memory @@ -3652,22 +4010,23 @@ #endif int isExclusive = (flags & SQLITE_OPEN_EXCLUSIVE); int isDelete = (flags & SQLITE_OPEN_DELETEONCLOSE); int isCreate = (flags & SQLITE_OPEN_CREATE); -#ifndef NDEBUG int isReadonly = (flags & SQLITE_OPEN_READONLY); -#endif int isReadWrite = (flags & SQLITE_OPEN_READWRITE); #ifndef NDEBUG int isOpenJournal = (isCreate && ( eType==SQLITE_OPEN_MASTER_JOURNAL || eType==SQLITE_OPEN_MAIN_JOURNAL || eType==SQLITE_OPEN_WAL )); #endif + + OSTRACE(("OPEN name=%s, pFile=%p, flags=%x, pOutFlags=%p\n", + zUtf8Name, id, flags, pOutFlags)); /* Check the following statements are true: ** ** (a) Exactly one of the READWRITE and READONLY flags must be set, and ** (b) if CREATE is set, then READWRITE must also be set, and @@ -3710,10 +4069,11 @@ if( !zUtf8Name ){ assert(isDelete && !isOpenJournal); memset(zTmpname, 0, MAX_PATH+2); rc = getTempname(MAX_PATH+2, zTmpname); if( rc!=SQLITE_OK ){ + OSTRACE(("OPEN name=%s, rc=%s", zUtf8Name, sqlite3ErrName(rc))); return rc; } zUtf8Name = zTmpname; } @@ -3725,15 +4085,17 @@ zUtf8Name[strlen(zUtf8Name)+1]==0 ); /* Convert the filename to the system encoding. */ zConverted = convertUtf8Filename(zUtf8Name); if( zConverted==0 ){ + OSTRACE(("OPEN name=%s, rc=SQLITE_IOERR_NOMEM", zUtf8Name)); return SQLITE_IOERR_NOMEM; } if( winIsDir(zConverted) ){ sqlite3_free(zConverted); + OSTRACE(("OPEN name=%s, rc=SQLITE_CANTOPEN_ISDIR", zUtf8Name)); return SQLITE_CANTOPEN_ISDIR; } if( isReadWrite ){ dwDesiredAccess = GENERIC_READ | GENERIC_WRITE; @@ -3820,13 +4182,12 @@ } } #endif logIoerr(cnt); - OSTRACE(("OPEN %d %s 0x%lx %s\n", - h, zName, dwDesiredAccess, - h==INVALID_HANDLE_VALUE ? "failed" : "ok")); + OSTRACE(("OPEN file=%p, name=%s, access=%lx, rc=%s\n", h, zUtf8Name, + dwDesiredAccess, (h==INVALID_HANDLE_VALUE) ? "failed" : "ok")); if( h==INVALID_HANDLE_VALUE ){ pFile->lastErrno = lastErrno; winLogError(SQLITE_CANTOPEN, pFile->lastErrno, "winOpen", zUtf8Name); sqlite3_free(zConverted); @@ -3846,16 +4207,21 @@ }else{ *pOutFlags = SQLITE_OPEN_READONLY; } } + OSTRACE(("OPEN file=%p, name=%s, access=%lx, pOutFlags=%p, *pOutFlags=%d, " + "rc=%s\n", h, zUtf8Name, dwDesiredAccess, pOutFlags, pOutFlags ? + *pOutFlags : 0, (h==INVALID_HANDLE_VALUE) ? "failed" : "ok")); + #if SQLITE_OS_WINCE if( isReadWrite && eType==SQLITE_OPEN_MAIN_DB && (rc = winceCreateLock(zName, pFile))!=SQLITE_OK ){ osCloseHandle(h); sqlite3_free(zConverted); + OSTRACE(("OPEN-CE-LOCK name=%s, rc=%s\n", zName, sqlite3ErrName(rc))); return rc; } if( isTemp ){ pFile->zDeleteOnClose = zConverted; }else @@ -3865,15 +4231,25 @@ } pFile->pMethod = &winIoMethod; pFile->pVfs = pVfs; pFile->h = h; + if( isReadonly ){ + pFile->ctrlFlags |= WINFILE_RDONLY; + } if( sqlite3_uri_boolean(zName, "psow", SQLITE_POWERSAFE_OVERWRITE) ){ pFile->ctrlFlags |= WINFILE_PSOW; } pFile->lastErrno = NO_ERROR; pFile->zPath = zName; +#if SQLITE_MAX_MMAP_SIZE>0 + pFile->hMap = NULL; + pFile->pMapRegion = 0; + pFile->mmapSize = 0; + pFile->mmapSizeActual = 0; + pFile->mmapSizeMax = sqlite3GlobalConfig.mxMmap; +#endif OpenCounter(+1); return rc; } @@ -3901,10 +4277,12 @@ void *zConverted; UNUSED_PARAMETER(pVfs); UNUSED_PARAMETER(syncDir); SimulateIOError(return SQLITE_IOERR_DELETE); + OSTRACE(("DELETE name=%s, syncDir=%d\n", zFilename, syncDir)); + zConverted = convertUtf8Filename(zFilename); if( zConverted==0 ){ return SQLITE_IOERR_NOMEM; } if( isNT() ){ @@ -3986,16 +4364,16 @@ "winDelete", zFilename); }else{ logIoerr(cnt); } sqlite3_free(zConverted); - OSTRACE(("DELETE \"%s\" %s\n", zFilename, (rc ? "failed" : "ok" ))); + OSTRACE(("DELETE name=%s, rc=%s\n", zFilename, sqlite3ErrName(rc))); return rc; } /* -** Check the existance and status of a file. +** Check the existence and status of a file. */ static int winAccess( sqlite3_vfs *pVfs, /* Not used on win32 */ const char *zFilename, /* Name of file to check */ int flags, /* Type of test to make on this file */ @@ -4006,12 +4384,16 @@ DWORD lastErrno; void *zConverted; UNUSED_PARAMETER(pVfs); SimulateIOError( return SQLITE_IOERR_ACCESS; ); + OSTRACE(("ACCESS name=%s, flags=%x, pResOut=%p\n", + zFilename, flags, pResOut)); + zConverted = convertUtf8Filename(zFilename); if( zConverted==0 ){ + OSTRACE(("ACCESS name=%s, rc=SQLITE_IOERR_NOMEM\n", zFilename)); return SQLITE_IOERR_NOMEM; } if( isNT() ){ int cnt = 0; WIN32_FILE_ATTRIBUTE_DATA sAttrData; @@ -4058,10 +4440,12 @@ break; default: assert(!"Invalid flags argument"); } *pResOut = rc; + OSTRACE(("ACCESS name=%s, pResOut=%p, *pResOut=%d, rc=SQLITE_OK\n", + zFilename, pResOut, *pResOut)); return SQLITE_OK; } /* @@ -4498,20 +4882,19 @@ /* Double-check that the aSyscall[] array has been constructed ** correctly. See ticket [bb3a86e890c8e96ab] */ assert( ArraySize(aSyscall)==74 ); -#ifndef SQLITE_OMIT_WAL /* get memory map allocation granularity */ memset(&winSysInfo, 0, sizeof(SYSTEM_INFO)); #if SQLITE_OS_WINRT osGetNativeSystemInfo(&winSysInfo); #else osGetSystemInfo(&winSysInfo); #endif - assert(winSysInfo.dwAllocationGranularity > 0); -#endif + assert( winSysInfo.dwAllocationGranularity>0 ); + assert( winSysInfo.dwPageSize>0 ); sqlite3_vfs_register(&winVfs, 1); return SQLITE_OK; } Index: src/pager.c ================================================================== --- src/pager.c +++ src/pager.c @@ -271,11 +271,11 @@ ** layer must either commit or rollback the transaction. ** ** * A write transaction is active. ** * An EXCLUSIVE or greater lock is held on the database file. ** * All writing and syncing of journal and database data has finished. -** If no error occured, all that remains is to finalize the journal to +** If no error occurred, all that remains is to finalize the journal to ** commit the transaction. If an error did occur, the caller will need ** to rollback the transaction. ** ** ERROR: ** @@ -519,11 +519,11 @@ ** journal file from being successfully finalized, the setMaster flag ** is cleared anyway (and the pager will move to ERROR state). ** ** doNotSpill, doNotSyncSpill ** -** These two boolean variables control the behaviour of cache-spills +** These two boolean variables control the behavior of cache-spills ** (calls made by the pcache module to the pagerStress() routine to ** write cached data to the file-system in order to free up memory). ** ** When doNotSpill is non-zero, writing to the database from pagerStress() ** is disabled altogether. This is done in a very obscure case that @@ -653,10 +653,15 @@ i64 journalHdr; /* Byte offset to previous journal header */ sqlite3_backup *pBackup; /* Pointer to list of ongoing backup processes */ PagerSavepoint *aSavepoint; /* Array of active savepoints */ int nSavepoint; /* Number of elements in aSavepoint[] */ char dbFileVers[16]; /* Changes whenever database file changes */ + + u8 bUseFetch; /* True to use xFetch() */ + int nMmapOut; /* Number of mmap pages currently outstanding */ + sqlite3_int64 szMmap; /* Desired maximum mmap size */ + PgHdr *pMmapFreelist; /* List of free mmap page headers (pDirty) */ /* ** End of the routinely-changing class members ***************************************************************************/ u16 nExtra; /* Add this many bytes to each in-memory page */ @@ -763,10 +768,20 @@ # define MEMDB 0 #else # define MEMDB pPager->memDb #endif +/* +** The macro USEFETCH is true if we are allowed to use the xFetch and xUnfetch +** interfaces to access the database using memory-mapped I/O. +*/ +#if SQLITE_MAX_MMAP_SIZE>0 +# define USEFETCH(x) ((x)->bUseFetch) +#else +# define USEFETCH(x) 0 +#endif + /* ** The maximum legal page number is (2^31 - 1). */ #define PAGER_MAX_PGNO 2147483647 @@ -1397,11 +1412,11 @@ put32bits(&zHeader[sizeof(aJournalMagic)], 0xffffffff); }else{ memset(zHeader, 0, sizeof(aJournalMagic)+4); } - /* The random check-hash initialiser */ + /* The random check-hash initializer */ sqlite3_randomness(sizeof(pPager->cksumInit), &pPager->cksumInit); put32bits(&zHeader[sizeof(aJournalMagic)+4], pPager->cksumInit); /* The initial database size */ put32bits(&zHeader[sizeof(aJournalMagic)+8], pPager->dbOrigSize); /* The assumed sector size for this process */ @@ -2250,11 +2265,11 @@ && isSynced ){ i64 ofst = (pgno-1)*(i64)pPager->pageSize; testcase( !isSavepnt && pPg!=0 && (pPg->flags&PGHDR_NEED_SYNC)!=0 ); assert( !pagerUseWal(pPager) ); - rc = sqlite3OsWrite(pPager->fd, (u8*)aData, pPager->pageSize, ofst); + rc = sqlite3OsWrite(pPager->fd, (u8 *)aData, pPager->pageSize, ofst); if( pgno>pPager->dbFileSize ){ pPager->dbFileSize = pgno; } if( pPager->pBackup ){ CODEC1(pPager, aData, pgno, 3, rc=SQLITE_NOMEM); @@ -2641,10 +2656,11 @@ Pgno mxPg = 0; /* Size of the original file in pages */ int rc; /* Result code of a subroutine */ int res = 1; /* Value returned by sqlite3OsAccess() */ char *zMaster = 0; /* Name of master journal file if any */ int needPagerReset; /* True to reset page prior to first page rollback */ + int nPlayback = 0; /* Total number of pages restored from journal */ /* Figure out how many records are in the journal. Abort early if ** the journal is empty. */ assert( isOpen(pPager->jfd) ); @@ -2741,11 +2757,13 @@ if( needPagerReset ){ pager_reset(pPager); needPagerReset = 0; } rc = pager_playback_one_page(pPager,&pPager->journalOff,0,1,0); - if( rc!=SQLITE_OK ){ + if( rc==SQLITE_OK ){ + nPlayback++; + }else{ if( rc==SQLITE_DONE ){ pPager->journalOff = szJ; break; }else if( rc==SQLITE_IOERR_SHORT_READ ){ /* If the journal has been truncated, simply stop reading and @@ -2811,10 +2829,14 @@ ** see if it is possible to delete the master journal. */ rc = pager_delmaster(pPager, zMaster); testcase( rc!=SQLITE_OK ); } + if( isHot && nPlayback ){ + sqlite3_log(SQLITE_NOTICE_RECOVER_ROLLBACK, "recovered %d pages from %s", + nPlayback, pPager->zJournal); + } /* The Pager.sectorSize variable may have been updated while rolling ** back a journal created by a process with a different sector size ** value. Reset it to the correct value for this process. */ @@ -2832,15 +2854,14 @@ ** the value read from the database file. ** ** If an IO error occurs, then the IO error is returned to the caller. ** Otherwise, SQLITE_OK is returned. */ -static int readDbPage(PgHdr *pPg){ +static int readDbPage(PgHdr *pPg, u32 iFrame){ Pager *pPager = pPg->pPager; /* Pager object associated with page pPg */ Pgno pgno = pPg->pgno; /* Page number to read */ int rc = SQLITE_OK; /* Return code */ - int isInWal = 0; /* True if page is in log file */ int pgsz = pPager->pageSize; /* Number of bytes to read */ assert( pPager->eState>=PAGER_READER && !MEMDB ); assert( isOpen(pPager->fd) ); @@ -2848,15 +2869,17 @@ assert( pPager->tempFile ); memset(pPg->pData, 0, pPager->pageSize); return SQLITE_OK; } - if( pagerUseWal(pPager) ){ +#ifndef SQLITE_OMIT_WAL + if( iFrame ){ /* Try to pull the page from the write-ahead log. */ - rc = sqlite3WalRead(pPager->pWal, pgno, &isInWal, pgsz, pPg->pData); - } - if( rc==SQLITE_OK && !isInWal ){ + rc = sqlite3WalReadFrame(pPager->pWal, iFrame, pgsz, pPg->pData); + }else +#endif + { i64 iOffset = (pgno-1)*(i64)pPager->pageSize; rc = sqlite3OsRead(pPager->fd, pPg->pData, pgsz, iOffset); if( rc==SQLITE_IOERR_SHORT_READ ){ rc = SQLITE_OK; } @@ -2931,16 +2954,21 @@ static int pagerUndoCallback(void *pCtx, Pgno iPg){ int rc = SQLITE_OK; Pager *pPager = (Pager *)pCtx; PgHdr *pPg; + assert( pagerUseWal(pPager) ); pPg = sqlite3PagerLookup(pPager, iPg); if( pPg ){ if( sqlite3PcachePageRefcount(pPg)==1 ){ sqlite3PcacheDrop(pPg); }else{ - rc = readDbPage(pPg); + u32 iFrame = 0; + rc = sqlite3WalFindFrame(pPager->pWal, pPg->pgno, &iFrame); + if( rc==SQLITE_OK ){ + rc = readDbPage(pPg, iFrame); + } if( rc==SQLITE_OK ){ pPager->xReiniter(pPg); } sqlite3PagerUnref(pPg); } @@ -3080,10 +3108,11 @@ sqlite3WalEndReadTransaction(pPager->pWal); rc = sqlite3WalBeginReadTransaction(pPager->pWal, &changed); if( rc!=SQLITE_OK || changed ){ pager_reset(pPager); + if( USEFETCH(pPager) ) sqlite3OsUnfetch(pPager->fd, 0, 0); } return rc; } #endif @@ -3340,10 +3369,33 @@ ** Change the maximum number of in-memory pages that are allowed. */ void sqlite3PagerSetCachesize(Pager *pPager, int mxPage){ sqlite3PcacheSetCachesize(pPager->pPCache, mxPage); } + +/* +** Invoke SQLITE_FCNTL_MMAP_SIZE based on the current value of szMmap. +*/ +static void pagerFixMaplimit(Pager *pPager){ +#if SQLITE_MAX_MMAP_SIZE>0 + sqlite3_file *fd = pPager->fd; + if( isOpen(fd) ){ + sqlite3_int64 sz; + pPager->bUseFetch = (fd->pMethods->iVersion>=3) && pPager->szMmap>0; + sz = pPager->szMmap; + sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_MMAP_SIZE, &sz); + } +#endif +} + +/* +** Change the maximum size of any memory mapping made of the database file. +*/ +void sqlite3PagerSetMmapLimit(Pager *pPager, sqlite3_int64 szMmap){ + pPager->szMmap = szMmap; + pagerFixMaplimit(pPager); +} /* ** Free as much memory as possible from the pager. */ void sqlite3PagerShrink(Pager *pPager){ @@ -3576,10 +3628,11 @@ if( rc==SQLITE_OK ){ if( nReserve<0 ) nReserve = pPager->nReserve; assert( nReserve>=0 && nReserve<1000 ); pPager->nReserve = (i16)nReserve; pagerReportSize(pPager); + pagerFixMaplimit(pPager); } return rc; } /* @@ -3729,11 +3782,11 @@ ** ** If the condition asserted by this function were not true, and the ** dirty page were to be discarded from the cache via the pagerStress() ** routine, pagerStress() would not write the current page content to ** the database file. If a savepoint transaction were rolled back after -** this happened, the correct behaviour would be to restore the current +** this happened, the correct behavior would be to restore the current ** content of the page. However, since this content is not present in either ** the database file or the portion of the rollback journal and ** sub-journal rolled back the content could not be restored and the ** database image would become corrupt. It is therefore fortunate that ** this circumstance cannot arise. @@ -3800,10 +3853,85 @@ if( rc==SQLITE_OK ){ rc = sqlite3OsFileSize(pPager->jfd, &pPager->journalHdr); } return rc; } + +/* +** Obtain a reference to a memory mapped page object for page number pgno. +** The new object will use the pointer pData, obtained from xFetch(). +** If successful, set *ppPage to point to the new page reference +** and return SQLITE_OK. Otherwise, return an SQLite error code and set +** *ppPage to zero. +** +** Page references obtained by calling this function should be released +** by calling pagerReleaseMapPage(). +*/ +static int pagerAcquireMapPage( + Pager *pPager, /* Pager object */ + Pgno pgno, /* Page number */ + void *pData, /* xFetch()'d data for this page */ + PgHdr **ppPage /* OUT: Acquired page object */ +){ + PgHdr *p; /* Memory mapped page to return */ + + if( pPager->pMmapFreelist ){ + *ppPage = p = pPager->pMmapFreelist; + pPager->pMmapFreelist = p->pDirty; + p->pDirty = 0; + memset(p->pExtra, 0, pPager->nExtra); + }else{ + *ppPage = p = (PgHdr *)sqlite3MallocZero(sizeof(PgHdr) + pPager->nExtra); + if( p==0 ){ + sqlite3OsUnfetch(pPager->fd, (i64)(pgno-1) * pPager->pageSize, pData); + return SQLITE_NOMEM; + } + p->pExtra = (void *)&p[1]; + p->flags = PGHDR_MMAP; + p->nRef = 1; + p->pPager = pPager; + } + + assert( p->pExtra==(void *)&p[1] ); + assert( p->pPage==0 ); + assert( p->flags==PGHDR_MMAP ); + assert( p->pPager==pPager ); + assert( p->nRef==1 ); + + p->pgno = pgno; + p->pData = pData; + pPager->nMmapOut++; + + return SQLITE_OK; +} + +/* +** Release a reference to page pPg. pPg must have been returned by an +** earlier call to pagerAcquireMapPage(). +*/ +static void pagerReleaseMapPage(PgHdr *pPg){ + Pager *pPager = pPg->pPager; + pPager->nMmapOut--; + pPg->pDirty = pPager->pMmapFreelist; + pPager->pMmapFreelist = pPg; + + assert( pPager->fd->pMethods->iVersion>=3 ); + sqlite3OsUnfetch(pPager->fd, (i64)(pPg->pgno-1)*pPager->pageSize, pPg->pData); +} + +/* +** Free all PgHdr objects stored in the Pager.pMmapFreelist list. +*/ +static void pagerFreeMapHdrs(Pager *pPager){ + PgHdr *p; + PgHdr *pNext; + for(p=pPager->pMmapFreelist; p; p=pNext){ + pNext = p->pDirty; + sqlite3_free(p); + } +} + /* ** Shutdown the page cache. Free all memory and close all files. ** ** If a transaction was in progress when this routine is called, that @@ -3821,10 +3949,11 @@ u8 *pTmp = (u8 *)pPager->pTmpSpace; assert( assert_pager_state(pPager) ); disable_simulated_io_errors(); sqlite3BeginBenignMalloc(); + pagerFreeMapHdrs(pPager); /* pPager->errCode = 0; */ pPager->exclusiveMode = 0; #ifndef SQLITE_OMIT_WAL sqlite3WalClose(pPager->pWal, pPager->ckptSyncFlags, pPager->pageSize, pTmp); pPager->pWal = 0; @@ -4082,11 +4211,13 @@ /* Before the first write, give the VFS a hint of what the final ** file size will be. */ assert( rc!=SQLITE_OK || isOpen(pPager->fd) ); - if( rc==SQLITE_OK && pPager->dbSize>pPager->dbHintSize ){ + if( rc==SQLITE_OK + && (pList->pDirty ? pPager->dbSize : pList->pgno+1)>pPager->dbHintSize + ){ sqlite3_int64 szFile = pPager->pageSize * (sqlite3_int64)pPager->dbSize; sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &szFile); pPager->dbHintSize = pPager->dbSize; } @@ -4636,10 +4767,11 @@ } /* pPager->xBusyHandler = 0; */ /* pPager->pBusyHandlerArg = 0; */ pPager->xReiniter = xReinit; /* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */ + /* pPager->szMmap = SQLITE_DEFAULT_MMAP_SIZE // will be set by btree.c */ *ppPager = pPager; return SQLITE_OK; } @@ -4927,13 +5059,15 @@ assert( (pPager->eLock==SHARED_LOCK) || (pPager->exclusiveMode && pPager->eLock>SHARED_LOCK) ); } - if( !pPager->tempFile - && (pPager->pBackup || sqlite3PcachePagecount(pPager->pPCache)>0) - ){ + if( !pPager->tempFile && ( + pPager->pBackup + || sqlite3PcachePagecount(pPager->pPCache)>0 + || USEFETCH(pPager) + )){ /* The shared-lock has just been acquired on the database file ** and there are already pages in the cache (from a previous ** read or write transaction). Check to see if the database ** has been modified. If the database has changed, flush the ** cache. @@ -4955,19 +5089,29 @@ if( rc ) goto failed; if( nPage>0 ){ IOTRACE(("CKVERS %p %d\n", pPager, sizeof(dbFileVers))); rc = sqlite3OsRead(pPager->fd, &dbFileVers, sizeof(dbFileVers), 24); - if( rc!=SQLITE_OK ){ + if( rc!=SQLITE_OK && rc!=SQLITE_IOERR_SHORT_READ ){ goto failed; } }else{ memset(dbFileVers, 0, sizeof(dbFileVers)); } if( memcmp(pPager->dbFileVers, dbFileVers, sizeof(dbFileVers))!=0 ){ pager_reset(pPager); + + /* Unmap the database file. It is possible that external processes + ** may have truncated the database file and then extended it back + ** to its original size while this process was not holding a lock. + ** In this case there may exist a Pager.pMap mapping that appears + ** to be the right size but is not actually valid. Avoid this + ** possibility by unmapping the db here. */ + if( USEFETCH(pPager) ){ + sqlite3OsUnfetch(pPager->fd, 0, 0); + } } } /* If there is a WAL file in the file-system, open this database in WAL ** mode. Otherwise, the following function call is a no-op. @@ -5005,11 +5149,11 @@ ** Except, in locking_mode=EXCLUSIVE when there is nothing to in ** the rollback journal, the unlock is not performed and there is ** nothing to rollback, so this routine is a no-op. */ static void pagerUnlockIfUnused(Pager *pPager){ - if( (sqlite3PcacheRefCount(pPager->pPCache)==0) ){ + if( pPager->nMmapOut==0 && (sqlite3PcacheRefCount(pPager->pPCache)==0) ){ pagerUnlockAndRollback(pPager); } } /* @@ -5064,17 +5208,31 @@ */ int sqlite3PagerAcquire( Pager *pPager, /* The pager open on the database file */ Pgno pgno, /* Page number to fetch */ DbPage **ppPage, /* Write a pointer to the page here */ - int noContent /* Do not bother reading content from disk if true */ + int flags /* PAGER_ACQUIRE_XXX flags */ ){ - int rc; - PgHdr *pPg; + int rc = SQLITE_OK; + PgHdr *pPg = 0; + u32 iFrame = 0; /* Frame to read from WAL file */ + const int noContent = (flags & PAGER_ACQUIRE_NOCONTENT); + + /* It is acceptable to use a read-only (mmap) page for any page except + ** page 1 if there is no write-transaction open or the ACQUIRE_READONLY + ** flag was specified by the caller. And so long as the db is not a + ** temporary or in-memory database. */ + const int bMmapOk = (pgno!=1 && USEFETCH(pPager) + && (pPager->eState==PAGER_READER || (flags & PAGER_ACQUIRE_READONLY)) +#ifdef SQLITE_HAS_CODEC + && pPager->xCodec==0 +#endif + ); assert( pPager->eState>=PAGER_READER ); assert( assert_pager_state(pPager) ); + assert( noContent==0 || bMmapOk==0 ); if( pgno==0 ){ return SQLITE_CORRUPT_BKPT; } @@ -5081,10 +5239,43 @@ /* If the pager is in the error state, return an error immediately. ** Otherwise, request the page from the PCache layer. */ if( pPager->errCode!=SQLITE_OK ){ rc = pPager->errCode; }else{ + + if( bMmapOk && pagerUseWal(pPager) ){ + rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iFrame); + if( rc!=SQLITE_OK ) goto pager_acquire_err; + } + + if( iFrame==0 && bMmapOk ){ + void *pData = 0; + + rc = sqlite3OsFetch(pPager->fd, + (i64)(pgno-1) * pPager->pageSize, pPager->pageSize, &pData + ); + + if( rc==SQLITE_OK && pData ){ + if( pPager->eState>PAGER_READER ){ + (void)sqlite3PcacheFetch(pPager->pPCache, pgno, 0, &pPg); + } + if( pPg==0 ){ + rc = pagerAcquireMapPage(pPager, pgno, pData, &pPg); + }else{ + sqlite3OsUnfetch(pPager->fd, (i64)(pgno-1)*pPager->pageSize, pData); + } + if( pPg ){ + assert( rc==SQLITE_OK ); + *ppPage = pPg; + return SQLITE_OK; + } + } + if( rc!=SQLITE_OK ){ + goto pager_acquire_err; + } + } + rc = sqlite3PcacheFetch(pPager->pPCache, pgno, 1, ppPage); } if( rc!=SQLITE_OK ){ /* Either the call to sqlite3PcacheFetch() returned an error or the @@ -5139,13 +5330,17 @@ sqlite3EndBenignMalloc(); } memset(pPg->pData, 0, pPager->pageSize); IOTRACE(("ZERO %p %d\n", pPager, pgno)); }else{ + if( pagerUseWal(pPager) && bMmapOk==0 ){ + rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iFrame); + if( rc!=SQLITE_OK ) goto pager_acquire_err; + } assert( pPg->pPager==pPager ); pPager->aStat[PAGER_STAT_MISS]++; - rc = readDbPage(pPg); + rc = readDbPage(pPg, iFrame); if( rc!=SQLITE_OK ){ goto pager_acquire_err; } } pager_set_pagehash(pPg); @@ -5194,11 +5389,15 @@ ** removed. */ void sqlite3PagerUnref(DbPage *pPg){ if( pPg ){ Pager *pPager = pPg->pPager; - sqlite3PcacheRelease(pPg); + if( pPg->flags & PGHDR_MMAP ){ + pagerReleaseMapPage(pPg); + }else{ + sqlite3PcacheRelease(pPg); + } pagerUnlockIfUnused(pPager); } } /* @@ -5529,10 +5728,11 @@ PgHdr *pPg = pDbPage; Pager *pPager = pPg->pPager; Pgno nPagePerSector = (pPager->sectorSize/pPager->pageSize); + assert( (pPg->flags & PGHDR_MMAP)==0 ); assert( pPager->eState>=PAGER_WRITER_LOCKED ); assert( pPager->eState!=PAGER_ERROR ); assert( assert_pager_state(pPager) ); if( nPagePerSector>1 ){ @@ -5728,10 +5928,15 @@ if( rc==SQLITE_OK ){ rc = sqlite3OsWrite(pPager->fd, zBuf, pPager->pageSize, 0); pPager->aStat[PAGER_STAT_WRITE]++; } if( rc==SQLITE_OK ){ + /* Update the pager's copy of the change-counter. Otherwise, the + ** next time a read transaction is opened the cache will be + ** flushed (as the change-counter values will not match). */ + const void *pCopy = (const void *)&((const char *)zBuf)[24]; + memcpy(&pPager->dbFileVers, pCopy, sizeof(pPager->dbFileVers)); pPager->changeCountDone = 1; } }else{ pPager->changeCountDone = 1; } @@ -6085,11 +6290,11 @@ }else{ rc = pager_playback(pPager, 0); } assert( pPager->eState==PAGER_READER || rc!=SQLITE_OK ); - assert( rc==SQLITE_OK || rc==SQLITE_FULL + assert( rc==SQLITE_OK || rc==SQLITE_FULL || rc==SQLITE_CORRUPT || rc==SQLITE_NOMEM || (rc&0xFF)==SQLITE_IOERR ); /* If an error occurs during a ROLLBACK, we can no longer trust the pager ** cache. So call pager_error() on the way out to make any error persistent. */ @@ -6819,15 +7024,16 @@ /* Open the connection to the log file. If this operation fails, ** (e.g. due to malloc() failure), return an error code. */ if( rc==SQLITE_OK ){ - rc = sqlite3WalOpen(pPager->pVfs, + rc = sqlite3WalOpen(pPager->pVfs, pPager->fd, pPager->zWal, pPager->exclusiveMode, pPager->journalSizeLimit, &pPager->pWal ); } + pagerFixMaplimit(pPager); return rc; } @@ -6914,10 +7120,11 @@ rc = pagerExclusiveLock(pPager); if( rc==SQLITE_OK ){ rc = sqlite3WalClose(pPager->pWal, pPager->ckptSyncFlags, pPager->pageSize, (u8*)pPager->pTmpSpace); pPager->pWal = 0; + pagerFixMaplimit(pPager); } } return rc; } Index: src/pager.h ================================================================== --- src/pager.h +++ src/pager.h @@ -76,10 +76,16 @@ #define PAGER_JOURNALMODE_OFF 2 /* Journal omitted. */ #define PAGER_JOURNALMODE_TRUNCATE 3 /* Commit by truncating journal */ #define PAGER_JOURNALMODE_MEMORY 4 /* In-memory journal file */ #define PAGER_JOURNALMODE_WAL 5 /* Use write-ahead logging */ +/* +** Flags that make up the mask passed to sqlite3PagerAcquire(). +*/ +#define PAGER_ACQUIRE_NOCONTENT 0x01 /* Do not load data from disk */ +#define PAGER_ACQUIRE_READONLY 0x02 /* Read-only page is acceptable */ + /* ** The remainder of this file contains the declarations of the functions ** that make up the Pager sub-system API. See source code comments for ** a detailed description of each routine. */ @@ -100,10 +106,11 @@ /* Functions used to configure a Pager object. */ void sqlite3PagerSetBusyhandler(Pager*, int(*)(void *), void *); int sqlite3PagerSetPagesize(Pager*, u32*, int); int sqlite3PagerMaxPageCount(Pager*, int); void sqlite3PagerSetCachesize(Pager*, int); +void sqlite3PagerSetMmapLimit(Pager *, sqlite3_int64); void sqlite3PagerShrink(Pager*); void sqlite3PagerSetSafetyLevel(Pager*,int,int,int); int sqlite3PagerLockingMode(Pager *, int); int sqlite3PagerSetJournalMode(Pager *, int); int sqlite3PagerGetJournalMode(Pager*); Index: src/parse.y ================================================================== --- src/parse.y +++ src/parse.y @@ -518,11 +518,13 @@ if( A ){ struct SrcList_item *pNew = &A->a[A->nSrc-1]; struct SrcList_item *pOld = F->a; pNew->zName = pOld->zName; pNew->zDatabase = pOld->zDatabase; + pNew->pSelect = pOld->pSelect; pOld->zName = pOld->zDatabase = 0; + pOld->pSelect = 0; } sqlite3SrcListDelete(pParse->db, F); }else{ Select *pSubquery; sqlite3SrcListShiftJoinType(F); Index: src/pcache.h ================================================================== --- src/pcache.h +++ src/pcache.h @@ -50,10 +50,12 @@ #define PGHDR_NEED_SYNC 0x004 /* Fsync the rollback journal before ** writing this page to the database */ #define PGHDR_NEED_READ 0x008 /* Content is unread */ #define PGHDR_REUSE_UNLIKELY 0x010 /* A hint that reuse is unlikely */ #define PGHDR_DONT_WRITE 0x020 /* Do not write content to disk */ + +#define PGHDR_MMAP 0x040 /* This is an mmap page object */ /* Initialize and shutdown the page cache subsystem */ int sqlite3PcacheInitialize(void); void sqlite3PcacheShutdown(void); Index: src/pragma.c ================================================================== --- src/pragma.c +++ src/pragma.c @@ -317,11 +317,11 @@ int iDb; /* Database index for */ char *aFcntl[4]; /* Argument to SQLITE_FCNTL_PRAGMA */ int rc; /* return value form SQLITE_FCNTL_PRAGMA */ sqlite3 *db = pParse->db; /* The database connection */ Db *pDb; /* The specific database being pragmaed */ - Vdbe *v = pParse->pVdbe = sqlite3VdbeCreate(db); /* Prepared statement */ + Vdbe *v = sqlite3GetVdbe(pParse); /* Prepared statement */ if( v==0 ) return; sqlite3VdbeRunOnlyOnce(v); pParse->nMem = 2; @@ -400,15 +400,16 @@ */ if( sqlite3StrICmp(zLeft,"default_cache_size")==0 ){ static const VdbeOpList getCacheSize[] = { { OP_Transaction, 0, 0, 0}, /* 0 */ { OP_ReadCookie, 0, 1, BTREE_DEFAULT_CACHE_SIZE}, /* 1 */ - { OP_IfPos, 1, 7, 0}, + { OP_IfPos, 1, 8, 0}, { OP_Integer, 0, 2, 0}, { OP_Subtract, 1, 2, 1}, - { OP_IfPos, 1, 7, 0}, + { OP_IfPos, 1, 8, 0}, { OP_Integer, 0, 1, 0}, /* 6 */ + { OP_Noop, 0, 0, 0}, { OP_ResultRow, 1, 1, 0}, }; int addr; if( sqlite3ReadSchema(pParse) ) goto pragma_out; sqlite3VdbeUsesBtree(v, iDb); @@ -742,10 +743,47 @@ pDb->pSchema->cache_size = size; sqlite3BtreeSetCacheSize(pDb->pBt, pDb->pSchema->cache_size); } }else + /* + ** PRAGMA [database.]mmap_size(N) + ** + ** Used to set mapping size limit. The mapping size limit is + ** used to limit the aggregate size of all memory mapped regions of the + ** database file. If this parameter is set to zero, then memory mapping + ** is not used at all. If N is negative, then the default memory map + ** limit determined by sqlite3_config(SQLITE_CONFIG_MMAP_SIZE) is set. + ** The parameter N is measured in bytes. + ** + ** This value is advisory. The underlying VFS is free to memory map + ** as little or as much as it wants. Except, if N is set to 0 then the + ** upper layers will never invoke the xFetch interfaces to the VFS. + */ + if( sqlite3StrICmp(zLeft,"mmap_size")==0 ){ + sqlite3_int64 sz; + assert( sqlite3SchemaMutexHeld(db, iDb, 0) ); + if( zRight ){ + int ii; + sqlite3Atoi64(zRight, &sz, 1000, SQLITE_UTF8); + if( sz<0 ) sz = sqlite3GlobalConfig.szMmap; + if( pId2->n==0 ) db->szMmap = sz; + for(ii=db->nDb-1; ii>=0; ii--){ + if( db->aDb[ii].pBt && (ii==iDb || pId2->n==0) ){ + sqlite3BtreeSetMmapLimit(db->aDb[ii].pBt, sz); + } + } + } + sz = -1; + if( sqlite3_file_control(db,zDb,SQLITE_FCNTL_MMAP_SIZE,&sz)==SQLITE_OK ){ +#if SQLITE_MAX_MMAP_SIZE==0 + sz = 0; +#endif + returnSingleInt(pParse, "mmap_size", sz); + } + }else + /* ** PRAGMA temp_store ** PRAGMA temp_store = "default"|"memory"|"file" ** ** Return or set the local value of the temp_store flag. Changing @@ -1526,10 +1564,15 @@ ** PRAGMA [database.]schema_version ** PRAGMA [database.]schema_version = ** ** PRAGMA [database.]user_version ** PRAGMA [database.]user_version = + ** + ** PRAGMA [database.]freelist_count = + ** + ** PRAGMA [database.]application_id + ** PRAGMA [database.]application_id = ** ** The pragma's schema_version and user_version are used to set or get ** the value of the schema-version and user-version, respectively. Both ** the schema-version and the user-version are 32-bit signed integers ** stored in the database header. @@ -1548,14 +1591,18 @@ ** applications for any purpose. */ if( sqlite3StrICmp(zLeft, "schema_version")==0 || sqlite3StrICmp(zLeft, "user_version")==0 || sqlite3StrICmp(zLeft, "freelist_count")==0 + || sqlite3StrICmp(zLeft, "application_id")==0 ){ int iCookie; /* Cookie index. 1 for schema-cookie, 6 for user-cookie. */ sqlite3VdbeUsesBtree(v, iDb); switch( zLeft[0] ){ + case 'a': case 'A': + iCookie = BTREE_APPLICATION_ID; + break; case 'f': case 'F': iCookie = BTREE_FREE_PAGE_COUNT; break; case 's': case 'S': iCookie = BTREE_SCHEMA_VERSION; Index: src/prepare.c ================================================================== --- src/prepare.c +++ src/prepare.c @@ -177,11 +177,11 @@ assert( sqlite3_mutex_held(db->mutex) ); assert( iDb==1 || sqlite3BtreeHoldsMutex(db->aDb[iDb].pBt) ); /* zMasterSchema and zInitScript are set to point at the master schema ** and initialisation script appropriate for the database being - ** initialised. zMasterName is the name of the master table. + ** initialized. zMasterName is the name of the master table. */ if( !OMIT_TEMPDB && iDb==1 ){ zMasterSchema = temp_master_schema; }else{ zMasterSchema = master_schema; @@ -402,11 +402,11 @@ if( rc ){ sqlite3ResetOneSchema(db, i); } } - /* Once all the other databases have been initialised, load the schema + /* Once all the other databases have been initialized, load the schema ** for the TEMP database. This is loaded last, as the TEMP database ** schema may contain references to objects in other databases. */ #ifndef SQLITE_OMIT_TEMPDB if( rc==SQLITE_OK && ALWAYS(db->nDb>1) @@ -425,11 +425,11 @@ return rc; } /* -** This routine is a no-op if the database schema is already initialised. +** This routine is a no-op if the database schema is already initialized. ** Otherwise, the schema is loaded. An error code is returned. */ int sqlite3ReadSchema(Parse *pParse){ int rc = SQLITE_OK; sqlite3 *db = pParse->db; @@ -652,11 +652,10 @@ azColName[i], SQLITE_STATIC); } } #endif - assert( db->init.busy==0 || saveSqlFlag==0 ); if( db->init.busy==0 ){ Vdbe *pVdbe = pParse->pVdbe; sqlite3VdbeSetSql(pVdbe, zSql, (int)(pParse->zTail-zSql), saveSqlFlag); } if( pParse->pVdbe && (rc!=SQLITE_OK || db->mallocFailed) ){ Index: src/resolve.c ================================================================== --- src/resolve.c +++ src/resolve.c @@ -386,11 +386,14 @@ ** In cases like this, replace pExpr with a copy of the expression that ** forms the result set entry ("a+b" in the example) and return immediately. ** Note that the expression in the result set should have already been ** resolved by the time the WHERE clause is resolved. */ - if( cnt==0 && (pEList = pNC->pEList)!=0 && zTab==0 ){ + if( (pEList = pNC->pEList)!=0 + && zTab==0 + && ((pNC->ncFlags & NC_AsMaybe)==0 || cnt==0) + ){ for(j=0; jnExpr; j++){ char *zAs = pEList->a[j].zName; if( zAs!=0 && sqlite3StrICmp(zAs, zCol)==0 ){ Expr *pOrig; assert( pExpr->pLeft==0 && pExpr->pRight==0 ); @@ -477,11 +480,13 @@ pExpr->pRight = 0; pExpr->op = (isTrigger ? TK_TRIGGER : TK_COLUMN); lookupname_end: if( cnt==1 ){ assert( pNC!=0 ); - sqlite3AuthRead(pParse, pExpr, pSchema, pNC->pSrcList); + if( pExpr->op!=TK_AS ){ + sqlite3AuthRead(pParse, pExpr, pSchema, pNC->pSrcList); + } /* Increment the nRef value on all name contexts from TopNC up to ** the point where the name matched. */ for(;;){ assert( pTopNC!=0 ); pTopNC->nRef++; @@ -1152,15 +1157,14 @@ ** ** Minor point: If this is the case, then the expression will be ** re-evaluated for each reference to it. */ sNC.pEList = p->pEList; - if( sqlite3ResolveExprNames(&sNC, p->pWhere) || - sqlite3ResolveExprNames(&sNC, p->pHaving) - ){ - return WRC_Abort; - } + if( sqlite3ResolveExprNames(&sNC, p->pHaving) ) return WRC_Abort; + sNC.ncFlags |= NC_AsMaybe; + if( sqlite3ResolveExprNames(&sNC, p->pWhere) ) return WRC_Abort; + sNC.ncFlags &= ~NC_AsMaybe; /* The ORDER BY and GROUP BY clauses may not refer to terms in ** outer queries */ sNC.pNext = 0; @@ -1277,10 +1281,11 @@ pParse->nHeight += pExpr->nHeight; } #endif savedHasAgg = pNC->ncFlags & NC_HasAgg; pNC->ncFlags &= ~NC_HasAgg; + memset(&w, 0, sizeof(w)); w.xExprCallback = resolveExprStep; w.xSelectCallback = resolveSelectStep; w.pParse = pNC->pParse; w.u.pNC = pNC; sqlite3WalkExpr(&w, pExpr); @@ -1317,11 +1322,12 @@ NameContext *pOuterNC /* Name context for parent SELECT statement */ ){ Walker w; assert( p!=0 ); + memset(&w, 0, sizeof(w)); w.xExprCallback = resolveExprStep; w.xSelectCallback = resolveSelectStep; w.pParse = pParse; w.u.pNC = pOuterNC; sqlite3WalkSelect(&w, p); } Index: src/select.c ================================================================== --- src/select.c +++ src/select.c @@ -3258,10 +3258,73 @@ } pFrom->pIndex = pIdx; } return SQLITE_OK; } +/* +** Detect compound SELECT statements that use an ORDER BY clause with +** an alternative collating sequence. +** +** SELECT ... FROM t1 EXCEPT SELECT ... FROM t2 ORDER BY .. COLLATE ... +** +** These are rewritten as a subquery: +** +** SELECT * FROM (SELECT ... FROM t1 EXCEPT SELECT ... FROM t2) +** ORDER BY ... COLLATE ... +** +** This transformation is necessary because the multiSelectOrderBy() routine +** above that generates the code for a compound SELECT with an ORDER BY clause +** uses a merge algorithm that requires the same collating sequence on the +** result columns as on the ORDER BY clause. See ticket +** http://www.sqlite.org/src/info/6709574d2a +** +** This transformation is only needed for EXCEPT, INTERSECT, and UNION. +** The UNION ALL operator works fine with multiSelectOrderBy() even when +** there are COLLATE terms in the ORDER BY. +*/ +static int convertCompoundSelectToSubquery(Walker *pWalker, Select *p){ + int i; + Select *pNew; + Select *pX; + sqlite3 *db; + struct ExprList_item *a; + SrcList *pNewSrc; + Parse *pParse; + Token dummy; + + if( p->pPrior==0 ) return WRC_Continue; + if( p->pOrderBy==0 ) return WRC_Continue; + for(pX=p; pX && (pX->op==TK_ALL || pX->op==TK_SELECT); pX=pX->pPrior){} + if( pX==0 ) return WRC_Continue; + a = p->pOrderBy->a; + for(i=p->pOrderBy->nExpr-1; i>=0; i--){ + if( a[i].pExpr->flags & EP_Collate ) break; + } + if( i<0 ) return WRC_Continue; + + /* If we reach this point, that means the transformation is required. */ + + pParse = pWalker->pParse; + db = pParse->db; + pNew = sqlite3DbMallocZero(db, sizeof(*pNew) ); + if( pNew==0 ) return WRC_Abort; + memset(&dummy, 0, sizeof(dummy)); + pNewSrc = sqlite3SrcListAppendFromTerm(pParse,0,0,0,&dummy,pNew,0,0); + if( pNewSrc==0 ) return WRC_Abort; + *pNew = *p; + p->pSrc = pNewSrc; + p->pEList = sqlite3ExprListAppend(pParse, 0, sqlite3Expr(db, TK_ALL, 0)); + p->op = TK_SELECT; + p->pWhere = 0; + pNew->pGroupBy = 0; + pNew->pHaving = 0; + pNew->pOrderBy = 0; + p->pPrior = 0; + pNew->pLimit = 0; + pNew->pOffset = 0; + return WRC_Continue; +} /* ** This routine is a Walker callback for "expanding" a SELECT statement. ** "Expanding" means to do the following: ** @@ -3574,13 +3637,16 @@ ** The calling function can detect the problem by looking at pParse->nErr ** and/or pParse->db->mallocFailed. */ static void sqlite3SelectExpand(Parse *pParse, Select *pSelect){ Walker w; - w.xSelectCallback = selectExpander; + memset(&w, 0, sizeof(w)); + w.xSelectCallback = convertCompoundSelectToSubquery; w.xExprCallback = exprWalkNoop; w.pParse = pParse; + sqlite3WalkSelect(&w, pSelect); + w.xSelectCallback = selectExpander; sqlite3WalkSelect(&w, pSelect); } #ifndef SQLITE_OMIT_SUBQUERY @@ -3632,13 +3698,15 @@ ** Use this routine after name resolution. */ static void sqlite3SelectAddTypeInfo(Parse *pParse, Select *pSelect){ #ifndef SQLITE_OMIT_SUBQUERY Walker w; + memset(&w, 0, sizeof(w)); w.xSelectCallback = selectAddSubqueryTypeInfo; w.xExprCallback = exprWalkNoop; w.pParse = pParse; + w.bSelectDepthFirst = 1; sqlite3WalkSelect(&w, pSelect); #endif } @@ -4045,11 +4113,11 @@ pItem->regReturn = ++pParse->nMem; topAddr = sqlite3VdbeAddOp2(v, OP_Integer, 0, pItem->regReturn); pItem->addrFillSub = topAddr+1; VdbeNoopComment((v, "materialize %s", pItem->pTab->zName)); if( pItem->isCorrelated==0 ){ - /* If the subquery is no correlated and if we are not inside of + /* If the subquery is not correlated and if we are not inside of ** a trigger, then we only need to compute the value of the subquery ** once. */ onceAddr = sqlite3CodeOnce(pParse); } sqlite3SelectDestInit(&dest, SRT_EphemTab, pItem->iCursor); @@ -4568,11 +4636,11 @@ ** first iteration (since the first iteration of the loop is ** guaranteed to operate on the row with the minimum or maximum ** value of x, the only row required). ** ** A special flag must be passed to sqlite3WhereBegin() to slightly - ** modify behaviour as follows: + ** modify behavior as follows: ** ** + If the query is a "SELECT min(x)", then the loop coded by ** where.c should not iterate over any values with a NULL value ** for x. ** Index: src/shell.c ================================================================== --- src/shell.c +++ src/shell.c @@ -1478,22 +1478,10 @@ p->zDbFilename, sqlite3_errmsg(db)); exit(1); } #ifndef SQLITE_OMIT_LOAD_EXTENSION sqlite3_enable_load_extension(p->db, 1); -#endif -#ifdef SQLITE_ENABLE_REGEXP - { - extern int sqlite3_add_regexp_func(sqlite3*); - sqlite3_add_regexp_func(db); - } -#endif -#ifdef SQLITE_ENABLE_SPELLFIX - { - extern int sqlite3_spellfix1_register(sqlite3*); - sqlite3_spellfix1_register(db); - } #endif } } /* @@ -1549,10 +1537,47 @@ } fprintf(stderr, "ERROR: Not a boolean value: \"%s\". Assuming \"no\".\n", zArg); return 0; } + +/* +** Interpret zArg as an integer value, possibly with suffixes. +*/ +static sqlite3_int64 integerValue(const char *zArg){ + sqlite3_int64 v = 0; + static const struct { char *zSuffix; int iMult; } aMult[] = { + { "KiB", 1024 }, + { "MiB", 1024*1024 }, + { "GiB", 1024*1024*1024 }, + { "KB", 1000 }, + { "MB", 1000000 }, + { "GB", 1000000000 }, + { "K", 1000 }, + { "M", 1000000 }, + { "G", 1000000000 }, + }; + int i; + int isNeg = 0; + if( zArg[0]=='-' ){ + isNeg = 1; + zArg++; + }else if( zArg[0]=='+' ){ + zArg++; + } + while( isdigit(zArg[0]) ){ + v = v*10 + zArg[0] - '0'; + zArg++; + } + for(i=0; idb, @@ -2263,12 +2287,11 @@ "SELECT sql FROM " " (SELECT sql sql, type type, tbl_name tbl_name, name name, rowid x" " FROM sqlite_master UNION ALL" " SELECT sql, type, tbl_name, name, rowid FROM sqlite_temp_master) " "WHERE type!='meta' AND sql NOTNULL AND name NOT LIKE 'sqlite_%'" - "ORDER BY substr(type,2,1)," - " CASE type WHEN 'view' THEN rowid ELSE name END", + "ORDER BY rowid", callback, &data, &zErrMsg ); } if( zErrMsg ){ fprintf(stderr,"Error: %s\n", zErrMsg); @@ -2386,13 +2409,13 @@ if( nPrintCol<1 ) nPrintCol = 1; nPrintRow = (nRow + nPrintCol - 1)/nPrintCol; for(i=0; iout, "%s%-*s", zSp, maxlen, azResult[j] ? azResult[j] : ""); } - printf("\n"); + fprintf(p->out, "\n"); } } for(ii=0; iidb, opt); - printf("%d (0x%08x)\n", rc, rc); + fprintf(p->out, "%d (0x%08x)\n", rc, rc); } else { fprintf(stderr,"Error: testctrl %s takes a single int option\n", azArg[1]); } break; @@ -2458,22 +2481,22 @@ case SQLITE_TESTCTRL_PRNG_SAVE: case SQLITE_TESTCTRL_PRNG_RESTORE: case SQLITE_TESTCTRL_PRNG_RESET: if( nArg==2 ){ rc = sqlite3_test_control(testctrl); - printf("%d (0x%08x)\n", rc, rc); + fprintf(p->out, "%d (0x%08x)\n", rc, rc); } else { fprintf(stderr,"Error: testctrl %s takes no options\n", azArg[1]); } break; /* sqlite3_test_control(int, uint) */ case SQLITE_TESTCTRL_PENDING_BYTE: if( nArg==3 ){ - unsigned int opt = (unsigned int)atoi(azArg[2]); + unsigned int opt = (unsigned int)integerValue(azArg[2]); rc = sqlite3_test_control(testctrl, opt); - printf("%d (0x%08x)\n", rc, rc); + fprintf(p->out, "%d (0x%08x)\n", rc, rc); } else { fprintf(stderr,"Error: testctrl %s takes a single unsigned" " int option\n", azArg[1]); } break; @@ -2482,11 +2505,11 @@ case SQLITE_TESTCTRL_ASSERT: case SQLITE_TESTCTRL_ALWAYS: if( nArg==3 ){ int opt = atoi(azArg[2]); rc = sqlite3_test_control(testctrl, opt); - printf("%d (0x%08x)\n", rc, rc); + fprintf(p->out, "%d (0x%08x)\n", rc, rc); } else { fprintf(stderr,"Error: testctrl %s takes a single int option\n", azArg[1]); } break; @@ -2495,11 +2518,11 @@ #ifdef SQLITE_N_KEYWORD case SQLITE_TESTCTRL_ISKEYWORD: if( nArg==3 ){ const char *opt = azArg[2]; rc = sqlite3_test_control(testctrl, opt); - printf("%d (0x%08x)\n", rc, rc); + fprintf(p->out, "%d (0x%08x)\n", rc, rc); } else { fprintf(stderr,"Error: testctrl %s takes a single char * option\n", azArg[1]); } break; @@ -2540,30 +2563,30 @@ } #endif }else if( c=='v' && strncmp(azArg[0], "version", n)==0 ){ - printf("SQLite %s %s\n" /*extra-version-info*/, + fprintf(p->out, "SQLite %s %s\n" /*extra-version-info*/, sqlite3_libversion(), sqlite3_sourceid()); }else if( c=='v' && strncmp(azArg[0], "vfsname", n)==0 ){ const char *zDbName = nArg==2 ? azArg[1] : "main"; char *zVfsName = 0; if( p->db ){ sqlite3_file_control(p->db, zDbName, SQLITE_FCNTL_VFSNAME, &zVfsName); if( zVfsName ){ - printf("%s\n", zVfsName); + fprintf(p->out, "%s\n", zVfsName); sqlite3_free(zVfsName); } } }else #if defined(SQLITE_DEBUG) && defined(SQLITE_ENABLE_WHERETRACE) if( c=='w' && strncmp(azArg[0], "wheretrace", n)==0 ){ extern int sqlite3WhereTrace; - sqlite3WhereTrace = atoi(azArg[1]); + sqlite3WhereTrace = booleanValue(azArg[1]); }else #endif if( c=='w' && strncmp(azArg[0], "width", n)==0 && nArg>1 ){ int j; @@ -2745,10 +2768,14 @@ errCnt++; } free(zSql); zSql = 0; nSql = 0; + }else if( zSql && _all_whitespace(zSql) ){ + free(zSql); + zSql = 0; + nSql = 0; } } if( zSql ){ if( !_all_whitespace(zSql) ){ fprintf(stderr, "Error: incomplete SQL: %s\n", zSql); @@ -2880,10 +2907,11 @@ " -help show this message\n" " -html set output mode to HTML\n" " -interactive force interactive I/O\n" " -line set output mode to 'line'\n" " -list set output mode to 'list'\n" + " -mmap N default mmap size set to N\n" #ifdef SQLITE_ENABLE_MULTIPLEX " -multiplex enable the multiplexor VFS\n" #endif " -nullvalue TEXT set text string for NULL values. Default ''\n" " -separator SEP set output field separator. Default: '|'\n" @@ -2999,16 +3027,11 @@ int j, c; const char *zSize; sqlite3_int64 szHeap; zSize = cmdline_option_value(argc, argv, ++i); - szHeap = atoi(zSize); - for(j=0; (c = zSize[j])!=0; j++){ - if( c=='M' ){ szHeap *= 1000000; break; } - if( c=='K' ){ szHeap *= 1000; break; } - if( c=='G' ){ szHeap *= 1000000000; break; } - } + szHeap = integerValue(zSize); if( szHeap>0x7fff0000 ) szHeap = 0x7fff0000; sqlite3_config(SQLITE_CONFIG_HEAP, malloc((int)szHeap), (int)szHeap, 64); #endif #ifdef SQLITE_ENABLE_VFSTRACE }else if( strcmp(z,"-vfstrace")==0 ){ @@ -3024,10 +3047,13 @@ #ifdef SQLITE_ENABLE_MULTIPLEX }else if( strcmp(z,"-multiplex")==0 ){ extern int sqlite3_multiple_initialize(const char*,int); sqlite3_multiplex_initialize(0, 1); #endif + }else if( strcmp(z,"-mmap")==0 ){ + sqlite3_int64 sz = integerValue(cmdline_option_value(argc,argv,++i)); + sqlite3_config(SQLITE_CONFIG_MMAP_SIZE, sz, sz); }else if( strcmp(z,"-vfs")==0 ){ sqlite3_vfs *pVfs = sqlite3_vfs_find(cmdline_option_value(argc,argv,++i)); if( pVfs ){ sqlite3_vfs_register(pVfs, 1); }else{ @@ -3109,10 +3135,12 @@ stdin_is_interactive = 1; }else if( strcmp(z,"-batch")==0 ){ stdin_is_interactive = 0; }else if( strcmp(z,"-heap")==0 ){ i++; + }else if( strcmp(z,"-mmap")==0 ){ + i++; }else if( strcmp(z,"-vfs")==0 ){ i++; #ifdef SQLITE_ENABLE_VFSTRACE }else if( strcmp(z,"-vfstrace")==0 ){ i++; @@ -3126,11 +3154,11 @@ }else if( strcmp(z,"-cmd")==0 ){ if( i==argc-1 ) break; z = cmdline_option_value(argc,argv,++i); if( z[0]=='.' ){ rc = do_meta_command(z, &data); - if( rc && bail_on_error ) return rc; + if( rc && bail_on_error ) return rc==2 ? 0 : rc; }else{ open_db(&data); rc = shell_exec(data.db, z, shell_callback, &data, &zErrMsg); if( zErrMsg!=0 ){ fprintf(stderr,"Error: %s\n", zErrMsg); @@ -3150,10 +3178,11 @@ if( zFirstCmd ){ /* Run just the command that follows the database name */ if( zFirstCmd[0]=='.' ){ rc = do_meta_command(zFirstCmd, &data); + if( rc==2 ) rc = 0; }else{ open_db(&data); rc = shell_exec(data.db, zFirstCmd, shell_callback, &data, &zErrMsg); if( zErrMsg!=0 ){ fprintf(stderr,"Error: %s\n", zErrMsg); Index: src/sqlite.h.in ================================================================== --- src/sqlite.h.in +++ src/sqlite.h.in @@ -418,10 +418,12 @@ #define SQLITE_NOLFS 22 /* Uses OS features not supported on host */ #define SQLITE_AUTH 23 /* Authorization denied */ #define SQLITE_FORMAT 24 /* Auxiliary database format error */ #define SQLITE_RANGE 25 /* 2nd parameter to sqlite3_bind out of range */ #define SQLITE_NOTADB 26 /* File opened that is not a database file */ +#define SQLITE_NOTICE 27 /* Notifications from sqlite3_log() */ +#define SQLITE_WARNING 28 /* Warnings from sqlite3_log() */ #define SQLITE_ROW 100 /* sqlite3_step() has another row ready */ #define SQLITE_DONE 101 /* sqlite3_step() has finished executing */ /* end-of-error-codes */ /* @@ -468,10 +470,11 @@ #define SQLITE_IOERR_SHMSIZE (SQLITE_IOERR | (19<<8)) #define SQLITE_IOERR_SHMLOCK (SQLITE_IOERR | (20<<8)) #define SQLITE_IOERR_SHMMAP (SQLITE_IOERR | (21<<8)) #define SQLITE_IOERR_SEEK (SQLITE_IOERR | (22<<8)) #define SQLITE_IOERR_DELETE_NOENT (SQLITE_IOERR | (23<<8)) +#define SQLITE_IOERR_MMAP (SQLITE_IOERR | (24<<8)) #define SQLITE_LOCKED_SHAREDCACHE (SQLITE_LOCKED | (1<<8)) #define SQLITE_BUSY_RECOVERY (SQLITE_BUSY | (1<<8)) #define SQLITE_CANTOPEN_NOTEMPDIR (SQLITE_CANTOPEN | (1<<8)) #define SQLITE_CANTOPEN_ISDIR (SQLITE_CANTOPEN | (2<<8)) #define SQLITE_CANTOPEN_FULLPATH (SQLITE_CANTOPEN | (3<<8)) @@ -487,10 +490,12 @@ #define SQLITE_CONSTRAINT_NOTNULL (SQLITE_CONSTRAINT | (5<<8)) #define SQLITE_CONSTRAINT_PRIMARYKEY (SQLITE_CONSTRAINT | (6<<8)) #define SQLITE_CONSTRAINT_TRIGGER (SQLITE_CONSTRAINT | (7<<8)) #define SQLITE_CONSTRAINT_UNIQUE (SQLITE_CONSTRAINT | (8<<8)) #define SQLITE_CONSTRAINT_VTAB (SQLITE_CONSTRAINT | (9<<8)) +#define SQLITE_NOTICE_RECOVER_WAL (SQLITE_NOTICE | (1<<8)) +#define SQLITE_NOTICE_RECOVER_ROLLBACK (SQLITE_NOTICE | (2<<8)) /* ** CAPI3REF: Flags For File Open Operations ** ** These bit values are intended for use in the @@ -726,10 +731,13 @@ int (*xShmMap)(sqlite3_file*, int iPg, int pgsz, int, void volatile**); int (*xShmLock)(sqlite3_file*, int offset, int n, int flags); void (*xShmBarrier)(sqlite3_file*); int (*xShmUnmap)(sqlite3_file*, int deleteFlag); /* Methods above are valid for version 2 */ + int (*xFetch)(sqlite3_file*, sqlite3_int64 iOfst, int iAmt, void **pp); + int (*xUnfetch)(sqlite3_file*, sqlite3_int64 iOfst, void *p); + /* Methods above are valid for version 3 */ /* Additional methods may be added in future releases */ }; /* ** CAPI3REF: Standard File Control Opcodes @@ -862,11 +870,12 @@ ** compilation of the PRAGMA fails with an error. ^The [SQLITE_FCNTL_PRAGMA] ** file control occurs at the beginning of pragma statement analysis and so ** it is able to override built-in [PRAGMA] statements. ** **
  • [[SQLITE_FCNTL_BUSYHANDLER]] -** ^This file-control may be invoked by SQLite on the database file handle +** ^The [SQLITE_FCNTL_BUSYHANDLER] +** file-control may be invoked by SQLite on the database file handle ** shortly after it is opened in order to provide a custom VFS with access ** to the connections busy-handler callback. The argument is of type (void **) ** - an array of two (void *) values. The first (void *) actually points ** to a function of type (int (*)(void *)). In order to invoke the connections ** busy-handler, this function should be invoked with the second (void *) in @@ -873,17 +882,28 @@ ** the array as the only argument. If it returns non-zero, then the operation ** should be retried. If it returns zero, the custom VFS should abandon the ** current operation. ** **
  • [[SQLITE_FCNTL_TEMPFILENAME]] -** ^Application can invoke this file-control to have SQLite generate a +** ^Application can invoke the [SQLITE_FCNTL_TEMPFILENAME] file-control +** to have SQLite generate a ** temporary filename using the same algorithm that is followed to generate ** temporary filenames for TEMP tables and other internal uses. The ** argument should be a char** which will be filled with the filename ** written into memory obtained from [sqlite3_malloc()]. The caller should ** invoke [sqlite3_free()] on the result to avoid a memory leak. ** +**
  • [[SQLITE_FCNTL_MMAP_SIZE]] +** The [SQLITE_FCNTL_MMAP_SIZE] file control is used to query or set the +** maximum number of bytes that will be used for memory-mapped I/O. +** The argument is a pointer to a value of type sqlite3_int64 that +** is an advisory maximum number of bytes in the file to memory map. The +** pointer is overwritten with the old value. The limit is not changed if +** the value originally pointed to is negative, and so the current limit +** can be queried by passing in a pointer to a negative number. This +** file-control is used internally to implement [PRAGMA mmap_size]. +** ** */ #define SQLITE_FCNTL_LOCKSTATE 1 #define SQLITE_GET_LOCKPROXYFILE 2 #define SQLITE_SET_LOCKPROXYFILE 3 @@ -898,10 +918,11 @@ #define SQLITE_FCNTL_VFSNAME 12 #define SQLITE_FCNTL_POWERSAFE_OVERWRITE 13 #define SQLITE_FCNTL_PRAGMA 14 #define SQLITE_FCNTL_BUSYHANDLER 15 #define SQLITE_FCNTL_TEMPFILENAME 16 +#define SQLITE_FCNTL_MMAP_SIZE 18 /* ** CAPI3REF: Mutex Handle ** ** The mutex module within SQLite defines [sqlite3_mutex] to be an @@ -1564,11 +1585,13 @@ **
    ^(This option takes a single argument which is a pointer to an ** [sqlite3_pcache_methods2] object. SQLite copies of the current ** page cache implementation into that object.)^
    ** ** [[SQLITE_CONFIG_LOG]]
    SQLITE_CONFIG_LOG
    -**
    ^The SQLITE_CONFIG_LOG option takes two arguments: a pointer to a +**
    The SQLITE_CONFIG_LOG option is used to configure the SQLite +** global [error log]. +** (^The SQLITE_CONFIG_LOG option takes two arguments: a pointer to a ** function with a call signature of void(*)(void*,int,const char*), ** and a pointer to void. ^If the function pointer is not NULL, it is ** invoked by [sqlite3_log()] to process each logging event. ^If the ** function pointer is NULL, the [sqlite3_log()] interface becomes a no-op. ** ^The void pointer that is the second argument to SQLITE_CONFIG_LOG is @@ -1610,26 +1633,42 @@ ** ** [[SQLITE_CONFIG_PCACHE]] [[SQLITE_CONFIG_GETPCACHE]] **
    SQLITE_CONFIG_PCACHE and SQLITE_CONFIG_GETPCACHE **
    These options are obsolete and should not be used by new code. ** They are retained for backwards compatibility but are now no-ops. -** +**
    ** ** [[SQLITE_CONFIG_SQLLOG]] **
    SQLITE_CONFIG_SQLLOG **
    This option is only available if sqlite is compiled with the -** SQLITE_ENABLE_SQLLOG pre-processor macro defined. The first argument should +** [SQLITE_ENABLE_SQLLOG] pre-processor macro defined. The first argument should ** be a pointer to a function of type void(*)(void*,sqlite3*,const char*, int). ** The second should be of type (void*). The callback is invoked by the library ** in three separate circumstances, identified by the value passed as the ** fourth parameter. If the fourth parameter is 0, then the database connection ** passed as the second argument has just been opened. The third argument ** points to a buffer containing the name of the main database file. If the ** fourth parameter is 1, then the SQL statement that the third parameter ** points to has just been executed. Or, if the fourth parameter is 2, then ** the connection being passed as the second parameter is being closed. The -** third parameter is passed NULL In this case. +** third parameter is passed NULL In this case. An example of using this +** configuration option can be seen in the "test_sqllog.c" source file in +** the canonical SQLite source tree.
    +** +** [[SQLITE_CONFIG_MMAP_SIZE]] +**
    SQLITE_CONFIG_MMAP_SIZE +**
    SQLITE_CONFIG_MMAP_SIZE takes two 64-bit integer (sqlite3_int64) values +** that are the default mmap size limit (the default setting for +** [PRAGMA mmap_size]) and the maximum allowed mmap size limit. +** The default setting can be overridden by each database connection using +** either the [PRAGMA mmap_size] command, or by using the +** [SQLITE_FCNTL_MMAP_SIZE] file control. The maximum allowed mmap size +** cannot be changed at run-time. Nor may the maximum allowed mmap size +** exceed the compile-time maximum mmap size set by the +** [SQLITE_MAX_MMAP_SIZE] compile-time option. +** If either argument to this option is negative, then that argument is +** changed to its compile-time default. ** */ #define SQLITE_CONFIG_SINGLETHREAD 1 /* nil */ #define SQLITE_CONFIG_MULTITHREAD 2 /* nil */ #define SQLITE_CONFIG_SERIALIZED 3 /* nil */ @@ -1649,10 +1688,11 @@ #define SQLITE_CONFIG_URI 17 /* int */ #define SQLITE_CONFIG_PCACHE2 18 /* sqlite3_pcache_methods2* */ #define SQLITE_CONFIG_GETPCACHE2 19 /* sqlite3_pcache_methods2* */ #define SQLITE_CONFIG_COVERING_INDEX_SCAN 20 /* int */ #define SQLITE_CONFIG_SQLLOG 21 /* xSqllog, void* */ +#define SQLITE_CONFIG_MMAP_SIZE 22 /* sqlite3_int64, sqlite3_int64 */ /* ** CAPI3REF: Database Connection Configuration Options ** ** These constants are the available integer configuration options that @@ -2482,10 +2522,13 @@ ** SQL statement text as the statement first begins executing. ** ^(Additional sqlite3_trace() callbacks might occur ** as each triggered subprogram is entered. The callbacks for triggers ** contain a UTF-8 SQL comment that identifies the trigger.)^ ** +** The [SQLITE_TRACE_SIZE_LIMIT] compile-time option can be used to limit +** the length of [bound parameter] expansion in the output of sqlite3_trace(). +** ** ^The callback function registered by sqlite3_profile() is invoked ** as each SQL statement finishes. ^The profile callback contains ** the original statement text and an estimate of wall-clock time ** of how long that statement took to run. ^The profile callback ** time is in units of nanoseconds, however the current implementation @@ -2673,11 +2716,11 @@ ** "private". ^Setting it to "shared" is equivalent to setting the ** SQLITE_OPEN_SHAREDCACHE bit in the flags argument passed to ** sqlite3_open_v2(). ^Setting the cache parameter to "private" is ** equivalent to setting the SQLITE_OPEN_PRIVATECACHE bit. ** ^If sqlite3_open_v2() is used and the "cache" parameter is present in -** a URI filename, its value overrides any behaviour requested by setting +** a URI filename, its value overrides any behavior requested by setting ** SQLITE_OPEN_PRIVATECACHE or SQLITE_OPEN_SHAREDCACHE flag. ** ** ** ^Specifying an unknown parameter in the query component of a URI is not an ** error. Future versions of SQLite might understand additional query @@ -3020,11 +3063,12 @@ ** **
      **
    1. ** ^If the database schema changes, instead of returning [SQLITE_SCHEMA] as it ** always used to do, [sqlite3_step()] will automatically recompile the SQL -** statement and try to run it again. +** statement and try to run it again. As many as [SQLITE_MAX_SCHEMA_RETRY] +** retries will occur before sqlite3_step() gives up and returns an error. **
      2. ** **
    2. ** ^When an error occurs, [sqlite3_step()] will return one of the detailed ** [error codes] or [extended error codes]. ^The legacy behavior was that @@ -3224,10 +3268,13 @@ ** for "?NNN" parameters is the value of NNN. ** ^The NNN value must be between 1 and the [sqlite3_limit()] ** parameter [SQLITE_LIMIT_VARIABLE_NUMBER] (default value: 999). ** ** ^The third argument is the value to bind to the parameter. +** ^If the third parameter to sqlite3_bind_text() or sqlite3_bind_text16() +** or sqlite3_bind_blob() is a NULL pointer then the fourth parameter +** is ignored and the end result is the same as sqlite3_bind_null(). ** ** ^(In those routines that have a fourth argument, its value is the ** number of bytes in the parameter. To be clear: the value is the ** number of bytes in the value, not the number of characters.)^ ** ^If the fourth parameter to sqlite3_bind_text() or sqlite3_bind_text16() @@ -3991,11 +4038,12 @@ SQLITE_DEPRECATED int sqlite3_aggregate_count(sqlite3_context*); SQLITE_DEPRECATED int sqlite3_expired(sqlite3_stmt*); SQLITE_DEPRECATED int sqlite3_transfer_bindings(sqlite3_stmt*, sqlite3_stmt*); SQLITE_DEPRECATED int sqlite3_global_recover(void); SQLITE_DEPRECATED void sqlite3_thread_cleanup(void); -SQLITE_DEPRECATED int sqlite3_memory_alarm(void(*)(void*,sqlite3_int64,int),void*,sqlite3_int64); +SQLITE_DEPRECATED int sqlite3_memory_alarm(void(*)(void*,sqlite3_int64,int), + void*,sqlite3_int64); #endif /* ** CAPI3REF: Obtaining SQL Function Parameter Values ** @@ -4071,18 +4119,21 @@ ** an aggregate query, the xStep() callback of the aggregate function ** implementation is never called and xFinal() is called exactly once. ** In those cases, sqlite3_aggregate_context() might be called for the ** first time from within xFinal().)^ ** -** ^The sqlite3_aggregate_context(C,N) routine returns a NULL pointer if N is -** less than or equal to zero or if a memory allocate error occurs. +** ^The sqlite3_aggregate_context(C,N) routine returns a NULL pointer +** when first called if N is less than or equal to zero or if a memory +** allocate error occurs. ** ** ^(The amount of space allocated by sqlite3_aggregate_context(C,N) is ** determined by the N parameter on first successful call. Changing the ** value of N in subsequent call to sqlite3_aggregate_context() within ** the same aggregate function instance will not resize the memory -** allocation.)^ +** allocation.)^ Within the xFinal callback, it is customary to set +** N=0 in calls to sqlite3_aggregate_context(C,N) so that no +** pointless memory allocations occur. ** ** ^SQLite automatically frees the memory allocated by ** sqlite3_aggregate_context() when the aggregate query concludes. ** ** The first parameter must be a copy of the @@ -4176,11 +4227,11 @@ ** SQLITE_TRANSIENT value means that the content will likely change in ** the near future and that SQLite should make its own private copy of ** the content before returning. ** ** The typedef is necessary to work around problems in certain -** C++ compilers. See ticket #2191. +** C++ compilers. */ typedef void (*sqlite3_destructor_type)(void*); #define SQLITE_STATIC ((sqlite3_destructor_type)0) #define SQLITE_TRANSIENT ((sqlite3_destructor_type)-1) @@ -4975,15 +5026,24 @@ ** CAPI3REF: Load An Extension ** ** ^This interface loads an SQLite extension library from the named file. ** ** ^The sqlite3_load_extension() interface attempts to load an -** SQLite extension library contained in the file zFile. +** [SQLite extension] library contained in the file zFile. If +** the file cannot be loaded directly, attempts are made to load +** with various operating-system specific extensions added. +** So for example, if "samplelib" cannot be loaded, then names like +** "samplelib.so" or "samplelib.dylib" or "samplelib.dll" might +** be tried also. ** ** ^The entry point is zProc. -** ^zProc may be 0, in which case the name of the entry point -** defaults to "sqlite3_extension_init". +** ^(zProc may be 0, in which case SQLite will try to come up with an +** entry point name on its own. It first tries "sqlite3_extension_init". +** If that does not work, it constructs a name "sqlite3_X_init" where the +** X is consists of the lower-case equivalent of all ASCII alphabetic +** characters in the filename from the last "/" to the first following +** "." and omitting any initial "lib".)^ ** ^The sqlite3_load_extension() interface returns ** [SQLITE_OK] on success and [SQLITE_ERROR] if something goes wrong. ** ^If an error occurs and pzErrMsg is not 0, then the ** [sqlite3_load_extension()] interface shall attempt to ** fill *pzErrMsg with error message text stored in memory @@ -5005,15 +5065,15 @@ /* ** CAPI3REF: Enable Or Disable Extension Loading ** ** ^So as not to open security holes in older applications that are -** unprepared to deal with extension loading, and as a means of disabling -** extension loading while evaluating user-entered SQL, the following API +** unprepared to deal with [extension loading], and as a means of disabling +** [extension loading] while evaluating user-entered SQL, the following API ** is provided to turn the [sqlite3_load_extension()] mechanism on and off. ** -** ^Extension loading is off by default. See ticket #1863. +** ^Extension loading is off by default. ** ^Call the sqlite3_enable_load_extension() routine with onoff==1 ** to turn extension loading on and call it with onoff==0 to turn ** it back off again. */ int sqlite3_enable_load_extension(sqlite3 *db, int onoff); @@ -5021,11 +5081,11 @@ /* ** CAPI3REF: Automatically Load Statically Linked Extensions ** ** ^This interface causes the xEntryPoint() function to be invoked for ** each new [database connection] that is created. The idea here is that -** xEntryPoint() is the entry point for a statically linked SQLite extension +** xEntryPoint() is the entry point for a statically linked [SQLite extension] ** that is to be automatically loaded into all new database connections. ** ** ^(Even though the function prototype shows that xEntryPoint() takes ** no arguments and returns void, SQLite invokes xEntryPoint() with three ** arguments and expects and integer result as if the signature of the @@ -6372,11 +6432,11 @@ ** intact. If the requested page is not already in the cache, then the ** cache implementation should use the value of the createFlag ** parameter to help it determined what action to take: ** ** -**
      createFlag Behaviour when page is not already in cache +**
      createFlag Behavior when page is not already in cache **
      0 Do not allocate a new page. Return NULL. **
      1 Allocate a new page if it easy and convenient to do so. ** Otherwise return NULL. **
      2 Make every effort to allocate a new page. Only return ** NULL if allocating a new page is effectively impossible. @@ -6801,14 +6861,29 @@ ** independence" that SQLite uses internally when comparing identifiers. */ int sqlite3_stricmp(const char *, const char *); int sqlite3_strnicmp(const char *, const char *, int); +/* +** CAPI3REF: String Globbing +* +** ^The [sqlite3_strglob(P,X)] interface returns zero if string X matches +** the glob pattern P, and it returns non-zero if string X does not match +** the glob pattern P. ^The definition of glob pattern matching used in +** [sqlite3_strglob(P,X)] is the same as for the "X GLOB P" operator in the +** SQL dialect used by SQLite. ^The sqlite3_strglob(P,X) function is case +** sensitive. +** +** Note that this routine returns zero on a match and non-zero if the strings +** do not match, the same as [sqlite3_stricmp()] and [sqlite3_strnicmp()]. +*/ +int sqlite3_strglob(const char *zGlob, const char *zStr); + /* ** CAPI3REF: Error Logging Interface ** -** ^The [sqlite3_log()] interface writes a message into the error log +** ^The [sqlite3_log()] interface writes a message into the [error log] ** established by the [SQLITE_CONFIG_LOG] option to [sqlite3_config()]. ** ^If logging is enabled, the zFormat string and subsequent arguments are ** used with [sqlite3_snprintf()] to generate the final output string. ** ** The sqlite3_log() interface is intended for use by extensions such as Index: src/sqlite3ext.h ================================================================== --- src/sqlite3ext.h +++ src/sqlite3ext.h @@ -467,9 +467,18 @@ #define sqlite3_uri_parameter sqlite3_api->uri_parameter #define sqlite3_uri_vsnprintf sqlite3_api->vsnprintf #define sqlite3_wal_checkpoint_v2 sqlite3_api->wal_checkpoint_v2 #endif /* SQLITE_CORE */ -#define SQLITE_EXTENSION_INIT1 const sqlite3_api_routines *sqlite3_api = 0; -#define SQLITE_EXTENSION_INIT2(v) sqlite3_api = v; +#ifndef SQLITE_CORE + /* This case when the file really is being compiled as a loadable + ** extension */ +# define SQLITE_EXTENSION_INIT1 const sqlite3_api_routines *sqlite3_api=0; +# define SQLITE_EXTENSION_INIT2(v) sqlite3_api=v; +#else + /* This case when the file is being statically linked into the + ** application */ +# define SQLITE_EXTENSION_INIT1 /*no-op*/ +# define SQLITE_EXTENSION_INIT2(v) (void)v; /* unused parameter */ +#endif #endif /* _SQLITE3EXT_H_ */ Index: src/sqliteInt.h ================================================================== --- src/sqliteInt.h +++ src/sqliteInt.h @@ -120,15 +120,15 @@ ** ** Older versions of SQLite used an optional THREADSAFE macro. ** We support that for legacy. */ #if !defined(SQLITE_THREADSAFE) -#if defined(THREADSAFE) -# define SQLITE_THREADSAFE THREADSAFE -#else -# define SQLITE_THREADSAFE 1 /* IMP: R-07272-22309 */ -#endif +# if defined(THREADSAFE) +# define SQLITE_THREADSAFE THREADSAFE +# else +# define SQLITE_THREADSAFE 1 /* IMP: R-07272-22309 */ +# endif #endif /* ** Powersafe overwrite is on by default. But can be turned off using ** the -DSQLITE_POWERSAFE_OVERWRITE=0 command-line option. @@ -392,10 +392,11 @@ ** Provide a default value for SQLITE_TEMP_STORE in case it is not specified ** on the command-line */ #ifndef SQLITE_TEMP_STORE # define SQLITE_TEMP_STORE 1 +# define SQLITE_TEMP_STORE_xc 1 /* Exclude from ctime.c */ #endif /* ** GCC does not define the offsetof() macro so we'll have to do it ** ourselves. @@ -539,10 +540,53 @@ # define EIGHT_BYTE_ALIGNMENT(X) ((((char*)(X) - (char*)0)&3)==0) #else # define EIGHT_BYTE_ALIGNMENT(X) ((((char*)(X) - (char*)0)&7)==0) #endif +/* +** Disable MMAP on platforms where it is known to not work +*/ +#if defined(__OpenBSD__) || defined(__QNXNTO__) +# undef SQLITE_MAX_MMAP_SIZE +# define SQLITE_MAX_MMAP_SIZE 0 +#endif + +/* +** Default maximum size of memory used by memory-mapped I/O in the VFS +*/ +#ifdef __APPLE__ +# include +# if TARGET_OS_IPHONE +# undef SQLITE_MAX_MMAP_SIZE +# define SQLITE_MAX_MMAP_SIZE 0 +# endif +#endif +#ifndef SQLITE_MAX_MMAP_SIZE +# if defined(__linux__) \ + || defined(_WIN32) \ + || (defined(__APPLE__) && defined(__MACH__)) \ + || defined(__sun) +# define SQLITE_MAX_MMAP_SIZE 0x7fff0000 /* 2147418112 */ +# else +# define SQLITE_MAX_MMAP_SIZE 0 +# endif +# define SQLITE_MAX_MMAP_SIZE_xc 1 /* exclude from ctime.c */ +#endif + +/* +** The default MMAP_SIZE is zero on all platforms. Or, even if a larger +** default MMAP_SIZE is specified at compile-time, make sure that it does +** not exceed the maximum mmap size. +*/ +#ifndef SQLITE_DEFAULT_MMAP_SIZE +# define SQLITE_DEFAULT_MMAP_SIZE 0 +# define SQLITE_DEFAULT_MMAP_SIZE_xc 1 /* Exclude from ctime.c */ +#endif +#if SQLITE_DEFAULT_MMAP_SIZE>SQLITE_MAX_MMAP_SIZE +# undef SQLITE_DEFAULT_MMAP_SIZE +# define SQLITE_DEFAULT_MMAP_SIZE SQLITE_MAX_MMAP_SIZE +#endif /* ** An instance of the following structure is used to store the busy-handler ** callback for a given sqlite handle. ** @@ -834,10 +878,11 @@ sqlite3_mutex *mutex; /* Connection mutex */ Db *aDb; /* All backends */ int nDb; /* Number of backends currently in use */ int flags; /* Miscellaneous flags. See below */ i64 lastRowid; /* ROWID of most recent insert (see above) */ + i64 szMmap; /* Default mmap_size setting */ unsigned int openFlags; /* Flags passed to sqlite3_vfs.xOpen() */ int errCode; /* Most recent error code (SQLITE_*) */ int errMask; /* & result codes with this before returning */ u16 dbOptFlags; /* Flags to enable/disable optimizations */ u8 autoCommit; /* The auto-commit flag. */ @@ -2070,10 +2115,12 @@ */ #define NC_AllowAgg 0x01 /* Aggregate functions are allowed here */ #define NC_HasAgg 0x02 /* One or more aggregate functions seen */ #define NC_IsCheck 0x04 /* True if resolving names in a CHECK constraint */ #define NC_InAggFunc 0x08 /* True if analyzing arguments to an agg func */ +#define NC_AsMaybe 0x10 /* Resolve to AS terms of the result set only + ** if no other resolution is available */ /* ** An instance of the following structure contains all information ** needed to generate code for a single SELECT statement. ** @@ -2505,10 +2552,12 @@ sqlite3_mutex_methods mutex; /* Low-level mutex interface */ sqlite3_pcache_methods2 pcache2; /* Low-level page-cache interface */ void *pHeap; /* Heap storage space */ int nHeap; /* Size of pHeap[] */ int mnReq, mxReq; /* Min and max heap requests sizes */ + sqlite3_int64 szMmap; /* mmap() space per open file */ + sqlite3_int64 mxMmap; /* Maximum value for szMmap */ void *pScratch; /* Scratch memory */ int szScratch; /* Size of each scratch buffer */ int nScratch; /* Number of scratch buffers */ void *pPage; /* Page cache memory */ int szPage; /* Size of each page in pPage[] */ @@ -2539,10 +2588,11 @@ struct Walker { int (*xExprCallback)(Walker*, Expr*); /* Callback for expressions */ int (*xSelectCallback)(Walker*,Select*); /* Callback for SELECTs */ Parse *pParse; /* Parser context. */ int walkerDepth; /* Number of subqueries */ + u8 bSelectDepthFirst; /* Do subqueries first */ union { /* Extra data for callback */ NameContext *pNC; /* Naming context */ int i; /* Integer value */ SrcList *pSrcList; /* FROM clause */ struct SrcCount *pSrcCount; /* Counting column references */ @@ -3042,10 +3092,16 @@ int sqlite3Atoi64(const char*, i64*, int, u8); void sqlite3Error(sqlite3*, int, const char*,...); void *sqlite3HexToBlob(sqlite3*, const char *z, int n); u8 sqlite3HexToInt(int h); int sqlite3TwoPartName(Parse *, Token *, Token *, Token **); + +#if defined(SQLITE_DEBUG) || defined(SQLITE_TEST) || \ + defined(SQLITE_DEBUG_OS_TRACE) +const char *sqlite3ErrName(int); +#endif + const char *sqlite3ErrStr(int); int sqlite3ReadSchema(Parse *pParse); CollSeq *sqlite3FindCollSeq(sqlite3*,u8 enc, const char*,int); CollSeq *sqlite3LocateCollSeq(Parse *pParse, const char*zName); CollSeq *sqlite3ExprCollSeq(Parse *pParse, Expr *pExpr); Index: src/tclsqlite.c ================================================================== --- src/tclsqlite.c +++ src/tclsqlite.c @@ -1003,11 +1003,11 @@ pDb->disableAuth++; if( sqlite3_exec(pDb->db, zEnd, 0, 0, 0) ){ /* This is a tricky scenario to handle. The most likely cause of an ** error is that the exec() above was an attempt to commit the ** top-level transaction that returned SQLITE_BUSY. Or, less likely, - ** that an IO-error has occured. In either case, throw a Tcl exception + ** that an IO-error has occurred. In either case, throw a Tcl exception ** and try to rollback the transaction. ** ** But it could also be that the user executed one or more BEGIN, ** COMMIT, SAVEPOINT, RELEASE or ROLLBACK commands that are confusing ** this method's logic. Not clear how this would be best handled. @@ -3681,13 +3681,10 @@ extern int Sqlitetestrtree_Init(Tcl_Interp*); extern int Sqlitequota_Init(Tcl_Interp*); extern int Sqlitemultiplex_Init(Tcl_Interp*); extern int SqliteSuperlock_Init(Tcl_Interp*); extern int SqlitetestSyscall_Init(Tcl_Interp*); - extern int Sqlitetestfuzzer_Init(Tcl_Interp*); - extern int Sqlitetestwholenumber_Init(Tcl_Interp*); - extern int Sqlitetestregexp_Init(Tcl_Interp*); #if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) extern int Sqlitetestfts3_Init(Tcl_Interp *interp); #endif @@ -3726,13 +3723,10 @@ Sqlitetestrtree_Init(interp); Sqlitequota_Init(interp); Sqlitemultiplex_Init(interp); SqliteSuperlock_Init(interp); SqlitetestSyscall_Init(interp); - Sqlitetestfuzzer_Init(interp); - Sqlitetestwholenumber_Init(interp); - Sqlitetestregexp_Init(interp); #if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) Sqlitetestfts3_Init(interp); #endif Index: src/test1.c ================================================================== --- src/test1.c +++ src/test1.c @@ -111,81 +111,12 @@ *ppDb = (sqlite3*)sqlite3TestTextToPtr(zA); } return TCL_OK; } - -const char *sqlite3TestErrorName(int rc){ - const char *zName = 0; - switch( rc ){ - case SQLITE_OK: zName = "SQLITE_OK"; break; - case SQLITE_ERROR: zName = "SQLITE_ERROR"; break; - case SQLITE_INTERNAL: zName = "SQLITE_INTERNAL"; break; - case SQLITE_PERM: zName = "SQLITE_PERM"; break; - case SQLITE_ABORT: zName = "SQLITE_ABORT"; break; - case SQLITE_BUSY: zName = "SQLITE_BUSY"; break; - case SQLITE_LOCKED: zName = "SQLITE_LOCKED"; break; - case SQLITE_LOCKED_SHAREDCACHE: zName = "SQLITE_LOCKED_SHAREDCACHE";break; - case SQLITE_NOMEM: zName = "SQLITE_NOMEM"; break; - case SQLITE_READONLY: zName = "SQLITE_READONLY"; break; - case SQLITE_INTERRUPT: zName = "SQLITE_INTERRUPT"; break; - case SQLITE_IOERR: zName = "SQLITE_IOERR"; break; - case SQLITE_CORRUPT: zName = "SQLITE_CORRUPT"; break; - case SQLITE_NOTFOUND: zName = "SQLITE_NOTFOUND"; break; - case SQLITE_FULL: zName = "SQLITE_FULL"; break; - case SQLITE_CANTOPEN: zName = "SQLITE_CANTOPEN"; break; - case SQLITE_PROTOCOL: zName = "SQLITE_PROTOCOL"; break; - case SQLITE_EMPTY: zName = "SQLITE_EMPTY"; break; - case SQLITE_SCHEMA: zName = "SQLITE_SCHEMA"; break; - case SQLITE_TOOBIG: zName = "SQLITE_TOOBIG"; break; - case SQLITE_CONSTRAINT: zName = "SQLITE_CONSTRAINT"; break; - case SQLITE_CONSTRAINT_UNIQUE: zName = "SQLITE_CONSTRAINT_UNIQUE"; break; - case SQLITE_CONSTRAINT_TRIGGER: zName = "SQLITE_CONSTRAINT_TRIGGER";break; - case SQLITE_CONSTRAINT_FOREIGNKEY: - zName = "SQLITE_CONSTRAINT_FOREIGNKEY"; break; - case SQLITE_CONSTRAINT_CHECK: zName = "SQLITE_CONSTRAINT_CHECK"; break; - case SQLITE_CONSTRAINT_PRIMARYKEY: - zName = "SQLITE_CONSTRAINT_PRIMARYKEY"; break; - case SQLITE_CONSTRAINT_NOTNULL: zName = "SQLITE_CONSTRAINT_NOTNULL";break; - case SQLITE_CONSTRAINT_COMMITHOOK: - zName = "SQLITE_CONSTRAINT_COMMITHOOK"; break; - case SQLITE_CONSTRAINT_VTAB: zName = "SQLITE_CONSTRAINT_VTAB"; break; - case SQLITE_CONSTRAINT_FUNCTION: zName = "SQLITE_CONSTRAINT_FUNCTION";break; - case SQLITE_MISMATCH: zName = "SQLITE_MISMATCH"; break; - case SQLITE_MISUSE: zName = "SQLITE_MISUSE"; break; - case SQLITE_NOLFS: zName = "SQLITE_NOLFS"; break; - case SQLITE_AUTH: zName = "SQLITE_AUTH"; break; - case SQLITE_FORMAT: zName = "SQLITE_FORMAT"; break; - case SQLITE_RANGE: zName = "SQLITE_RANGE"; break; - case SQLITE_NOTADB: zName = "SQLITE_NOTADB"; break; - case SQLITE_ROW: zName = "SQLITE_ROW"; break; - case SQLITE_DONE: zName = "SQLITE_DONE"; break; - case SQLITE_IOERR_READ: zName = "SQLITE_IOERR_READ"; break; - case SQLITE_IOERR_SHORT_READ: zName = "SQLITE_IOERR_SHORT_READ"; break; - case SQLITE_IOERR_WRITE: zName = "SQLITE_IOERR_WRITE"; break; - case SQLITE_IOERR_FSYNC: zName = "SQLITE_IOERR_FSYNC"; break; - case SQLITE_IOERR_DIR_FSYNC: zName = "SQLITE_IOERR_DIR_FSYNC"; break; - case SQLITE_IOERR_TRUNCATE: zName = "SQLITE_IOERR_TRUNCATE"; break; - case SQLITE_IOERR_FSTAT: zName = "SQLITE_IOERR_FSTAT"; break; - case SQLITE_IOERR_UNLOCK: zName = "SQLITE_IOERR_UNLOCK"; break; - case SQLITE_IOERR_RDLOCK: zName = "SQLITE_IOERR_RDLOCK"; break; - case SQLITE_IOERR_DELETE: zName = "SQLITE_IOERR_DELETE"; break; - case SQLITE_IOERR_BLOCKED: zName = "SQLITE_IOERR_BLOCKED"; break; - case SQLITE_IOERR_NOMEM: zName = "SQLITE_IOERR_NOMEM"; break; - case SQLITE_IOERR_ACCESS: zName = "SQLITE_IOERR_ACCESS"; break; - case SQLITE_IOERR_CHECKRESERVEDLOCK: - zName = "SQLITE_IOERR_CHECKRESERVEDLOCK"; break; - case SQLITE_IOERR_LOCK: zName = "SQLITE_IOERR_LOCK"; break; - case SQLITE_CORRUPT_VTAB: zName = "SQLITE_CORRUPT_VTAB"; break; - case SQLITE_READONLY_RECOVERY: zName = "SQLITE_READONLY_RECOVERY"; break; - case SQLITE_READONLY_CANTLOCK: zName = "SQLITE_READONLY_CANTLOCK"; break; - case SQLITE_READONLY_ROLLBACK: zName = "SQLITE_READONLY_ROLLBACK"; break; - default: zName = "SQLITE_Unknown"; break; - } - return zName; -} -#define t1ErrorName sqlite3TestErrorName +extern const char *sqlite3ErrName(int); +#define t1ErrorName sqlite3ErrName /* ** Convert an sqlite3_stmt* into an sqlite3*. This depends on the ** fact that the sqlite3* is the first field in the Vdbe structure. */ @@ -1726,11 +1657,11 @@ } rc = sqlite3_blob_read(pBlob, zBuf, nByte, iOffset); if( rc==SQLITE_OK ){ Tcl_SetObjResult(interp, Tcl_NewByteArrayObj(zBuf, nByte)); }else{ - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); } Tcl_Free((char *)zBuf); return (rc==SQLITE_OK ? TCL_OK : TCL_ERROR); } @@ -1776,11 +1707,11 @@ if( objc==5 && Tcl_GetIntFromObj(interp, objv[4], &nBuf) ){ return TCL_ERROR; } rc = sqlite3_blob_write(pBlob, zBuf, nBuf, iOffset); if( rc!=SQLITE_OK ){ - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); } return (rc==SQLITE_OK ? TCL_OK : TCL_ERROR); } @@ -1802,11 +1733,11 @@ if( blobHandleFromObj(interp, objv[1], &pBlob) ) return TCL_ERROR; if( Tcl_GetWideIntFromObj(interp, objv[2], &iRowid) ) return TCL_ERROR; rc = sqlite3_blob_reopen(pBlob, iRowid); if( rc!=SQLITE_OK ){ - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); } return (rc==SQLITE_OK ? TCL_OK : TCL_ERROR); } @@ -2012,11 +1943,11 @@ (p->pFinal ? cf2Final : 0), cf2Destroy ); if( rc!=SQLITE_OK ){ Tcl_ResetResult(interp); - Tcl_AppendResult(interp, sqlite3TestErrorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -2688,11 +2619,11 @@ sqlite3_mutex_leave(db->mutex); } if( sqlite3TestErrCode(interp, db, rc) ) return TCL_ERROR; if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, sqlite3TestErrorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; bad_args: @@ -3246,11 +3177,11 @@ if( Tcl_GetIntFromObj(interp, objv[4], &bytes) ) return TCL_ERROR; rc = sqlite3_bind_text(pStmt, idx, value, bytes, SQLITE_TRANSIENT); if( sqlite3TestErrCode(interp, StmtToDb(pStmt), rc) ) return TCL_ERROR; if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, sqlite3TestErrorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -3294,11 +3225,11 @@ if( Tcl_GetIntFromObj(interp, oBytes, &bytes) ) return TCL_ERROR; rc = sqlite3_bind_text16(pStmt, idx, (void *)value, bytes, xDel); if( sqlite3TestErrCode(interp, StmtToDb(pStmt), rc) ) return TCL_ERROR; if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, sqlite3TestErrorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } #endif /* SQLITE_OMIT_UTF16 */ return TCL_OK; @@ -4568,11 +4499,11 @@ return TCL_ERROR; } if( getDbPointer(interp, argv[1], &db) ) return TCL_ERROR; if( Tcl_GetInt(interp, argv[2], &ms) ) return TCL_ERROR; rc = sqlite3_busy_timeout(db, ms); - Tcl_AppendResult(interp, sqlite3TestErrorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_OK; } /* ** Usage: tcl_variable_type VARIABLENAME @@ -5089,11 +5020,11 @@ zDb = Tcl_GetString(objv[2]); if( zDb[0]=='\0' ) zDb = NULL; rc = sqlite3_file_control(db, zDb, SQLITE_FCNTL_CHUNK_SIZE, (void *)&nSize); if( rc ){ - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_ERROR; } return TCL_OK; } @@ -5126,11 +5057,11 @@ zDb = Tcl_GetString(objv[2]); if( zDb[0]=='\0' ) zDb = NULL; rc = sqlite3_file_control(db, zDb, SQLITE_FCNTL_SIZE_HINT, (void *)&nSize); if( rc ){ - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_ERROR; } return TCL_OK; } @@ -5670,11 +5601,11 @@ } logcallback = {0, 0}; static void xLogcallback(void *unused, int err, char *zMsg){ Tcl_Obj *pNew = Tcl_DuplicateObj(logcallback.pObj); Tcl_IncrRefCount(pNew); Tcl_ListObjAppendElement( - 0, pNew, Tcl_NewStringObj(sqlite3TestErrorName(err), -1) + 0, pNew, Tcl_NewStringObj(sqlite3ErrName(err), -1) ); Tcl_ListObjAppendElement(0, pNew, Tcl_NewStringObj(zMsg, -1)); Tcl_EvalObjEx(logcallback.pInterp, pNew, TCL_EVAL_GLOBAL|TCL_EVAL_DIRECT); Tcl_DecrRefCount(pNew); } @@ -5841,10 +5772,35 @@ } Tcl_ResetResult(interp); return TCL_OK; } + +#if SQLITE_OS_UNIX +#include +#include + +static int test_getrusage( + void * clientData, + Tcl_Interp *interp, + int objc, + Tcl_Obj *CONST objv[] +){ + char buf[1024]; + struct rusage r; + memset(&r, 0, sizeof(r)); + getrusage(RUSAGE_SELF, &r); + + sprintf(buf, "ru_utime=%d.%06d ru_stime=%d.%06d ru_minflt=%d ru_majflt=%d", + (int)r.ru_utime.tv_sec, (int)r.ru_utime.tv_usec, + (int)r.ru_stime.tv_sec, (int)r.ru_stime.tv_usec, + (int)r.ru_minflt, (int)r.ru_majflt + ); + Tcl_SetObjResult(interp, Tcl_NewStringObj(buf, -1)); + return TCL_OK; +} +#endif #if SQLITE_OS_WIN /* ** Information passed from the main thread into the windows file locker ** background thread. @@ -6012,10 +5968,73 @@ return TCL_ERROR; } sqlite3_test_control(SQLITE_TESTCTRL_OPTIMIZATIONS, db, mask); return TCL_OK; } + +typedef struct sqlite3_api_routines sqlite3_api_routines; +/* +** load_static_extension DB NAME ... +** +** Load one or more statically linked extensions. +*/ +static int tclLoadStaticExtensionCmd( + void * clientData, + Tcl_Interp *interp, + int objc, + Tcl_Obj *CONST objv[] +){ + extern int sqlite3_amatch_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_closure_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_fuzzer_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_ieee_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_nextchar_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_regexp_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_spellfix_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_wholenumber_init(sqlite3*,char**,const sqlite3_api_routines*); + static const struct { + const char *zExtName; + int (*pInit)(sqlite3*,char**,const sqlite3_api_routines*); + } aExtension[] = { + { "amatch", sqlite3_amatch_init }, + { "closure", sqlite3_closure_init }, + { "fuzzer", sqlite3_fuzzer_init }, + { "ieee754", sqlite3_ieee_init }, + { "nextchar", sqlite3_nextchar_init }, + { "regexp", sqlite3_regexp_init }, + { "spellfix", sqlite3_spellfix_init }, + { "wholenumber", sqlite3_wholenumber_init }, + }; + sqlite3 *db; + const char *zName; + int i, j, rc; + char *zErrMsg = 0; + if( objc<3 ){ + Tcl_WrongNumArgs(interp, 1, objv, "DB NAME ..."); + return TCL_ERROR; + } + if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; + for(j=2; j=ArraySize(aExtension) ){ + Tcl_AppendResult(interp, "no such extension: ", zName, (char*)0); + return TCL_ERROR; + } + rc = aExtension[i].pInit(db, &zErrMsg, 0); + if( rc!=SQLITE_OK || zErrMsg ){ + Tcl_AppendResult(interp, "initialization of ", zName, " failed: ", zErrMsg, + (char*)0); + sqlite3_free(zErrMsg); + return TCL_ERROR; + } + } + return TCL_OK; +} + /* ** Register commands with the TCL interpreter. */ int Sqlitetest1_Init(Tcl_Interp *interp){ @@ -6231,10 +6250,14 @@ { "test_sqlite3_log", test_sqlite3_log, 0 }, #ifndef SQLITE_OMIT_EXPLAIN { "print_explain_query_plan", test_print_eqp, 0 }, #endif { "sqlite3_test_control", test_test_control }, +#if SQLITE_OS_UNIX + { "getrusage", test_getrusage }, +#endif + { "load_static_extension", tclLoadStaticExtensionCmd }, }; static int bitmask_size = sizeof(Bitmask)*8; int i; extern int sqlite3_sync_count, sqlite3_fullsync_count; extern int sqlite3_opentemp_count; Index: src/test2.c ================================================================== --- src/test2.c +++ src/test2.c @@ -17,39 +17,11 @@ #include "tcl.h" #include #include #include -/* -** Interpret an SQLite error number -*/ -static char *errorName(int rc){ - char *zName; - switch( rc ){ - case SQLITE_OK: zName = "SQLITE_OK"; break; - case SQLITE_ERROR: zName = "SQLITE_ERROR"; break; - case SQLITE_PERM: zName = "SQLITE_PERM"; break; - case SQLITE_ABORT: zName = "SQLITE_ABORT"; break; - case SQLITE_BUSY: zName = "SQLITE_BUSY"; break; - case SQLITE_NOMEM: zName = "SQLITE_NOMEM"; break; - case SQLITE_READONLY: zName = "SQLITE_READONLY"; break; - case SQLITE_INTERRUPT: zName = "SQLITE_INTERRUPT"; break; - case SQLITE_IOERR: zName = "SQLITE_IOERR"; break; - case SQLITE_CORRUPT: zName = "SQLITE_CORRUPT"; break; - case SQLITE_FULL: zName = "SQLITE_FULL"; break; - case SQLITE_CANTOPEN: zName = "SQLITE_CANTOPEN"; break; - case SQLITE_PROTOCOL: zName = "SQLITE_PROTOCOL"; break; - case SQLITE_EMPTY: zName = "SQLITE_EMPTY"; break; - case SQLITE_SCHEMA: zName = "SQLITE_SCHEMA"; break; - case SQLITE_CONSTRAINT: zName = "SQLITE_CONSTRAINT"; break; - case SQLITE_MISMATCH: zName = "SQLITE_MISMATCH"; break; - case SQLITE_MISUSE: zName = "SQLITE_MISUSE"; break; - case SQLITE_NOLFS: zName = "SQLITE_NOLFS"; break; - default: zName = "SQLITE_Unknown"; break; - } - return zName; -} +extern const char *sqlite3ErrName(int); /* ** Page size and reserved size used for testing. */ static int test_pagesize = 1024; @@ -85,11 +57,11 @@ if( Tcl_GetInt(interp, argv[2], &nPage) ) return TCL_ERROR; rc = sqlite3PagerOpen(sqlite3_vfs_find(0), &pPager, argv[1], 0, 0, SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE | SQLITE_OPEN_MAIN_DB, pager_test_reiniter); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } sqlite3PagerSetCachesize(pPager, nPage); pageSize = test_pagesize; sqlite3PagerSetPagesize(pPager, &pageSize, -1); @@ -117,11 +89,11 @@ return TCL_ERROR; } pPager = sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerClose(pPager); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -144,11 +116,11 @@ return TCL_ERROR; } pPager = sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerRollback(pPager); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -171,16 +143,16 @@ return TCL_ERROR; } pPager = sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerCommitPhaseOne(pPager, 0, 0); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } rc = sqlite3PagerCommitPhaseTwo(pPager); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -203,11 +175,11 @@ return TCL_ERROR; } pPager = sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerOpenSavepoint(pPager, 1); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -231,11 +203,11 @@ } pPager = sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerSavepoint(pPager, SAVEPOINT_ROLLBACK, 0); sqlite3PagerSavepoint(pPager, SAVEPOINT_RELEASE, 0); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -258,11 +230,11 @@ return TCL_ERROR; } pPager = sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerSavepoint(pPager, SAVEPOINT_RELEASE, 0); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -351,11 +323,11 @@ rc = sqlite3PagerSharedLock(pPager); if( rc==SQLITE_OK ){ rc = sqlite3PagerGet(pPager, pgno, &pPage); } if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } sqlite3_snprintf(sizeof(zBuf),zBuf,"%p",pPage); Tcl_AppendResult(interp, zBuf, 0); return TCL_OK; @@ -505,11 +477,11 @@ return TCL_ERROR; } pPage = (DbPage *)sqlite3TestTextToPtr(argv[1]); rc = sqlite3PagerWrite(pPage); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } pData = sqlite3PagerGetData(pPage); strncpy(pData, argv[2], test_pagesize-1); pData[test_pagesize-1] = 0; @@ -554,21 +526,21 @@ zFile[nFile+1] = 0; rc = sqlite3OsOpenMalloc(pVfs, zFile, &fd, (SQLITE_OPEN_CREATE|SQLITE_OPEN_READWRITE|SQLITE_OPEN_MAIN_DB), 0 ); if( rc ){ - Tcl_AppendResult(interp, "open failed: ", errorName(rc), 0); + Tcl_AppendResult(interp, "open failed: ", sqlite3ErrName(rc), 0); sqlite3_free(zFile); return TCL_ERROR; } offset = n; offset *= 1024*1024; rc = sqlite3OsWrite(fd, "Hello, World!", 14, offset); sqlite3OsCloseFree(fd); sqlite3_free(zFile); if( rc ){ - Tcl_AppendResult(interp, "write failed: ", errorName(rc), 0); + Tcl_AppendResult(interp, "write failed: ", sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } #endif Index: src/test3.c ================================================================== --- src/test3.c +++ src/test3.c @@ -17,35 +17,11 @@ #include "btreeInt.h" #include "tcl.h" #include #include -/* -** Interpret an SQLite error number -*/ -static char *errorName(int rc){ - char *zName; - switch( rc ){ - case SQLITE_OK: zName = "SQLITE_OK"; break; - case SQLITE_ERROR: zName = "SQLITE_ERROR"; break; - case SQLITE_PERM: zName = "SQLITE_PERM"; break; - case SQLITE_ABORT: zName = "SQLITE_ABORT"; break; - case SQLITE_BUSY: zName = "SQLITE_BUSY"; break; - case SQLITE_NOMEM: zName = "SQLITE_NOMEM"; break; - case SQLITE_READONLY: zName = "SQLITE_READONLY"; break; - case SQLITE_INTERRUPT: zName = "SQLITE_INTERRUPT"; break; - case SQLITE_IOERR: zName = "SQLITE_IOERR"; break; - case SQLITE_CORRUPT: zName = "SQLITE_CORRUPT"; break; - case SQLITE_FULL: zName = "SQLITE_FULL"; break; - case SQLITE_CANTOPEN: zName = "SQLITE_CANTOPEN"; break; - case SQLITE_PROTOCOL: zName = "SQLITE_PROTOCOL"; break; - case SQLITE_EMPTY: zName = "SQLITE_EMPTY"; break; - case SQLITE_LOCKED: zName = "SQLITE_LOCKED"; break; - default: zName = "SQLITE_Unknown"; break; - } - return zName; -} +extern const char *sqlite3ErrName(int); /* ** A bogus sqlite3 connection structure for use in the btree ** tests. */ @@ -87,11 +63,11 @@ zFilename[n+1] = 0; rc = sqlite3BtreeOpen(sDb.pVfs, zFilename, &sDb, &pBt, 0, SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE | SQLITE_OPEN_MAIN_DB); sqlite3_free(zFilename); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } sqlite3BtreeSetCacheSize(pBt, nCache); sqlite3_snprintf(sizeof(zBuf), zBuf,"%p", pBt); Tcl_AppendResult(interp, zBuf, 0); @@ -117,11 +93,11 @@ return TCL_ERROR; } pBt = sqlite3TestTextToPtr(argv[1]); rc = sqlite3BtreeClose(pBt); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } nRefSqlite3--; if( nRefSqlite3==0 ){ sqlite3_mutex_leave(sDb.mutex); @@ -154,11 +130,11 @@ pBt = sqlite3TestTextToPtr(argv[1]); sqlite3BtreeEnter(pBt); rc = sqlite3BtreeBeginTrans(pBt, 1); sqlite3BtreeLeave(pBt); if( rc!=SQLITE_OK ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return TCL_OK; } @@ -248,11 +224,11 @@ rc = sqlite3BtreeCursor(pBt, iTable, wrFlag, 0, pCur); } sqlite3BtreeLeave(pBt); if( rc ){ ckfree((char *)pCur); - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } sqlite3_snprintf(sizeof(zBuf), zBuf,"%p", pCur); Tcl_AppendResult(interp, zBuf, 0); return SQLITE_OK; @@ -283,11 +259,11 @@ sqlite3BtreeEnter(pBt); rc = sqlite3BtreeCloseCursor(pCur); sqlite3BtreeLeave(pBt); ckfree((char *)pCur); if( rc ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } return SQLITE_OK; } @@ -317,11 +293,11 @@ pCur = sqlite3TestTextToPtr(argv[1]); sqlite3BtreeEnter(pCur->pBtree); rc = sqlite3BtreeNext(pCur, &res); sqlite3BtreeLeave(pCur->pBtree); if( rc ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } sqlite3_snprintf(sizeof(zBuf),zBuf,"%d",res); Tcl_AppendResult(interp, zBuf, 0); return SQLITE_OK; @@ -352,11 +328,11 @@ pCur = sqlite3TestTextToPtr(argv[1]); sqlite3BtreeEnter(pCur->pBtree); rc = sqlite3BtreeFirst(pCur, &res); sqlite3BtreeLeave(pCur->pBtree); if( rc ){ - Tcl_AppendResult(interp, errorName(rc), 0); + Tcl_AppendResult(interp, sqlite3ErrName(rc), 0); return TCL_ERROR; } sqlite3_snprintf(sizeof(zBuf),zBuf,"%d",res); Tcl_AppendResult(interp, zBuf, 0); return SQLITE_OK; Index: src/test4.c ================================================================== --- src/test4.c +++ src/test4.c @@ -18,10 +18,12 @@ #include #include #include #include +extern const char *sqlite3ErrName(int); + /* ** Each thread is controlled by an instance of the following ** structure. */ typedef struct Thread Thread; @@ -370,38 +372,11 @@ if( !threadset[i].busy ){ Tcl_AppendResult(interp, "no such thread", 0); return TCL_ERROR; } thread_wait(&threadset[i]); - switch( threadset[i].rc ){ - case SQLITE_OK: zName = "SQLITE_OK"; break; - case SQLITE_ERROR: zName = "SQLITE_ERROR"; break; - case SQLITE_PERM: zName = "SQLITE_PERM"; break; - case SQLITE_ABORT: zName = "SQLITE_ABORT"; break; - case SQLITE_BUSY: zName = "SQLITE_BUSY"; break; - case SQLITE_LOCKED: zName = "SQLITE_LOCKED"; break; - case SQLITE_NOMEM: zName = "SQLITE_NOMEM"; break; - case SQLITE_READONLY: zName = "SQLITE_READONLY"; break; - case SQLITE_INTERRUPT: zName = "SQLITE_INTERRUPT"; break; - case SQLITE_IOERR: zName = "SQLITE_IOERR"; break; - case SQLITE_CORRUPT: zName = "SQLITE_CORRUPT"; break; - case SQLITE_FULL: zName = "SQLITE_FULL"; break; - case SQLITE_CANTOPEN: zName = "SQLITE_CANTOPEN"; break; - case SQLITE_PROTOCOL: zName = "SQLITE_PROTOCOL"; break; - case SQLITE_EMPTY: zName = "SQLITE_EMPTY"; break; - case SQLITE_SCHEMA: zName = "SQLITE_SCHEMA"; break; - case SQLITE_CONSTRAINT: zName = "SQLITE_CONSTRAINT"; break; - case SQLITE_MISMATCH: zName = "SQLITE_MISMATCH"; break; - case SQLITE_MISUSE: zName = "SQLITE_MISUSE"; break; - case SQLITE_NOLFS: zName = "SQLITE_NOLFS"; break; - case SQLITE_AUTH: zName = "SQLITE_AUTH"; break; - case SQLITE_FORMAT: zName = "SQLITE_FORMAT"; break; - case SQLITE_RANGE: zName = "SQLITE_RANGE"; break; - case SQLITE_ROW: zName = "SQLITE_ROW"; break; - case SQLITE_DONE: zName = "SQLITE_DONE"; break; - default: zName = "SQLITE_Unknown"; break; - } + zName = sqlite3ErrName(threadset[i].rc); Tcl_AppendResult(interp, zName, 0); return TCL_OK; } /* Index: src/test6.c ================================================================== --- src/test6.c +++ src/test6.c @@ -85,11 +85,11 @@ ** ** If the IOCAP_ATOMIC512 flag is set, and the WriteBuffer represents ** an aligned write() of an integer number of 512 byte regions, then ** option (3) above is never selected. Instead, each 512 byte region ** is either correctly written or left completely untouched. Similar -** logic governs the behaviour if any of the other ATOMICXXX flags +** logic governs the behavior if any of the other ATOMICXXX flags ** is set. ** ** If either the IOCAP_SAFEAPPEND or IOCAP_SEQUENTIAL flags are set ** and a crash is being simulated, then an entry of the write-list is ** selected at random. Everything in the list after the selected entry Index: src/test7.c ================================================================== --- src/test7.c +++ src/test7.c @@ -374,10 +374,12 @@ } Tcl_AppendResult(interp, threadset[i].colv[n], 0); return TCL_OK; } +extern const char *sqlite3ErrName(int); + /* ** Usage: client_result ID ** ** Wait on the most recent operation to complete, then return the ** result code from that operation. @@ -401,38 +403,11 @@ if( !threadset[i].busy ){ Tcl_AppendResult(interp, "no such thread", 0); return TCL_ERROR; } client_wait(&threadset[i]); - switch( threadset[i].rc ){ - case SQLITE_OK: zName = "SQLITE_OK"; break; - case SQLITE_ERROR: zName = "SQLITE_ERROR"; break; - case SQLITE_PERM: zName = "SQLITE_PERM"; break; - case SQLITE_ABORT: zName = "SQLITE_ABORT"; break; - case SQLITE_BUSY: zName = "SQLITE_BUSY"; break; - case SQLITE_LOCKED: zName = "SQLITE_LOCKED"; break; - case SQLITE_NOMEM: zName = "SQLITE_NOMEM"; break; - case SQLITE_READONLY: zName = "SQLITE_READONLY"; break; - case SQLITE_INTERRUPT: zName = "SQLITE_INTERRUPT"; break; - case SQLITE_IOERR: zName = "SQLITE_IOERR"; break; - case SQLITE_CORRUPT: zName = "SQLITE_CORRUPT"; break; - case SQLITE_FULL: zName = "SQLITE_FULL"; break; - case SQLITE_CANTOPEN: zName = "SQLITE_CANTOPEN"; break; - case SQLITE_PROTOCOL: zName = "SQLITE_PROTOCOL"; break; - case SQLITE_EMPTY: zName = "SQLITE_EMPTY"; break; - case SQLITE_SCHEMA: zName = "SQLITE_SCHEMA"; break; - case SQLITE_CONSTRAINT: zName = "SQLITE_CONSTRAINT"; break; - case SQLITE_MISMATCH: zName = "SQLITE_MISMATCH"; break; - case SQLITE_MISUSE: zName = "SQLITE_MISUSE"; break; - case SQLITE_NOLFS: zName = "SQLITE_NOLFS"; break; - case SQLITE_AUTH: zName = "SQLITE_AUTH"; break; - case SQLITE_FORMAT: zName = "SQLITE_FORMAT"; break; - case SQLITE_RANGE: zName = "SQLITE_RANGE"; break; - case SQLITE_ROW: zName = "SQLITE_ROW"; break; - case SQLITE_DONE: zName = "SQLITE_DONE"; break; - default: zName = "SQLITE_Unknown"; break; - } + zName = sqlite3ErrName(threadset[i].rc); Tcl_AppendResult(interp, zName, 0); return TCL_OK; } /* Index: src/test8.c ================================================================== --- src/test8.c +++ src/test8.c @@ -1298,11 +1298,11 @@ /* ** Decode a pointer to an sqlite3 object. */ extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); -extern const char *sqlite3TestErrorName(int rc); +extern const char *sqlite3ErrName(int); static void moduleDestroy(void *p){ sqlite3_free(p); } @@ -1338,11 +1338,11 @@ rc = sqlite3_create_module_v2(db, "echo_v2", &echoModuleV2, (void*)pMod, moduleDestroy ); } - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* ** Tcl interface to sqlite3_declare_vtab, invoked as follows from Tcl: @@ -1368,33 +1368,10 @@ return TCL_ERROR; } return TCL_OK; } -#include "test_spellfix.c" - -/* -** Register the spellfix virtual table module. -*/ -static int register_spellfix_module( - ClientData clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ - sqlite3 *db; - - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; - - sqlite3_spellfix1_register(db); - return TCL_OK; -} - #endif /* ifndef SQLITE_OMIT_VIRTUALTABLE */ /* ** Register commands with the TCL interpreter. */ @@ -1404,11 +1381,10 @@ char *zName; Tcl_ObjCmdProc *xProc; void *clientData; } aObjCmd[] = { { "register_echo_module", register_echo_module, 0 }, - { "register_spellfix_module", register_spellfix_module, 0 }, { "sqlite3_declare_vtab", declare_vtab, 0 }, }; int i; for(i=0; i -/* From test1.c */ -const char *sqlite3TestErrorName(int); +/* From main.c */ +extern const char *sqlite3ErrName(int); struct TestAsyncGlobal { int isInstalled; /* True when async VFS is installed */ } testasync_g = { 0 }; @@ -58,11 +58,11 @@ return TCL_ERROR; } rc = sqlite3async_initialize(zParent, isDefault); if( rc!=SQLITE_OK ){ - Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3TestErrorName(rc), -1)); + Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3ErrName(rc), -1)); return TCL_ERROR; } return TCL_OK; } @@ -206,11 +206,11 @@ eOpt==SQLITEASYNC_DELAY ? SQLITEASYNC_GET_DELAY : SQLITEASYNC_GET_LOCKFILES, &iVal); } if( rc!=SQLITE_OK ){ - Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3TestErrorName(rc), -1)); + Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3ErrName(rc), -1)); return TCL_ERROR; } if( eOpt==SQLITEASYNC_HALT ){ Tcl_SetObjResult(interp, Tcl_NewStringObj(az[iVal], -1)); Index: src/test_autoext.c ================================================================== --- src/test_autoext.c +++ src/test_autoext.c @@ -13,11 +13,11 @@ */ #include "tcl.h" #include "sqlite3ext.h" #ifndef SQLITE_OMIT_LOAD_EXTENSION -static SQLITE_EXTENSION_INIT1 +SQLITE_EXTENSION_INIT1 /* ** The sqr() SQL function returns the square of its input value. */ static void sqrFunc( Index: src/test_backup.c ================================================================== --- src/test_backup.c +++ src/test_backup.c @@ -14,14 +14,16 @@ */ #include "tcl.h" #include #include + +/* These functions are implemented in main.c. */ +extern const char *sqlite3ErrName(int); /* These functions are implemented in test1.c. */ -int getDbPointer(Tcl_Interp *, const char *, sqlite3 **); -const char *sqlite3TestErrorName(int); +extern int getDbPointer(Tcl_Interp *, const char *, sqlite3 **); static int backupTestCmd( ClientData clientData, Tcl_Interp *interp, int objc, @@ -68,21 +70,21 @@ cmdInfo.deleteProc = 0; Tcl_SetCommandInfo(interp, zCmdName, &cmdInfo); Tcl_DeleteCommand(interp, zCmdName); rc = sqlite3_backup_finish(p); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); break; } case BACKUP_STEP: { int nPage; if( TCL_OK!=Tcl_GetIntFromObj(interp, objv[2], &nPage) ){ return TCL_ERROR; } rc = sqlite3_backup_step(p, nPage); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); break; } case BACKUP_REMAINING: Tcl_SetObjResult(interp, Tcl_NewIntObj(sqlite3_backup_remaining(p))); Index: src/test_config.c ================================================================== --- src/test_config.c +++ src/test_config.c @@ -55,11 +55,11 @@ Tcl_SetVar2(interp, "sqlite_options","casesensitivelike","1",TCL_GLOBAL_ONLY); #else Tcl_SetVar2(interp, "sqlite_options","casesensitivelike","0",TCL_GLOBAL_ONLY); #endif -#ifdef SQLITE_CURDIR +#if !SQLITE_OS_WINCE && !SQLITE_OS_WINRT Tcl_SetVar2(interp, "sqlite_options", "curdir", "1", TCL_GLOBAL_ONLY); #else Tcl_SetVar2(interp, "sqlite_options", "curdir", "0", TCL_GLOBAL_ONLY); #endif @@ -84,10 +84,16 @@ #ifdef SQLITE_DISABLE_LFS Tcl_SetVar2(interp, "sqlite_options", "lfs", "0", TCL_GLOBAL_ONLY); #else Tcl_SetVar2(interp, "sqlite_options", "lfs", "1", TCL_GLOBAL_ONLY); #endif + +#if SQLITE_MAX_MMAP_SIZE>0 + Tcl_SetVar2(interp, "sqlite_options", "mmap", "1", TCL_GLOBAL_ONLY); +#else + Tcl_SetVar2(interp, "sqlite_options", "mmap", "0", TCL_GLOBAL_ONLY); +#endif #if 1 /* def SQLITE_MEMDEBUG */ Tcl_SetVar2(interp, "sqlite_options", "memdebug", "1", TCL_GLOBAL_ONLY); #else Tcl_SetVar2(interp, "sqlite_options", "memdebug", "0", TCL_GLOBAL_ONLY); @@ -393,15 +399,11 @@ Tcl_SetVar2(interp, "sqlite_options", "memorymanage", "1", TCL_GLOBAL_ONLY); #else Tcl_SetVar2(interp, "sqlite_options", "memorymanage", "0", TCL_GLOBAL_ONLY); #endif -#ifdef SQLITE_OMIT_MERGE_SORT - Tcl_SetVar2(interp, "sqlite_options", "mergesort", "0", TCL_GLOBAL_ONLY); -#else - Tcl_SetVar2(interp, "sqlite_options", "mergesort", "1", TCL_GLOBAL_ONLY); -#endif +Tcl_SetVar2(interp, "sqlite_options", "mergesort", "1", TCL_GLOBAL_ONLY); #ifdef SQLITE_OMIT_OR_OPTIMIZATION Tcl_SetVar2(interp, "sqlite_options", "or_opt", "0", TCL_GLOBAL_ONLY); #else Tcl_SetVar2(interp, "sqlite_options", "or_opt", "1", TCL_GLOBAL_ONLY); Index: src/test_fs.c ================================================================== --- src/test_fs.c +++ src/test_fs.c @@ -97,11 +97,11 @@ *pzErr = sqlite3_mprintf("wrong number of arguments"); return SQLITE_ERROR; } zTbl = argv[3]; - nByte = sizeof(fs_vtab) + strlen(zTbl) + 1 + strlen(zDb) + 1; + nByte = sizeof(fs_vtab) + (int)strlen(zTbl) + 1 + (int)strlen(zDb) + 1; pVtab = (fs_vtab *)sqlite3MallocZero( nByte ); if( !pVtab ) return SQLITE_NOMEM; pVtab->zTbl = (char *)&pVtab[1]; pVtab->zDb = &pVtab->zTbl[strlen(zTbl)+1]; DELETED src/test_fuzzer.c Index: src/test_fuzzer.c ================================================================== --- src/test_fuzzer.c +++ src/test_fuzzer.c @@ -1,1215 +0,0 @@ -/* -** 2011 March 24 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** -** Code for a demonstration virtual table that generates variations -** on an input word at increasing edit distances from the original. -** -** A fuzzer virtual table is created like this: -** -** CREATE VIRTUAL TABLE f USING fuzzer(); -** -** When it is created, the new fuzzer table must be supplied with the -** name of a "fuzzer data table", which must reside in the same database -** file as the new fuzzer table. The fuzzer data table contains the various -** transformations and their costs that the fuzzer logic uses to generate -** variations. -** -** The fuzzer data table must contain exactly four columns (more precisely, -** the statement "SELECT * FROM " must return records -** that consist of four columns). It does not matter what the columns are -** named. -** -** Each row in the fuzzer data table represents a single character -** transformation. The left most column of the row (column 0) contains an -** integer value - the identifier of the ruleset to which the transformation -** rule belongs (see "MULTIPLE RULE SETS" below). The second column of the -** row (column 0) contains the input character or characters. The third -** column contains the output character or characters. And the fourth column -** contains the integer cost of making the transformation. For example: -** -** CREATE TABLE f_data(ruleset, cFrom, cTo, Cost); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, '', 'a', 100); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'b', '', 87); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'o', 'oe', 38); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'oe', 'o', 40); -** -** The first row inserted into the fuzzer data table by the SQL script -** above indicates that the cost of inserting a letter 'a' is 100. (All -** costs are integers. We recommend that costs be scaled so that the -** average cost is around 100.) The second INSERT statement creates a rule -** saying that the cost of deleting a single letter 'b' is 87. The third -** and fourth INSERT statements mean that the cost of transforming a -** single letter "o" into the two-letter sequence "oe" is 38 and that the -** cost of transforming "oe" back into "o" is 40. -** -** The contents of the fuzzer data table are loaded into main memory when -** a fuzzer table is first created, and may be internally reloaded by the -** system at any subsequent time. Therefore, the fuzzer data table should be -** populated before the fuzzer table is created and not modified thereafter. -** If you do need to modify the contents of the fuzzer data table, it is -** recommended that the associated fuzzer table be dropped, the fuzzer data -** table edited, and the fuzzer table recreated within a single transaction. -** Alternatively, the fuzzer data table can be edited then the database -** connection can be closed and reopened. -** -** Once it has been created, the fuzzer table can be queried as follows: -** -** SELECT word, distance FROM f -** WHERE word MATCH 'abcdefg' -** AND distance<200; -** -** This first query outputs the string "abcdefg" and all strings that -** can be derived from that string by appling the specified transformations. -** The strings are output together with their total transformation cost -** (called "distance") and appear in order of increasing cost. No string -** is output more than once. If there are multiple ways to transform the -** target string into the output string then the lowest cost transform is -** the one that is returned. In the example, the search is limited to -** strings with a total distance of less than 200. -** -** The fuzzer is a read-only table. Any attempt to DELETE, INSERT, or -** UPDATE on a fuzzer table will throw an error. -** -** It is important to put some kind of a limit on the fuzzer output. This -** can be either in the form of a LIMIT clause at the end of the query, -** or better, a "distance -#include -#include -#include - -#ifndef SQLITE_OMIT_VIRTUALTABLE - -/* -** Forward declaration of objects used by this implementation -*/ -typedef struct fuzzer_vtab fuzzer_vtab; -typedef struct fuzzer_cursor fuzzer_cursor; -typedef struct fuzzer_rule fuzzer_rule; -typedef struct fuzzer_seen fuzzer_seen; -typedef struct fuzzer_stem fuzzer_stem; - -/* -** Various types. -** -** fuzzer_cost is the "cost" of an edit operation. -** -** fuzzer_len is the length of a matching string. -** -** fuzzer_ruleid is an ruleset identifier. -*/ -typedef int fuzzer_cost; -typedef signed char fuzzer_len; -typedef int fuzzer_ruleid; - -/* -** Limits -*/ -#define FUZZER_MX_LENGTH 50 /* Maximum length of a rule string */ -#define FUZZER_MX_RULEID 2147483647 /* Maximum rule ID */ -#define FUZZER_MX_COST 1000 /* Maximum single-rule cost */ -#define FUZZER_MX_OUTPUT_LENGTH 100 /* Maximum length of an output string */ - - -/* -** Each transformation rule is stored as an instance of this object. -** All rules are kept on a linked list sorted by rCost. -*/ -struct fuzzer_rule { - fuzzer_rule *pNext; /* Next rule in order of increasing rCost */ - char *zFrom; /* Transform from */ - fuzzer_cost rCost; /* Cost of this transformation */ - fuzzer_len nFrom, nTo; /* Length of the zFrom and zTo strings */ - fuzzer_ruleid iRuleset; /* The rule set to which this rule belongs */ - char zTo[4]; /* Transform to (extra space appended) */ -}; - -/* -** A stem object is used to generate variants. It is also used to record -** previously generated outputs. -** -** Every stem is added to a hash table as it is output. Generation of -** duplicate stems is suppressed. -** -** Active stems (those that might generate new outputs) are kepts on a linked -** list sorted by increasing cost. The cost is the sum of rBaseCost and -** pRule->rCost. -*/ -struct fuzzer_stem { - char *zBasis; /* Word being fuzzed */ - const fuzzer_rule *pRule; /* Current rule to apply */ - fuzzer_stem *pNext; /* Next stem in rCost order */ - fuzzer_stem *pHash; /* Next stem with same hash on zBasis */ - fuzzer_cost rBaseCost; /* Base cost of getting to zBasis */ - fuzzer_cost rCostX; /* Precomputed rBaseCost + pRule->rCost */ - fuzzer_len nBasis; /* Length of the zBasis string */ - fuzzer_len n; /* Apply pRule at this character offset */ -}; - -/* -** A fuzzer virtual-table object -*/ -struct fuzzer_vtab { - sqlite3_vtab base; /* Base class - must be first */ - char *zClassName; /* Name of this class. Default: "fuzzer" */ - fuzzer_rule *pRule; /* All active rules in this fuzzer */ - int nCursor; /* Number of active cursors */ -}; - -#define FUZZER_HASH 4001 /* Hash table size */ -#define FUZZER_NQUEUE 20 /* Number of slots on the stem queue */ - -/* A fuzzer cursor object */ -struct fuzzer_cursor { - sqlite3_vtab_cursor base; /* Base class - must be first */ - sqlite3_int64 iRowid; /* The rowid of the current word */ - fuzzer_vtab *pVtab; /* The virtual table this cursor belongs to */ - fuzzer_cost rLimit; /* Maximum cost of any term */ - fuzzer_stem *pStem; /* Stem with smallest rCostX */ - fuzzer_stem *pDone; /* Stems already processed to completion */ - fuzzer_stem *aQueue[FUZZER_NQUEUE]; /* Queue of stems with higher rCostX */ - int mxQueue; /* Largest used index in aQueue[] */ - char *zBuf; /* Temporary use buffer */ - int nBuf; /* Bytes allocated for zBuf */ - int nStem; /* Number of stems allocated */ - int iRuleset; /* Only process rules from this ruleset */ - fuzzer_rule nullRule; /* Null rule used first */ - fuzzer_stem *apHash[FUZZER_HASH]; /* Hash of previously generated terms */ -}; - -/* -** The two input rule lists are both sorted in order of increasing -** cost. Merge them together into a single list, sorted by cost, and -** return a pointer to the head of that list. -*/ -static fuzzer_rule *fuzzerMergeRules(fuzzer_rule *pA, fuzzer_rule *pB){ - fuzzer_rule head; - fuzzer_rule *pTail; - - pTail = &head; - while( pA && pB ){ - if( pA->rCost<=pB->rCost ){ - pTail->pNext = pA; - pTail = pA; - pA = pA->pNext; - }else{ - pTail->pNext = pB; - pTail = pB; - pB = pB->pNext; - } - } - if( pA==0 ){ - pTail->pNext = pB; - }else{ - pTail->pNext = pA; - } - return head.pNext; -} - -/* -** Statement pStmt currently points to a row in the fuzzer data table. This -** function allocates and populates a fuzzer_rule structure according to -** the content of the row. -** -** If successful, *ppRule is set to point to the new object and SQLITE_OK -** is returned. Otherwise, *ppRule is zeroed, *pzErr may be set to point -** to an error message and an SQLite error code returned. -*/ -static int fuzzerLoadOneRule( - fuzzer_vtab *p, /* Fuzzer virtual table handle */ - sqlite3_stmt *pStmt, /* Base rule on statements current row */ - fuzzer_rule **ppRule, /* OUT: New rule object */ - char **pzErr /* OUT: Error message */ -){ - sqlite3_int64 iRuleset = sqlite3_column_int64(pStmt, 0); - const char *zFrom = (const char *)sqlite3_column_text(pStmt, 1); - const char *zTo = (const char *)sqlite3_column_text(pStmt, 2); - int nCost = sqlite3_column_int(pStmt, 3); - - int rc = SQLITE_OK; /* Return code */ - int nFrom; /* Size of string zFrom, in bytes */ - int nTo; /* Size of string zTo, in bytes */ - fuzzer_rule *pRule = 0; /* New rule object to return */ - - if( zFrom==0 ) zFrom = ""; - if( zTo==0 ) zTo = ""; - nFrom = (int)strlen(zFrom); - nTo = (int)strlen(zTo); - - /* Silently ignore null transformations */ - if( strcmp(zFrom, zTo)==0 ){ - *ppRule = 0; - return SQLITE_OK; - } - - if( nCost<=0 || nCost>FUZZER_MX_COST ){ - *pzErr = sqlite3_mprintf("%s: cost must be between 1 and %d", - p->zClassName, FUZZER_MX_COST - ); - rc = SQLITE_ERROR; - }else - if( nFrom>FUZZER_MX_LENGTH || nTo>FUZZER_MX_LENGTH ){ - *pzErr = sqlite3_mprintf("%s: maximum string length is %d", - p->zClassName, FUZZER_MX_LENGTH - ); - rc = SQLITE_ERROR; - }else - if( iRuleset<0 || iRuleset>FUZZER_MX_RULEID ){ - *pzErr = sqlite3_mprintf("%s: ruleset must be between 0 and %d", - p->zClassName, FUZZER_MX_RULEID - ); - rc = SQLITE_ERROR; - }else{ - - pRule = sqlite3_malloc( sizeof(*pRule) + nFrom + nTo ); - if( pRule==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(pRule, 0, sizeof(*pRule)); - pRule->zFrom = &pRule->zTo[nTo+1]; - pRule->nFrom = nFrom; - memcpy(pRule->zFrom, zFrom, nFrom+1); - memcpy(pRule->zTo, zTo, nTo+1); - pRule->nTo = nTo; - pRule->rCost = nCost; - pRule->iRuleset = (int)iRuleset; - } - } - - *ppRule = pRule; - return rc; -} - -/* -** Load the content of the fuzzer data table into memory. -*/ -static int fuzzerLoadRules( - sqlite3 *db, /* Database handle */ - fuzzer_vtab *p, /* Virtual fuzzer table to configure */ - const char *zDb, /* Database containing rules data */ - const char *zData, /* Table containing rules data */ - char **pzErr /* OUT: Error message */ -){ - int rc = SQLITE_OK; /* Return code */ - char *zSql; /* SELECT used to read from rules table */ - fuzzer_rule *pHead = 0; - - zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zData); - if( zSql==0 ){ - rc = SQLITE_NOMEM; - }else{ - int rc2; /* finalize() return code */ - sqlite3_stmt *pStmt = 0; - rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); - if( rc!=SQLITE_OK ){ - *pzErr = sqlite3_mprintf("%s: %s", p->zClassName, sqlite3_errmsg(db)); - }else if( sqlite3_column_count(pStmt)!=4 ){ - *pzErr = sqlite3_mprintf("%s: %s has %d columns, expected 4", - p->zClassName, zData, sqlite3_column_count(pStmt) - ); - rc = SQLITE_ERROR; - }else{ - while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ - fuzzer_rule *pRule = 0; - rc = fuzzerLoadOneRule(p, pStmt, &pRule, pzErr); - if( pRule ){ - pRule->pNext = pHead; - pHead = pRule; - } - } - } - rc2 = sqlite3_finalize(pStmt); - if( rc==SQLITE_OK ) rc = rc2; - } - sqlite3_free(zSql); - - /* All rules are now in a singly linked list starting at pHead. This - ** block sorts them by cost and then sets fuzzer_vtab.pRule to point to - ** point to the head of the sorted list. - */ - if( rc==SQLITE_OK ){ - unsigned int i; - fuzzer_rule *pX; - fuzzer_rule *a[15]; - for(i=0; ipNext; - pX->pNext = 0; - for(i=0; a[i] && ipRule = fuzzerMergeRules(p->pRule, pX); - }else{ - /* An error has occurred. Setting p->pRule to point to the head of the - ** allocated list ensures that the list will be cleaned up in this case. - */ - assert( p->pRule==0 ); - p->pRule = pHead; - } - - return rc; -} - -/* -** This function converts an SQL quoted string into an unquoted string -** and returns a pointer to a buffer allocated using sqlite3_malloc() -** containing the result. The caller should eventually free this buffer -** using sqlite3_free. -** -** Examples: -** -** "abc" becomes abc -** 'xyz' becomes xyz -** [pqr] becomes pqr -** `mno` becomes mno -*/ -static char *fuzzerDequote(const char *zIn){ - int nIn; /* Size of input string, in bytes */ - char *zOut; /* Output (dequoted) string */ - - nIn = (int)strlen(zIn); - zOut = sqlite3_malloc(nIn+1); - if( zOut ){ - char q = zIn[0]; /* Quote character (if any ) */ - - if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ - memcpy(zOut, zIn, nIn+1); - }else{ - int iOut = 0; /* Index of next byte to write to output */ - int iIn; /* Index of next byte to read from input */ - - if( q=='[' ) q = ']'; - for(iIn=1; iInnCursor==0 ); - while( p->pRule ){ - fuzzer_rule *pRule = p->pRule; - p->pRule = pRule->pNext; - sqlite3_free(pRule); - } - sqlite3_free(p); - return SQLITE_OK; -} - -/* -** xConnect/xCreate method for the fuzzer module. Arguments are: -** -** argv[0] -> module name ("fuzzer") -** argv[1] -> database name -** argv[2] -> table name -** argv[3] -> fuzzer rule table name -*/ -static int fuzzerConnect( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVtab, - char **pzErr -){ - int rc = SQLITE_OK; /* Return code */ - fuzzer_vtab *pNew = 0; /* New virtual table */ - const char *zModule = argv[0]; - const char *zDb = argv[1]; - - if( argc!=4 ){ - *pzErr = sqlite3_mprintf( - "%s: wrong number of CREATE VIRTUAL TABLE arguments", zModule - ); - rc = SQLITE_ERROR; - }else{ - int nModule; /* Length of zModule, in bytes */ - - nModule = (int)strlen(zModule); - pNew = sqlite3_malloc( sizeof(*pNew) + nModule + 1); - if( pNew==0 ){ - rc = SQLITE_NOMEM; - }else{ - char *zTab; /* Dequoted name of fuzzer data table */ - - memset(pNew, 0, sizeof(*pNew)); - pNew->zClassName = (char*)&pNew[1]; - memcpy(pNew->zClassName, zModule, nModule+1); - - zTab = fuzzerDequote(argv[3]); - if( zTab==0 ){ - rc = SQLITE_NOMEM; - }else{ - rc = fuzzerLoadRules(db, pNew, zDb, zTab, pzErr); - sqlite3_free(zTab); - } - - if( rc==SQLITE_OK ){ - rc = sqlite3_declare_vtab(db, "CREATE TABLE x(word,distance,ruleset)"); - } - if( rc!=SQLITE_OK ){ - fuzzerDisconnect((sqlite3_vtab *)pNew); - pNew = 0; - } - } - } - - *ppVtab = (sqlite3_vtab *)pNew; - return rc; -} - -/* -** Open a new fuzzer cursor. -*/ -static int fuzzerOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ - fuzzer_vtab *p = (fuzzer_vtab*)pVTab; - fuzzer_cursor *pCur; - pCur = sqlite3_malloc( sizeof(*pCur) ); - if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); - pCur->pVtab = p; - *ppCursor = &pCur->base; - p->nCursor++; - return SQLITE_OK; -} - -/* -** Free all stems in a list. -*/ -static void fuzzerClearStemList(fuzzer_stem *pStem){ - while( pStem ){ - fuzzer_stem *pNext = pStem->pNext; - sqlite3_free(pStem); - pStem = pNext; - } -} - -/* -** Free up all the memory allocated by a cursor. Set it rLimit to 0 -** to indicate that it is at EOF. -*/ -static void fuzzerClearCursor(fuzzer_cursor *pCur, int clearHash){ - int i; - fuzzerClearStemList(pCur->pStem); - fuzzerClearStemList(pCur->pDone); - for(i=0; iaQueue[i]); - pCur->rLimit = (fuzzer_cost)0; - if( clearHash && pCur->nStem ){ - pCur->mxQueue = 0; - pCur->pStem = 0; - pCur->pDone = 0; - memset(pCur->aQueue, 0, sizeof(pCur->aQueue)); - memset(pCur->apHash, 0, sizeof(pCur->apHash)); - } - pCur->nStem = 0; -} - -/* -** Close a fuzzer cursor. -*/ -static int fuzzerClose(sqlite3_vtab_cursor *cur){ - fuzzer_cursor *pCur = (fuzzer_cursor *)cur; - fuzzerClearCursor(pCur, 0); - sqlite3_free(pCur->zBuf); - pCur->pVtab->nCursor--; - sqlite3_free(pCur); - return SQLITE_OK; -} - -/* -** Compute the current output term for a fuzzer_stem. -*/ -static int fuzzerRender( - fuzzer_stem *pStem, /* The stem to be rendered */ - char **pzBuf, /* Write results into this buffer. realloc if needed */ - int *pnBuf /* Size of the buffer */ -){ - const fuzzer_rule *pRule = pStem->pRule; - int n; /* Size of output term without nul-term */ - char *z; /* Buffer to assemble output term in */ - - n = pStem->nBasis + pRule->nTo - pRule->nFrom; - if( (*pnBuf)n; - z = *pzBuf; - if( n<0 ){ - memcpy(z, pStem->zBasis, pStem->nBasis+1); - }else{ - memcpy(z, pStem->zBasis, n); - memcpy(&z[n], pRule->zTo, pRule->nTo); - memcpy(&z[n+pRule->nTo], &pStem->zBasis[n+pRule->nFrom], - pStem->nBasis-n-pRule->nFrom+1); - } - - assert( z[pStem->nBasis + pRule->nTo - pRule->nFrom]==0 ); - return SQLITE_OK; -} - -/* -** Compute a hash on zBasis. -*/ -static unsigned int fuzzerHash(const char *z){ - unsigned int h = 0; - while( *z ){ h = (h<<3) ^ (h>>29) ^ *(z++); } - return h % FUZZER_HASH; -} - -/* -** Current cost of a stem -*/ -static fuzzer_cost fuzzerCost(fuzzer_stem *pStem){ - return pStem->rCostX = pStem->rBaseCost + pStem->pRule->rCost; -} - -#if 0 -/* -** Print a description of a fuzzer_stem on stderr. -*/ -static void fuzzerStemPrint( - const char *zPrefix, - fuzzer_stem *pStem, - const char *zSuffix -){ - if( pStem->n<0 ){ - fprintf(stderr, "%s[%s](%d)-->self%s", - zPrefix, - pStem->zBasis, pStem->rBaseCost, - zSuffix - ); - }else{ - char *zBuf = 0; - int nBuf = 0; - if( fuzzerRender(pStem, &zBuf, &nBuf)!=SQLITE_OK ) return; - fprintf(stderr, "%s[%s](%d)-->{%s}(%d)%s", - zPrefix, - pStem->zBasis, pStem->rBaseCost, zBuf, pStem->, - zSuffix - ); - sqlite3_free(zBuf); - } -} -#endif - -/* -** Return 1 if the string to which the cursor is point has already -** been emitted. Return 0 if not. Return -1 on a memory allocation -** failures. -*/ -static int fuzzerSeen(fuzzer_cursor *pCur, fuzzer_stem *pStem){ - unsigned int h; - fuzzer_stem *pLookup; - - if( fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ - return -1; - } - h = fuzzerHash(pCur->zBuf); - pLookup = pCur->apHash[h]; - while( pLookup && strcmp(pLookup->zBasis, pCur->zBuf)!=0 ){ - pLookup = pLookup->pHash; - } - return pLookup!=0; -} - -/* -** If argument pRule is NULL, this function returns false. -** -** Otherwise, it returns true if rule pRule should be skipped. A rule -** should be skipped if it does not belong to rule-set iRuleset, or if -** applying it to stem pStem would create a string longer than -** FUZZER_MX_OUTPUT_LENGTH bytes. -*/ -static int fuzzerSkipRule( - const fuzzer_rule *pRule, /* Determine whether or not to skip this */ - fuzzer_stem *pStem, /* Stem rule may be applied to */ - int iRuleset /* Rule-set used by the current query */ -){ - return pRule && ( - (pRule->iRuleset!=iRuleset) - || (pStem->nBasis + pRule->nTo - pRule->nFrom)>FUZZER_MX_OUTPUT_LENGTH - ); -} - -/* -** Advance a fuzzer_stem to its next value. Return 0 if there are -** no more values that can be generated by this fuzzer_stem. Return -** -1 on a memory allocation failure. -*/ -static int fuzzerAdvance(fuzzer_cursor *pCur, fuzzer_stem *pStem){ - const fuzzer_rule *pRule; - while( (pRule = pStem->pRule)!=0 ){ - assert( pRule==&pCur->nullRule || pRule->iRuleset==pCur->iRuleset ); - while( pStem->n < pStem->nBasis - pRule->nFrom ){ - pStem->n++; - if( pRule->nFrom==0 - || memcmp(&pStem->zBasis[pStem->n], pRule->zFrom, pRule->nFrom)==0 - ){ - /* Found a rewrite case. Make sure it is not a duplicate */ - int rc = fuzzerSeen(pCur, pStem); - if( rc<0 ) return -1; - if( rc==0 ){ - fuzzerCost(pStem); - return 1; - } - } - } - pStem->n = -1; - do{ - pRule = pRule->pNext; - }while( fuzzerSkipRule(pRule, pStem, pCur->iRuleset) ); - pStem->pRule = pRule; - if( pRule && fuzzerCost(pStem)>pCur->rLimit ) pStem->pRule = 0; - } - return 0; -} - -/* -** The two input stem lists are both sorted in order of increasing -** rCostX. Merge them together into a single list, sorted by rCostX, and -** return a pointer to the head of that new list. -*/ -static fuzzer_stem *fuzzerMergeStems(fuzzer_stem *pA, fuzzer_stem *pB){ - fuzzer_stem head; - fuzzer_stem *pTail; - - pTail = &head; - while( pA && pB ){ - if( pA->rCostX<=pB->rCostX ){ - pTail->pNext = pA; - pTail = pA; - pA = pA->pNext; - }else{ - pTail->pNext = pB; - pTail = pB; - pB = pB->pNext; - } - } - if( pA==0 ){ - pTail->pNext = pB; - }else{ - pTail->pNext = pA; - } - return head.pNext; -} - -/* -** Load pCur->pStem with the lowest-cost stem. Return a pointer -** to the lowest-cost stem. -*/ -static fuzzer_stem *fuzzerLowestCostStem(fuzzer_cursor *pCur){ - fuzzer_stem *pBest, *pX; - int iBest; - int i; - - if( pCur->pStem==0 ){ - iBest = -1; - pBest = 0; - for(i=0; i<=pCur->mxQueue; i++){ - pX = pCur->aQueue[i]; - if( pX==0 ) continue; - if( pBest==0 || pBest->rCostX>pX->rCostX ){ - pBest = pX; - iBest = i; - } - } - if( pBest ){ - pCur->aQueue[iBest] = pBest->pNext; - pBest->pNext = 0; - pCur->pStem = pBest; - } - } - return pCur->pStem; -} - -/* -** Insert pNew into queue of pending stems. Then find the stem -** with the lowest rCostX and move it into pCur->pStem. -** list. The insert is done such the pNew is in the correct order -** according to fuzzer_stem.zBaseCost+fuzzer_stem.pRule->rCost. -*/ -static fuzzer_stem *fuzzerInsert(fuzzer_cursor *pCur, fuzzer_stem *pNew){ - fuzzer_stem *pX; - int i; - - /* If pCur->pStem exists and is greater than pNew, then make pNew - ** the new pCur->pStem and insert the old pCur->pStem instead. - */ - if( (pX = pCur->pStem)!=0 && pX->rCostX>pNew->rCostX ){ - pNew->pNext = 0; - pCur->pStem = pNew; - pNew = pX; - } - - /* Insert the new value */ - pNew->pNext = 0; - pX = pNew; - for(i=0; i<=pCur->mxQueue; i++){ - if( pCur->aQueue[i] ){ - pX = fuzzerMergeStems(pX, pCur->aQueue[i]); - pCur->aQueue[i] = 0; - }else{ - pCur->aQueue[i] = pX; - break; - } - } - if( i>pCur->mxQueue ){ - if( imxQueue = i; - pCur->aQueue[i] = pX; - }else{ - assert( pCur->mxQueue==FUZZER_NQUEUE-1 ); - pX = fuzzerMergeStems(pX, pCur->aQueue[FUZZER_NQUEUE-1]); - pCur->aQueue[FUZZER_NQUEUE-1] = pX; - } - } - - return fuzzerLowestCostStem(pCur); -} - -/* -** Allocate a new fuzzer_stem. Add it to the hash table but do not -** link it into either the pCur->pStem or pCur->pDone lists. -*/ -static fuzzer_stem *fuzzerNewStem( - fuzzer_cursor *pCur, - const char *zWord, - fuzzer_cost rBaseCost -){ - fuzzer_stem *pNew; - fuzzer_rule *pRule; - unsigned int h; - - pNew = sqlite3_malloc( sizeof(*pNew) + (int)strlen(zWord) + 1 ); - if( pNew==0 ) return 0; - memset(pNew, 0, sizeof(*pNew)); - pNew->zBasis = (char*)&pNew[1]; - pNew->nBasis = (int)strlen(zWord); - memcpy(pNew->zBasis, zWord, pNew->nBasis+1); - pRule = pCur->pVtab->pRule; - while( fuzzerSkipRule(pRule, pNew, pCur->iRuleset) ){ - pRule = pRule->pNext; - } - pNew->pRule = pRule; - pNew->n = -1; - pNew->rBaseCost = pNew->rCostX = rBaseCost; - h = fuzzerHash(pNew->zBasis); - pNew->pHash = pCur->apHash[h]; - pCur->apHash[h] = pNew; - pCur->nStem++; - return pNew; -} - - -/* -** Advance a cursor to its next row of output -*/ -static int fuzzerNext(sqlite3_vtab_cursor *cur){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - int rc; - fuzzer_stem *pStem, *pNew; - - pCur->iRowid++; - - /* Use the element the cursor is currently point to to create - ** a new stem and insert the new stem into the priority queue. - */ - pStem = pCur->pStem; - if( pStem->rCostX>0 ){ - rc = fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf); - if( rc==SQLITE_NOMEM ) return SQLITE_NOMEM; - pNew = fuzzerNewStem(pCur, pCur->zBuf, pStem->rCostX); - if( pNew ){ - if( fuzzerAdvance(pCur, pNew)==0 ){ - pNew->pNext = pCur->pDone; - pCur->pDone = pNew; - }else{ - if( fuzzerInsert(pCur, pNew)==pNew ){ - return SQLITE_OK; - } - } - }else{ - return SQLITE_NOMEM; - } - } - - /* Adjust the priority queue so that the first element of the - ** stem list is the next lowest cost word. - */ - while( (pStem = pCur->pStem)!=0 ){ - int res = fuzzerAdvance(pCur, pStem); - if( res<0 ){ - return SQLITE_NOMEM; - }else if( res>0 ){ - pCur->pStem = 0; - pStem = fuzzerInsert(pCur, pStem); - if( (rc = fuzzerSeen(pCur, pStem))!=0 ){ - if( rc<0 ) return SQLITE_NOMEM; - continue; - } - return SQLITE_OK; /* New word found */ - } - pCur->pStem = 0; - pStem->pNext = pCur->pDone; - pCur->pDone = pStem; - if( fuzzerLowestCostStem(pCur) ){ - rc = fuzzerSeen(pCur, pCur->pStem); - if( rc<0 ) return SQLITE_NOMEM; - if( rc==0 ){ - return SQLITE_OK; - } - } - } - - /* Reach this point only if queue has been exhausted and there is - ** nothing left to be output. */ - pCur->rLimit = (fuzzer_cost)0; - return SQLITE_OK; -} - -/* -** Called to "rewind" a cursor back to the beginning so that -** it starts its output over again. Always called at least once -** prior to any fuzzerColumn, fuzzerRowid, or fuzzerEof call. -*/ -static int fuzzerFilter( - sqlite3_vtab_cursor *pVtabCursor, - int idxNum, const char *idxStr, - int argc, sqlite3_value **argv -){ - fuzzer_cursor *pCur = (fuzzer_cursor *)pVtabCursor; - const char *zWord = ""; - fuzzer_stem *pStem; - int idx; - - fuzzerClearCursor(pCur, 1); - pCur->rLimit = 2147483647; - idx = 0; - if( idxNum & 1 ){ - zWord = (const char*)sqlite3_value_text(argv[0]); - idx++; - } - if( idxNum & 2 ){ - pCur->rLimit = (fuzzer_cost)sqlite3_value_int(argv[idx]); - idx++; - } - if( idxNum & 4 ){ - pCur->iRuleset = (fuzzer_cost)sqlite3_value_int(argv[idx]); - idx++; - } - pCur->nullRule.pNext = pCur->pVtab->pRule; - pCur->nullRule.rCost = 0; - pCur->nullRule.nFrom = 0; - pCur->nullRule.nTo = 0; - pCur->nullRule.zFrom = ""; - pCur->iRowid = 1; - assert( pCur->pStem==0 ); - - /* If the query term is longer than FUZZER_MX_OUTPUT_LENGTH bytes, this - ** query will return zero rows. */ - if( (int)strlen(zWord)pStem = pStem = fuzzerNewStem(pCur, zWord, (fuzzer_cost)0); - if( pStem==0 ) return SQLITE_NOMEM; - pStem->pRule = &pCur->nullRule; - pStem->n = pStem->nBasis; - }else{ - pCur->rLimit = 0; - } - - return SQLITE_OK; -} - -/* -** Only the word and distance columns have values. All other columns -** return NULL -*/ -static int fuzzerColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - if( i==0 ){ - /* the "word" column */ - if( fuzzerRender(pCur->pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ - return SQLITE_NOMEM; - } - sqlite3_result_text(ctx, pCur->zBuf, -1, SQLITE_TRANSIENT); - }else if( i==1 ){ - /* the "distance" column */ - sqlite3_result_int(ctx, pCur->pStem->rCostX); - }else{ - /* All other columns are NULL */ - sqlite3_result_null(ctx); - } - return SQLITE_OK; -} - -/* -** The rowid. -*/ -static int fuzzerRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - *pRowid = pCur->iRowid; - return SQLITE_OK; -} - -/* -** When the fuzzer_cursor.rLimit value is 0 or less, that is a signal -** that the cursor has nothing more to output. -*/ -static int fuzzerEof(sqlite3_vtab_cursor *cur){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - return pCur->rLimit<=(fuzzer_cost)0; -} - -/* -** Search for terms of these forms: -** -** (A) word MATCH $str -** (B1) distance < $value -** (B2) distance <= $value -** (C) ruleid == $ruleid -** -** The distance< and distance<= are both treated as distance<=. -** The query plan number is a bit vector: -** -** bit 1: Term of the form (A) found -** bit 2: Term like (B1) or (B2) found -** bit 3: Term like (C) found -** -** If bit-1 is set, $str is always in filter.argv[0]. If bit-2 is set -** then $value is in filter.argv[0] if bit-1 is clear and is in -** filter.argv[1] if bit-1 is set. If bit-3 is set, then $ruleid is -** in filter.argv[0] if bit-1 and bit-2 are both zero, is in -** filter.argv[1] if exactly one of bit-1 and bit-2 are set, and is in -** filter.argv[2] if both bit-1 and bit-2 are set. -*/ -static int fuzzerBestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ - int iPlan = 0; - int iDistTerm = -1; - int iRulesetTerm = -1; - int i; - const struct sqlite3_index_constraint *pConstraint; - pConstraint = pIdxInfo->aConstraint; - for(i=0; inConstraint; i++, pConstraint++){ - if( pConstraint->usable==0 ) continue; - if( (iPlan & 1)==0 - && pConstraint->iColumn==0 - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH - ){ - iPlan |= 1; - pIdxInfo->aConstraintUsage[i].argvIndex = 1; - pIdxInfo->aConstraintUsage[i].omit = 1; - } - if( (iPlan & 2)==0 - && pConstraint->iColumn==1 - && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT - || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) - ){ - iPlan |= 2; - iDistTerm = i; - } - if( (iPlan & 4)==0 - && pConstraint->iColumn==2 - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 4; - pIdxInfo->aConstraintUsage[i].omit = 1; - iRulesetTerm = i; - } - } - if( iPlan & 2 ){ - pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = 1+((iPlan&1)!=0); - } - if( iPlan & 4 ){ - int idx = 1; - if( iPlan & 1 ) idx++; - if( iPlan & 2 ) idx++; - pIdxInfo->aConstraintUsage[iRulesetTerm].argvIndex = idx; - } - pIdxInfo->idxNum = iPlan; - if( pIdxInfo->nOrderBy==1 - && pIdxInfo->aOrderBy[0].iColumn==1 - && pIdxInfo->aOrderBy[0].desc==0 - ){ - pIdxInfo->orderByConsumed = 1; - } - pIdxInfo->estimatedCost = (double)10000; - - return SQLITE_OK; -} - -/* -** A virtual table module that implements the "fuzzer". -*/ -static sqlite3_module fuzzerModule = { - 0, /* iVersion */ - fuzzerConnect, - fuzzerConnect, - fuzzerBestIndex, - fuzzerDisconnect, - fuzzerDisconnect, - fuzzerOpen, /* xOpen - open a cursor */ - fuzzerClose, /* xClose - close a cursor */ - fuzzerFilter, /* xFilter - configure scan constraints */ - fuzzerNext, /* xNext - advance a cursor */ - fuzzerEof, /* xEof - check for end of scan */ - fuzzerColumn, /* xColumn - read data */ - fuzzerRowid, /* xRowid - read data */ - 0, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindMethod */ - 0, /* xRename */ -}; - -#endif /* SQLITE_OMIT_VIRTUALTABLE */ - - -/* -** Register the fuzzer virtual table -*/ -int fuzzer_register(sqlite3 *db){ - int rc = SQLITE_OK; -#ifndef SQLITE_OMIT_VIRTUALTABLE - rc = sqlite3_create_module(db, "fuzzer", &fuzzerModule, 0); -#endif - return rc; -} - -#ifdef SQLITE_TEST -#include -/* -** Decode a pointer to an sqlite3 object. -*/ -extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); - -/* -** Register the echo virtual table module. -*/ -static int register_fuzzer_module( - ClientData clientData, /* Pointer to sqlite3_enable_XXX function */ - Tcl_Interp *interp, /* The TCL interpreter that invoked this command */ - int objc, /* Number of arguments */ - Tcl_Obj *CONST objv[] /* Command arguments */ -){ - sqlite3 *db; - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - getDbPointer(interp, Tcl_GetString(objv[1]), &db); - fuzzer_register(db); - return TCL_OK; -} - - -/* -** Register commands with the TCL interpreter. -*/ -int Sqlitetestfuzzer_Init(Tcl_Interp *interp){ - static struct { - char *zName; - Tcl_ObjCmdProc *xProc; - void *clientData; - } aObjCmd[] = { - { "register_fuzzer_module", register_fuzzer_module, 0 }, - }; - int i; - for(i=0; i "SQLITE_ERROR" +** sqlite3ErrName(1) -> "SQLITE_ERROR" */ -const char *sqlite3TestErrorName(int); +extern const char *sqlite3ErrName(int); /* ** Transform pointers to text and back again */ static void pointerToText(void *p, char *z){ @@ -1070,11 +1070,11 @@ int rc; int sz, cnt; sqlite3 *db; int bufid; static char azBuf[2][10000]; - int getDbPointer(Tcl_Interp*, const char*, sqlite3**); + extern int getDbPointer(Tcl_Interp*, const char*, sqlite3**); if( objc!=5 ){ Tcl_WrongNumArgs(interp, 1, objv, "BUFID SIZE COUNT"); return TCL_ERROR; } if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; @@ -1124,11 +1124,11 @@ }else{ zBuf = realloc(zBuf, nByte); rc = sqlite3_config(SQLITE_CONFIG_HEAP, zBuf, nByte, nMinAlloc); } - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* ** Usage: sqlite3_config_error [DB] @@ -1141,11 +1141,11 @@ Tcl_Interp *interp, int objc, Tcl_Obj *CONST objv[] ){ sqlite3 *db; - int getDbPointer(Tcl_Interp*, const char*, sqlite3**); + extern int getDbPointer(Tcl_Interp*, const char*, sqlite3**); if( objc!=2 && objc!=1 ){ Tcl_WrongNumArgs(interp, 1, objv, "[DB]"); return TCL_ERROR; } @@ -1190,11 +1190,11 @@ if( Tcl_GetBooleanFromObj(interp, objv[1], &bOpenUri) ){ return TCL_ERROR; } rc = sqlite3_config(SQLITE_CONFIG_URI, bOpenUri); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* @@ -1219,11 +1219,11 @@ if( Tcl_GetBooleanFromObj(interp, objv[1], &bUseCis) ){ return TCL_ERROR; } rc = sqlite3_config(SQLITE_CONFIG_COVERING_INDEX_SCAN, bUseCis); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* @@ -1333,11 +1333,11 @@ ){ int rc, iValue, mxValue; int i, op, resetFlag; const char *zOpName; sqlite3 *db; - int getDbPointer(Tcl_Interp*, const char*, sqlite3**); + extern int getDbPointer(Tcl_Interp*, const char*, sqlite3**); static const struct { const char *zName; int op; } aOp[] = { { "LOOKASIDE_USED", SQLITE_DBSTATUS_LOOKASIDE_USED }, @@ -1399,11 +1399,11 @@ } if( TCL_OK!=Tcl_GetBooleanFromObj(interp, objv[1], &isInstall) ){ return TCL_ERROR; } rc = faultsimInstall(isInstall); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* ** sqlite3_install_memsys3 @@ -1417,11 +1417,11 @@ int rc = SQLITE_MISUSE; #ifdef SQLITE_ENABLE_MEMSYS3 const sqlite3_mem_methods *sqlite3MemGetMemsys3(void); rc = sqlite3_config(SQLITE_CONFIG_MALLOC, sqlite3MemGetMemsys3()); #endif - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } static int test_vfs_oom_test( void * clientData, Index: src/test_multiplex.c ================================================================== --- src/test_multiplex.c +++ src/test_multiplex.c @@ -58,11 +58,11 @@ #endif #include "sqlite3ext.h" /* ** These should be defined to be the same as the values in -** sqliteInt.h. They are defined seperately here so that +** sqliteInt.h. They are defined separately here so that ** the multiplex VFS shim can be built as a loadable ** module. */ #define UNUSED_PARAMETER(x) (void)(x) #define MAX_PAGE_SIZE 0x10000 @@ -1181,11 +1181,11 @@ } /***************************** Test Code ***********************************/ #ifdef SQLITE_TEST #include -extern const char *sqlite3TestErrorName(int); +extern const char *sqlite3ErrName(int); /* ** tclcmd: sqlite3_multiplex_initialize NAME MAKEDEFAULT */ @@ -1210,11 +1210,11 @@ if( Tcl_GetBooleanFromObj(interp, objv[2], &makeDefault) ) return TCL_ERROR; if( zName[0]=='\0' ) zName = 0; /* Call sqlite3_multiplex_initialize() */ rc = sqlite3_multiplex_initialize(zName, makeDefault); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* @@ -1235,11 +1235,11 @@ return TCL_ERROR; } /* Call sqlite3_multiplex_shutdown() */ rc = sqlite3_multiplex_shutdown(); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* @@ -1353,11 +1353,11 @@ Tcl_WrongNumArgs(interp, 4, objv, "SUB-COMMAND"); return TCL_ERROR; } rc = sqlite3_file_control(db, Tcl_GetString(objv[2]), aSub[idx].op, pArg); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return (rc==SQLITE_OK) ? TCL_OK : TCL_ERROR; } /* ** This routine registers the custom TCL commands defined in this Index: src/test_mutex.c ================================================================== --- src/test_mutex.c +++ src/test_mutex.c @@ -17,12 +17,12 @@ #include "sqliteInt.h" #include #include #include -/* defined in test1.c */ -const char *sqlite3TestErrorName(int); +/* defined in main.c */ +extern const char *sqlite3ErrName(int); /* A countable mutex */ struct sqlite3_mutex { sqlite3_mutex *pReal; int eType; @@ -146,11 +146,11 @@ Tcl_WrongNumArgs(interp, 1, objv, ""); return TCL_ERROR; } rc = sqlite3_shutdown(); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* ** sqlite3_initialize @@ -167,11 +167,11 @@ Tcl_WrongNumArgs(interp, 1, objv, ""); return TCL_ERROR; } rc = sqlite3_initialize(); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* ** install_mutex_counters BOOLEAN @@ -228,11 +228,11 @@ if( rc==SQLITE_OK ){ g.isInstalled = isInstall; } - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } /* ** read_mutex_counters @@ -352,11 +352,11 @@ }else{ i = aOpt[i].iValue; } rc = sqlite3_config(i); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_VOLATILE); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_VOLATILE); return TCL_OK; } static sqlite3 *getDbPointer(Tcl_Interp *pInterp, Tcl_Obj *pObj){ sqlite3 *db; Index: src/test_quota.c ================================================================== --- src/test_quota.c +++ src/test_quota.c @@ -1323,11 +1323,11 @@ struct TclQuotaCallback { Tcl_Interp *interp; /* Interpreter in which to run the script */ Tcl_Obj *pScript; /* Script to be run */ }; -extern const char *sqlite3TestErrorName(int); +extern const char *sqlite3ErrName(int); /* ** This is the callback from a quota-over-limit. */ @@ -1405,11 +1405,11 @@ if( Tcl_GetBooleanFromObj(interp, objv[2], &makeDefault) ) return TCL_ERROR; if( zName[0]=='\0' ) zName = 0; /* Call sqlite3_quota_initialize() */ rc = sqlite3_quota_initialize(zName, makeDefault); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* @@ -1428,11 +1428,11 @@ return TCL_ERROR; } /* Call sqlite3_quota_shutdown() */ rc = sqlite3_quota_shutdown(); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* @@ -1483,11 +1483,11 @@ } /* Invoke sqlite3_quota_set() */ rc = sqlite3_quota_set(zPattern, iLimit, xCallback, (void*)p, xDestroy); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* ** tclcmd: sqlite3_quota_file FILENAME @@ -1509,11 +1509,11 @@ zFilename = Tcl_GetString(objv[1]); /* Invoke sqlite3_quota_file() */ rc = sqlite3_quota_file(zFilename); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); return TCL_OK; } /* ** tclcmd: sqlite3_quota_dump DELETED src/test_regexp.c Index: src/test_regexp.c ================================================================== --- src/test_regexp.c +++ src/test_regexp.c @@ -1,770 +0,0 @@ -/* -** 2012-11-13 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** The code in this file implements a compact but reasonably -** efficient regular-expression matcher for posix extended regular -** expressions against UTF8 text. The following syntax is supported: -** -** X* zero or more occurrences of X -** X+ one or more occurrences of X -** X? zero or one occurrences of X -** X{p,q} between p and q occurrences of X -** (X) match X -** X|Y X or Y -** ^X X occurring at the beginning of the string -** X$ X occurring at the end of the string -** . Match any single character -** \c Character c where c is one of \{}()[]|*+?. -** \c C-language escapes for c in afnrtv. ex: \t or \n -** \uXXXX Where XXXX is exactly 4 hex digits, unicode value XXXX -** \xXX Where XX is exactly 2 hex digits, unicode value XX -** [abc] Any single character from the set abc -** [^abc] Any single character not in the set abc -** [a-z] Any single character in the range a-z -** [^a-z] Any single character not in the range a-z -** \b Word boundary -** \w Word character. [A-Za-z0-9_] -** \W Non-word character -** \d Digit -** \D Non-digit -** \s Whitespace character -** \S Non-whitespace character -** -** A nondeterministic finite automaton (NFA) is used for matching, so the -** performance is bounded by O(N*M) where N is the size of the regular -** expression and M is the size of the input string. The matcher never -** exhibits exponential behavior. Note that the X{p,q} operator expands -** to p copies of X following by q-p copies of X? and that the size of the -** regular expression in the O(N*M) performance bound is computed after -** this expansion. -*/ -#include -#include -#include "sqlite3.h" - -/* The end-of-input character */ -#define RE_EOF 0 /* End of input */ - -/* The NFA is implemented as sequence of opcodes taken from the following -** set. Each opcode has a single integer argument. -*/ -#define RE_OP_MATCH 1 /* Match the one character in the argument */ -#define RE_OP_ANY 2 /* Match any one character. (Implements ".") */ -#define RE_OP_ANYSTAR 3 /* Special optimized version of .* */ -#define RE_OP_FORK 4 /* Continue to both next and opcode at iArg */ -#define RE_OP_GOTO 5 /* Jump to opcode at iArg */ -#define RE_OP_ACCEPT 6 /* Halt and indicate a successful match */ -#define RE_OP_CC_INC 7 /* Beginning of a [...] character class */ -#define RE_OP_CC_EXC 8 /* Beginning of a [^...] character class */ -#define RE_OP_CC_VALUE 9 /* Single value in a character class */ -#define RE_OP_CC_RANGE 10 /* Range of values in a character class */ -#define RE_OP_WORD 11 /* Perl word character [A-Za-z0-9_] */ -#define RE_OP_NOTWORD 12 /* Not a perl word character */ -#define RE_OP_DIGIT 13 /* digit: [0-9] */ -#define RE_OP_NOTDIGIT 14 /* Not a digit */ -#define RE_OP_SPACE 15 /* space: [ \t\n\r\v\f] */ -#define RE_OP_NOTSPACE 16 /* Not a digit */ -#define RE_OP_BOUNDARY 17 /* Boundary between word and non-word */ - -/* Each opcode is a "state" in the NFA */ -typedef unsigned short ReStateNumber; - -/* Because this is an NFA and not a DFA, multiple states can be active at -** once. An instance of the following object records all active states in -** the NFA. The implementation is optimized for the common case where the -** number of actives states is small. -*/ -typedef struct ReStateSet { - unsigned nState; /* Number of current states */ - ReStateNumber *aState; /* Current states */ -} ReStateSet; - -/* An input string read one character at a time. -*/ -typedef struct ReInput ReInput; -struct ReInput { - const unsigned char *z; /* All text */ - int i; /* Next byte to read */ - int mx; /* EOF when i>=mx */ -}; - -/* A compiled NFA (or an NFA that is in the process of being compiled) is -** an instance of the following object. -*/ -typedef struct ReCompiled ReCompiled; -struct ReCompiled { - ReInput sIn; /* Regular expression text */ - const char *zErr; /* Error message to return */ - char *aOp; /* Operators for the virtual machine */ - int *aArg; /* Arguments to each operator */ - unsigned (*xNextChar)(ReInput*); /* Next character function */ - unsigned char zInit[12]; /* Initial text to match */ - int nInit; /* Number of characters in zInit */ - unsigned nState; /* Number of entries in aOp[] and aArg[] */ - unsigned nAlloc; /* Slots allocated for aOp[] and aArg[] */ -}; - -/* Add a state to the given state set if it is not already there */ -static void re_add_state(ReStateSet *pSet, int newState){ - unsigned i; - for(i=0; inState; i++) if( pSet->aState[i]==newState ) return; - pSet->aState[pSet->nState++] = newState; -} - -/* Extract the next unicode character from *pzIn and return it. Advance -** *pzIn to the first byte past the end of the character returned. To -** be clear: this routine converts utf8 to unicode. This routine is -** optimized for the common case where the next character is a single byte. -*/ -static unsigned re_next_char(ReInput *p){ - unsigned c; - if( p->i>=p->mx ) return 0; - c = p->z[p->i++]; - if( c>=0x80 ){ - if( (c&0xe0)==0xc0 && p->imx && (p->z[p->i]&0xc0)==0x80 ){ - c = (c&0x1f)<<6 | (p->z[p->i++]&0x3f); - if( c<0x80 ) c = 0xfffd; - }else if( (c&0xf0)==0xe0 && p->i+1mx && (p->z[p->i]&0xc0)==0x80 - && (p->z[p->i+1]&0xc0)==0x80 ){ - c = (c&0x0f)<<12 | ((p->z[p->i]&0x3f)<<6) | (p->z[p->i+1]&0x3f); - p->i += 2; - if( c<=0x3ff || (c>=0xd800 && c<=0xdfff) ) c = 0xfffd; - }else if( (c&0xf8)==0xf0 && p->i+3mx && (p->z[p->i]&0xc0)==0x80 - && (p->z[p->i+1]&0xc0)==0x80 && (p->z[p->i+2]&0xc0)==0x80 ){ - c = (c&0x07)<<18 | ((p->z[p->i]&0x3f)<<12) | ((p->z[p->i+1]&0x3f)<<6) - | (p->z[p->i+2]&0x3f); - p->i += 3; - if( c<=0xffff || c>0x10ffff ) c = 0xfffd; - }else{ - c = 0xfffd; - } - } - return c; -} -static unsigned re_next_char_nocase(ReInput *p){ - unsigned c = re_next_char(p); - if( c>='A' && c<='Z' ) c += 'a' - 'A'; - return c; -} - -/* Return true if c is a perl "word" character: [A-Za-z0-9_] */ -static int re_word_char(int c){ - return (c>='0' && c<='9') || (c>='a' && c<='z') - || (c>='A' && c<='Z') || c=='_'; -} - -/* Return true if c is a "digit" character: [0-9] */ -static int re_digit_char(int c){ - return (c>='0' && c<='9'); -} - -/* Return true if c is a perl "space" character: [ \t\r\n\v\f] */ -static int re_space_char(int c){ - return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f'; -} - -/* Run a compiled regular expression on the zero-terminated input -** string zIn[]. Return true on a match and false if there is no match. -*/ -int re_match(ReCompiled *pRe, const unsigned char *zIn, int nIn){ - ReStateSet aStateSet[2], *pThis, *pNext; - ReStateNumber aSpace[100]; - ReStateNumber *pToFree; - unsigned int i = 0; - unsigned int iSwap = 0; - int c = RE_EOF+1; - int cPrev = 0; - int rc = 0; - ReInput in; - - in.z = zIn; - in.i = 0; - in.mx = nIn>=0 ? nIn : strlen((char const*)zIn); - - /* Look for the initial prefix match, if there is one. */ - if( pRe->nInit ){ - unsigned char x = pRe->zInit[0]; - while( in.i+pRe->nInit<=in.mx - && (zIn[in.i]!=x || - strncmp((const char*)zIn+in.i, (const char*)pRe->zInit, pRe->nInit)!=0) - ){ - in.i++; - } - if( in.i+pRe->nInit>in.mx ) return 0; - } - - if( pRe->nState<=(sizeof(aSpace)/(sizeof(aSpace[0])*2)) ){ - pToFree = 0; - aStateSet[0].aState = aSpace; - }else{ - pToFree = sqlite3_malloc( sizeof(ReStateNumber)*2*pRe->nState ); - if( pToFree==0 ) return -1; - aStateSet[0].aState = pToFree; - } - aStateSet[1].aState = &aStateSet[0].aState[pRe->nState]; - pNext = &aStateSet[1]; - pNext->nState = 0; - re_add_state(pNext, 0); - while( c!=RE_EOF && pNext->nState>0 ){ - cPrev = c; - c = pRe->xNextChar(&in); - pThis = pNext; - pNext = &aStateSet[iSwap]; - iSwap = 1 - iSwap; - pNext->nState = 0; - for(i=0; inState; i++){ - int x = pThis->aState[i]; - switch( pRe->aOp[x] ){ - case RE_OP_MATCH: { - if( pRe->aArg[x]==c ) re_add_state(pNext, x+1); - break; - } - case RE_OP_ANY: { - re_add_state(pNext, x+1); - break; - } - case RE_OP_WORD: { - if( re_word_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_NOTWORD: { - if( !re_word_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_DIGIT: { - if( re_digit_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_NOTDIGIT: { - if( !re_digit_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_SPACE: { - if( re_space_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_NOTSPACE: { - if( !re_space_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_BOUNDARY: { - if( re_word_char(c)!=re_word_char(cPrev) ) re_add_state(pThis, x+1); - break; - } - case RE_OP_ANYSTAR: { - re_add_state(pNext, x); - re_add_state(pThis, x+1); - break; - } - case RE_OP_FORK: { - re_add_state(pThis, x+pRe->aArg[x]); - re_add_state(pThis, x+1); - break; - } - case RE_OP_GOTO: { - re_add_state(pThis, x+pRe->aArg[x]); - break; - } - case RE_OP_ACCEPT: { - rc = 1; - goto re_match_end; - } - case RE_OP_CC_INC: - case RE_OP_CC_EXC: { - int j = 1; - int n = pRe->aArg[x]; - int hit = 0; - for(j=1; j>0 && jaOp[x+j]==RE_OP_CC_VALUE ){ - if( pRe->aArg[x+j]==c ){ - hit = 1; - j = -1; - } - }else{ - if( pRe->aArg[x+j]<=c && pRe->aArg[x+j+1]>=c ){ - hit = 1; - j = -1; - }else{ - j++; - } - } - } - if( pRe->aOp[x]==RE_OP_CC_EXC ) hit = !hit; - if( hit ) re_add_state(pNext, x+n); - break; - } - } - } - } - for(i=0; inState; i++){ - if( pRe->aOp[pNext->aState[i]]==RE_OP_ACCEPT ){ rc = 1; break; } - } -re_match_end: - sqlite3_free(pToFree); - return rc; -} - -/* Resize the opcode and argument arrays for an RE under construction. -*/ -static int re_resize(ReCompiled *p, int N){ - char *aOp; - int *aArg; - aOp = sqlite3_realloc(p->aOp, N*sizeof(p->aOp[0])); - if( aOp==0 ) return 1; - p->aOp = aOp; - aArg = sqlite3_realloc(p->aArg, N*sizeof(p->aArg[0])); - if( aArg==0 ) return 1; - p->aArg = aArg; - p->nAlloc = N; - return 0; -} - -/* Insert a new opcode and argument into an RE under construction. The -** insertion point is just prior to existing opcode iBefore. -*/ -static int re_insert(ReCompiled *p, int iBefore, int op, int arg){ - int i; - if( p->nAlloc<=p->nState && re_resize(p, p->nAlloc*2) ) return 0; - for(i=p->nState; i>iBefore; i--){ - p->aOp[i] = p->aOp[i-1]; - p->aArg[i] = p->aArg[i-1]; - } - p->nState++; - p->aOp[iBefore] = op; - p->aArg[iBefore] = arg; - return iBefore; -} - -/* Append a new opcode and argument to the end of the RE under construction. -*/ -static int re_append(ReCompiled *p, int op, int arg){ - return re_insert(p, p->nState, op, arg); -} - -/* Make a copy of N opcodes starting at iStart onto the end of the RE -** under construction. -*/ -static void re_copy(ReCompiled *p, int iStart, int N){ - if( p->nState+N>=p->nAlloc && re_resize(p, p->nAlloc*2+N) ) return; - memcpy(&p->aOp[p->nState], &p->aOp[iStart], N*sizeof(p->aOp[0])); - memcpy(&p->aArg[p->nState], &p->aArg[iStart], N*sizeof(p->aArg[0])); - p->nState += N; -} - -/* Return true if c is a hexadecimal digit character: [0-9a-fA-F] -** If c is a hex digit, also set *pV = (*pV)*16 + valueof(c). If -** c is not a hex digit *pV is unchanged. -*/ -static int re_hex(int c, int *pV){ - if( c>='0' && c<='9' ){ - c -= '0'; - }else if( c>='a' && c<='f' ){ - c -= 'a' - 10; - }else if( c>='A' && c<='F' ){ - c -= 'A' - 10; - }else{ - return 0; - } - *pV = (*pV)*16 + (c & 0xff); - return 1; -} - -/* A backslash character has been seen, read the next character and -** return its interpretation. -*/ -static unsigned re_esc_char(ReCompiled *p){ - static const char zEsc[] = "afnrtv\\()*.+?[$^{|}]"; - static const char zTrans[] = "\a\f\n\r\t\v"; - int i, v = 0; - char c; - if( p->sIn.i>=p->sIn.mx ) return 0; - c = p->sIn.z[p->sIn.i]; - if( c=='u' && p->sIn.i+4sIn.mx ){ - const unsigned char *zIn = p->sIn.z + p->sIn.i; - if( re_hex(zIn[1],&v) - && re_hex(zIn[2],&v) - && re_hex(zIn[3],&v) - && re_hex(zIn[4],&v) - ){ - p->sIn.i += 5; - return v; - } - } - if( c=='x' && p->sIn.i+2sIn.mx ){ - const unsigned char *zIn = p->sIn.z + p->sIn.i; - if( re_hex(zIn[1],&v) - && re_hex(zIn[2],&v) - ){ - p->sIn.i += 3; - return v; - } - } - for(i=0; zEsc[i] && zEsc[i]!=c; i++){} - if( zEsc[i] ){ - if( i<6 ) c = zTrans[i]; - p->sIn.i++; - }else{ - p->zErr = "unknown \\ escape"; - } - return c; -} - -/* Forward declaration */ -static const char *re_subcompile_string(ReCompiled*); - -/* Peek at the next byte of input */ -static unsigned char rePeek(ReCompiled *p){ - return p->sIn.isIn.mx ? p->sIn.z[p->sIn.i] : 0; -} - -/* Compile RE text into a sequence of opcodes. Continue up to the -** first unmatched ")" character, then return. If an error is found, -** return a pointer to the error message string. -*/ -static const char *re_subcompile_re(ReCompiled *p){ - const char *zErr; - int iStart, iEnd, iGoto; - iStart = p->nState; - zErr = re_subcompile_string(p); - if( zErr ) return zErr; - while( rePeek(p)=='|' ){ - iEnd = p->nState; - re_insert(p, iStart, RE_OP_FORK, iEnd + 2 - iStart); - iGoto = re_append(p, RE_OP_GOTO, 0); - p->sIn.i++; - zErr = re_subcompile_string(p); - if( zErr ) return zErr; - p->aArg[iGoto] = p->nState - iGoto; - } - return 0; -} - -/* Compile an element of regular expression text (anything that can be -** an operand to the "|" operator). Return NULL on success or a pointer -** to the error message if there is a problem. -*/ -static const char *re_subcompile_string(ReCompiled *p){ - int iPrev = -1; - int iStart; - unsigned c; - const char *zErr; - while( (c = p->xNextChar(&p->sIn))!=0 ){ - iStart = p->nState; - switch( c ){ - case '|': - case '$': - case ')': { - p->sIn.i--; - return 0; - } - case '(': { - zErr = re_subcompile_re(p); - if( zErr ) return zErr; - if( rePeek(p)!=')' ) return "unmatched '('"; - p->sIn.i++; - break; - } - case '.': { - if( rePeek(p)=='*' ){ - re_append(p, RE_OP_ANYSTAR, 0); - p->sIn.i++; - }else{ - re_append(p, RE_OP_ANY, 0); - } - break; - } - case '*': { - if( iPrev<0 ) return "'*' without operand"; - re_insert(p, iPrev, RE_OP_GOTO, p->nState - iPrev + 1); - re_append(p, RE_OP_FORK, iPrev - p->nState + 1); - break; - } - case '+': { - if( iPrev<0 ) return "'+' without operand"; - re_append(p, RE_OP_FORK, iPrev - p->nState); - break; - } - case '?': { - if( iPrev<0 ) return "'?' without operand"; - re_insert(p, iPrev, RE_OP_FORK, p->nState - iPrev+1); - break; - } - case '{': { - int m = 0, n = 0; - int sz, j; - if( iPrev<0 ) return "'{m,n}' without operand"; - while( (c=rePeek(p))>='0' && c<='9' ){ m = m*10 + c - '0'; p->sIn.i++; } - n = m; - if( c==',' ){ - p->sIn.i++; - n = 0; - while( (c=rePeek(p))>='0' && c<='9' ){ n = n*10 + c-'0'; p->sIn.i++; } - } - if( c!='}' ) return "unmatched '{'"; - if( n>0 && nsIn.i++; - sz = p->nState - iPrev; - if( m==0 ){ - if( n==0 ) return "both m and n are zero in '{m,n}'"; - re_insert(p, iPrev, RE_OP_FORK, sz+1); - n--; - }else{ - for(j=1; j0 ){ - re_append(p, RE_OP_FORK, -sz); - } - break; - } - case '[': { - int iFirst = p->nState; - if( rePeek(p)=='^' ){ - re_append(p, RE_OP_CC_EXC, 0); - p->sIn.i++; - }else{ - re_append(p, RE_OP_CC_INC, 0); - } - while( (c = p->xNextChar(&p->sIn))!=0 ){ - if( c=='[' && rePeek(p)==':' ){ - return "POSIX character classes not supported"; - } - if( c=='\\' ) c = re_esc_char(p); - if( rePeek(p)=='-' ){ - re_append(p, RE_OP_CC_RANGE, c); - p->sIn.i++; - c = p->xNextChar(&p->sIn); - if( c=='\\' ) c = re_esc_char(p); - re_append(p, RE_OP_CC_RANGE, c); - }else{ - re_append(p, RE_OP_CC_VALUE, c); - } - if( rePeek(p)==']' ){ p->sIn.i++; break; } - } - if( c==0 ) return "unclosed '['"; - p->aArg[iFirst] = p->nState - iFirst; - break; - } - case '\\': { - int specialOp = 0; - switch( rePeek(p) ){ - case 'b': specialOp = RE_OP_BOUNDARY; break; - case 'd': specialOp = RE_OP_DIGIT; break; - case 'D': specialOp = RE_OP_NOTDIGIT; break; - case 's': specialOp = RE_OP_SPACE; break; - case 'S': specialOp = RE_OP_NOTSPACE; break; - case 'w': specialOp = RE_OP_WORD; break; - case 'W': specialOp = RE_OP_NOTWORD; break; - } - if( specialOp ){ - p->sIn.i++; - re_append(p, specialOp, 0); - }else{ - c = re_esc_char(p); - re_append(p, RE_OP_MATCH, c); - } - break; - } - default: { - re_append(p, RE_OP_MATCH, c); - break; - } - } - iPrev = iStart; - } - return 0; -} - -/* Free and reclaim all the memory used by a previously compiled -** regular expression. Applications should invoke this routine once -** for every call to re_compile() to avoid memory leaks. -*/ -void re_free(ReCompiled *pRe){ - if( pRe ){ - sqlite3_free(pRe->aOp); - sqlite3_free(pRe->aArg); - sqlite3_free(pRe); - } -} - -/* -** Compile a textual regular expression in zIn[] into a compiled regular -** expression suitable for us by re_match() and return a pointer to the -** compiled regular expression in *ppRe. Return NULL on success or an -** error message if something goes wrong. -*/ -const char *re_compile(ReCompiled **ppRe, const char *zIn, int noCase){ - ReCompiled *pRe; - const char *zErr; - int i, j; - - *ppRe = 0; - pRe = sqlite3_malloc( sizeof(*pRe) ); - if( pRe==0 ){ - return "out of memory"; - } - memset(pRe, 0, sizeof(*pRe)); - pRe->xNextChar = noCase ? re_next_char_nocase : re_next_char; - if( re_resize(pRe, 30) ){ - re_free(pRe); - return "out of memory"; - } - if( zIn[0]=='^' ){ - zIn++; - }else{ - re_append(pRe, RE_OP_ANYSTAR, 0); - } - pRe->sIn.z = (unsigned char*)zIn; - pRe->sIn.i = 0; - pRe->sIn.mx = strlen(zIn); - zErr = re_subcompile_re(pRe); - if( zErr ){ - re_free(pRe); - return zErr; - } - if( rePeek(pRe)=='$' && pRe->sIn.i+1>=pRe->sIn.mx ){ - re_append(pRe, RE_OP_MATCH, RE_EOF); - re_append(pRe, RE_OP_ACCEPT, 0); - *ppRe = pRe; - }else if( pRe->sIn.i>=pRe->sIn.mx ){ - re_append(pRe, RE_OP_ACCEPT, 0); - *ppRe = pRe; - }else{ - re_free(pRe); - return "unrecognized character"; - } - - /* The following is a performance optimization. If the regex begins with - ** ".*" (if the input regex lacks an initial "^") and afterwards there are - ** one or more matching characters, enter those matching characters into - ** zInit[]. The re_match() routine can then search ahead in the input - ** string looking for the initial match without having to run the whole - ** regex engine over the string. Do not worry able trying to match - ** unicode characters beyond plane 0 - those are very rare and this is - ** just an optimization. */ - if( pRe->aOp[0]==RE_OP_ANYSTAR ){ - for(j=0, i=1; jzInit)-2 && pRe->aOp[i]==RE_OP_MATCH; i++){ - unsigned x = pRe->aArg[i]; - if( x<=127 ){ - pRe->zInit[j++] = x; - }else if( x<=0xfff ){ - pRe->zInit[j++] = 0xc0 | (x>>6); - pRe->zInit[j++] = 0x80 | (x&0x3f); - }else if( x<=0xffff ){ - pRe->zInit[j++] = 0xd0 | (x>>12); - pRe->zInit[j++] = 0x80 | ((x>>6)&0x3f); - pRe->zInit[j++] = 0x80 | (x&0x3f); - }else{ - break; - } - } - if( j>0 && pRe->zInit[j-1]==0 ) j--; - pRe->nInit = j; - } - return pRe->zErr; -} - -/* -** Implementation of the regexp() SQL function. This function implements -** the build-in REGEXP operator. The first argument to the function is the -** pattern and the second argument is the string. So, the SQL statements: -** -** A REGEXP B -** -** is implemented as regexp(B,A). -*/ -static void re_sql_func( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - ReCompiled *pRe; /* Compiled regular expression */ - const char *zPattern; /* The regular expression */ - const unsigned char *zStr;/* String being searched */ - const char *zErr; /* Compile error message */ - - pRe = sqlite3_get_auxdata(context, 0); - if( pRe==0 ){ - zPattern = (const char*)sqlite3_value_text(argv[0]); - if( zPattern==0 ) return; - zErr = re_compile(&pRe, zPattern, 0); - if( zErr ){ - re_free(pRe); - sqlite3_result_error(context, zErr, -1); - return; - } - if( pRe==0 ){ - sqlite3_result_error_nomem(context); - return; - } - sqlite3_set_auxdata(context, 0, pRe, (void(*)(void*))re_free); - } - zStr = (const unsigned char*)sqlite3_value_text(argv[1]); - if( zStr!=0 ){ - sqlite3_result_int(context, re_match(pRe, zStr, -1)); - } -} - -/* -** Invoke this routine in order to install the REGEXP function in an -** SQLite database connection. -** -** Use: -** -** sqlite3_auto_extension(sqlite3_add_regexp_func); -** -** to cause this extension to be automatically loaded into each new -** database connection. -*/ -int sqlite3_add_regexp_func(sqlite3 *db){ - return sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, 0, - re_sql_func, 0, 0); -} - - -/***************************** Test Code ***********************************/ -#ifdef SQLITE_TEST -#include -extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); - -/* Implementation of the TCL command: -** -** sqlite3_add_regexp_func $DB -*/ -static int tclSqlite3AddRegexpFunc( - void * clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ - sqlite3 *db; - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; - sqlite3_add_regexp_func(db); - return TCL_OK; -} - -/* Register the sqlite3_add_regexp_func TCL command with the TCL interpreter. -*/ -int Sqlitetestregexp_Init(Tcl_Interp *interp){ - Tcl_CreateObjCommand(interp, "sqlite3_add_regexp_func", - tclSqlite3AddRegexpFunc, 0, 0); - return TCL_OK; -} -#endif /* SQLITE_TEST */ -/**************************** End Of Test Code *******************************/ Index: src/test_rtree.c ================================================================== --- src/test_rtree.c +++ src/test_rtree.c @@ -252,21 +252,21 @@ UNUSED_PARAMETER(interp); UNUSED_PARAMETER(objc); UNUSED_PARAMETER(objv); #else extern int getDbPointer(Tcl_Interp*, const char*, sqlite3**); - extern const char *sqlite3TestErrorName(int); + extern const char *sqlite3ErrName(int); sqlite3 *db; int rc; if( objc!=2 ){ Tcl_WrongNumArgs(interp, 1, objv, "DB"); return TCL_ERROR; } if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; rc = sqlite3_rtree_geometry_callback(db, "cube", cube_geom, (void *)&gHere); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); #endif return TCL_OK; } static int register_circle_geom( @@ -280,25 +280,25 @@ UNUSED_PARAMETER(interp); UNUSED_PARAMETER(objc); UNUSED_PARAMETER(objv); #else extern int getDbPointer(Tcl_Interp*, const char*, sqlite3**); - extern const char *sqlite3TestErrorName(int); + extern const char *sqlite3ErrName(int); sqlite3 *db; int rc; if( objc!=2 ){ Tcl_WrongNumArgs(interp, 1, objv, "DB"); return TCL_ERROR; } if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; rc = sqlite3_rtree_geometry_callback(db, "circle", circle_geom, 0); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), TCL_STATIC); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), TCL_STATIC); #endif return TCL_OK; } int Sqlitetestrtree_Init(Tcl_Interp *interp){ Tcl_CreateObjCommand(interp, "register_cube_geom", register_cube_geom, 0, 0); Tcl_CreateObjCommand(interp, "register_circle_geom",register_circle_geom,0,0); return TCL_OK; } DELETED src/test_spellfix.c Index: src/test_spellfix.c ================================================================== --- src/test_spellfix.c +++ src/test_spellfix.c @@ -1,2847 +0,0 @@ -/* -** 2012 April 10 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** -** This module implements the spellfix1 VIRTUAL TABLE that can be used -** to search a large vocabulary for close matches. See separate -** documentation files (spellfix1.wiki and editdist3.wiki) for details. -*/ -#if SQLITE_CORE -# include "sqliteInt.h" -#else -# include -# include -# include -# include "sqlite3ext.h" -# include -# define ALWAYS(X) 1 -# define NEVER(X) 0 - typedef unsigned char u8; - typedef unsigned short u16; - SQLITE_EXTENSION_INIT1 -#endif /* !SQLITE_CORE */ -#include - -/* -** Character classes for ASCII characters: -** -** 0 '' Silent letters: H W -** 1 'A' Any vowel: A E I O U (Y) -** 2 'B' A bilabeal stop or fricative: B F P V W -** 3 'C' Other fricatives or back stops: C G J K Q S X Z -** 4 'D' Alveolar stops: D T -** 5 'H' Letter H at the beginning of a word -** 6 'L' Glide: L -** 7 'R' Semivowel: R -** 8 'M' Nasals: M N -** 9 'Y' Letter Y at the beginning of a word. -** 10 '9' Digits: 0 1 2 3 4 5 6 7 8 9 -** 11 ' ' White space -** 12 '?' Other. -*/ -#define CCLASS_SILENT 0 -#define CCLASS_VOWEL 1 -#define CCLASS_B 2 -#define CCLASS_C 3 -#define CCLASS_D 4 -#define CCLASS_H 5 -#define CCLASS_L 6 -#define CCLASS_R 7 -#define CCLASS_M 8 -#define CCLASS_Y 9 -#define CCLASS_DIGIT 10 -#define CCLASS_SPACE 11 -#define CCLASS_OTHER 12 - -/* -** The following table gives the character class for non-initial ASCII -** characters. -*/ -static const unsigned char midClass[] = { - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, - /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, - /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, - /* ' */ CCLASS_SILENT, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, - /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, - /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, - /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, - /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, - /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, - /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, - /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, - /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, - /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, - /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, - /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, - /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, - /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, - /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, - /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, - /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_VOWEL, - /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, - /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, - /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, - /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, - /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, - /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, - /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, - /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, - /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, - /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, - /* x */ CCLASS_C, /* y */ CCLASS_VOWEL, /* z */ CCLASS_C, - /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, - /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, -}; -/* -** This tables gives the character class for ASCII characters that form the -** initial character of a word. The only difference from midClass is with -** the letters H, W, and Y. -*/ -static const unsigned char initClass[] = { - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, - /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, - /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, - /* ' */ CCLASS_OTHER, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, - /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, - /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, - /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, - /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, - /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, - /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, - /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, - /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, - /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, - /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, - /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, - /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, - /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, - /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, - /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, - /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_Y, - /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, - /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, - /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, - /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, - /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, - /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, - /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, - /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, - /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, - /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, - /* x */ CCLASS_C, /* y */ CCLASS_Y, /* z */ CCLASS_C, - /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, - /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, -}; - -/* -** Mapping from the character class number (0-13) to a symbol for each -** character class. Note that initClass[] can be used to map the class -** symbol back into the class number. -*/ -static const unsigned char className[] = ".ABCDHLRMY9 ?"; - -/* -** Generate a "phonetic hash" from a string of ASCII characters -** in zIn[0..nIn-1]. -** -** * Map characters by character class as defined above. -** * Omit double-letters -** * Omit vowels beside R and L -** * Omit T when followed by CH -** * Omit W when followed by R -** * Omit D when followed by J or G -** * Omit K in KN or G in GN at the beginning of a word -** -** Space to hold the result is obtained from sqlite3_malloc() -** -** Return NULL if memory allocation fails. -*/ -static unsigned char *phoneticHash(const unsigned char *zIn, int nIn){ - unsigned char *zOut = sqlite3_malloc( nIn + 1 ); - int i; - int nOut = 0; - char cPrev = 0x77; - char cPrevX = 0x77; - const unsigned char *aClass = initClass; - - if( zOut==0 ) return 0; - if( nIn>2 ){ - switch( zIn[0] ){ - case 'g': - case 'k': { - if( zIn[1]=='n' ){ zIn++; nIn--; } - break; - } - } - } - for(i=0; i=0 ); - if( nOut==0 || c!=zOut[nOut-1] ) zOut[nOut++] = c; - } - zOut[nOut] = 0; - return zOut; -} - -/* -** This is an SQL function wrapper around phoneticHash(). See -** the description of phoneticHash() for additional information. -*/ -static void phoneticHashSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - const unsigned char *zIn; - unsigned char *zOut; - - zIn = sqlite3_value_text(argv[0]); - if( zIn==0 ) return; - zOut = phoneticHash(zIn, sqlite3_value_bytes(argv[0])); - if( zOut==0 ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); - } -} - -/* -** Return the character class number for a character given its -** context. -*/ -static char characterClass(char cPrev, char c){ - return cPrev==0 ? initClass[c&0x7f] : midClass[c&0x7f]; -} - -/* -** Return the cost of inserting or deleting character c immediately -** following character cPrev. If cPrev==0, that means c is the first -** character of the word. -*/ -static int insertOrDeleteCost(char cPrev, char c, char cNext){ - char classC = characterClass(cPrev, c); - char classCprev; - - if( classC==CCLASS_SILENT ){ - /* Insert or delete "silent" characters such as H or W */ - return 1; - } - if( cPrev==c ){ - /* Repeated characters, or miss a repeat */ - return 10; - } - if( classC==CCLASS_VOWEL && (cPrev=='r' || cNext=='r') ){ - return 20; /* Insert a vowel before or after 'r' */ - } - classCprev = characterClass(cPrev, cPrev); - if( classC==classCprev ){ - if( classC==CCLASS_VOWEL ){ - /* Remove or add a new vowel to a vowel cluster */ - return 15; - }else{ - /* Remove or add a consonant not in the same class */ - return 50; - } - } - - /* any other character insertion or deletion */ - return 100; -} - -/* -** Divide the insertion cost by this factor when appending to the -** end of the word. -*/ -#define FINAL_INS_COST_DIV 4 - -/* -** Return the cost of substituting cTo in place of cFrom assuming -** the previous character is cPrev. If cPrev==0 then cTo is the first -** character of the word. -*/ -static int substituteCost(char cPrev, char cFrom, char cTo){ - char classFrom, classTo; - if( cFrom==cTo ){ - /* Exact match */ - return 0; - } - if( cFrom==(cTo^0x20) && ((cTo>='A' && cTo<='Z') || (cTo>='a' && cTo<='z')) ){ - /* differ only in case */ - return 0; - } - classFrom = characterClass(cPrev, cFrom); - classTo = characterClass(cPrev, cTo); - if( classFrom==classTo ){ - /* Same character class */ - return 40; - } - if( classFrom>=CCLASS_B && classFrom<=CCLASS_Y - && classTo>=CCLASS_B && classTo<=CCLASS_Y ){ - /* Convert from one consonant to another, but in a different class */ - return 75; - } - /* Any other subsitution */ - return 100; -} - -/* -** Given two strings zA and zB which are pure ASCII, return the cost -** of transforming zA into zB. If zA ends with '*' assume that it is -** a prefix of zB and give only minimal penalty for extra characters -** on the end of zB. -** -** Smaller numbers mean a closer match. -** -** Negative values indicate an error: -** -1 One of the inputs is NULL -** -2 Non-ASCII characters on input -** -3 Unable to allocate memory -** -** If pnMatch is not NULL, then *pnMatch is set to the number of bytes -** of zB that matched the pattern in zA. If zA does not end with a '*', -** then this value is always the number of bytes in zB (i.e. strlen(zB)). -** If zA does end in a '*', then it is the number of bytes in the prefix -** of zB that was deemed to match zA. -*/ -static int editdist1(const char *zA, const char *zB, int *pnMatch){ - int nA, nB; /* Number of characters in zA[] and zB[] */ - int xA, xB; /* Loop counters for zA[] and zB[] */ - char cA, cB; /* Current character of zA and zB */ - char cAprev, cBprev; /* Previous character of zA and zB */ - char cAnext, cBnext; /* Next character in zA and zB */ - int d; /* North-west cost value */ - int dc = 0; /* North-west character value */ - int res; /* Final result */ - int *m; /* The cost matrix */ - char *cx; /* Corresponding character values */ - int *toFree = 0; /* Malloced space */ - int mStack[60+15]; /* Stack space to use if not too much is needed */ - int nMatch = 0; - - /* Early out if either input is NULL */ - if( zA==0 || zB==0 ) return -1; - - /* Skip any common prefix */ - while( zA[0] && zA[0]==zB[0] ){ dc = zA[0]; zA++; zB++; nMatch++; } - if( pnMatch ) *pnMatch = nMatch; - if( zA[0]==0 && zB[0]==0 ) return 0; - -#if 0 - printf("A=\"%s\" B=\"%s\" dc=%c\n", zA, zB, dc?dc:' '); -#endif - - /* Verify input strings and measure their lengths */ - for(nA=0; zA[nA]; nA++){ - if( zA[nA]&0x80 ) return -2; - } - for(nB=0; zB[nB]; nB++){ - if( zB[nB]&0x80 ) return -2; - } - - /* Special processing if either string is empty */ - if( nA==0 ){ - cBprev = dc; - for(xB=res=0; (cB = zB[xB])!=0; xB++){ - res += insertOrDeleteCost(cBprev, cB, zB[xB+1])/FINAL_INS_COST_DIV; - cBprev = cB; - } - return res; - } - if( nB==0 ){ - cAprev = dc; - for(xA=res=0; (cA = zA[xA])!=0; xA++){ - res += insertOrDeleteCost(cAprev, cA, zA[xA+1]); - cAprev = cA; - } - return res; - } - - /* A is a prefix of B */ - if( zA[0]=='*' && zA[1]==0 ) return 0; - - /* Allocate and initialize the Wagner matrix */ - if( nB<(sizeof(mStack)*4)/(sizeof(mStack[0])*5) ){ - m = mStack; - }else{ - m = toFree = sqlite3_malloc( (nB+1)*5*sizeof(m[0])/4 ); - if( m==0 ) return -3; - } - cx = (char*)&m[nB+1]; - - /* Compute the Wagner edit distance */ - m[0] = 0; - cx[0] = dc; - cBprev = dc; - for(xB=1; xB<=nB; xB++){ - cBnext = zB[xB]; - cB = zB[xB-1]; - cx[xB] = cB; - m[xB] = m[xB-1] + insertOrDeleteCost(cBprev, cB, cBnext); - cBprev = cB; - } - cAprev = dc; - for(xA=1; xA<=nA; xA++){ - int lastA = (xA==nA); - cA = zA[xA-1]; - cAnext = zA[xA]; - if( cA=='*' && lastA ) break; - d = m[0]; - dc = cx[0]; - m[0] = d + insertOrDeleteCost(cAprev, cA, cAnext); - cBprev = 0; - for(xB=1; xB<=nB; xB++){ - int totalCost, insCost, delCost, subCost, ncx; - cB = zB[xB-1]; - cBnext = zB[xB]; - - /* Cost to insert cB */ - insCost = insertOrDeleteCost(cx[xB-1], cB, cBnext); - if( lastA ) insCost /= FINAL_INS_COST_DIV; - - /* Cost to delete cA */ - delCost = insertOrDeleteCost(cx[xB], cA, cBnext); - - /* Cost to substitute cA->cB */ - subCost = substituteCost(cx[xB-1], cA, cB); - - /* Best cost */ - totalCost = insCost + m[xB-1]; - ncx = cB; - if( (delCost + m[xB])nLang; i++){ - EditDist3Cost *pCost, *pNext; - pCost = p->a[i].pCost; - while( pCost ){ - pNext = pCost->pNext; - sqlite3_free(pCost); - pCost = pNext; - } - } - sqlite3_free(p->a); - memset(p, 0, sizeof(*p)); -} -static void editDist3ConfigDelete(void *pIn){ - EditDist3Config *p = (EditDist3Config*)pIn; - editDist3ConfigClear(p); - sqlite3_free(p); -} - -/* -** Load all edit-distance weights from a table. -*/ -static int editDist3ConfigLoad( - EditDist3Config *p, /* The edit distance configuration to load */ - sqlite3 *db, /* Load from this database */ - const char *zTable /* Name of the table from which to load */ -){ - sqlite3_stmt *pStmt; - int rc, rc2; - char *zSql; - int iLangPrev = -9999; - EditDist3Lang *pLang = 0; - - zSql = sqlite3_mprintf("SELECT iLang, cFrom, cTo, iCost" - " FROM \"%w\" WHERE iLang>=0 ORDER BY iLang", zTable); - if( zSql==0 ) return SQLITE_NOMEM; - rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0); - sqlite3_free(zSql); - if( rc ) return rc; - editDist3ConfigClear(p); - while( sqlite3_step(pStmt)==SQLITE_ROW ){ - int iLang = sqlite3_column_int(pStmt, 0); - const char *zFrom = (const char*)sqlite3_column_text(pStmt, 1); - int nFrom = zFrom ? sqlite3_column_bytes(pStmt, 1) : 0; - const char *zTo = (const char*)sqlite3_column_text(pStmt, 2); - int nTo = zTo ? sqlite3_column_bytes(pStmt, 2) : 0; - int iCost = sqlite3_column_int(pStmt, 3); - - assert( zFrom!=0 || nFrom==0 ); - assert( zTo!=0 || nTo==0 ); - if( nFrom>100 || nTo>100 ) continue; - if( iCost<0 ) continue; - if( pLang==0 || iLang!=iLangPrev ){ - EditDist3Lang *pNew; - pNew = sqlite3_realloc(p->a, (p->nLang+1)*sizeof(p->a[0])); - if( pNew==0 ){ rc = SQLITE_NOMEM; break; } - p->a = pNew; - pLang = &p->a[p->nLang]; - p->nLang++; - pLang->iLang = iLang; - pLang->iInsCost = 100; - pLang->iDelCost = 100; - pLang->iSubCost = 150; - pLang->pCost = 0; - iLangPrev = iLang; - } - if( nFrom==1 && zFrom[0]=='?' && nTo==0 ){ - pLang->iDelCost = iCost; - }else if( nFrom==0 && nTo==1 && zTo[0]=='?' ){ - pLang->iInsCost = iCost; - }else if( nFrom==1 && nTo==1 && zFrom[0]=='?' && zTo[0]=='?' ){ - pLang->iSubCost = iCost; - }else{ - EditDist3Cost *pCost; - int nExtra = nFrom + nTo - 4; - if( nExtra<0 ) nExtra = 0; - pCost = sqlite3_malloc( sizeof(*pCost) + nExtra ); - if( pCost==0 ){ rc = SQLITE_NOMEM; break; } - pCost->nFrom = nFrom; - pCost->nTo = nTo; - pCost->iCost = iCost; - memcpy(pCost->a, zFrom, nFrom); - memcpy(pCost->a + nFrom, zTo, nTo); - pCost->pNext = pLang->pCost; - pLang->pCost = pCost; - } - } - rc2 = sqlite3_finalize(pStmt); - if( rc==SQLITE_OK ) rc = rc2; - return rc; -} - -/* -** Return the length (in bytes) of a utf-8 character. Or return a maximum -** of N. -*/ -static int utf8Len(unsigned char c, int N){ - int len = 1; - if( c>0x7f ){ - if( (c&0xe0)==0xc0 ){ - len = 2; - }else if( (c&0xf0)==0xe0 ){ - len = 3; - }else{ - len = 4; - } - } - if( len>N ) len = N; - return len; -} - -/* -** Return TRUE (non-zero) if the To side of the given cost matches -** the given string. -*/ -static int matchTo(EditDist3Cost *p, const char *z, int n){ - if( p->nTo>n ) return 0; - if( strncmp(p->a+p->nFrom, z, p->nTo)!=0 ) return 0; - return 1; -} - -/* -** Return TRUE (non-zero) if the From side of the given cost matches -** the given string. -*/ -static int matchFrom(EditDist3Cost *p, const char *z, int n){ - assert( p->nFrom<=n ); - if( strncmp(p->a, z, p->nFrom)!=0 ) return 0; - return 1; -} - -/* -** Return TRUE (non-zero) of the next FROM character and the next TO -** character are the same. -*/ -static int matchFromTo( - EditDist3FromString *pStr, /* Left hand string */ - int n1, /* Index of comparison character on the left */ - const char *z2, /* Right-handl comparison character */ - int n2 /* Bytes remaining in z2[] */ -){ - int b1 = pStr->a[n1].nByte; - if( b1>n2 ) return 0; - if( memcmp(pStr->z+n1, z2, b1)!=0 ) return 0; - return 1; -} - -/* -** Delete an EditDist3FromString objecct -*/ -static void editDist3FromStringDelete(EditDist3FromString *p){ - int i; - if( p ){ - for(i=0; in; i++){ - sqlite3_free(p->a[i].apDel); - sqlite3_free(p->a[i].apSubst); - } - sqlite3_free(p); - } -} - -/* -** Create a EditDist3FromString object. -*/ -static EditDist3FromString *editDist3FromStringNew( - const EditDist3Lang *pLang, - const char *z, - int n -){ - EditDist3FromString *pStr; - EditDist3Cost *p; - int i; - - if( z==0 ) return 0; - if( n<0 ) n = (int)strlen(z); - pStr = sqlite3_malloc( sizeof(*pStr) + sizeof(pStr->a[0])*n + n + 1 ); - if( pStr==0 ) return 0; - pStr->a = (EditDist3From*)&pStr[1]; - memset(pStr->a, 0, sizeof(pStr->a[0])*n); - pStr->n = n; - pStr->z = (char*)&pStr->a[n]; - memcpy(pStr->z, z, n+1); - if( n && z[n-1]=='*' ){ - pStr->isPrefix = 1; - n--; - pStr->n--; - pStr->z[n] = 0; - }else{ - pStr->isPrefix = 0; - } - - for(i=0; ia[i]; - memset(pFrom, 0, sizeof(*pFrom)); - pFrom->nByte = utf8Len((unsigned char)z[i], n-i); - for(p=pLang->pCost; p; p=p->pNext){ - EditDist3Cost **apNew; - if( i+p->nFrom>n ) continue; - if( matchFrom(p, z+i, n-i)==0 ) continue; - if( p->nTo==0 ){ - apNew = sqlite3_realloc(pFrom->apDel, - sizeof(*apNew)*(pFrom->nDel+1)); - if( apNew==0 ) break; - pFrom->apDel = apNew; - apNew[pFrom->nDel++] = p; - }else{ - apNew = sqlite3_realloc(pFrom->apSubst, - sizeof(*apNew)*(pFrom->nSubst+1)); - if( apNew==0 ) break; - pFrom->apSubst = apNew; - apNew[pFrom->nSubst++] = p; - } - } - if( p ){ - editDist3FromStringDelete(pStr); - pStr = 0; - break; - } - } - return pStr; -} - -/* -** Update entry m[i] such that it is the minimum of its current value -** and m[j]+iCost. -** -** If the iCost is 1,000,000 or greater, then consider the cost to be -** infinite and skip the update. -*/ -static void updateCost( - unsigned int *m, - int i, - int j, - int iCost -){ - assert( iCost>=0 ); - if( iCost<10000 ){ - unsigned int b = m[j] + iCost; - if( bpCost; p; p=p->pNext){ - EditDist3Cost **apNew; - if( p->nFrom>0 ) continue; - if( i2+p->nTo>n2 ) continue; - if( matchTo(p, z2+i2, n2-i2)==0 ) continue; - a2[i2].nIns++; - apNew = sqlite3_realloc(a2[i2].apIns, sizeof(*apNew)*a2[i2].nIns); - if( apNew==0 ){ - res = -1; /* Out of memory */ - goto editDist3Abort; - } - a2[i2].apIns = apNew; - a2[i2].apIns[a2[i2].nIns-1] = p; - } - } - - /* Prepare to compute the minimum edit distance */ - szRow = f.n+1; - memset(m, 0x01, (n2+1)*szRow*sizeof(m[0])); - m[0] = 0; - - /* First fill in the top-row of the matrix with FROM deletion costs */ - for(i1=0; i1iDelCost); - for(k=0; knFrom, i1, p->iCost); - } - } - - /* Fill in all subsequent rows, top-to-bottom, left-to-right */ - for(i2=0; i2iInsCost); - for(k=0; knTo), rxp, p->iCost); - } - for(i1=0; i1iDelCost); - for(k=0; knFrom, cxp, p->iCost); - } - updateCost(m, cx, cxu, pLang->iInsCost); - if( matchFromTo(&f, i1, z2+i2, n2-i2) ){ - updateCost(m, cx, cxd, 0); - } - updateCost(m, cx, cxd, pLang->iSubCost); - for(k=0; knFrom+szRow*p->nTo, cxd, p->iCost); - } - } - } - } - -#if 0 /* Enable for debugging */ - printf(" ^"); - for(i1=0; i19999 ) printf(" ****"); - else printf(" %4d", v); - } - printf("\n"); - for(i2=0; i29999 ) printf(" ****"); - else printf(" %4d", v); - } - printf("\n"); - } -#endif - - /* Free memory allocations and return the result */ - res = (int)m[szRow*(n2+1)-1]; - n = n2; - if( f.isPrefix ){ - for(i2=1; i2<=n2; i2++){ - int b = m[szRow*i2-1]; - if( b<=res ){ - res = b; - n = i2 - 1; - } - } - } - if( pnMatch ){ - int nExtra = 0; - for(k=0; knLang; i++){ - if( pConfig->a[i].iLang==iLang ) return &pConfig->a[i]; - } - return &editDist3Lang; -} - -/* -** Function: editdist3(A,B,iLang) -** editdist3(tablename) -** -** Return the cost of transforming string A into string B using edit -** weights for iLang. -** -** The second form loads edit weights into memory from a table. -*/ -static void editDist3SqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - EditDist3Config *pConfig = (EditDist3Config*)sqlite3_user_data(context); - sqlite3 *db = sqlite3_context_db_handle(context); - int rc; - if( argc==1 ){ - const char *zTable = (const char*)sqlite3_value_text(argv[0]); - rc = editDist3ConfigLoad(pConfig, db, zTable); - if( rc ) sqlite3_result_error_code(context, rc); - }else{ - const char *zA = (const char*)sqlite3_value_text(argv[0]); - const char *zB = (const char*)sqlite3_value_text(argv[1]); - int nA = sqlite3_value_bytes(argv[0]); - int nB = sqlite3_value_bytes(argv[1]); - int iLang = argc==3 ? sqlite3_value_int(argv[2]) : 0; - const EditDist3Lang *pLang = editDist3FindLang(pConfig, iLang); - EditDist3FromString *pFrom; - int dist; - - pFrom = editDist3FromStringNew(pLang, zA, nA); - if( pFrom==0 ){ - sqlite3_result_error_nomem(context); - return; - } - dist = editDist3Core(pFrom, zB, nB, pLang, 0); - editDist3FromStringDelete(pFrom); - if( dist==(-1) ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_int(context, dist); - } - } -} - -/* -** Register the editDist3 function with SQLite -*/ -static int editDist3Install(sqlite3 *db){ - int rc; - EditDist3Config *pConfig = sqlite3_malloc( sizeof(*pConfig) ); - if( pConfig==0 ) return SQLITE_NOMEM; - memset(pConfig, 0, sizeof(*pConfig)); - rc = sqlite3_create_function_v2(db, "editdist3", - 2, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function_v2(db, "editdist3", - 3, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function_v2(db, "editdist3", - 1, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, - editDist3ConfigDelete); - }else{ - sqlite3_free(pConfig); - } - return rc; -} -/* End configurable cost unicode edit distance routines -****************************************************************************** -****************************************************************************** -** Begin transliterate unicode-to-ascii implementation -*/ - -#if !SQLITE_AMALGAMATION -/* -** This lookup table is used to help decode the first byte of -** a multi-byte UTF8 character. -*/ -static const unsigned char sqlite3Utf8Trans1[] = { - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, - 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, - 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00, -}; -#endif - -/* -** Return the value of the first UTF-8 character in the string. -*/ -static int utf8Read(const unsigned char *z, int n, int *pSize){ - int c, i; - - /* All callers to this routine (in the current implementation) - ** always have n>0. */ - if( NEVER(n==0) ){ - c = i = 0; - }else{ - c = z[0]; - i = 1; - if( c>=0xc0 ){ - c = sqlite3Utf8Trans1[c-0xc0]; - while( i0 ){ - c = utf8Read(zIn, nIn, &sz); - zIn += sz; - nIn -= sz; - if( c<=127 ){ - zOut[nOut++] = c; - }else{ - int xTop, xBtm, x; - xTop = sizeof(translit)/sizeof(translit[0]) - 1; - xBtm = 0; - while( xTop>=xBtm ){ - x = (xTop + xBtm)/2; - if( translit[x].cFrom==c ){ - zOut[nOut++] = translit[x].cTo0; - if( translit[x].cTo1 ){ - zOut[nOut++] = translit[x].cTo1; - /* Add an extra "ch" after the "sh" for Щ and щ */ - if( c==0x0429 || c== 0x0449 ){ - zOut[nOut++] = 'c'; - zOut[nOut++] = 'h'; - } - } - c = 0; - break; - }else if( translit[x].cFrom>c ){ - xTop = x-1; - }else{ - xBtm = x+1; - } - } - if( c ) zOut[nOut++] = '?'; - } - } - zOut[nOut] = 0; - return zOut; -} - -/* -** Return the number of characters in the shortest prefix of the input -** string that transliterates to an ASCII string nTrans bytes or longer. -** Or, if the transliteration of the input string is less than nTrans -** bytes in size, return the number of characters in the input string. -*/ -static int translen_to_charlen(const char *zIn, int nIn, int nTrans){ - int i, c, sz, nOut; - int nChar; - - i = nOut = 0; - for(nChar=0; i=128 ){ - int xTop, xBtm, x; - xTop = sizeof(translit)/sizeof(translit[0]) - 1; - xBtm = 0; - while( xTop>=xBtm ){ - x = (xTop + xBtm)/2; - if( translit[x].cFrom==c ){ - if( translit[x].cTo1 ) nOut++; - if( c==0x0429 || c== 0x0449 ) nOut += 2; - break; - }else if( translit[x].cFrom>c ){ - xTop = x-1; - }else{ - xBtm = x+1; - } - } - } - } - - return nChar; -} - - -/* -** spellfix1_translit(X) -** -** Convert a string that contains non-ASCII Roman characters into -** pure ASCII. -*/ -static void transliterateSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - const unsigned char *zIn = sqlite3_value_text(argv[0]); - int nIn = sqlite3_value_bytes(argv[0]); - unsigned char *zOut = transliterate(zIn, nIn); - if( zOut==0 ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); - } -} - -/* -** spellfix1_scriptcode(X) -** -** Try to determine the dominant script used by the word X and return -** its ISO 15924 numeric code. -** -** The current implementation only understands the following scripts: -** -** 215 (Latin) -** 220 (Cyrillic) -** 200 (Greek) -** -** This routine will return 998 if the input X contains characters from -** two or more of the above scripts or 999 if X contains no characters -** from any of the above scripts. -*/ -static void scriptCodeSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - const unsigned char *zIn = sqlite3_value_text(argv[0]); - int nIn = sqlite3_value_bytes(argv[0]); - int c, sz; - int scriptMask = 0; - int res; -# define SCRIPT_LATIN 0x0001 -# define SCRIPT_CYRILLIC 0x0002 -# define SCRIPT_GREEK 0x0004 - - while( nIn>0 ){ - c = utf8Read(zIn, nIn, &sz); - zIn += sz; - nIn -= sz; - if( c<0x02af ){ - scriptMask |= SCRIPT_LATIN; - }else if( c>=0x0400 && c<=0x04ff ){ - scriptMask |= SCRIPT_CYRILLIC; - }else if( c>=0x0386 && c<=0x03ce ){ - scriptMask |= SCRIPT_GREEK; - } - } - switch( scriptMask ){ - case 0: res = 999; break; - case SCRIPT_LATIN: res = 215; break; - case SCRIPT_CYRILLIC: res = 220; break; - case SCRIPT_GREEK: res = 200; break; - default: res = 998; break; - } - sqlite3_result_int(context, res); -} - -/* End transliterate -****************************************************************************** -****************************************************************************** -** Begin spellfix1 virtual table. -*/ - -/* Maximum length of a phonehash used for querying the shadow table */ -#define SPELLFIX_MX_HASH 8 - -/* Maximum number of hash strings to examine per query */ -#define SPELLFIX_MX_RUN 1 - -typedef struct spellfix1_vtab spellfix1_vtab; -typedef struct spellfix1_cursor spellfix1_cursor; - -/* Fuzzy-search virtual table object */ -struct spellfix1_vtab { - sqlite3_vtab base; /* Base class - must be first */ - sqlite3 *db; /* Database connection */ - char *zDbName; /* Name of database holding this table */ - char *zTableName; /* Name of the virtual table */ - char *zCostTable; /* Table holding edit-distance cost numbers */ - EditDist3Config *pConfig3; /* Parsed edit distance costs */ -}; - -/* Fuzzy-search cursor object */ -struct spellfix1_cursor { - sqlite3_vtab_cursor base; /* Base class - must be first */ - spellfix1_vtab *pVTab; /* The table to which this cursor belongs */ - char *zPattern; /* rhs of MATCH clause */ - int nRow; /* Number of rows of content */ - int nAlloc; /* Number of allocated rows */ - int iRow; /* Current row of content */ - int iLang; /* Value of the langid= constraint */ - int iTop; /* Value of the top= constraint */ - int iScope; /* Value of the scope= constraint */ - int nSearch; /* Number of vocabulary items checked */ - sqlite3_stmt *pFullScan; /* Shadow query for a full table scan */ - struct spellfix1_row { /* For each row of content */ - sqlite3_int64 iRowid; /* Rowid for this row */ - char *zWord; /* Text for this row */ - int iRank; /* Rank for this row */ - int iDistance; /* Distance from pattern for this row */ - int iScore; /* Score for sorting */ - int iMatchlen; /* Value of matchlen column (or -1) */ - char zHash[SPELLFIX_MX_HASH]; /* the phonehash used for this match */ - } *a; -}; - -/* -** Construct one or more SQL statements from the format string given -** and then evaluate those statements. The success code is written -** into *pRc. -** -** If *pRc is initially non-zero then this routine is a no-op. -*/ -static void spellfix1DbExec( - int *pRc, /* Success code */ - sqlite3 *db, /* Database in which to run SQL */ - const char *zFormat, /* Format string for SQL */ - ... /* Arguments to the format string */ -){ - va_list ap; - char *zSql; - if( *pRc ) return; - va_start(ap, zFormat); - zSql = sqlite3_vmprintf(zFormat, ap); - va_end(ap); - if( zSql==0 ){ - *pRc = SQLITE_NOMEM; - }else{ - *pRc = sqlite3_exec(db, zSql, 0, 0, 0); - sqlite3_free(zSql); - } -} - -/* -** xDisconnect/xDestroy method for the fuzzy-search module. -*/ -static int spellfix1Uninit(int isDestroy, sqlite3_vtab *pVTab){ - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - int rc = SQLITE_OK; - if( isDestroy ){ - sqlite3 *db = p->db; - spellfix1DbExec(&rc, db, "DROP TABLE IF EXISTS \"%w\".\"%w_vocab\"", - p->zDbName, p->zTableName); - } - if( rc==SQLITE_OK ){ - sqlite3_free(p->zTableName); - editDist3ConfigDelete(p->pConfig3); - sqlite3_free(p->zCostTable); - sqlite3_free(p); - } - return rc; -} -static int spellfix1Disconnect(sqlite3_vtab *pVTab){ - return spellfix1Uninit(0, pVTab); -} -static int spellfix1Destroy(sqlite3_vtab *pVTab){ - return spellfix1Uninit(1, pVTab); -} - -/* -** Make a copy of a string. Remove leading and trailing whitespace -** and dequote it. -*/ -static char *spellfix1Dequote(const char *zIn){ - char *zOut; - int i, j; - char c; - while( isspace(zIn[0]) ) zIn++; - zOut = sqlite3_mprintf("%s", zIn); - if( zOut==0 ) return 0; - i = (int)strlen(zOut); -#if 0 /* The parser will never leave spaces at the end */ - while( i>0 && isspace(zOut[i-1]) ){ i--; } -#endif - zOut[i] = 0; - c = zOut[0]; - if( c=='\'' || c=='"' ){ - for(i=1, j=0; ALWAYS(zOut[i]); i++){ - zOut[j++] = zOut[i]; - if( zOut[i]==c ){ - if( zOut[i+1]==c ){ - i++; - }else{ - zOut[j-1] = 0; - break; - } - } - } - } - return zOut; -} - - -/* -** xConnect/xCreate method for the spellfix1 module. Arguments are: -** -** argv[0] -> module name ("spellfix1") -** argv[1] -> database name -** argv[2] -> table name -** argv[3].. -> optional arguments (i.e. "edit_cost_table" parameter) -*/ -static int spellfix1Init( - int isCreate, - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVTab, - char **pzErr -){ - spellfix1_vtab *pNew = 0; - const char *zModule = argv[0]; - const char *zDbName = argv[1]; - const char *zTableName = argv[2]; - int nDbName; - int rc = SQLITE_OK; - int i; - - nDbName = (int)strlen(zDbName); - pNew = sqlite3_malloc( sizeof(*pNew) + nDbName + 1); - if( pNew==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(pNew, 0, sizeof(*pNew)); - pNew->zDbName = (char*)&pNew[1]; - memcpy(pNew->zDbName, zDbName, nDbName+1); - pNew->zTableName = sqlite3_mprintf("%s", zTableName); - pNew->db = db; - if( pNew->zTableName==0 ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_declare_vtab(db, - "CREATE TABLE x(word,rank,distance,langid, " - "score, matchlen, phonehash HIDDEN, " - "top HIDDEN, scope HIDDEN, srchcnt HIDDEN, " - "soundslike HIDDEN, command HIDDEN)" - ); -#define SPELLFIX_COL_WORD 0 -#define SPELLFIX_COL_RANK 1 -#define SPELLFIX_COL_DISTANCE 2 -#define SPELLFIX_COL_LANGID 3 -#define SPELLFIX_COL_SCORE 4 -#define SPELLFIX_COL_MATCHLEN 5 -#define SPELLFIX_COL_PHONEHASH 6 -#define SPELLFIX_COL_TOP 7 -#define SPELLFIX_COL_SCOPE 8 -#define SPELLFIX_COL_SRCHCNT 9 -#define SPELLFIX_COL_SOUNDSLIKE 10 -#define SPELLFIX_COL_COMMAND 11 - } - if( rc==SQLITE_OK && isCreate ){ - sqlite3_uint64 r; - spellfix1DbExec(&rc, db, - "CREATE TABLE IF NOT EXISTS \"%w\".\"%w_vocab\"(\n" - " id INTEGER PRIMARY KEY,\n" - " rank INT,\n" - " langid INT,\n" - " word TEXT,\n" - " k1 TEXT,\n" - " k2 TEXT\n" - ");\n", - zDbName, zTableName - ); - sqlite3_randomness(sizeof(r), &r); - spellfix1DbExec(&rc, db, - "CREATE INDEX IF NOT EXISTS \"%w\".\"%w_index_%llx\" " - "ON \"%w_vocab\"(langid,k2);", - zDbName, zModule, r, zTableName - ); - } - for(i=3; rc==SQLITE_OK && ibase); - }else{ - *ppVTab = (sqlite3_vtab *)pNew; - } - return rc; -} - -/* -** The xConnect and xCreate methods -*/ -static int spellfix1Connect( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVTab, - char **pzErr -){ - return spellfix1Init(0, db, pAux, argc, argv, ppVTab, pzErr); -} -static int spellfix1Create( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVTab, - char **pzErr -){ - return spellfix1Init(1, db, pAux, argc, argv, ppVTab, pzErr); -} - -/* -** Clear all of the content from a cursor. -*/ -static void spellfix1ResetCursor(spellfix1_cursor *pCur){ - int i; - for(i=0; inRow; i++){ - sqlite3_free(pCur->a[i].zWord); - } - pCur->nRow = 0; - pCur->iRow = 0; - pCur->nSearch = 0; - if( pCur->pFullScan ){ - sqlite3_finalize(pCur->pFullScan); - pCur->pFullScan = 0; - } -} - -/* -** Resize the cursor to hold up to N rows of content -*/ -static void spellfix1ResizeCursor(spellfix1_cursor *pCur, int N){ - struct spellfix1_row *aNew; - assert( N>=pCur->nRow ); - aNew = sqlite3_realloc(pCur->a, sizeof(pCur->a[0])*N); - if( aNew==0 && N>0 ){ - spellfix1ResetCursor(pCur); - sqlite3_free(pCur->a); - pCur->nAlloc = 0; - pCur->a = 0; - }else{ - pCur->nAlloc = N; - pCur->a = aNew; - } -} - - -/* -** Close a fuzzy-search cursor. -*/ -static int spellfix1Close(sqlite3_vtab_cursor *cur){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - spellfix1ResetCursor(pCur); - spellfix1ResizeCursor(pCur, 0); - sqlite3_free(pCur->zPattern); - sqlite3_free(pCur); - return SQLITE_OK; -} - -/* -** Search for terms of these forms: -** -** (A) word MATCH $str -** (B) langid == $langid -** (C) top = $top -** (D) scope = $scope -** (E) distance < $distance -** (F) distance <= $distance -** -** The plan number is a bit mask formed with these bits: -** -** 0x01 (A) is found -** 0x02 (B) is found -** 0x04 (C) is found -** 0x08 (D) is found -** 0x10 (E) is found -** 0x20 (F) is found -** -** filter.argv[*] values contains $str, $langid, $top, and $scope, -** if specified and in that order. -*/ -static int spellfix1BestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ - int iPlan = 0; - int iLangTerm = -1; - int iTopTerm = -1; - int iScopeTerm = -1; - int iDistTerm = -1; - int i; - const struct sqlite3_index_constraint *pConstraint; - pConstraint = pIdxInfo->aConstraint; - for(i=0; inConstraint; i++, pConstraint++){ - if( pConstraint->usable==0 ) continue; - - /* Terms of the form: word MATCH $str */ - if( (iPlan & 1)==0 - && pConstraint->iColumn==SPELLFIX_COL_WORD - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH - ){ - iPlan |= 1; - pIdxInfo->aConstraintUsage[i].argvIndex = 1; - pIdxInfo->aConstraintUsage[i].omit = 1; - } - - /* Terms of the form: langid = $langid */ - if( (iPlan & 2)==0 - && pConstraint->iColumn==SPELLFIX_COL_LANGID - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 2; - iLangTerm = i; - } - - /* Terms of the form: top = $top */ - if( (iPlan & 4)==0 - && pConstraint->iColumn==SPELLFIX_COL_TOP - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 4; - iTopTerm = i; - } - - /* Terms of the form: scope = $scope */ - if( (iPlan & 8)==0 - && pConstraint->iColumn==SPELLFIX_COL_SCOPE - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 8; - iScopeTerm = i; - } - - /* Terms of the form: distance < $dist or distance <= $dist */ - if( (iPlan & (16|32))==0 - && pConstraint->iColumn==SPELLFIX_COL_DISTANCE - && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT - || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) - ){ - iPlan |= pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ? 16 : 32; - iDistTerm = i; - } - } - if( iPlan&1 ){ - int idx = 2; - pIdxInfo->idxNum = iPlan; - if( pIdxInfo->nOrderBy==1 - && pIdxInfo->aOrderBy[0].iColumn==SPELLFIX_COL_SCORE - && pIdxInfo->aOrderBy[0].desc==0 - ){ - pIdxInfo->orderByConsumed = 1; /* Default order by iScore */ - } - if( iPlan&2 ){ - pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iLangTerm].omit = 1; - } - if( iPlan&4 ){ - pIdxInfo->aConstraintUsage[iTopTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iTopTerm].omit = 1; - } - if( iPlan&8 ){ - pIdxInfo->aConstraintUsage[iScopeTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iScopeTerm].omit = 1; - } - if( iPlan&(16|32) ){ - pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iDistTerm].omit = 1; - } - pIdxInfo->estimatedCost = (double)10000; - }else{ - pIdxInfo->idxNum = 0; - pIdxInfo->estimatedCost = (double)10000000; - } - return SQLITE_OK; -} - -/* -** Open a new fuzzy-search cursor. -*/ -static int spellfix1Open(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - spellfix1_cursor *pCur; - pCur = sqlite3_malloc( sizeof(*pCur) ); - if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); - pCur->pVTab = p; - *ppCursor = &pCur->base; - return SQLITE_OK; -} - -/* -** Adjust a distance measurement by the words rank in order to show -** preference to common words. -*/ -static int spellfix1Score(int iDistance, int iRank){ - int iLog2; - for(iLog2=0; iRank>0; iLog2++, iRank>>=1){} - return iDistance + 32 - iLog2; -} - -/* -** Compare two spellfix1_row objects for sorting purposes in qsort() such -** that they sort in order of increasing distance. -*/ -static int spellfix1RowCompare(const void *A, const void *B){ - const struct spellfix1_row *a = (const struct spellfix1_row*)A; - const struct spellfix1_row *b = (const struct spellfix1_row*)B; - return a->iScore - b->iScore; -} - -/* -** A structure used to pass information from spellfix1FilterForMatch() -** into spellfix1RunQuery(). -*/ -typedef struct MatchQuery { - spellfix1_cursor *pCur; /* The cursor being queried */ - sqlite3_stmt *pStmt; /* shadow table query statment */ - char zHash[SPELLFIX_MX_HASH]; /* The current phonehash for zPattern */ - const char *zPattern; /* Transliterated input string */ - int nPattern; /* Length of zPattern */ - EditDist3FromString *pMatchStr3; /* Original unicode string */ - EditDist3Config *pConfig3; /* Edit-distance cost coefficients */ - const EditDist3Lang *pLang; /* The selected language coefficients */ - int iLang; /* The language id */ - int iScope; /* Default scope */ - int iMaxDist; /* Maximum allowed edit distance, or -1 */ - int rc; /* Error code */ - int nRun; /* Number of prior runs for the same zPattern */ - char azPrior[SPELLFIX_MX_RUN][SPELLFIX_MX_HASH]; /* Prior hashes */ -} MatchQuery; - -/* -** Run a query looking for the best matches against zPattern using -** zHash as the character class seed hash. -*/ -static void spellfix1RunQuery(MatchQuery *p, const char *zQuery, int nQuery){ - const char *zK1; - const char *zWord; - int iDist; - int iRank; - int iScore; - int iWorst = 0; - int idx; - int idxWorst = -1; - int i; - int iScope = p->iScope; - spellfix1_cursor *pCur = p->pCur; - sqlite3_stmt *pStmt = p->pStmt; - char zHash1[SPELLFIX_MX_HASH]; - char zHash2[SPELLFIX_MX_HASH]; - char *zClass; - int nClass; - int rc; - - if( pCur->a==0 || p->rc ) return; /* Prior memory allocation failure */ - zClass = (char*)phoneticHash((unsigned char*)zQuery, nQuery); - if( zClass==0 ){ - p->rc = SQLITE_NOMEM; - return; - } - nClass = (int)strlen(zClass); - if( nClass>SPELLFIX_MX_HASH-2 ){ - nClass = SPELLFIX_MX_HASH-2; - zClass[nClass] = 0; - } - if( nClass<=iScope ){ - if( nClass>2 ){ - iScope = nClass-1; - }else{ - iScope = nClass; - } - } - memcpy(zHash1, zClass, iScope); - sqlite3_free(zClass); - zHash1[iScope] = 0; - memcpy(zHash2, zHash1, iScope); - zHash2[iScope] = 'Z'; - zHash2[iScope+1] = 0; -#if SPELLFIX_MX_RUN>1 - for(i=0; inRun; i++){ - if( strcmp(p->azPrior[i], zHash1)==0 ) return; - } -#endif - assert( p->nRunazPrior[p->nRun++], zHash1, iScope+1); - if( sqlite3_bind_text(pStmt, 1, zHash1, -1, SQLITE_STATIC)==SQLITE_NOMEM - || sqlite3_bind_text(pStmt, 2, zHash2, -1, SQLITE_STATIC)==SQLITE_NOMEM - ){ - p->rc = SQLITE_NOMEM; - return; - } -#if SPELLFIX_MX_RUN>1 - for(i=0; inRow; i++){ - if( pCur->a[i].iScore>iWorst ){ - iWorst = pCur->a[i].iScore; - idxWorst = i; - } - } -#endif - while( sqlite3_step(pStmt)==SQLITE_ROW ){ - int iMatchlen = -1; - iRank = sqlite3_column_int(pStmt, 2); - if( p->pMatchStr3 ){ - int nWord = sqlite3_column_bytes(pStmt, 1); - zWord = (const char*)sqlite3_column_text(pStmt, 1); - iDist = editDist3Core(p->pMatchStr3, zWord, nWord, p->pLang, &iMatchlen); - }else{ - zK1 = (const char*)sqlite3_column_text(pStmt, 3); - if( zK1==0 ) continue; - iDist = editdist1(p->zPattern, zK1, 0); - } - if( iDist<0 ){ - p->rc = SQLITE_NOMEM; - break; - } - pCur->nSearch++; - iScore = spellfix1Score(iDist,iRank); - if( p->iMaxDist>=0 ){ - if( iDist>p->iMaxDist ) continue; - if( pCur->nRow>=pCur->nAlloc-1 ){ - spellfix1ResizeCursor(pCur, pCur->nAlloc*2 + 10); - if( pCur->a==0 ) break; - } - idx = pCur->nRow; - }else if( pCur->nRownAlloc ){ - idx = pCur->nRow; - }else if( iScorea[idx].zWord); - }else{ - continue; - } - pCur->a[idx].zWord = sqlite3_mprintf("%s", sqlite3_column_text(pStmt, 1)); - if( pCur->a[idx].zWord==0 ){ - p->rc = SQLITE_NOMEM; - break; - } - pCur->a[idx].iRowid = sqlite3_column_int64(pStmt, 0); - pCur->a[idx].iRank = iRank; - pCur->a[idx].iDistance = iDist; - pCur->a[idx].iScore = iScore; - pCur->a[idx].iMatchlen = iMatchlen; - memcpy(pCur->a[idx].zHash, zHash1, iScope+1); - if( pCur->nRownAlloc ) pCur->nRow++; - if( pCur->nRow==pCur->nAlloc ){ - iWorst = pCur->a[0].iScore; - idxWorst = 0; - for(i=1; inRow; i++){ - iScore = pCur->a[i].iScore; - if( iWorstrc = rc; -} - -/* -** This version of the xFilter method work if the MATCH term is present -** and we are doing a scan. -*/ -static int spellfix1FilterForMatch( - spellfix1_cursor *pCur, - int idxNum, - int argc, - sqlite3_value **argv -){ - const unsigned char *zMatchThis; /* RHS of the MATCH operator */ - EditDist3FromString *pMatchStr3 = 0; /* zMatchThis as an editdist string */ - char *zPattern; /* Transliteration of zMatchThis */ - int nPattern; /* Length of zPattern */ - int iLimit = 20; /* Max number of rows of output */ - int iScope = 3; /* Use this many characters of zClass */ - int iLang = 0; /* Language code */ - char *zSql; /* SQL of shadow table query */ - sqlite3_stmt *pStmt = 0; /* Shadow table query */ - int rc; /* Result code */ - int idx = 1; /* Next available filter parameter */ - spellfix1_vtab *p = pCur->pVTab; /* The virtual table that owns pCur */ - MatchQuery x; /* For passing info to RunQuery() */ - - /* Load the cost table if we have not already done so */ - if( p->zCostTable!=0 && p->pConfig3==0 ){ - p->pConfig3 = sqlite3_malloc( sizeof(p->pConfig3[0]) ); - if( p->pConfig3==0 ) return SQLITE_NOMEM; - memset(p->pConfig3, 0, sizeof(p->pConfig3[0])); - rc = editDist3ConfigLoad(p->pConfig3, p->db, p->zCostTable); - if( rc ) return rc; - } - memset(&x, 0, sizeof(x)); - x.iScope = 3; /* Default scope if none specified by "WHERE scope=N" */ - x.iMaxDist = -1; /* Maximum allowed edit distance */ - - if( idxNum&2 ){ - iLang = sqlite3_value_int(argv[idx++]); - } - if( idxNum&4 ){ - iLimit = sqlite3_value_int(argv[idx++]); - if( iLimit<1 ) iLimit = 1; - } - if( idxNum&8 ){ - x.iScope = sqlite3_value_int(argv[idx++]); - if( x.iScope<1 ) x.iScope = 1; - if( x.iScope>SPELLFIX_MX_HASH-2 ) x.iScope = SPELLFIX_MX_HASH-2; - } - if( idxNum&(16|32) ){ - x.iMaxDist = sqlite3_value_int(argv[idx++]); - if( idxNum&16 ) x.iMaxDist--; - if( x.iMaxDist<0 ) x.iMaxDist = 0; - } - spellfix1ResetCursor(pCur); - spellfix1ResizeCursor(pCur, iLimit); - zMatchThis = sqlite3_value_text(argv[0]); - if( zMatchThis==0 ) return SQLITE_OK; - if( p->pConfig3 ){ - x.pLang = editDist3FindLang(p->pConfig3, iLang); - pMatchStr3 = editDist3FromStringNew(x.pLang, (const char*)zMatchThis, -1); - if( pMatchStr3==0 ){ - x.rc = SQLITE_NOMEM; - goto filter_exit; - } - }else{ - x.pLang = 0; - } - zPattern = (char*)transliterate(zMatchThis, sqlite3_value_bytes(argv[0])); - sqlite3_free(pCur->zPattern); - pCur->zPattern = zPattern; - if( zPattern==0 ){ - x.rc = SQLITE_NOMEM; - goto filter_exit; - } - nPattern = (int)strlen(zPattern); - if( zPattern[nPattern-1]=='*' ) nPattern--; - zSql = sqlite3_mprintf( - "SELECT id, word, rank, k1" - " FROM \"%w\".\"%w_vocab\"" - " WHERE langid=%d AND k2>=?1 AND k2zDbName, p->zTableName, iLang - ); - if( zSql==0 ){ - x.rc = SQLITE_NOMEM; - pStmt = 0; - goto filter_exit; - } - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0); - sqlite3_free(zSql); - pCur->iLang = iLang; - x.pCur = pCur; - x.pStmt = pStmt; - x.zPattern = zPattern; - x.nPattern = nPattern; - x.pMatchStr3 = pMatchStr3; - x.iLang = iLang; - x.rc = rc; - x.pConfig3 = p->pConfig3; - if( x.rc==SQLITE_OK ){ - spellfix1RunQuery(&x, zPattern, nPattern); - } - - if( pCur->a ){ - qsort(pCur->a, pCur->nRow, sizeof(pCur->a[0]), spellfix1RowCompare); - pCur->iTop = iLimit; - pCur->iScope = iScope; - }else{ - x.rc = SQLITE_NOMEM; - } - -filter_exit: - sqlite3_finalize(pStmt); - editDist3FromStringDelete(pMatchStr3); - return x.rc; -} - -/* -** This version of xFilter handles a full-table scan case -*/ -static int spellfix1FilterForFullScan( - spellfix1_cursor *pCur, - int idxNum, - int argc, - sqlite3_value **argv -){ - int rc; - char *zSql; - spellfix1_vtab *pVTab = pCur->pVTab; - spellfix1ResetCursor(pCur); - zSql = sqlite3_mprintf( - "SELECT word, rank, NULL, langid, id FROM \"%w\".\"%w_vocab\"", - pVTab->zDbName, pVTab->zTableName); - if( zSql==0 ) return SQLITE_NOMEM; - rc = sqlite3_prepare_v2(pVTab->db, zSql, -1, &pCur->pFullScan, 0); - sqlite3_free(zSql); - pCur->nRow = pCur->iRow = 0; - if( rc==SQLITE_OK ){ - rc = sqlite3_step(pCur->pFullScan); - if( rc==SQLITE_ROW ){ pCur->iRow = -1; rc = SQLITE_OK; } - if( rc==SQLITE_DONE ){ rc = SQLITE_OK; } - }else{ - pCur->iRow = 0; - } - return rc; -} - - -/* -** Called to "rewind" a cursor back to the beginning so that -** it starts its output over again. Always called at least once -** prior to any spellfix1Column, spellfix1Rowid, or spellfix1Eof call. -*/ -static int spellfix1Filter( - sqlite3_vtab_cursor *cur, - int idxNum, const char *idxStr, - int argc, sqlite3_value **argv -){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - int rc; - if( idxNum & 1 ){ - rc = spellfix1FilterForMatch(pCur, idxNum, argc, argv); - }else{ - rc = spellfix1FilterForFullScan(pCur, idxNum, argc, argv); - } - return rc; -} - - -/* -** Advance a cursor to its next row of output -*/ -static int spellfix1Next(sqlite3_vtab_cursor *cur){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - int rc = SQLITE_OK; - if( pCur->iRow < pCur->nRow ){ - if( pCur->pFullScan ){ - rc = sqlite3_step(pCur->pFullScan); - if( rc!=SQLITE_ROW ) pCur->iRow = pCur->nRow; - if( rc==SQLITE_ROW || rc==SQLITE_DONE ) rc = SQLITE_OK; - }else{ - pCur->iRow++; - } - } - return rc; -} - -/* -** Return TRUE if we are at the end-of-file -*/ -static int spellfix1Eof(sqlite3_vtab_cursor *cur){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - return pCur->iRow>=pCur->nRow; -} - -/* -** Return columns from the current row. -*/ -static int spellfix1Column( - sqlite3_vtab_cursor *cur, - sqlite3_context *ctx, - int i -){ - spellfix1_cursor *pCur = (spellfix1_cursor*)cur; - if( pCur->pFullScan ){ - if( i<=SPELLFIX_COL_LANGID ){ - sqlite3_result_value(ctx, sqlite3_column_value(pCur->pFullScan, i)); - }else{ - sqlite3_result_null(ctx); - } - return SQLITE_OK; - } - switch( i ){ - case SPELLFIX_COL_WORD: { - sqlite3_result_text(ctx, pCur->a[pCur->iRow].zWord, -1, SQLITE_STATIC); - break; - } - case SPELLFIX_COL_RANK: { - sqlite3_result_int(ctx, pCur->a[pCur->iRow].iRank); - break; - } - case SPELLFIX_COL_DISTANCE: { - sqlite3_result_int(ctx, pCur->a[pCur->iRow].iDistance); - break; - } - case SPELLFIX_COL_LANGID: { - sqlite3_result_int(ctx, pCur->iLang); - break; - } - case SPELLFIX_COL_SCORE: { - sqlite3_result_int(ctx, pCur->a[pCur->iRow].iScore); - break; - } - case SPELLFIX_COL_MATCHLEN: { - int iMatchlen = pCur->a[pCur->iRow].iMatchlen; - if( iMatchlen<0 ){ - int nPattern = (int)strlen(pCur->zPattern); - char *zWord = pCur->a[pCur->iRow].zWord; - int nWord = (int)strlen(zWord); - - if( nPattern>0 && pCur->zPattern[nPattern-1]=='*' ){ - char *zTranslit; - int res; - zTranslit = (char *)transliterate((unsigned char *)zWord, nWord); - if( !zTranslit ) return SQLITE_NOMEM; - res = editdist1(pCur->zPattern, zTranslit, &iMatchlen); - sqlite3_free(zTranslit); - if( res<0 ) return SQLITE_NOMEM; - iMatchlen = translen_to_charlen(zWord, nWord, iMatchlen); - }else{ - iMatchlen = utf8Charlen(zWord, nWord); - } - } - - sqlite3_result_int(ctx, iMatchlen); - break; - } - case SPELLFIX_COL_PHONEHASH: { - sqlite3_result_text(ctx, pCur->a[pCur->iRow].zHash, -1, SQLITE_STATIC); - break; - } - case SPELLFIX_COL_TOP: { - sqlite3_result_int(ctx, pCur->iTop); - break; - } - case SPELLFIX_COL_SCOPE: { - sqlite3_result_int(ctx, pCur->iScope); - break; - } - case SPELLFIX_COL_SRCHCNT: { - sqlite3_result_int(ctx, pCur->nSearch); - break; - } - default: { - sqlite3_result_null(ctx); - break; - } - } - return SQLITE_OK; -} - -/* -** The rowid. -*/ -static int spellfix1Rowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ - spellfix1_cursor *pCur = (spellfix1_cursor*)cur; - if( pCur->pFullScan ){ - *pRowid = sqlite3_column_int64(pCur->pFullScan, 4); - }else{ - *pRowid = pCur->a[pCur->iRow].iRowid; - } - return SQLITE_OK; -} - -/* -** The xUpdate() method. -*/ -static int spellfix1Update( - sqlite3_vtab *pVTab, - int argc, - sqlite3_value **argv, - sqlite_int64 *pRowid -){ - int rc = SQLITE_OK; - sqlite3_int64 rowid, newRowid; - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - sqlite3 *db = p->db; - - if( argc==1 ){ - /* A delete operation on the rowid given by argv[0] */ - rowid = *pRowid = sqlite3_value_int64(argv[0]); - spellfix1DbExec(&rc, db, "DELETE FROM \"%w\".\"%w_vocab\" " - " WHERE id=%lld", - p->zDbName, p->zTableName, rowid); - }else{ - const unsigned char *zWord = sqlite3_value_text(argv[SPELLFIX_COL_WORD+2]); - int nWord = sqlite3_value_bytes(argv[SPELLFIX_COL_WORD+2]); - int iLang = sqlite3_value_int(argv[SPELLFIX_COL_LANGID+2]); - int iRank = sqlite3_value_int(argv[SPELLFIX_COL_RANK+2]); - const unsigned char *zSoundslike = - sqlite3_value_text(argv[SPELLFIX_COL_SOUNDSLIKE+2]); - int nSoundslike = sqlite3_value_bytes(argv[SPELLFIX_COL_SOUNDSLIKE+2]); - char *zK1, *zK2; - int i; - char c; - - if( zWord==0 ){ - /* Inserts of the form: INSERT INTO table(command) VALUES('xyzzy'); - ** cause zWord to be NULL, so we look at the "command" column to see - ** what special actions to take */ - const char *zCmd = - (const char*)sqlite3_value_text(argv[SPELLFIX_COL_COMMAND+2]); - if( zCmd==0 ){ - pVTab->zErrMsg = sqlite3_mprintf("%s.word may not be NULL", - p->zTableName); - return SQLITE_CONSTRAINT_NOTNULL; - } - if( strcmp(zCmd,"reset")==0 ){ - /* Reset the edit cost table (if there is one). */ - editDist3ConfigDelete(p->pConfig3); - p->pConfig3 = 0; - return SQLITE_OK; - } - if( strncmp(zCmd,"edit_cost_table=",16)==0 ){ - editDist3ConfigDelete(p->pConfig3); - p->pConfig3 = 0; - sqlite3_free(p->zCostTable); - p->zCostTable = spellfix1Dequote(zCmd+16); - if( p->zCostTable==0 ) return SQLITE_NOMEM; - if( p->zCostTable[0]==0 || sqlite3_stricmp(p->zCostTable,"null")==0 ){ - sqlite3_free(p->zCostTable); - p->zCostTable = 0; - } - return SQLITE_OK; - } - pVTab->zErrMsg = sqlite3_mprintf("unknown value for %s.command: \"%w\"", - p->zTableName, zCmd); - return SQLITE_ERROR; - } - if( iRank<1 ) iRank = 1; - if( zSoundslike ){ - zK1 = (char*)transliterate(zSoundslike, nSoundslike); - }else{ - zK1 = (char*)transliterate(zWord, nWord); - } - if( zK1==0 ) return SQLITE_NOMEM; - for(i=0; (c = zK1[i])!=0; i++){ - if( c>='A' && c<='Z' ) zK1[i] += 'a' - 'A'; - } - zK2 = (char*)phoneticHash((const unsigned char*)zK1, i); - if( zK2==0 ){ - sqlite3_free(zK1); - return SQLITE_NOMEM; - } - if( sqlite3_value_type(argv[0])==SQLITE_NULL ){ - spellfix1DbExec(&rc, db, - "INSERT INTO \"%w\".\"%w_vocab\"(rank,langid,word,k1,k2) " - "VALUES(%d,%d,%Q,%Q,%Q)", - p->zDbName, p->zTableName, - iRank, iLang, zWord, zK1, zK2 - ); - *pRowid = sqlite3_last_insert_rowid(db); - }else{ - rowid = sqlite3_value_int64(argv[0]); - newRowid = *pRowid = sqlite3_value_int64(argv[1]); - spellfix1DbExec(&rc, db, - "UPDATE \"%w\".\"%w_vocab\" SET id=%lld, rank=%d, langid=%d," - " word=%Q, k1=%Q, k2=%Q WHERE id=%lld", - p->zDbName, p->zTableName, newRowid, iRank, iLang, - zWord, zK1, zK2, rowid - ); - } - sqlite3_free(zK1); - sqlite3_free(zK2); - } - return rc; -} - -/* -** Rename the spellfix1 table. -*/ -static int spellfix1Rename(sqlite3_vtab *pVTab, const char *zNew){ - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - sqlite3 *db = p->db; - int rc = SQLITE_OK; - char *zNewName = sqlite3_mprintf("%s", zNew); - if( zNewName==0 ){ - return SQLITE_NOMEM; - } - spellfix1DbExec(&rc, db, - "ALTER TABLE \"%w\".\"%w_vocab\" RENAME TO \"%w_vocab\"", - p->zDbName, p->zTableName, zNewName - ); - if( rc==SQLITE_OK ){ - sqlite3_free(p->zTableName); - p->zTableName = zNewName; - }else{ - sqlite3_free(zNewName); - } - return rc; -} - - -/* -** A virtual table module that provides fuzzy search. -*/ -static sqlite3_module spellfix1Module = { - 0, /* iVersion */ - spellfix1Create, /* xCreate - handle CREATE VIRTUAL TABLE */ - spellfix1Connect, /* xConnect - reconnected to an existing table */ - spellfix1BestIndex, /* xBestIndex - figure out how to do a query */ - spellfix1Disconnect, /* xDisconnect - close a connection */ - spellfix1Destroy, /* xDestroy - handle DROP TABLE */ - spellfix1Open, /* xOpen - open a cursor */ - spellfix1Close, /* xClose - close a cursor */ - spellfix1Filter, /* xFilter - configure scan constraints */ - spellfix1Next, /* xNext - advance a cursor */ - spellfix1Eof, /* xEof - check for end of scan */ - spellfix1Column, /* xColumn - read data */ - spellfix1Rowid, /* xRowid - read data */ - spellfix1Update, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindMethod */ - spellfix1Rename, /* xRename */ -}; - -/* -** Register the various functions and the virtual table. -*/ -static int spellfix1Register(sqlite3 *db){ - int rc = SQLITE_OK; - int i; - rc = sqlite3_create_function(db, "spellfix1_translit", 1, SQLITE_UTF8, 0, - transliterateSqlFunc, 0, 0); - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function(db, "spellfix1_editdist", 2, SQLITE_UTF8, 0, - editdistSqlFunc, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function(db, "spellfix1_phonehash", 1, SQLITE_UTF8, 0, - phoneticHashSqlFunc, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function(db, "spellfix1_scriptcode", 1, SQLITE_UTF8, 0, - scriptCodeSqlFunc, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_module(db, "spellfix1", &spellfix1Module, 0); - } - if( rc==SQLITE_OK ){ - rc = editDist3Install(db); - } - - /* Verify sanity of the translit[] table */ - for(i=0; i +#include +#include +#include -#include "sys/types.h" -#include "unistd.h" +#include +#include static int getProcessId(void){ #if SQLITE_OS_WIN return (int)_getpid(); #else return (int)getpid(); #endif } - +/* Names of environment variables to be used */ #define ENVIRONMENT_VARIABLE1_NAME "SQLITE_SQLLOG_DIR" #define ENVIRONMENT_VARIABLE2_NAME "SQLITE_SQLLOG_REUSE_FILES" /* Assume that all database and database file names are shorted than this. */ #define SQLLOG_NAMESZ 512 @@ -84,18 +96,23 @@ ** open (if any more are opened an error is logged using sqlite3_log() ** and processing is halted). */ #define MAX_CONNECTIONS 256 +/* There is one instance of this object for each SQLite database connection +** that is being logged. +*/ struct SLConn { int isErr; /* True if an error has occurred */ sqlite3 *db; /* Connection handle */ int iLog; /* First integer value used in file names */ FILE *fd; /* File descriptor for log file */ }; -struct SLGlobal { +/* This object is a singleton that keeps track of all data loggers. +*/ +static struct SLGlobal { /* Protected by MUTEX_STATIC_MASTER */ sqlite3_mutex *mutex; /* Recursive mutex */ int nConn; /* Size of aConn[] array */ /* Protected by SLGlobal.mutex */ @@ -386,10 +403,28 @@ } } /* ** The SQLITE_CONFIG_SQLLOG callback registered by sqlite3_init_sqllog(). +** +** The eType parameter has the following values: +** +** 0: Opening a new database connection. zSql is the name of the +** file being opened. db is a pointer to the newly created database +** connection. +** +** 1: An SQL statement has run to completion. zSql is the text of the +** SQL statement with all parameters expanded to their actual values. +** +** 2: Closing a database connection. zSql is NULL. The db pointer to +** the database connection being closed has already been shut down +** and cannot be used for any further SQL. +** +** The pCtx parameter is a copy of the pointer that was originally passed +** into the sqlite3_config(SQLITE_CONFIG_SQLLOG) statement. In this +** particular implementation, pCtx is always a pointer to the +** sqllogglobal global variable define above. */ static void testSqllog(void *pCtx, sqlite3 *db, const char *zSql, int eType){ struct SLConn *p = 0; sqlite3_mutex *master = sqlite3_mutex_alloc(SQLITE_MUTEX_STATIC_MASTER); Index: src/test_syscall.c ================================================================== --- src/test_syscall.c +++ src/test_syscall.c @@ -21,11 +21,11 @@ ** LIST must be a list consisting of zero or more of the following ** literal values: ** ** open close access getcwd stat fstat ** ftruncate fcntl read pread pread64 write -** pwrite pwrite64 fchmod fallocate +** pwrite pwrite64 fchmod fallocate mmap ** ** test_syscall uninstall ** Uninstall all wrapper functions. ** ** test_syscall fault ?COUNT PERSIST? @@ -76,13 +76,14 @@ #include #include "sqliteInt.h" #if SQLITE_OS_UNIX -/* From test1.c */ -extern const char *sqlite3TestErrorName(int); +/* From main.c */ +extern const char *sqlite3ErrName(int); +#include #include #include static struct TestSyscallGlobal { int bPersist; /* 1 for persistent errors, 0 for transient */ @@ -104,11 +105,12 @@ static int ts_write(int fd, const void *aBuf, size_t nBuf); static int ts_pwrite(int fd, const void *aBuf, size_t nBuf, off_t off); static int ts_pwrite64(int fd, const void *aBuf, size_t nBuf, off_t off); static int ts_fchmod(int fd, mode_t mode); static int ts_fallocate(int fd, off_t off, off_t len); - +static void *ts_mmap(void *, size_t, int, int, int, off_t); +static void *ts_mremap(void*, size_t, size_t, int, ...); struct TestSyscallArray { const char *zName; sqlite3_syscall_ptr xTest; sqlite3_syscall_ptr xOrig; @@ -129,10 +131,12 @@ /* 11 */ { "write", (sqlite3_syscall_ptr)ts_write, 0, 0, 0 }, /* 12 */ { "pwrite", (sqlite3_syscall_ptr)ts_pwrite, 0, 0, 0 }, /* 13 */ { "pwrite64", (sqlite3_syscall_ptr)ts_pwrite64, 0, 0, 0 }, /* 14 */ { "fchmod", (sqlite3_syscall_ptr)ts_fchmod, 0, 0, 0 }, /* 15 */ { "fallocate", (sqlite3_syscall_ptr)ts_fallocate, 0, 0, 0 }, + /* 16 */ { "mmap", (sqlite3_syscall_ptr)ts_mmap, 0, 0, 0 }, + /* 17 */ { "mremap", (sqlite3_syscall_ptr)ts_mremap, 0, 0, 0 }, { 0, 0, 0, 0, 0 } }; #define orig_open ((int(*)(const char *, int, int))aSyscall[0].xOrig) #define orig_close ((int(*)(int))aSyscall[1].xOrig) @@ -150,10 +154,12 @@ aSyscall[12].xOrig) #define orig_pwrite64 ((ssize_t(*)(int,const void*,size_t,off_t))\ aSyscall[13].xOrig) #define orig_fchmod ((int(*)(int,mode_t))aSyscall[14].xOrig) #define orig_fallocate ((int(*)(int,off_t,off_t))aSyscall[15].xOrig) +#define orig_mmap ((void*(*)(void*,size_t,int,int,int,off_t))aSyscall[16].xOrig) +#define orig_mremap ((void*(*)(void*,size_t,size_t,int,...))aSyscall[17].xOrig) /* ** This function is called exactly once from within each invocation of a ** system call wrapper in this file. It returns 1 if the function should ** fail, or 0 if it should succeed. @@ -374,10 +380,35 @@ if( tsIsFail() ){ return tsErrno("fallocate"); } return orig_fallocate(fd, off, len); } + +static void *ts_mmap( + void *pAddr, + size_t nByte, + int prot, + int flags, + int fd, + off_t iOff +){ + if( tsIsFailErrno("mmap") ){ + return MAP_FAILED; + } + return orig_mmap(pAddr, nByte, prot, flags, fd, iOff); +} + +static void *ts_mremap(void *a, size_t b, size_t c, int d, ...){ + va_list ap; + void *pArg; + if( tsIsFailErrno("mremap") ){ + return MAP_FAILED; + } + va_start(ap, d); + pArg = va_arg(ap, void *); + return orig_mremap(a, b, c, d, pArg); +} static int test_syscall_install( void * clientData, Tcl_Interp *interp, int objc, @@ -465,11 +496,11 @@ if( memcmp(aSyscall[i].zName, zFunc, nFunc) ) continue; aSyscall[i].xOrig = 0; } } if( rc!=SQLITE_OK ){ - Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3TestErrorName(rc), -1)); + Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3ErrName(rc), -1)); return TCL_ERROR; } Tcl_ResetResult(interp); return TCL_OK; Index: src/test_thread.c ================================================================== --- src/test_thread.c +++ src/test_thread.c @@ -58,16 +58,18 @@ static Tcl_ObjCmdProc blocking_prepare_v2_proc; #endif int Sqlitetest1_Init(Tcl_Interp *); int Sqlite3_Init(Tcl_Interp *); +/* Functions from main.c */ +extern const char *sqlite3ErrName(int); + /* Functions from test1.c */ -void *sqlite3TestTextToPtr(const char *); -const char *sqlite3TestErrorName(int); -int getDbPointer(Tcl_Interp *, const char *, sqlite3 **); -int sqlite3TestMakePointerStr(Tcl_Interp *, char *, void *); -int sqlite3TestErrCode(Tcl_Interp *, sqlite3 *, int); +extern void *sqlite3TestTextToPtr(const char *); +extern int getDbPointer(Tcl_Interp *, const char *, sqlite3 **); +extern int sqlite3TestMakePointerStr(Tcl_Interp *, char *, void *); +extern int sqlite3TestErrCode(Tcl_Interp *, sqlite3 *, int); /* ** Handler for events of type EvalEvent. */ static int tclScriptEvent(Tcl_Event *evPtr, int flags){ @@ -557,11 +559,11 @@ } pStmt = (sqlite3_stmt*)sqlite3TestTextToPtr(Tcl_GetString(objv[1])); rc = sqlite3_blocking_step(pStmt); - Tcl_SetResult(interp, (char *)sqlite3TestErrorName(rc), 0); + Tcl_SetResult(interp, (char *)sqlite3ErrName(rc), 0); return TCL_OK; } /* ** Usage: sqlite3_blocking_prepare_v2 DB sql bytes ?tailvar? @@ -604,11 +606,11 @@ } Tcl_ObjSetVar2(interp, objv[4], 0, Tcl_NewStringObj(zTail, bytes), 0); } if( rc!=SQLITE_OK ){ assert( pStmt==0 ); - sprintf(zBuf, "%s ", (char *)sqlite3TestErrorName(rc)); + sprintf(zBuf, "%s ", (char *)sqlite3ErrName(rc)); Tcl_AppendResult(interp, zBuf, sqlite3_errmsg(db), 0); return TCL_ERROR; } if( pStmt ){ Index: src/test_vfs.c ================================================================== --- src/test_vfs.c +++ src/test_vfs.c @@ -123,12 +123,13 @@ #define TESTVFS_WRITE_MASK 0x00001000 #define TESTVFS_TRUNCATE_MASK 0x00002000 #define TESTVFS_ACCESS_MASK 0x00004000 #define TESTVFS_FULLPATHNAME_MASK 0x00008000 #define TESTVFS_READ_MASK 0x00010000 +#define TESTVFS_UNLOCK_MASK 0x00020000 -#define TESTVFS_ALL_MASK 0x0001FFFF +#define TESTVFS_ALL_MASK 0x0003FFFF #define TESTVFS_MAX_PAGES 1024 /* @@ -465,12 +466,16 @@ /* ** Unlock an tvfs-file. */ static int tvfsUnlock(sqlite3_file *pFile, int eLock){ - TestvfsFd *p = tvfsGetFd(pFile); - return sqlite3OsUnlock(p->pReal, eLock); + TestvfsFd *pFd = tvfsGetFd(pFile); + Testvfs *p = (Testvfs *)pFd->pVfs->pAppData; + if( p->mask&TESTVFS_WRITE_MASK && tvfsInjectIoerr(p) ){ + return SQLITE_IOERR_UNLOCK; + } + return sqlite3OsUnlock(pFd->pReal, eLock); } /* ** Check if another file-handle holds a RESERVED lock on an tvfs-file. */ @@ -1099,10 +1104,11 @@ { "xTruncate", TESTVFS_TRUNCATE_MASK }, { "xOpen", TESTVFS_OPEN_MASK }, { "xClose", TESTVFS_CLOSE_MASK }, { "xAccess", TESTVFS_ACCESS_MASK }, { "xFullPathname", TESTVFS_FULLPATHNAME_MASK }, + { "xUnlock", TESTVFS_UNLOCK_MASK }, }; Tcl_Obj **apElem = 0; int nElem = 0; int i; int mask = 0; DELETED src/test_wholenumber.c Index: src/test_wholenumber.c ================================================================== --- src/test_wholenumber.c +++ src/test_wholenumber.c @@ -1,311 +0,0 @@ -/* -** 2011 April 02 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** -** This file implements a virtual table that returns the whole numbers -** between 1 and 4294967295, inclusive. -** -** Example: -** -** CREATE VIRTUAL TABLE nums USING wholenumber; -** SELECT value FROM nums WHERE value<10; -** -** Results in: -** -** 1 2 3 4 5 6 7 8 9 -*/ -#include "sqlite3.h" -#include -#include - -#ifndef SQLITE_OMIT_VIRTUALTABLE - - -/* A wholenumber cursor object */ -typedef struct wholenumber_cursor wholenumber_cursor; -struct wholenumber_cursor { - sqlite3_vtab_cursor base; /* Base class - must be first */ - sqlite3_int64 iValue; /* Current value */ - sqlite3_int64 mxValue; /* Maximum value */ -}; - -/* Methods for the wholenumber module */ -static int wholenumberConnect( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVtab, - char **pzErr -){ - sqlite3_vtab *pNew; - pNew = *ppVtab = sqlite3_malloc( sizeof(*pNew) ); - if( pNew==0 ) return SQLITE_NOMEM; - sqlite3_declare_vtab(db, "CREATE TABLE x(value)"); - memset(pNew, 0, sizeof(*pNew)); - return SQLITE_OK; -} -/* Note that for this virtual table, the xCreate and xConnect -** methods are identical. */ - -static int wholenumberDisconnect(sqlite3_vtab *pVtab){ - sqlite3_free(pVtab); - return SQLITE_OK; -} -/* The xDisconnect and xDestroy methods are also the same */ - - -/* -** Open a new wholenumber cursor. -*/ -static int wholenumberOpen(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ - wholenumber_cursor *pCur; - pCur = sqlite3_malloc( sizeof(*pCur) ); - if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); - *ppCursor = &pCur->base; - return SQLITE_OK; -} - -/* -** Close a wholenumber cursor. -*/ -static int wholenumberClose(sqlite3_vtab_cursor *cur){ - sqlite3_free(cur); - return SQLITE_OK; -} - - -/* -** Advance a cursor to its next row of output -*/ -static int wholenumberNext(sqlite3_vtab_cursor *cur){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - pCur->iValue++; - return SQLITE_OK; -} - -/* -** Return the value associated with a wholenumber. -*/ -static int wholenumberColumn( - sqlite3_vtab_cursor *cur, - sqlite3_context *ctx, - int i -){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - sqlite3_result_int64(ctx, pCur->iValue); - return SQLITE_OK; -} - -/* -** The rowid. -*/ -static int wholenumberRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - *pRowid = pCur->iValue; - return SQLITE_OK; -} - -/* -** When the wholenumber_cursor.rLimit value is 0 or less, that is a signal -** that the cursor has nothing more to output. -*/ -static int wholenumberEof(sqlite3_vtab_cursor *cur){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - return pCur->iValue>pCur->mxValue || pCur->iValue==0; -} - -/* -** Called to "rewind" a cursor back to the beginning so that -** it starts its output over again. Always called at least once -** prior to any wholenumberColumn, wholenumberRowid, or wholenumberEof call. -** -** idxNum Constraints -** ------ --------------------- -** 0 (none) -** 1 value > $argv0 -** 2 value >= $argv0 -** 4 value < $argv0 -** 8 value <= $argv0 -** -** 5 value > $argv0 AND value < $argv1 -** 6 value >= $argv0 AND value < $argv1 -** 9 value > $argv0 AND value <= $argv1 -** 10 value >= $argv0 AND value <= $argv1 -*/ -static int wholenumberFilter( - sqlite3_vtab_cursor *pVtabCursor, - int idxNum, const char *idxStr, - int argc, sqlite3_value **argv -){ - wholenumber_cursor *pCur = (wholenumber_cursor *)pVtabCursor; - sqlite3_int64 v; - int i = 0; - pCur->iValue = 1; - pCur->mxValue = 0xffffffff; /* 4294967295 */ - if( idxNum & 3 ){ - v = sqlite3_value_int64(argv[0]) + (idxNum&1); - if( v>pCur->iValue && v<=pCur->mxValue ) pCur->iValue = v; - i++; - } - if( idxNum & 12 ){ - v = sqlite3_value_int64(argv[i]) - ((idxNum>>2)&1); - if( v>=pCur->iValue && vmxValue ) pCur->mxValue = v; - } - return SQLITE_OK; -} - -/* -** Search for terms of these forms: -** -** (1) value > $value -** (2) value >= $value -** (4) value < $value -** (8) value <= $value -** -** idxNum is an ORed combination of 1 or 2 with 4 or 8. -*/ -static int wholenumberBestIndex( - sqlite3_vtab *tab, - sqlite3_index_info *pIdxInfo -){ - int i; - int idxNum = 0; - int argvIdx = 1; - int ltIdx = -1; - int gtIdx = -1; - const struct sqlite3_index_constraint *pConstraint; - pConstraint = pIdxInfo->aConstraint; - for(i=0; inConstraint; i++, pConstraint++){ - if( pConstraint->usable==0 ) continue; - if( (idxNum & 3)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_GT ){ - idxNum |= 1; - ltIdx = i; - } - if( (idxNum & 3)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_GE ){ - idxNum |= 2; - ltIdx = i; - } - if( (idxNum & 12)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ){ - idxNum |= 4; - gtIdx = i; - } - if( (idxNum & 12)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE ){ - idxNum |= 8; - gtIdx = i; - } - } - pIdxInfo->idxNum = idxNum; - if( ltIdx>=0 ){ - pIdxInfo->aConstraintUsage[ltIdx].argvIndex = argvIdx++; - pIdxInfo->aConstraintUsage[ltIdx].omit = 1; - } - if( gtIdx>=0 ){ - pIdxInfo->aConstraintUsage[gtIdx].argvIndex = argvIdx; - pIdxInfo->aConstraintUsage[gtIdx].omit = 1; - } - if( pIdxInfo->nOrderBy==1 - && pIdxInfo->aOrderBy[0].desc==0 - ){ - pIdxInfo->orderByConsumed = 1; - } - pIdxInfo->estimatedCost = (double)1; - return SQLITE_OK; -} - -/* -** A virtual table module that provides read-only access to a -** Tcl global variable namespace. -*/ -static sqlite3_module wholenumberModule = { - 0, /* iVersion */ - wholenumberConnect, - wholenumberConnect, - wholenumberBestIndex, - wholenumberDisconnect, - wholenumberDisconnect, - wholenumberOpen, /* xOpen - open a cursor */ - wholenumberClose, /* xClose - close a cursor */ - wholenumberFilter, /* xFilter - configure scan constraints */ - wholenumberNext, /* xNext - advance a cursor */ - wholenumberEof, /* xEof - check for end of scan */ - wholenumberColumn, /* xColumn - read data */ - wholenumberRowid, /* xRowid - read data */ - 0, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindMethod */ - 0, /* xRename */ -}; - -#endif /* SQLITE_OMIT_VIRTUALTABLE */ - - -/* -** Register the wholenumber virtual table -*/ -int wholenumber_register(sqlite3 *db){ - int rc = SQLITE_OK; -#ifndef SQLITE_OMIT_VIRTUALTABLE - rc = sqlite3_create_module(db, "wholenumber", &wholenumberModule, 0); -#endif - return rc; -} - -#ifdef SQLITE_TEST -#include -/* -** Decode a pointer to an sqlite3 object. -*/ -extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); - -/* -** Register the echo virtual table module. -*/ -static int register_wholenumber_module( - ClientData clientData, /* Pointer to sqlite3_enable_XXX function */ - Tcl_Interp *interp, /* The TCL interpreter that invoked this command */ - int objc, /* Number of arguments */ - Tcl_Obj *CONST objv[] /* Command arguments */ -){ - sqlite3 *db; - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; - wholenumber_register(db); - return TCL_OK; -} - - -/* -** Register commands with the TCL interpreter. -*/ -int Sqlitetestwholenumber_Init(Tcl_Interp *interp){ - static struct { - char *zName; - Tcl_ObjCmdProc *xProc; - void *clientData; - } aObjCmd[] = { - { "register_wholenumber_module", register_wholenumber_module, 0 }, - }; - int i; - for(i=0; i=pTab->nCol ){ if( sqlite3IsRowid(pChanges->a[i].zName) ){ + j = -1; chngRowid = 1; pRowidExpr = pChanges->a[i].pExpr; }else{ sqlite3ErrorMsg(pParse, "no such column: %s", pChanges->a[i].zName); pParse->checkSchema = 1; @@ -218,11 +219,12 @@ } #ifndef SQLITE_OMIT_AUTHORIZATION { int rc; rc = sqlite3AuthCheck(pParse, SQLITE_UPDATE, pTab->zName, - pTab->aCol[j].zName, db->aDb[iDb].zName); + j<0 ? "ROWID" : pTab->aCol[j].zName, + db->aDb[iDb].zName); if( rc==SQLITE_DENY ){ goto update_cleanup; }else if( rc==SQLITE_IGNORE ){ aXRef[j] = -1; } @@ -456,11 +458,11 @@ sqlite3CodeRowTrigger(pParse, pTrigger, TK_UPDATE, pChanges, TRIGGER_BEFORE, pTab, regOldRowid, onError, addr); /* The row-trigger may have deleted the row being updated. In this ** case, jump to the next row. No updates or AFTER triggers are - ** required. This behaviour - what happens when the row being updated + ** required. This behavior - what happens when the row being updated ** is deleted or renamed by a BEFORE trigger - is left undefined in the ** documentation. */ sqlite3VdbeAddOp3(v, OP_NotExists, iCur, addr, regOldRowid); Index: src/util.c ================================================================== --- src/util.c +++ src/util.c @@ -259,11 +259,11 @@ ** returns FALSE but it still converts the prefix and writes the result ** into *pResult. */ int sqlite3AtoF(const char *z, double *pResult, int length, u8 enc){ #ifndef SQLITE_OMIT_FLOATING_POINT - int incr = (enc==SQLITE_UTF8?1:2); + int incr; const char *zEnd = z + length; /* sign * significand * (10 ^ (esign * exponent)) */ int sign = 1; /* sign of significand */ i64 s = 0; /* significand */ int d = 0; /* adjust exponent for shifting decimal point */ @@ -270,14 +270,26 @@ int esign = 1; /* sign of exponent */ int e = 0; /* exponent */ int eValid = 1; /* True exponent is either not used or is well-formed */ double result; int nDigits = 0; + int nonNum = 0; + assert( enc==SQLITE_UTF8 || enc==SQLITE_UTF16LE || enc==SQLITE_UTF16BE ); *pResult = 0.0; /* Default return value, in case of an error */ - if( enc==SQLITE_UTF16BE ) z++; + if( enc==SQLITE_UTF8 ){ + incr = 1; + }else{ + int i; + incr = 2; + assert( SQLITE_UTF16LE==2 && SQLITE_UTF16BE==3 ); + for(i=3-enc; i=zEnd ) return 0; @@ -406,11 +418,11 @@ /* store the result */ *pResult = result; /* return true if number and no extra non-whitespace chracters after */ - return z>=zEnd && nDigits>0 && eValid; + return z>=zEnd && nDigits>0 && eValid && nonNum==0; #else return !sqlite3Atoi64(z, pResult, length, enc); #endif /* SQLITE_OMIT_FLOATING_POINT */ } @@ -455,25 +467,37 @@ ** If zNum is exactly 9223372036854665808, return 2. This special ** case is broken out because while 9223372036854665808 cannot be a ** signed 64-bit integer, its negative -9223372036854665808 can be. ** ** If zNum is too big for a 64-bit integer and is not -** 9223372036854665808 then return 1. +** 9223372036854665808 or if zNum contains any non-numeric text, +** then return 1. ** ** length is the number of bytes in the string (bytes, not characters). ** The string is not necessarily zero-terminated. The encoding is ** given by enc. */ int sqlite3Atoi64(const char *zNum, i64 *pNum, int length, u8 enc){ - int incr = (enc==SQLITE_UTF8?1:2); + int incr; u64 u = 0; int neg = 0; /* assume positive */ int i; int c = 0; + int nonNum = 0; const char *zStart; const char *zEnd = zNum + length; - if( enc==SQLITE_UTF16BE ) zNum++; + assert( enc==SQLITE_UTF8 || enc==SQLITE_UTF16LE || enc==SQLITE_UTF16BE ); + if( enc==SQLITE_UTF8 ){ + incr = 1; + }else{ + incr = 2; + assert( SQLITE_UTF16LE==2 && SQLITE_UTF16BE==3 ); + for(i=3-enc; i19*incr ){ + if( (c!=0 && &zNum[i]19*incr || nonNum ){ /* zNum is empty or contains non-numeric text or is longer ** than 19 digits (thus guaranteeing that it is too large) */ return 1; }else if( i<19*incr ){ /* Less than 19 digits, so we know that it fits in 64 bits */ Index: src/vacuum.c ================================================================== --- src/vacuum.c +++ src/vacuum.c @@ -287,10 +287,11 @@ static const unsigned char aCopy[] = { BTREE_SCHEMA_VERSION, 1, /* Add one to the old schema cookie */ BTREE_DEFAULT_CACHE_SIZE, 0, /* Preserve the default page cache size */ BTREE_TEXT_ENCODING, 0, /* Preserve the text encoding */ BTREE_USER_VERSION, 0, /* Preserve the user version */ + BTREE_APPLICATION_ID, 0, /* Preserve the application id */ }; assert( 1==sqlite3BtreeIsInTrans(pTemp) ); assert( 1==sqlite3BtreeIsInTrans(pMain) ); Index: src/vdbe.c ================================================================== --- src/vdbe.c +++ src/vdbe.c @@ -150,15 +150,11 @@ #define Deephemeralize(P) \ if( ((P)->flags&MEM_Ephem)!=0 \ && sqlite3VdbeMemMakeWriteable(P) ){ goto no_mem;} /* Return true if the cursor was opened using the OP_OpenSorter opcode. */ -#ifdef SQLITE_OMIT_MERGE_SORT -# define isSorter(x) 0 -#else # define isSorter(x) ((x)->pSorter!=0) -#endif /* ** Argument pMem points at a register that will be passed to a ** user-defined function or returned to the user as the result of a query. ** This routine sets the pMem->type variable used by the sqlite3_value_*() @@ -3319,21 +3315,16 @@ ** tables using an external merge-sort algorithm. */ case OP_SorterOpen: { VdbeCursor *pCx; -#ifndef SQLITE_OMIT_MERGE_SORT pCx = allocateCursor(p, pOp->p1, pOp->p2, -1, 1); if( pCx==0 ) goto no_mem; pCx->pKeyInfo = pOp->p4.pKeyInfo; pCx->pKeyInfo->enc = ENC(p->db); pCx->isSorter = 1; rc = sqlite3VdbeSorterInit(db, pCx); -#else - pOp->opcode = OP_OpenEphemeral; - pc--; -#endif break; } /* Opcode: OpenPseudo P1 P2 P3 * P5 ** @@ -3522,11 +3513,11 @@ ** r.flags = UNPACKED_INCRKEY; ** }else{ ** r.flags = 0; ** } */ - r.flags = (u16)(UNPACKED_INCRKEY * (1 & (oc - OP_SeekLt))); + r.flags = (u8)(UNPACKED_INCRKEY * (1 & (oc - OP_SeekLt))); assert( oc!=OP_SeekGt || r.flags==UNPACKED_INCRKEY ); assert( oc!=OP_SeekLe || r.flags==UNPACKED_INCRKEY ); assert( oc!=OP_SeekGe || r.flags==0 ); assert( oc!=OP_SeekLt || r.flags==0 ); @@ -4212,19 +4203,14 @@ ** Write into register P2 the current sorter data for sorter cursor P1. */ case OP_SorterData: { VdbeCursor *pC; -#ifndef SQLITE_OMIT_MERGE_SORT pOut = &aMem[pOp->p2]; pC = p->apCsr[pOp->p1]; assert( pC->isSorter ); rc = sqlite3VdbeSorterRowkey(pC, pOut); -#else - pOp->opcode = OP_RowKey; - pc--; -#endif break; } /* Opcode: RowData P1 P2 * * * ** @@ -4419,13 +4405,10 @@ ** rewinding so that the global variable will be incremented and ** regression tests can determine whether or not the optimizer is ** correctly optimizing out sorts. */ case OP_SorterSort: /* jump */ -#ifdef SQLITE_OMIT_MERGE_SORT - pOp->opcode = OP_Sort; -#endif case OP_Sort: { /* jump */ #ifdef SQLITE_TEST sqlite3_sort_count++; sqlite3_search_count--; #endif @@ -4500,13 +4483,10 @@ ** ** If P5 is positive and the jump is taken, then event counter ** number P5-1 in the prepared statement is incremented. */ case OP_SorterNext: /* jump */ -#ifdef SQLITE_OMIT_MERGE_SORT - pOp->opcode = OP_Next; -#endif case OP_Prev: /* jump */ case OP_Next: { /* jump */ VdbeCursor *pC; int res; @@ -4553,13 +4533,10 @@ ** ** This instruction only works for indices. The equivalent instruction ** for tables is OP_Insert. */ case OP_SorterInsert: /* in2 */ -#ifdef SQLITE_OMIT_MERGE_SORT - pOp->opcode = OP_IdxInsert; -#endif case OP_IdxInsert: { /* in2 */ VdbeCursor *pC; BtCursor *pCrsr; int nKey; const char *zKey; @@ -5782,11 +5759,11 @@ importVtabErrMsg(p, pVtab); if( SQLITE_OK==rc ){ /* Initialize sqlite3_vtab_cursor base class */ pVtabCursor->pVtab = pVtab; - /* Initialise vdbe cursor object */ + /* Initialize vdbe cursor object */ pCur = allocateCursor(p, pOp->p1, 0, -1, 0); if( pCur ){ pCur->pVtabCursor = pVtabCursor; pCur->pModule = pVtabCursor->pVtab->pModule; }else{ Index: src/vdbeInt.h ================================================================== --- src/vdbeInt.h +++ src/vdbeInt.h @@ -16,10 +16,18 @@ ** this header information was factored out. */ #ifndef _VDBEINT_H_ #define _VDBEINT_H_ +/* +** The maximum number of times that a statement will try to reparse +** itself before giving up and returning SQLITE_SCHEMA. +*/ +#ifndef SQLITE_MAX_SCHEMA_RETRY +# define SQLITE_MAX_SCHEMA_RETRY 50 +#endif + /* ** SQL is translated into a sequence of instructions to be ** executed by a virtual machine. Each instruction is an instance ** of the following structure. */ @@ -427,27 +435,17 @@ void sqlite3VdbeFrameDelete(VdbeFrame*); int sqlite3VdbeFrameRestore(VdbeFrame *); void sqlite3VdbeMemStoreType(Mem *pMem); int sqlite3VdbeTransferError(Vdbe *p); -#ifdef SQLITE_OMIT_MERGE_SORT -# define sqlite3VdbeSorterInit(Y,Z) SQLITE_OK -# define sqlite3VdbeSorterWrite(X,Y,Z) SQLITE_OK -# define sqlite3VdbeSorterClose(Y,Z) -# define sqlite3VdbeSorterRowkey(Y,Z) SQLITE_OK -# define sqlite3VdbeSorterRewind(X,Y,Z) SQLITE_OK -# define sqlite3VdbeSorterNext(X,Y,Z) SQLITE_OK -# define sqlite3VdbeSorterCompare(X,Y,Z) SQLITE_OK -#else int sqlite3VdbeSorterInit(sqlite3 *, VdbeCursor *); void sqlite3VdbeSorterClose(sqlite3 *, VdbeCursor *); int sqlite3VdbeSorterRowkey(const VdbeCursor *, Mem *); int sqlite3VdbeSorterNext(sqlite3 *, const VdbeCursor *, int *); int sqlite3VdbeSorterRewind(sqlite3 *, const VdbeCursor *, int *); int sqlite3VdbeSorterWrite(sqlite3 *, const VdbeCursor *, Mem *); int sqlite3VdbeSorterCompare(const VdbeCursor *, Mem *, int *); -#endif #if !defined(SQLITE_OMIT_SHARED_CACHE) && SQLITE_THREADSAFE>0 void sqlite3VdbeEnter(Vdbe*); void sqlite3VdbeLeave(Vdbe*); #else Index: src/vdbeapi.c ================================================================== --- src/vdbeapi.c +++ src/vdbeapi.c @@ -443,26 +443,18 @@ || rc==SQLITE_BUSY || rc==SQLITE_MISUSE ); assert( p->rc!=SQLITE_ROW && p->rc!=SQLITE_DONE ); if( p->isPrepareV2 && rc!=SQLITE_ROW && rc!=SQLITE_DONE ){ /* If this statement was prepared using sqlite3_prepare_v2(), and an - ** error has occured, then return the error code in p->rc to the + ** error has occurred, then return the error code in p->rc to the ** caller. Set the error code in the database handle to the same value. */ rc = sqlite3VdbeTransferError(p); } return (rc&db->errMask); } -/* -** The maximum number of times that a statement will try to reparse -** itself before giving up and returning SQLITE_SCHEMA. -*/ -#ifndef SQLITE_MAX_SCHEMA_RETRY -# define SQLITE_MAX_SCHEMA_RETRY 5 -#endif - /* ** This is the top-level implementation of sqlite3_step(). Call ** sqlite3Step() to do most of the work. If a schema error occurs, ** call sqlite3Reprepare() and try again. */ Index: src/vdbeaux.c ================================================================== --- src/vdbeaux.c +++ src/vdbeaux.c @@ -373,11 +373,11 @@ break; } } sqlite3DbFree(v->db, sIter.apSub); - /* Return true if hasAbort==mayAbort. Or if a malloc failure occured. + /* Return true if hasAbort==mayAbort. Or if a malloc failure occurred. ** If malloc failed, then the while() loop above may not have iterated ** through all opcodes and hasAbort may be set incorrectly. Return ** true for this case to prevent the assert() in the callers frame ** from failing. */ return ( v->db->mallocFailed || hasAbort==mayAbort ); @@ -2005,11 +2005,11 @@ sqlite3 *const db = p->db; int rc = SQLITE_OK; /* If p->iStatement is greater than zero, then this Vdbe opened a ** statement transaction that should be closed here. The only exception - ** is that an IO error may have occured, causing an emergency rollback. + ** is that an IO error may have occurred, causing an emergency rollback. ** In this case (db->nStatement==0), and there is nothing to do. */ if( db->nStatement && p->iStatement ){ int i; const int iSavepoint = p->iStatement-1; @@ -2141,11 +2141,11 @@ ** transaction must be rolled back to restore the database to a ** consistent state. ** ** Even if the statement is read-only, it is important to perform ** a statement or transaction rollback operation. If the error - ** occured while writing to the journal, sub-journal or database + ** occurred while writing to the journal, sub-journal or database ** file as part of an effort to free up cache space (see function ** pagerStress() in pager.c), the rollback is required to restore ** the pager to a consistent state. */ if( !p->readOnly || mrc!=SQLITE_INTERRUPT ){ @@ -2555,11 +2555,11 @@ ** ** In an SQLite index record, the serial type is stored directly before ** the blob of data that it corresponds to. In a table record, all serial ** types are stored at the start of the record, and the blobs of data at ** the end. Hence these functions allow the caller to handle the -** serial-type and data blob seperately. +** serial-type and data blob separately. ** ** The following table describes the various storage classes for data: ** ** serial type bytes of data type ** -------------- --------------- --------------- Index: src/vdbeblob.c ================================================================== --- src/vdbeblob.c +++ src/vdbeblob.c @@ -311,11 +311,11 @@ if( db->mallocFailed ){ goto blob_open_out; } sqlite3_bind_int64(pBlob->pStmt, 1, iRow); rc = blobSeekToRow(pBlob, iRow, &zErr); - } while( (++nAttempt)<5 && rc==SQLITE_SCHEMA ); + } while( (++nAttempt)mallocFailed==0 ){ *ppBlob = (sqlite3_blob *)pBlob; }else{ Index: src/vdbesort.c ================================================================== --- src/vdbesort.c +++ src/vdbesort.c @@ -16,11 +16,10 @@ */ #include "sqliteInt.h" #include "vdbeInt.h" -#ifndef SQLITE_OMIT_MERGE_SORT typedef struct VdbeSorterIter VdbeSorterIter; typedef struct SorterRecord SorterRecord; typedef struct FileWriter FileWriter; @@ -1035,7 +1034,5 @@ pKey = vdbeSorterRowkey(pSorter, &nKey); vdbeSorterCompare(pCsr, 1, pVal->z, pVal->n, pKey, nKey, pRes); return SQLITE_OK; } - -#endif /* #ifndef SQLITE_OMIT_MERGE_SORT */ Index: src/vdbetrace.c ================================================================== --- src/vdbetrace.c +++ src/vdbetrace.c @@ -50,10 +50,15 @@ ** obtained from sqlite3DbMalloc(). If sqlite3.vdbeExecCnt is 1, then the ** string contains a copy of zRawSql but with host parameters expanded to ** their current bindings. Or, if sqlite3.vdbeExecCnt is greater than 1, ** then the returned string holds a copy of zRawSql with "-- " prepended ** to each line of text. +** +** If the SQLITE_TRACE_SIZE_LIMIT macro is defined to an integer, then +** then long strings and blobs are truncated to that many bytes. This +** can be used to prevent unreasonably large trace strings when dealing +** with large (multi-megabyte) strings and blobs. ** ** The calling function is responsible for making sure the memory returned ** is eventually freed. ** ** ALGORITHM: Scan the input string looking for host parameters in any of @@ -121,34 +126,53 @@ }else if( pVar->flags & MEM_Int ){ sqlite3XPrintf(&out, "%lld", pVar->u.i); }else if( pVar->flags & MEM_Real ){ sqlite3XPrintf(&out, "%!.15g", pVar->r); }else if( pVar->flags & MEM_Str ){ + int nOut; /* Number of bytes of the string text to include in output */ #ifndef SQLITE_OMIT_UTF16 u8 enc = ENC(db); + Mem utf8; if( enc!=SQLITE_UTF8 ){ - Mem utf8; memset(&utf8, 0, sizeof(utf8)); utf8.db = db; sqlite3VdbeMemSetStr(&utf8, pVar->z, pVar->n, enc, SQLITE_STATIC); sqlite3VdbeChangeEncoding(&utf8, SQLITE_UTF8); - sqlite3XPrintf(&out, "'%.*q'", utf8.n, utf8.z); - sqlite3VdbeMemRelease(&utf8); - }else + pVar = &utf8; + } #endif - { - sqlite3XPrintf(&out, "'%.*q'", pVar->n, pVar->z); + nOut = pVar->n; +#ifdef SQLITE_TRACE_SIZE_LIMIT + if( n>SQLITE_TRACE_SIZE_LIMIT ){ + nOut = SQLITE_TRACE_SIZE_LIMIT; + while( nOutn && (pVar->z[n]&0xc0)==0x80 ){ n++; } } +#endif + sqlite3XPrintf(&out, "'%.*q'", nOut, pVar->z); +#ifdef SQLITE_TRACE_SIZE_LIMIT + if( nOutn ) sqlite3XPrintf(&out, "/*+%d bytes*/", pVar->n-n); +#endif +#ifndef SQLITE_OMIT_UTF16 + if( enc!=SQLITE_UTF8 ) sqlite3VdbeMemRelease(&utf8); +#endif }else if( pVar->flags & MEM_Zero ){ sqlite3XPrintf(&out, "zeroblob(%d)", pVar->u.nZero); }else{ + int nOut; /* Number of bytes of the blob to include in output */ assert( pVar->flags & MEM_Blob ); sqlite3StrAccumAppend(&out, "x'", 2); - for(i=0; in; i++){ + nOut = pVar->n; +#ifdef SQLITE_TRACE_SIZE_LIMIT + if( nOut>SQLITE_TRACE_SIZE_LIMIT ) nOut = SQLITE_TRACE_SIZE_LIMIT; +#endif + for(i=0; iz[i]&0xff); } sqlite3StrAccumAppend(&out, "'", 1); +#ifdef SQLITE_TRACE_SIZE_LIMIT + if( nOutn ) sqlite3XPrintf(&out, "/*+%d bytes*/", pVar->n-n); +#endif } } } return sqlite3StrAccumFinish(&out); } Index: src/wal.c ================================================================== --- src/wal.c +++ src/wal.c @@ -1205,12 +1205,13 @@ ** event via sqlite3_log(). This is to help with identifying performance ** problems caused by applications routinely shutting down without ** checkpointing the log file. */ if( pWal->hdr.nPage ){ - sqlite3_log(SQLITE_OK, "Recovered %d frames from WAL file %s", - pWal->hdr.nPage, pWal->zWalName + sqlite3_log(SQLITE_NOTICE_RECOVER_WAL, + "recovered %d frames from WAL file %s", + pWal->hdr.mxFrame, pWal->zWalName ); } } recovery_error: @@ -1720,20 +1721,21 @@ /* Sync the WAL to disk */ if( sync_flags ){ rc = sqlite3OsSync(pWal->pWalFd, sync_flags); } - /* If the database file may grow as a result of this checkpoint, hint - ** about the eventual size of the db file to the VFS layer. + /* If the database may grow as a result of this checkpoint, hint + ** about the eventual size of the db file to the VFS layer. */ if( rc==SQLITE_OK ){ i64 nReq = ((i64)mxPage * szPage); rc = sqlite3OsFileSize(pWal->pDbFd, &nSize); if( rc==SQLITE_OK && nSizepDbFd, SQLITE_FCNTL_SIZE_HINT, &nReq); } } + /* Iterate through the contents of the WAL, copying data to the db file. */ while( rc==SQLITE_OK && 0==walIteratorNext(pIter, &iDbpage, &iFrame) ){ i64 iOffset; assert( walFramePgno(pWal, iFrame)==iDbpage ); @@ -2285,23 +2287,21 @@ pWal->readLock = -1; } } /* -** Read a page from the WAL, if it is present in the WAL and if the -** current read transaction is configured to use the WAL. +** Search the wal file for page pgno. If found, set *piRead to the frame that +** contains the page. Otherwise, if pgno is not in the wal file, set *piRead +** to zero. ** -** The *pInWal is set to 1 if the requested page is in the WAL and -** has been loaded. Or *pInWal is set to 0 if the page was not in -** the WAL and needs to be read out of the database. +** Return SQLITE_OK if successful, or an error code if an error occurs. If an +** error does occur, the final value of *piRead is undefined. */ -int sqlite3WalRead( +int sqlite3WalFindFrame( Wal *pWal, /* WAL handle */ Pgno pgno, /* Database page number to read data for */ - int *pInWal, /* OUT: True if data is read from WAL */ - int nOut, /* Size of buffer pOut in bytes */ - u8 *pOut /* Buffer to write page data to */ + u32 *piRead /* OUT: Frame number (or zero) */ ){ u32 iRead = 0; /* If !=0, WAL frame to return data from */ u32 iLast = pWal->hdr.mxFrame; /* Last page in WAL for this reader */ int iHash; /* Used to loop through N hash tables */ @@ -2313,11 +2313,11 @@ ** in this case as an optimization. Likewise, if pWal->readLock==0, ** then the WAL is ignored by the reader so return early, as if the ** WAL were empty. */ if( iLast==0 || pWal->readLock==0 ){ - *pInWal = 0; + *piRead = 0; return SQLITE_OK; } /* Search the hash table or tables for an entry matching page number ** pgno. Each iteration of the following for() loop searches one @@ -2384,30 +2384,35 @@ } assert( iRead==iRead2 ); } #endif - /* If iRead is non-zero, then it is the log frame number that contains the - ** required page. Read and return data from the log file. - */ - if( iRead ){ - int sz; - i64 iOffset; - sz = pWal->hdr.szPage; - sz = (sz&0xfe00) + ((sz&0x0001)<<16); - testcase( sz<=32768 ); - testcase( sz>=65536 ); - iOffset = walFrameOffset(iRead, sz) + WAL_FRAME_HDRSIZE; - *pInWal = 1; - /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL */ - return sqlite3OsRead(pWal->pWalFd, pOut, (nOut>sz ? sz : nOut), iOffset); - } - - *pInWal = 0; + *piRead = iRead; return SQLITE_OK; } +/* +** Read the contents of frame iRead from the wal file into buffer pOut +** (which is nOut bytes in size). Return SQLITE_OK if successful, or an +** error code otherwise. +*/ +int sqlite3WalReadFrame( + Wal *pWal, /* WAL handle */ + u32 iRead, /* Frame to read */ + int nOut, /* Size of buffer pOut in bytes */ + u8 *pOut /* Buffer to write page data to */ +){ + int sz; + i64 iOffset; + sz = pWal->hdr.szPage; + sz = (sz&0xfe00) + ((sz&0x0001)<<16); + testcase( sz<=32768 ); + testcase( sz>=65536 ); + iOffset = walFrameOffset(iRead, sz) + WAL_FRAME_HDRSIZE; + /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL */ + return sqlite3OsRead(pWal->pWalFd, pOut, (nOut>sz ? sz : nOut), iOffset); +} /* ** Return the size of the database in pages (or zero, if unknown). */ Pgno sqlite3WalDbsize(Wal *pWal){ @@ -2950,10 +2955,13 @@ } /* Read the wal-index header. */ if( rc==SQLITE_OK ){ rc = walIndexReadHdr(pWal, &isChanged); + if( isChanged && pWal->pDbFd->pMethods->iVersion>=3 ){ + sqlite3OsUnfetch(pWal->pDbFd, 0, 0); + } } /* Copy data from the log to the database file. */ if( rc==SQLITE_OK ){ if( pWal->hdr.mxFrame && walPagesize(pWal)!=nBuf ){ Index: src/wal.h ================================================================== --- src/wal.h +++ src/wal.h @@ -29,11 +29,10 @@ # define sqlite3WalOpen(x,y,z) 0 # define sqlite3WalLimit(x,y) # define sqlite3WalClose(w,x,y,z) 0 # define sqlite3WalBeginReadTransaction(y,z) 0 # define sqlite3WalEndReadTransaction(z) -# define sqlite3WalRead(v,w,x,y,z) 0 # define sqlite3WalDbsize(y) 0 # define sqlite3WalBeginWriteTransaction(y) 0 # define sqlite3WalEndWriteTransaction(x) 0 # define sqlite3WalUndo(x,y,z) 0 # define sqlite3WalSavepoint(y,z) @@ -42,10 +41,11 @@ # define sqlite3WalCheckpoint(r,s,t,u,v,w,x,y,z) 0 # define sqlite3WalCallback(z) 0 # define sqlite3WalExclusiveMode(y,z) 0 # define sqlite3WalHeapMemory(z) 0 # define sqlite3WalFramesize(z) 0 +# define sqlite3WalFindFrame(x,y,z) 0 #else #define WAL_SAVEPOINT_NDATA 4 /* Connection to a write-ahead log (WAL) file. @@ -69,11 +69,12 @@ */ int sqlite3WalBeginReadTransaction(Wal *pWal, int *); void sqlite3WalEndReadTransaction(Wal *pWal); /* Read a page from the write-ahead log, if it is present. */ -int sqlite3WalRead(Wal *pWal, Pgno pgno, int *pInWal, int nOut, u8 *pOut); +int sqlite3WalFindFrame(Wal *, Pgno, u32 *); +int sqlite3WalReadFrame(Wal *, u32, int, u8 *); /* If the WAL is not empty, return the size of the database. */ Pgno sqlite3WalDbsize(Wal *pWal); /* Obtain or release the WRITER lock. */ Index: src/walker.c ================================================================== --- src/walker.c +++ src/walker.c @@ -111,11 +111,13 @@ } /* ** Call sqlite3WalkExpr() for every expression in Select statement p. ** Invoke sqlite3WalkSelect() for subqueries in the FROM clause and -** on the compound select chain, p->pPrior. +** on the compound select chain, p->pPrior. Invoke the xSelectCallback() +** either before or after the walk of expressions and FROM clause, depending +** on whether pWalker->bSelectDepthFirst is false or true, respectively. ** ** Return WRC_Continue under normal conditions. Return WRC_Abort if ** there is an abort request. ** ** If the Walker does not have an xSelectCallback() then this routine @@ -125,18 +127,27 @@ int rc; if( p==0 || pWalker->xSelectCallback==0 ) return WRC_Continue; rc = WRC_Continue; pWalker->walkerDepth++; while( p ){ - rc = pWalker->xSelectCallback(pWalker, p); - if( rc ) break; + if( !pWalker->bSelectDepthFirst ){ + rc = pWalker->xSelectCallback(pWalker, p); + if( rc ) break; + } if( sqlite3WalkSelectExpr(pWalker, p) || sqlite3WalkSelectFrom(pWalker, p) ){ pWalker->walkerDepth--; return WRC_Abort; + } + if( pWalker->bSelectDepthFirst ){ + rc = pWalker->xSelectCallback(pWalker, p); + /* Depth-first search is currently only used for + ** selectAddSubqueryTypeInfo() and that routine always returns + ** WRC_Continue (0). So the following branch is never taken. */ + if( NEVER(rc) ) break; } p = p->pPrior; } pWalker->walkerDepth--; return rc & WRC_Abort; } Index: src/where.c ================================================================== --- src/where.c +++ src/where.c @@ -260,10 +260,12 @@ #define WHERE_IDX_ONLY 0x00400000 /* Use index only - omit table */ #define WHERE_ORDERED 0x00800000 /* Output will appear in correct order */ #define WHERE_REVERSE 0x01000000 /* Scan in reverse order */ #define WHERE_UNIQUE 0x02000000 /* Selects no more than one row */ #define WHERE_ALL_UNIQUE 0x04000000 /* This and all prior have one row */ +#define WHERE_OB_UNIQUE 0x00004000 /* Values in ORDER BY columns are + ** different for every output row */ #define WHERE_VIRTUALTABLE 0x08000000 /* Use virtual-table processing */ #define WHERE_MULTI_OR 0x10000000 /* OR using multiple indices */ #define WHERE_TEMP_INDEX 0x20000000 /* Uses an ephemeral index */ #define WHERE_DISTINCT 0x40000000 /* Correct order for DISTINCT */ #define WHERE_COVER_SCAN 0x80000000 /* Full scan of a covering index */ @@ -560,11 +562,11 @@ /* ** Commute a comparison operator. Expressions of the form "X op Y" ** are converted into "Y op X". ** -** If left/right precendence rules come into play when determining the +** If left/right precedence rules come into play when determining the ** collating ** side of the comparison, it remains associated with the same side after ** the commutation. So "Y collate NOCASE op X" becomes ** "X op Y". This is because any collation sequence on ** the left hand side of a comparison overrides any collation sequence @@ -701,11 +703,11 @@ } if( sqlite3StrICmp(pColl->zName, pIdx->azColl[j]) ){ continue; } } - if( pTerm->prereqRight==0 ){ + if( pTerm->prereqRight==0 && (pTerm->eOperator&WO_EQ)!=0 ){ pResult = pTerm; goto findTerm_success; }else if( pResult==0 ){ pResult = pTerm; } @@ -2271,13 +2273,12 @@ Table *pTab = pSrc->pTab; sqlite3_index_info *pIdxInfo; struct sqlite3_index_constraint *pIdxCons; struct sqlite3_index_constraint_usage *pUsage; WhereTerm *pTerm; - int i, j, k; + int i, j; int nOrderBy; - int sortOrder; /* Sort order for IN clauses */ int bAllowIN; /* Allow IN optimizations */ double rCost; /* Make sure wsFlags is initialized to some sane value. Otherwise, if the ** malloc in allocateIndexInfo() fails and this function returns leaving @@ -2372,11 +2373,10 @@ if( vtabBestIndex(pParse, pTab, pIdxInfo) ){ return; } - sortOrder = SQLITE_SO_ASC; pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; for(i=0; inConstraint; i++, pIdxCons++){ if( pUsage[i].argvIndex>0 ){ j = pIdxCons->iTermOffset; pTerm = &pWC->a[j]; @@ -2387,21 +2387,32 @@ ** says that the equivalent EQ constraint cannot be safely omitted. ** If we do attempt to use such a constraint, some rows might be ** repeated in the output. */ break; } - for(k=0; knOrderBy; k++){ - if( pIdxInfo->aOrderBy[k].iColumn==pIdxCons->iColumn ){ - sortOrder = pIdxInfo->aOrderBy[k].desc; - break; - } - } + /* A virtual table that is constrained by an IN clause may not + ** consume the ORDER BY clause because (1) the order of IN terms + ** is not necessarily related to the order of output terms and + ** (2) Multiple outputs from a single IN value will not merge + ** together. */ + pIdxInfo->orderByConsumed = 0; } } } if( i>=pIdxInfo->nConstraint ) break; } + + /* The orderByConsumed signal is only valid if all outer loops collectively + ** generate just a single row of output. + */ + if( pIdxInfo->orderByConsumed ){ + for(i=0; ii; i++){ + if( (p->aLevel[i].plan.wsFlags & WHERE_UNIQUE)==0 ){ + pIdxInfo->orderByConsumed = 0; + } + } + } /* If there is an ORDER BY clause, and the selected virtual table index ** does not satisfy it, increase the cost of the scan accordingly. This ** matches the processing for non-virtual tables in bestBtreeIndex(). */ @@ -2422,12 +2433,11 @@ }else{ p->cost.rCost = rCost; } p->cost.plan.u.pVtabIdx = pIdxInfo; if( pIdxInfo->orderByConsumed ){ - assert( sortOrder==0 || sortOrder==1 ); - p->cost.plan.wsFlags |= WHERE_ORDERED + sortOrder*WHERE_REVERSE; + p->cost.plan.wsFlags |= WHERE_ORDERED; p->cost.plan.nOBSat = nOrderBy; }else{ p->cost.plan.nOBSat = p->i ? p->aLevel[p->i-1].plan.nOBSat : 0; } p->cost.plan.nEq = 0; @@ -2901,11 +2911,12 @@ */ static int isSortingIndex( WhereBestIdx *p, /* Best index search context */ Index *pIdx, /* The index we are testing */ int base, /* Cursor number for the table to be sorted */ - int *pbRev /* Set to 1 for reverse-order scan of pIdx */ + int *pbRev, /* Set to 1 for reverse-order scan of pIdx */ + int *pbObUnique /* ORDER BY column values will different in every row */ ){ int i; /* Number of pIdx terms used */ int j; /* Number of ORDER BY terms satisfied */ int sortOrder = 2; /* 0: forward. 1: backward. 2: unknown */ int nTerm; /* Number of ORDER BY terms */ @@ -2915,25 +2926,32 @@ Parse *pParse = p->pParse; /* Parser context */ sqlite3 *db = pParse->db; /* Database connection */ int nPriorSat; /* ORDER BY terms satisfied by outer loops */ int seenRowid = 0; /* True if an ORDER BY rowid term is seen */ int uniqueNotNull; /* pIdx is UNIQUE with all terms are NOT NULL */ + int outerObUnique; /* Outer loops generate different values in + ** every row for the ORDER BY columns */ if( p->i==0 ){ nPriorSat = 0; + outerObUnique = 1; }else{ + u32 wsFlags = p->aLevel[p->i-1].plan.wsFlags; nPriorSat = p->aLevel[p->i-1].plan.nOBSat; - if( (p->aLevel[p->i-1].plan.wsFlags & WHERE_ORDERED)==0 ){ + if( (wsFlags & WHERE_ORDERED)==0 ){ /* This loop cannot be ordered unless the next outer loop is ** also ordered */ return nPriorSat; } if( OptimizationDisabled(db, SQLITE_OrderByIdxJoin) ){ /* Only look at the outer-most loop if the OrderByIdxJoin ** optimization is disabled */ return nPriorSat; } + testcase( wsFlags & WHERE_OB_UNIQUE ); + testcase( wsFlags & WHERE_ALL_UNIQUE ); + outerObUnique = (wsFlags & (WHERE_OB_UNIQUE|WHERE_ALL_UNIQUE))!=0; } pOrderBy = p->pOrderBy; assert( pOrderBy!=0 ); if( pIdx->bUnordered ){ /* Hash indices (indicated by the "unordered" tag on sqlite_stat1) cannot @@ -3071,23 +3089,38 @@ testcase( isEq==2 ); testcase( isEq==3 ); uniqueNotNull = 0; } } + if( seenRowid ){ + uniqueNotNull = 1; + }else if( uniqueNotNull==0 || inColumn ){ + uniqueNotNull = 0; + } /* If we have not found at least one ORDER BY term that matches the ** index, then show no progress. */ if( pOBItem==&pOrderBy->a[nPriorSat] ) return nPriorSat; + + /* Either the outer queries must generate rows where there are no two + ** rows with the same values in all ORDER BY columns, or else this + ** loop must generate just a single row of output. Example: Suppose + ** the outer loops generate A=1 and A=1, and this loop generates B=3 + ** and B=4. Then without the following test, ORDER BY A,B would + ** generate the wrong order output: 1,3 1,4 1,3 1,4 + */ + if( outerObUnique==0 && uniqueNotNull==0 ) return nPriorSat; + *pbObUnique = uniqueNotNull; /* Return the necessary scan order back to the caller */ *pbRev = sortOrder & 1; /* If there was an "ORDER BY rowid" term that matched, or it is only ** possible for a single row from this table to match, then skip over ** any additional ORDER BY terms dealing with this table. */ - if( seenRowid || (uniqueNotNull && i>=pIdx->nColumn) ){ + if( uniqueNotNull ){ /* Advance j over additional ORDER BY terms associated with base */ WhereMaskSet *pMS = p->pWC->pMaskSet; Bitmask m = ~getMask(pMS, base); while( ja[j].pExpr)&m)==0 ){ j++; @@ -3367,16 +3400,18 @@ ** in pc.plan.wsFlags. Otherwise, if there is an ORDER BY clause but ** the index will scan rows in a different order, set the bSort ** variable. */ if( bSort && (pSrc->jointype & JT_LEFT)==0 ){ int bRev = 2; - WHERETRACE((" --> before isSortingIndex: nPriorSat=%d\n",nPriorSat)); - pc.plan.nOBSat = isSortingIndex(p, pProbe, iCur, &bRev); - WHERETRACE((" --> after isSortingIndex: bRev=%d nOBSat=%d\n", - bRev, pc.plan.nOBSat)); + int bObUnique = 0; + WHERETRACE((" --> before isSortIndex: nPriorSat=%d\n",nPriorSat)); + pc.plan.nOBSat = isSortingIndex(p, pProbe, iCur, &bRev, &bObUnique); + WHERETRACE((" --> after isSortIndex: bRev=%d bObU=%d nOBSat=%d\n", + bRev, bObUnique, pc.plan.nOBSat)); if( nPriorSatwctrlFlags & WHERE_ONEPASS_DESIRED)==0 && sqlite3GlobalConfig.bUseCis && OptimizationEnabled(pParse->db, SQLITE_CoverIdxScan) ){ /* This index is not useful for indexing, but it is a covering index. @@ -3626,11 +3662,11 @@ } /* If there is no ORDER BY clause and the SQLITE_ReverseOrder flag ** is set, then reverse the order that the index will be scanned ** in. This is used for application testing, to help find cases - ** where application behaviour depends on the (undefined) order that + ** where application behavior depends on the (undefined) order that ** SQLite outputs rows in in the absence of an ORDER BY clause. */ if( !p->pOrderBy && pParse->db->flags & SQLITE_ReverseOrder ){ p->cost.plan.wsFlags |= WHERE_REVERSE; } @@ -4134,10 +4170,11 @@ struct SrcList_item *pTabItem; /* FROM clause term being coded */ int addrBrk; /* Jump here to break out of the loop */ int addrCont; /* Jump here to continue with next cycle */ int iRowidReg = 0; /* Rowid is stored in this register, if not zero */ int iReleaseReg = 0; /* Temp register to free before returning */ + Bitmask newNotReady; /* Return value */ pParse = pWInfo->pParse; v = pParse->pVdbe; pWC = pWInfo->pWC; pLevel = &pWInfo->a[iLevel]; @@ -4144,10 +4181,11 @@ pTabItem = &pWInfo->pTabList->a[pLevel->iFrom]; iCur = pTabItem->iCursor; bRev = (pLevel->plan.wsFlags & WHERE_REVERSE)!=0; omitTable = (pLevel->plan.wsFlags & WHERE_IDX_ONLY)!=0 && (wctrlFlags & WHERE_FORCE_TABLE)==0; + VdbeNoopComment((v, "Begin Join Loop %d", iLevel)); /* Create labels for the "break" and "continue" instructions ** for the current loop. Jump to addrBrk to break out of a loop. ** Jump to cont to go immediately to the next iteration of the ** loop. @@ -4686,10 +4724,14 @@ ** ** Actually, each subexpression is converted to "xN AND w" where w is ** the "interesting" terms of z - terms that did not originate in the ** ON or USING clause of a LEFT JOIN, and terms that are usable as ** indices. + ** + ** This optimization also only applies if the (x1 OR x2 OR ...) term + ** is not contained in the ON clause of a LEFT JOIN. + ** See ticket http://www.sqlite.org/src/info/f2369304e4 */ if( pWC->nTerm>1 ){ int iTerm; for(iTerm=0; iTermnTerm; iTerm++){ Expr *pExpr = pWC->a[iTerm].pExpr; @@ -4707,11 +4749,11 @@ for(ii=0; iinTerm; ii++){ WhereTerm *pOrTerm = &pOrWc->a[ii]; if( pOrTerm->leftCursor==iCur || (pOrTerm->eOperator & WO_AND)!=0 ){ WhereInfo *pSubWInfo; /* Info for single OR-term scan */ Expr *pOrExpr = pOrTerm->pExpr; - if( pAndExpr ){ + if( pAndExpr && !ExprHasProperty(pOrExpr, EP_FromJoin) ){ pAndExpr->pLeft = pOrExpr; pOrExpr = pAndExpr; } /* Loop through table entries that match term pOrTerm. */ pSubWInfo = sqlite3WhereBegin(pParse, pOrTab, pOrExpr, 0, 0, @@ -4794,11 +4836,11 @@ pLevel->op = aStep[bRev]; pLevel->p1 = iCur; pLevel->p2 = 1 + sqlite3VdbeAddOp2(v, aStart[bRev], iCur, addrBrk); pLevel->p5 = SQLITE_STMTSTATUS_FULLSCAN_STEP; } - notReady &= ~getMask(pWC->pMaskSet, iCur); + newNotReady = notReady & ~getMask(pWC->pMaskSet, iCur); /* Insert code to test every subexpression that can be completely ** computed using the current set of tables. ** ** IMPLEMENTATION-OF: R-49525-50935 Terms that cannot be satisfied through @@ -4808,11 +4850,11 @@ for(pTerm=pWC->a, j=pWC->nTerm; j>0; j--, pTerm++){ Expr *pE; testcase( pTerm->wtFlags & TERM_VIRTUAL ); /* IMP: R-30575-11662 */ testcase( pTerm->wtFlags & TERM_CODED ); if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue; - if( (pTerm->prereqAll & notReady)!=0 ){ + if( (pTerm->prereqAll & newNotReady)!=0 ){ testcase( pWInfo->untestedTerms==0 && (pWInfo->wctrlFlags & WHERE_ONETABLE_ONLY)!=0 ); pWInfo->untestedTerms = 1; continue; } @@ -4822,10 +4864,37 @@ continue; } sqlite3ExprIfFalse(pParse, pE, addrCont, SQLITE_JUMPIFNULL); pTerm->wtFlags |= TERM_CODED; } + + /* Insert code to test for implied constraints based on transitivity + ** of the "==" operator. + ** + ** Example: If the WHERE clause contains "t1.a=t2.b" and "t2.b=123" + ** and we are coding the t1 loop and the t2 loop has not yet coded, + ** then we cannot use the "t1.a=t2.b" constraint, but we can code + ** the implied "t1.a=123" constraint. + */ + for(pTerm=pWC->a, j=pWC->nTerm; j>0; j--, pTerm++){ + Expr *pE; + WhereTerm *pAlt; + Expr sEq; + if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue; + if( pTerm->eOperator!=(WO_EQUIV|WO_EQ) ) continue; + if( pTerm->leftCursor!=iCur ) continue; + pE = pTerm->pExpr; + assert( !ExprHasProperty(pE, EP_FromJoin) ); + assert( (pTerm->prereqRight & newNotReady)!=0 ); + pAlt = findTerm(pWC, iCur, pTerm->u.leftColumn, notReady, WO_EQ|WO_IN, 0); + if( pAlt==0 ) continue; + if( pAlt->wtFlags & (TERM_CODED) ) continue; + VdbeNoopComment((v, "begin transitive constraint")); + sEq = *pAlt->pExpr; + sEq.pLeft = pE->pLeft; + sqlite3ExprIfFalse(pParse, &sEq, addrCont, SQLITE_JUMPIFNULL); + } /* For a LEFT OUTER JOIN, generate code that will record the fact that ** at least one row of the right table has matched the left table. */ if( pLevel->iLeftJoin ){ @@ -4835,11 +4904,11 @@ sqlite3ExprCacheClear(pParse); for(pTerm=pWC->a, j=0; jnTerm; j++, pTerm++){ testcase( pTerm->wtFlags & TERM_VIRTUAL ); /* IMP: R-30575-11662 */ testcase( pTerm->wtFlags & TERM_CODED ); if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue; - if( (pTerm->prereqAll & notReady)!=0 ){ + if( (pTerm->prereqAll & newNotReady)!=0 ){ assert( pWInfo->untestedTerms ); continue; } assert( pTerm->pExpr ); sqlite3ExprIfFalse(pParse, pTerm->pExpr, addrCont, SQLITE_JUMPIFNULL); @@ -4846,11 +4915,11 @@ pTerm->wtFlags |= TERM_CODED; } } sqlite3ReleaseTempReg(pParse, iReleaseReg); - return notReady; + return newNotReady; } #if defined(SQLITE_TEST) /* ** The following variable holds a text description of query plan generated Index: test/8_3_names.test ================================================================== --- test/8_3_names.test +++ test/8_3_names.test @@ -148,21 +148,21 @@ } db close forcedelete test.db do_test 8_3_names-5.0 { sqlite3 db file:./test.db?8_3_names=1 - register_wholenumber_module db + load_static_extension db wholenumber db eval { PRAGMA journal_mode=WAL; CREATE TABLE t1(x); CREATE VIRTUAL TABLE nums USING wholenumber; INSERT INTO t1 SELECT value FROM nums WHERE value BETWEEN 1 AND 1000; BEGIN; UPDATE t1 SET x=x*2; } sqlite3 db2 file:./test.db?8_3_names=1 - register_wholenumber_module db2 + load_static_extension db2 wholenumber db2 eval { BEGIN; SELECT sum(x) FROM t1; } } {500500} Index: test/analyze7.test ================================================================== --- test/analyze7.test +++ test/analyze7.test @@ -24,11 +24,11 @@ } # Generate some test data # do_test analyze7-1.0 { - register_wholenumber_module db + load_static_extension db wholenumber execsql { CREATE TABLE t1(a,b,c,d); CREATE INDEX t1a ON t1(a); CREATE INDEX t1b ON t1(b); CREATE INDEX t1cd ON t1(c,d); Index: test/auth.test ================================================================== --- test/auth.test +++ test/auth.test @@ -2366,11 +2366,34 @@ do_test auth-5.3.2 { execsql { SELECT * FROM t5 } } {1} } +# Ticket [0eb70d77cb05bb22720]: Invalid pointer passsed to the authorizer +# callback when updating a ROWID. +# +do_test auth-6.1 { + execsql { + CREATE TABLE t6(a,b,c,d,e,f,g,h); + INSERT INTO t6 VALUES(1,2,3,4,5,6,7,8); + } +} {} +set ::authargs [list] +proc auth {args} { + eval lappend ::authargs $args + return SQLITE_OK +} +do_test auth-6.2 { + execsql {UPDATE t6 SET rowID=rowID+100} + set ::authargs +} [list SQLITE_READ t6 ROWID main {} \ + SQLITE_UPDATE t6 ROWID main {} \ +] +do_test auth-6.3 { + execsql {SELECT rowid, * FROM t6} +} {101 1 2 3 4 5 6 7 8} rename proc {} rename proc_real proc finish_test Index: test/backup_ioerr.test ================================================================== --- test/backup_ioerr.test +++ test/backup_ioerr.test @@ -113,11 +113,11 @@ # # 5) Step the backup process to finish the backup. If an IO error is # reported, then the backup process is concluded with a call to # backup_finish(). # -# Test that if an IO error occurs, or if one occured while updating +# Test that if an IO error occurs, or if one occurred while updating # the backup database during step 4, then the conditions listed # under step 3 are all true. # # 6) Finish the backup process. # @@ -212,11 +212,11 @@ # Step 4: Write to the source database. set rc [catchsql { UPDATE t1 SET b = randstr(1000,1000) WHERE a < 50 } sdb] if {[lindex $rc 0] && $::sqlite_io_error_persist==0} { - # The IO error occured while updating the source database. In this + # The IO error occurred while updating the source database. In this # case the backup should be able to continue. set rc [B step 5000] if { $rc != "SQLITE_IOERR_UNLOCK" } { do_test backup_ioerr-$iTest.$iError.7 { list [B step 5000] [B finish] ADDED test/btreefault.test Index: test/btreefault.test ================================================================== --- test/btreefault.test +++ test/btreefault.test @@ -0,0 +1,50 @@ +# 2013 April 02 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# This file contains fault injection tests designed to test the btree.c +# module. +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +source $testdir/malloc_common.tcl +set testprefix btreefault + +do_test 1-pre1 { + execsql { + PRAGMA auto_vacuum = incremental; + PRAGMA journal_mode = DELETE; + CREATE TABLE t1(a PRIMARY KEY, b); + INSERT INTO t1 VALUES(randomblob(1000), randomblob(100)); + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + DELETE FROM t1 WHERE rowid%2; + } + faultsim_save_and_close +} {} + +do_faultsim_test 1 -prep { + faultsim_restore_and_reopen + set ::STMT [sqlite3_prepare db "SELECT * FROM t1 ORDER BY a" -1 DUMMY] + sqlite3_step $::STMT + sqlite3_step $::STMT +} -body { + execsql { PRAGMA incremental_vacuum = 10 } +} -test { + sqlite3_finalize $::STMT + faultsim_test_result {0 {}} + faultsim_integrity_check +} + +finish_test + Index: test/cache.test ================================================================== --- test/cache.test +++ test/cache.test @@ -44,11 +44,11 @@ # At one point, repeatedly locking and unlocking the cache was causing # a resource leak of one page per repetition. The page wasn't actually # leaked, but would not be reused until the pager-cache was full (i.e. # 2000 pages by default). # -# This tests that once the pager-cache is initialised, it can be locked +# This tests that once the pager-cache is initialized, it can be locked # and unlocked repeatedly without internally allocating any new pages. # set cache_size [pager_cache_size db] for {set ii 0} {$ii < 10} {incr ii} { do_test cache-1.3.$ii { ADDED test/closure01.test Index: test/closure01.test ================================================================== --- test/closure01.test +++ test/closure01.test @@ -0,0 +1,224 @@ +# 2013-04-25 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# Test cases for transitive_closure virtual table. + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +set testprefix closure01 + +load_static_extension db closure + +do_execsql_test 1.0 { + BEGIN; + CREATE TABLE t1(x INTEGER PRIMARY KEY, y INTEGER); + CREATE INDEX t1y ON t1(y); + INSERT INTO t1(x) VALUES(1),(2); + INSERT INTO t1(x) SELECT x+2 FROM t1; + INSERT INTO t1(x) SELECT x+4 FROM t1; + INSERT INTO t1(x) SELECT x+8 FROM t1; + INSERT INTO t1(x) SELECT x+16 FROM t1; + INSERT INTO t1(x) SELECT x+32 FROM t1; + INSERT INTO t1(x) SELECT x+64 FROM t1; + INSERT INTO t1(x) SELECT x+128 FROM t1; + INSERT INTO t1(x) SELECT x+256 FROM t1; + INSERT INTO t1(x) SELECT x+512 FROM t1; + INSERT INTO t1(x) SELECT x+1024 FROM t1; + INSERT INTO t1(x) SELECT x+2048 FROM t1; + INSERT INTO t1(x) SELECT x+4096 FROM t1; + INSERT INTO t1(x) SELECT x+8192 FROM t1; + INSERT INTO t1(x) SELECT x+16384 FROM t1; + INSERT INTO t1(x) SELECT x+32768 FROM t1; + INSERT INTO t1(x) SELECT x+65536 FROM t1; + UPDATE t1 SET y=x/2 WHERE x>1; + COMMIT; + CREATE VIRTUAL TABLE cx + USING transitive_closure(tablename=t1, idcolumn=x, parentcolumn=y); +} {} + +# The entire table +do_execsql_test 1.1 { + SELECT count(*), depth FROM cx WHERE root=1 GROUP BY depth ORDER BY 1; +} {/1 0 1 17 2 1 4 2 8 3 16 4 .* 65536 16/} + +# descendents of 32768 +do_execsql_test 1.2 { + SELECT * FROM cx WHERE root=32768 ORDER BY id; +} {32768 0 65536 1 65537 1 131072 2} + +# descendents of 16384 +do_execsql_test 1.3 { + SELECT * FROM cx WHERE root=16384 AND depth<=2 ORDER BY id; +} {16384 0 32768 1 32769 1 65536 2 65537 2 65538 2 65539 2} + +# children of 16384 +do_execsql_test 1.4 { + SELECT id, depth, root, tablename, idcolumn, parentcolumn FROM cx + WHERE root=16384 + AND depth=1 + ORDER BY id; +} {32768 1 {} t1 x y 32769 1 {} t1 x y} + +# great-grandparent of 16384 +do_execsql_test 1.5 { + SELECT id, depth, root, tablename, idcolumn, parentcolumn FROM cx + WHERE root=16384 + AND depth=3 + AND idcolumn='Y' + AND parentcolumn='X'; +} {2048 3 {} t1 Y X} + +# depth<5 +do_execsql_test 1.6 { + SELECT count(*), depth FROM cx WHERE root=1 AND depth<5 + GROUP BY depth ORDER BY 1; +} {1 0 2 1 4 2 8 3 16 4} + +# depth<=5 +do_execsql_test 1.7 { + SELECT count(*), depth FROM cx WHERE root=1 AND depth<=5 + GROUP BY depth ORDER BY 1; +} {1 0 2 1 4 2 8 3 16 4 32 5} + +# depth==5 +do_execsql_test 1.8 { + SELECT count(*), depth FROM cx WHERE root=1 AND depth=5 + GROUP BY depth ORDER BY 1; +} {32 5} + +# depth BETWEEN 3 AND 5 +do_execsql_test 1.9 { + SELECT count(*), depth FROM cx WHERE root=1 AND depth BETWEEN 3 AND 5 + GROUP BY depth ORDER BY 1; +} {8 3 16 4 32 5} + +# depth==5 with min() and max() +do_execsql_test 1.10 { + SELECT count(*), min(id), max(id) FROM cx WHERE root=1 AND depth=5; +} {32 32 63} + +# Create a much smaller table t2 with only 32 elements +db eval { + CREATE TABLE t2(x INTEGER PRIMARY KEY, y INTEGER); + INSERT INTO t2 SELECT x, y FROM t1 WHERE x<32; + CREATE INDEX t2y ON t2(y); + CREATE VIRTUAL TABLE c2 + USING transitive_closure(tablename=t2, idcolumn=x, parentcolumn=y); +} + +# t2 full-table +do_execsql_test 2.1 { + SELECT count(*), min(id), max(id) FROM c2 WHERE root=1; +} {31 1 31} +# t2 root=10 +do_execsql_test 2.2 { + SELECT id FROM c2 WHERE root=10; +} {10 20 21} +# t2 root=11 +do_execsql_test 2.3 { + SELECT id FROM c2 WHERE root=12; +} {12 24 25} +# t2 root IN [10,12] +do_execsql_test 2.4 { + SELECT id FROM c2 WHERE root IN (10,12) ORDER BY id; +} {10 12 20 21 24 25} +# t2 root IN [10,12] (sorted) +do_execsql_test 2.5 { + SELECT id FROM c2 WHERE root IN (10,12) ORDER BY +id; +} {10 12 20 21 24 25} + +# t2 c2up from 20 +do_execsql_test 3.0 { + CREATE VIRTUAL TABLE c2up USING transitive_closure( + tablename = t2, + idcolumn = y, + parentcolumn = x + ); + SELECT id FROM c2up WHERE root=20; +} {1 2 5 10 20} + +# cx as c2up +do_execsql_test 3.1 { + SELECT id FROM cx + WHERE root=20 + AND tablename='t2' + AND idcolumn='y' + AND parentcolumn='x'; +} {1 2 5 10 20} + +# t2 first cousins of 20 +do_execsql_test 3.2 { + SELECT DISTINCT id FROM c2 + WHERE root IN (SELECT id FROM c2up + WHERE root=20 AND depth<=2) + ORDER BY id; +} {5 10 11 20 21 22 23} + +# t2 first cousins of 20 +do_execsql_test 3.3 { + SELECT id FROM c2 + WHERE root=(SELECT id FROM c2up + WHERE root=20 AND depth=2) + AND depth=2 + EXCEPT + SELECT id FROM c2 + WHERE root=(SELECT id FROM c2up + WHERE root=20 AND depth=1) + AND depth<=1 + ORDER BY id; +} {22 23} + +# missing tablename. +do_test 4.1 { + catchsql { + SELECT id FROM cx + WHERE root=20 + AND tablename='t3' + AND idcolumn='y' + AND parentcolumn='x'; + } +} {1 {no such table: t3}} + +# missing idcolumn +do_test 4.2 { + catchsql { + SELECT id FROM cx + WHERE root=20 + AND tablename='t2' + AND idcolumn='xyz' + AND parentcolumn='x'; + } +} {1 {no such column: t2.xyz}} + +# missing parentcolumn +do_test 4.3 { + catchsql { + SELECT id FROM cx + WHERE root=20 + AND tablename='t2' + AND idcolumn='x' + AND parentcolumn='pqr'; + } +} {1 {no such column: t2.pqr}} + +# generic closure +do_execsql_test 5.1 { + CREATE VIRTUAL TABLE temp.closure USING transitive_closure; + SELECT id FROM closure + WHERE root=1 + AND depth=3 + AND tablename='t1' + AND idcolumn='x' + AND parentcolumn='y' + ORDER BY id; +} {8 9 10 11 12 13 14 15} + +finish_test Index: test/collate4.test ================================================================== --- test/collate4.test +++ test/collate4.test @@ -58,11 +58,11 @@ # selected or not selected to implement ORDER BY clauses when # user defined collation sequences are involved. # # Because these tests also exercise all the different ways indices # can be created, they also serve to verify that indices are correctly -# initialised with user-defined collation sequences when they are +# initialized with user-defined collation sequences when they are # created. # # Tests named collate4-1.1.* use indices with a single column. Tests # collate4-1.2.* use indices with two columns. # Index: test/crash5.test ================================================================== --- test/crash5.test +++ test/crash5.test @@ -63,11 +63,11 @@ sqlite3_memdebug_fail $iFail -repeat 0 catch {db eval { CREATE UNIQUE INDEX i1 ON t1(a); }} msg # puts "$n $msg ac=[sqlite3_get_autocommit db]" # If the transaction is still active (it may not be if the malloc() - # failure occured in the OS layer), write to the database. Make sure + # failure occurred in the OS layer), write to the database. Make sure # page 4 is among those written. # if {![sqlite3_get_autocommit db]} { db eval { DELETE FROM t1; -- This will put page 4 on the free list. Index: test/dbstatus2.test ================================================================== --- test/dbstatus2.test +++ test/dbstatus2.test @@ -38,10 +38,11 @@ } do_test 1.1 { db close sqlite3 db test.db + execsql { PRAGMA mmap_size = 0 } expr {[file size test.db] / 1024} } 6 do_test 1.2 { execsql { SELECT b FROM t1 WHERE a=2 } Index: test/e_createtable.test ================================================================== --- test/e_createtable.test +++ test/e_createtable.test @@ -1255,11 +1255,11 @@ # EVIDENCE-OF: R-61866-38053 Unless the column is an INTEGER PRIMARY KEY # SQLite allows NULL values in a PRIMARY KEY column. # # If the column is an integer primary key, attempting to insert a NULL -# into the column triggers the auto-increment behaviour. Attempting +# into the column triggers the auto-increment behavior. Attempting # to use UPDATE to set an ipk column to a NULL value is an error. # do_createtable_tests 4.5.1 { 1 "SELECT count(*) FROM t1 WHERE x IS NULL" 3 2 "SELECT count(*) FROM t2 WHERE x IS NULL" 6 Index: test/e_fkey.test ================================================================== --- test/e_fkey.test +++ test/e_fkey.test @@ -2739,12 +2739,12 @@ execsql { DROP TABLE c2 } execsql { DELETE FROM p } } {} #------------------------------------------------------------------------- -# Test that the special behaviours of ALTER and DROP TABLE are only -# activated when foreign keys are enabled. Special behaviours are: +# Test that the special behaviors of ALTER and DROP TABLE are only +# activated when foreign keys are enabled. Special behaviors are: # # 1. ADD COLUMN not allowing a REFERENCES clause with a non-NULL # default value. # 2. Modifying foreign key definitions when a parent table is RENAMEd. # 3. Running an implicit DELETE FROM command as part of DROP TABLE. @@ -2835,11 +2835,11 @@ " } {} do_test e_fkey-62.$zMatch.2 { execsql { INSERT INTO p VALUES(1, 2, 3) } - # MATCH SIMPLE behaviour: Allow any child key that contains one or more + # MATCH SIMPLE behavior: Allow any child key that contains one or more # NULL value to be inserted. Non-NULL values do not have to map to any # parent key values, so long as at least one field of the child key is # NULL. execsql { INSERT INTO c VALUES('w', 2, 3) } execsql { INSERT INTO c VALUES('x', 'x', NULL) } Index: test/e_select.test ================================================================== --- test/e_select.test +++ test/e_select.test @@ -1225,11 +1225,11 @@ # EVIDENCE-OF: R-08861-34280 If the simple SELECT is a SELECT ALL, then # the entire set of result rows are returned by the SELECT. # # EVIDENCE-OF: R-47911-02086 If neither ALL or DISTINCT are present, -# then the behaviour is as if ALL were specified. +# then the behavior is as if ALL were specified. # # EVIDENCE-OF: R-14442-41305 If the simple SELECT is a SELECT DISTINCT, # then duplicate rows are removed from the set of result rows before it # is returned. # Index: test/e_uri.test ================================================================== --- test/e_uri.test +++ test/e_uri.test @@ -359,11 +359,11 @@ # EVIDENCE-OF: R-49793-28525 Setting the cache parameter to "private" is # equivalent to setting the SQLITE_OPEN_PRIVATECACHE bit. # # EVIDENCE-OF: R-19510-48080 If sqlite3_open_v2() is used and the # "cache" parameter is present in a URI filename, its value overrides -# any behaviour requested by setting SQLITE_OPEN_PRIVATECACHE or +# any behavior requested by setting SQLITE_OPEN_PRIVATECACHE or # SQLITE_OPEN_SHAREDCACHE flag. # set orig [sqlite3_enable_shared_cache] foreach {tn uri flags shared_default isshared} { 1.1 "file:test.db" "" 0 0 Index: test/enc2.test ================================================================== --- test/enc2.test +++ test/enc2.test @@ -30,11 +30,11 @@ # enc2.1.*: Simple tests with a UTF-8 db. # enc2.2.*: Simple tests with a UTF-16LE db. # enc2.3.*: Simple tests with a UTF-16BE db. # enc2.4.*: Test that attached databases must have the same text encoding # as the main database. -# enc2.5.*: Test the behaviour of the library when a collation sequence is +# enc2.5.*: Test the behavior of the library when a collation sequence is # not available for the most desirable text encoding. # enc2.6.*: Similar test for user functions. # enc2.7.*: Test that the VerifyCookie opcode protects against assuming the # wrong text encoding for the database. # enc2.8.*: Test sqlite3_complete16() Index: test/exclusive2.test ================================================================== --- test/exclusive2.test +++ test/exclusive2.test @@ -22,10 +22,18 @@ ifcapable {!pager_pragmas} { finish_test return } + +# Tests in this file verify that locking_mode=exclusive causes SQLite to +# use cached pages even if the database is changed on disk. This doesn't +# work with mmap. +if {[permutation]=="mmap"} { + finish_test + return +} # This module does not work right if the cache spills at unexpected # moments. So disable the soft-heap-limit. # sqlite3_soft_heap_limit 0 Index: test/fts3aux1.test ================================================================== --- test/fts3aux1.test +++ test/fts3aux1.test @@ -352,14 +352,14 @@ CREATE VIRTUAL TABLE t2 USING fts4; } do_catchsql_test 3.1.2 { CREATE VIRTUAL TABLE terms2 USING fts4aux; -} {1 {wrong number of arguments to fts4aux constructor}} +} {1 {invalid arguments to fts4aux constructor}} do_catchsql_test 3.1.3 { CREATE VIRTUAL TABLE terms2 USING fts4aux(t2, t2); -} {1 {wrong number of arguments to fts4aux constructor}} +} {1 {invalid arguments to fts4aux constructor}} do_execsql_test 3.2.1 { CREATE VIRTUAL TABLE terms3 USING fts4aux(does_not_exist) } do_catchsql_test 3.2.2 { @@ -442,11 +442,10 @@ #------------------------------------------------------------------------- # The following tests check that fts4aux can handle an fts table with an # odd name (one that requires quoting for use in SQL statements). And that # the argument to the fts4aux constructor is properly dequoted before use. # -# do_execsql_test 5.1 { CREATE VIRTUAL TABLE "abc '!' def" USING fts4(x, y); INSERT INTO "abc '!' def" VALUES('XX', 'YY'); CREATE VIRTUAL TABLE terms3 USING fts4aux("abc '!' def"); @@ -456,7 +455,68 @@ do_execsql_test 5.2 { CREATE VIRTUAL TABLE "%%^^%%" USING fts4aux('abc ''!'' def'); SELECT * FROM "%%^^%%"; } {xx * 1 1 xx 0 1 1 yy * 1 1 yy 1 1 1} +#------------------------------------------------------------------------- +# Test that we can create an fts4aux table in the temp database. +# +forcedelete test.db2 +do_execsql_test 6.1 { + CREATE VIRTUAL TABLE ft1 USING fts4(x, y); + INSERT INTO ft1 VALUES('a b', 'c d'); + INSERT INTO ft1 VALUES('e e', 'c d'); + INSERT INTO ft1 VALUES('a a', 'b b'); + CREATE VIRTUAL TABLE temp.aux1 USING fts4aux(main, ft1); + SELECT * FROM aux1; +} { + a * 2 3 a 0 2 3 + b * 2 3 b 0 1 1 b 1 1 2 + c * 2 2 c 1 2 2 + d * 2 2 d 1 2 2 + e * 1 2 e 0 1 2 +} + +do_execsql_test 6.2 { + ATTACH 'test.db2' AS att; + CREATE VIRTUAL TABLE att.ft1 USING fts4(x, y); + INSERT INTO att.ft1 VALUES('v w', 'x y'); + INSERT INTO att.ft1 VALUES('z z', 'x y'); + INSERT INTO att.ft1 VALUES('v v', 'w w'); + CREATE VIRTUAL TABLE temp.aux2 USING fts4aux(att, ft1); + SELECT * FROM aux2; +} { + v * 2 3 v 0 2 3 + w * 2 3 w 0 1 1 w 1 1 2 + x * 2 2 x 1 2 2 + y * 2 2 y 1 2 2 + z * 1 2 z 0 1 2 +} + +foreach {tn q res1 res2} { + 1 { SELECT * FROM %%% WHERE term = 'a' } {a * 2 3 a 0 2 3} {} + 2 { SELECT * FROM %%% WHERE term = 'x' } {} {x * 2 2 x 1 2 2} + + 3 { SELECT * FROM %%% WHERE term >= 'y' } + {} {y * 2 2 y 1 2 2 z * 1 2 z 0 1 2} + + 4 { SELECT * FROM %%% WHERE term <= 'c' } + {a * 2 3 a 0 2 3 b * 2 3 b 0 1 1 b 1 1 2 c * 2 2 c 1 2 2} {} +} { + set sql1 [string map {%%% aux1} $q] + set sql2 [string map {%%% aux2} $q] + + do_execsql_test 7.$tn.1 $sql1 $res1 + do_execsql_test 7.$tn.2 $sql2 $res2 +} + +do_test 8.1 { + catchsql { CREATE VIRTUAL TABLE att.aux3 USING fts4aux(main, ft1) } +} {1 {invalid arguments to fts4aux constructor}} + +do_test 8.2 { + execsql {DETACH att} + catchsql { SELECT * FROM aux2 } +} {1 {SQL logic error or missing database}} finish_test + ADDED test/fts3expr3.test Index: test/fts3expr3.test ================================================================== --- test/fts3expr3.test +++ test/fts3expr3.test @@ -0,0 +1,210 @@ +# 2009 January 1 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#************************************************************************* +# This file implements regression tests for SQLite library. The +# focus of this script is testing the part of the FTS3 expression +# parser that rebalances large expressions. +# +# $Id: fts3expr2.test,v 1.2 2009/06/05 17:09:12 drh Exp $ +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +source $testdir/malloc_common.tcl +set ::testprefix fts3expr3 + +# If SQLITE_ENABLE_FTS3 is defined, omit this file. +ifcapable !fts3 { + finish_test + return +} + +set sqlite_fts3_enable_parentheses 1 + +proc strip_phrase_data {L} { + if {[lindex $L 0] eq "PHRASE"} { + return [list P [lrange $L 3 end]] + } + return [list \ + [lindex $L 0] \ + [strip_phrase_data [lindex $L 1]] \ + [strip_phrase_data [lindex $L 2]] \ + ] +} +proc test_fts3expr2 {expr} { + strip_phrase_data [ + db one {SELECT fts3_exprtest_rebalance('simple', $expr, 'a', 'b', 'c')} + ] +} + +proc balanced_exprtree_structure {nEntry} { + set L [list] + for {set i 1} {$i <= $nEntry} {incr i} { + lappend L xxx + } + while {[llength $L] > 1} { + set N [list] + if {[llength $L] % 2} { + foreach {a b} [lrange $L 0 end-1] { lappend N [list AND $a $b] } + lappend N [lindex $L end] + } else { + foreach {a b} $L { lappend N [list AND $a $b] } + } + set L $N + } + return [lindex $L 0] +} + +proc balanced_and_tree {nEntry} { + set query [balanced_exprtree_structure $nEntry] + if {$query == "xxx"} { + return "P 1" + } + for {set i 1} {$i <= $nEntry} {incr i} { + regsub xxx $query "{P $i}" query + } + return $query +} + +proc random_tree_structure {nEntry bParen op} { + set query xxx + for {set i 1} {$i < $nEntry} {incr i} { + set x1 [expr int(rand()*4.0)] + set x2 [expr int(rand()*2.0)] + if {$x1==0 && $bParen} { + set query "($query)" + } + if {$x2} { + set query "xxx $op $query" + } else { + set query "$query $op xxx" + } + } + return $query +} + +proc random_and_query {nEntry {bParen 0}} { + set query [random_tree_structure $nEntry $bParen AND] + for {set i 1} {$i <= $nEntry} {incr i} { + regsub xxx $query $i query + } + return $query +} + +proc random_or_query {nEntry} { + set query [random_tree_structure $nEntry 1 OR] + for {set i 1} {$i <= $nEntry} {incr i} { + regsub xxx $query $i query + } + return $query +} + +proc random_andor_query {nEntry} { + set query [random_tree_structure $nEntry 1 AND] + for {set i 1} {$i <= $nEntry} {incr i} { + regsub xxx $query "([random_or_query $nEntry])" query + } + return $query +} + +proc balanced_andor_tree {nEntry} { + set tree [balanced_exprtree_structure $nEntry] + set node "{[balanced_and_tree $nEntry]}" + regsub -all AND $node OR node + regsub -all xxx $tree $node tree + return $tree +} + +# Test that queries like "1 AND 2 AND 3 AND 4..." are transformed to +# balanced trees by FTS. +# +for {set i 1} {$i < 100} {incr i} { + do_test 1.$i { + test_fts3expr2 [random_and_query $i] + } [balanced_and_tree $i] +} + +# Same again, except with parenthesis inserted at arbitrary points. +# +for {set i 1} {$i < 100} {incr i} { + do_test 2.$i { + test_fts3expr2 [random_and_query $i 1] + } [balanced_and_tree $i] +} + +# Now attempt to balance two AND trees joined by an OR. +# +for {set i 1} {$i < 100} {incr i} { + do_test 3.$i { + test_fts3expr2 "[random_and_query $i 1] OR [random_and_query $i 1]" + } [list OR [balanced_and_tree $i] [balanced_and_tree $i]] +} + +# Try trees of AND nodes with leaves that are themselves trees of OR nodes. +# +for {set i 2} {$i < 64} {incr i 4} { + do_test 3.$i { + test_fts3expr2 [random_andor_query $i] + } [balanced_andor_tree $i] +} + +# These exceed the depth limit. +# +for {set i 65} {$i < 70} {incr i} { + do_test 3.$i { + list [catch {test_fts3expr2 [random_andor_query $i]} msg] $msg + } {1 {Error parsing expression}} +} + +# This also exceeds the depth limit. +# + +do_test 4.1.1 { + set q "1" + for {set i 2} {$i < 5000} {incr i} { + append q " AND $i" + } + list [catch {test_fts3expr2 $q} msg] $msg +} {1 {Error parsing expression}} +do_test 4.1.2 { + set q "1" + for {set i 2} {$i < 4000} {incr i} { + append q " AND $i" + } + catch {test_fts3expr2 $q} +} {0} + +proc create_toggle_tree {nDepth} { + if {$nDepth == 0} { return xxx } + set nNew [expr $nDepth-1] + if {$nDepth % 2} { + return "([create_toggle_tree $nNew]) OR ([create_toggle_tree $nNew])" + } + return "([create_toggle_tree $nNew]) AND ([create_toggle_tree $nNew])" +} + +do_test 4.2 { + list [catch {test_fts3expr2 [create_toggle_tree 17]} msg] $msg +} {1 {Error parsing expression}} + +set query [random_andor_query 12] +set result [balanced_andor_tree 12] +do_faultsim_test fts3expr3-fault-1 -faults oom-* -body { + test_fts3expr2 $::query +} -test { + faultsim_test_result [list 0 $::result] +} + +set sqlite_fts3_enable_parentheses 0 +finish_test + + + + Index: test/fts3near.test ================================================================== --- test/fts3near.test +++ test/fts3near.test @@ -577,8 +577,22 @@ do_test fts3near-6.5 { execsql { SELECT docid FROM t1 WHERE content MATCH 'abbrev NEAR/10000 zygosis' } } {3} + +# Ticket 38b1ae018f. +# +do_execsql_test fts3near-7.1 { + CREATE VIRTUAL TABLE x USING fts4(y,z); + INSERT INTO x VALUES('aaa bbb ccc ddd', 'bbb ddd aaa ccc'); + SELECT * FROM x where y MATCH 'bbb NEAR/6 aaa'; +} {{aaa bbb ccc ddd} {bbb ddd aaa ccc}} + +do_execsql_test fts3near-7.2 { + CREATE VIRTUAL TABLE t2 USING fts4(a, b); + INSERT INTO t2 VALUES('A B C', 'A D E'); + SELECT * FROM t2 where t2 MATCH 'a:A NEAR E' +} {} finish_test ADDED test/fts3tok1.test Index: test/fts3tok1.test ================================================================== --- test/fts3tok1.test +++ test/fts3tok1.test @@ -0,0 +1,117 @@ +# 2013 April 22 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#************************************************************************* +# This file implements regression tests for SQLite library. The +# focus of this script is testing the "fts3tokenize" virtual table +# that is part of the FTS3 module. +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +ifcapable !fts3 { finish_test ; return } +set ::testprefix fts3tok1 + +#------------------------------------------------------------------------- +# Simple test cases. Using the default (simple) tokenizer. +# +do_execsql_test 1.0 { + CREATE VIRTUAL TABLE t1 USING fts3tokenize(simple); + CREATE VIRTUAL TABLE t2 USING fts3tokenize(); + CREATE VIRTUAL TABLE t3 USING fts3tokenize(simple, '', 'xyz '); +} + +foreach {tn tbl} {1 t1 2 t2 3 t3} { + do_execsql_test 1.$tn.1 "SELECT * FROM $tbl WHERE input = 'one two three'" { + {one two three} one 0 3 0 + {one two three} two 4 7 1 + {one two three} three 8 13 2 + } + + do_execsql_test 1.$tn.2 " + SELECT token FROM $tbl WHERE input = 'OnE tWo tHrEe' + " { + one two three + } +} + +do_execsql_test 1.4 { + SELECT token FROM t3 WHERE input = '1x2x3x' +} {1 2 3} + +do_execsql_test 1.5 { + SELECT token FROM t1 WHERE input = '1x2x3x' +} {1x2x3x} + +do_execsql_test 1.6 { + SELECT token FROM t3 WHERE input = '1''2x3x' +} {1'2 3} + +do_execsql_test 1.7 { + SELECT token FROM t3 WHERE input = '' +} {} + +do_execsql_test 1.8 { + SELECT token FROM t3 WHERE input = NULL +} {} + +do_execsql_test 1.9 { + SELECT * FROM t3 WHERE input = 123 +} {123 123 0 3 0} + +do_execsql_test 1.10 { + SELECT * FROM t1 WHERE input = 'a b c' AND token = 'b'; +} { + {a b c} b 2 3 1 +} + +do_execsql_test 1.11 { + SELECT * FROM t1 WHERE token = 'b' AND input = 'a b c'; +} { + {a b c} b 2 3 1 +} + +do_execsql_test 1.12 { + SELECT * FROM t1 WHERE input < 'b' AND input = 'a b c'; +} { + {a b c} a 0 1 0 + {a b c} b 2 3 1 + {a b c} c 4 5 2 +} + +do_execsql_test 1.13.1 { + CREATE TABLE c1(x); + INSERT INTO c1(x) VALUES('a b c'); + INSERT INTO c1(x) VALUES('d e f'); +} +breakpoint +do_execsql_test 1.13.2 { + SELECT * FROM c1, t1 WHERE input = x AND c1.rowid=t1.rowid; +} { + {a b c} {a b c} a 0 1 0 + {d e f} {d e f} e 2 3 1 +} + + +#------------------------------------------------------------------------- +# Error cases. +# +do_catchsql_test 2.0 { + CREATE VIRTUAL TABLE tX USING fts3tokenize(nosuchtokenizer); +} {1 {unknown tokenizer: nosuchtokenizer}} + +do_catchsql_test 2.1 { + CREATE VIRTUAL TABLE t4 USING fts3tokenize; + SELECT * FROM t4; +} {1 {SQL logic error or missing database}} + + +finish_test + + ADDED test/fts3tok_err.test Index: test/fts3tok_err.test ================================================================== --- test/fts3tok_err.test +++ test/fts3tok_err.test @@ -0,0 +1,49 @@ +# 2013 April 22 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#************************************************************************* +# This file implements regression tests for SQLite library. The +# focus of this script is testing the "fts3tokenize" virtual table +# that is part of the FTS3 module. +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +source $testdir/malloc_common.tcl +ifcapable !fts3 { finish_test ; return } +set ::testprefix fts3tok_err + + +faultsim_save_and_close +do_faultsim_test fts3tok_err-1 -faults oom* -prep { + faultsim_restore_and_reopen +} -body { + execsql { CREATE VIRTUAL TABLE t1 USING fts3tokenize("simple"); } +} -test { + faultsim_test_result {0 {}} +} + +do_test fts3tok_err-2.prep { + faultsim_delete_and_reopen + execsql { CREATE VIRTUAL TABLE t1 USING fts3tokenize("simple"); } + faultsim_save_and_close +} {} + +do_faultsim_test fts3tok_err-2 -faults oom* -prep { + faultsim_restore_and_reopen +} -body { + execsql { SELECT token FROM t1 WHERE input = 'A galaxy far, far away' } +} -test { + faultsim_test_result {0 {a galaxy far far away}} +} + + +finish_test + + Index: test/func.test ================================================================== --- test/func.test +++ test/func.test @@ -1271,15 +1271,17 @@ db close sqlite3 db test.db sqlite3_db_status db CACHE_MISS 1 db eval {SELECT typeof(+x) FROM t29 ORDER BY id} } {integer null real blob text} -do_test func-29.4 { - set x [lindex [sqlite3_db_status db CACHE_MISS 1] 1] - if {$x>100} {set x many} - set x -} {many} +if {[permutation] != "mmap"} { + do_test func-29.4 { + set x [lindex [sqlite3_db_status db CACHE_MISS 1] 1] + if {$x>100} {set x many} + set x + } {many} +} do_test func-29.5 { db close sqlite3 db test.db sqlite3_db_status db CACHE_MISS 1 db eval {SELECT sum(length(x)) FROM t29} Index: test/fuzzer1.test ================================================================== --- test/fuzzer1.test +++ test/fuzzer1.test @@ -22,16 +22,11 @@ return } set ::testprefix fuzzer1 -# Test of test code. Only here to make the coverage metric better. -do_test 0.1 { - list [catch { register_fuzzer_module a b c } msg] $msg -} {1 {wrong # args: should be "register_fuzzer_module DB"}} - -register_fuzzer_module db +load_static_extension db fuzzer # Check configuration errors. # do_catchsql_test fuzzer1-1.1 { CREATE VIRTUAL TABLE f USING fuzzer; Index: test/fuzzerfault.test ================================================================== --- test/fuzzerfault.test +++ test/fuzzerfault.test @@ -15,11 +15,11 @@ set testdir [file dirname $argv0] source $testdir/tester.tcl ifcapable !vtab { finish_test ; return } set ::testprefix fuzzerfault -register_fuzzer_module db +load_static_extension db fuzzer do_test 1-pre1 { execsql { CREATE TABLE x1_rules(ruleset, cFrom, cTo, cost); INSERT INTO x1_rules VALUES(0, 'a', 'b', 1); @@ -28,11 +28,11 @@ } faultsim_save_and_close } {} do_faultsim_test 1 -prep { faultsim_restore_and_reopen - register_fuzzer_module db + load_static_extension db fuzzer } -body { execsql { CREATE VIRTUAL TABLE x1 USING fuzzer(x1_rules); SELECT word FROM x1 WHERE word MATCH 'xax'; } @@ -41,11 +41,11 @@ {1 {vtable constructor failed: x1}} } do_test 2-pre1 { faultsim_delete_and_reopen - register_fuzzer_module db + load_static_extension db fuzzer execsql { CREATE TABLE x2_rules(ruleset, cFrom, cTo, cost); INSERT INTO x2_rules VALUES(0, 'a', 'x', 1); INSERT INTO x2_rules VALUES(0, 'b', 'x', 2); INSERT INTO x2_rules VALUES(0, 'c', 'x', 3); @@ -54,11 +54,11 @@ faultsim_save_and_close } {} do_faultsim_test 2 -prep { faultsim_restore_and_reopen - register_fuzzer_module db + load_static_extension db fuzzer } -body { execsql { SELECT count(*) FROM x2 WHERE word MATCH 'abc'; } } -test { @@ -76,11 +76,11 @@ faultsim_save_and_close } {} do_faultsim_test 3 -prep { faultsim_restore_and_reopen - register_fuzzer_module db + load_static_extension db fuzzer } -body { execsql { CREATE VIRTUAL TABLE x1 USING fuzzer(x1_rules); SELECT count(*) FROM (SELECT * FROM x1 WHERE word MATCH 'a' LIMIT 2); } Index: test/incrblob.test ================================================================== --- test/incrblob.test +++ test/incrblob.test @@ -121,10 +121,11 @@ db close forcedelete test.db test.db-journal sqlite3 db test.db + execsql "PRAGMA mmap_size = 0" execsql "PRAGMA auto_vacuum = $AutoVacuumMode" do_test incrblob-2.$AutoVacuumMode.1 { set ::str [string repeat abcdefghij 2900] execsql { @@ -147,10 +148,11 @@ do_test incrblob-2.$AutoVacuumMode.3 { # Open and close the db to make sure the page cache is empty. db close sqlite3 db test.db + execsql "PRAGMA mmap_size = 0" # Read the last 20 bytes of the blob via a blob handle. set ::blob [db incrblob blobs v 1] seek $::blob -20 end set ::fragment [read $::blob] @@ -169,10 +171,11 @@ do_test incrblob-2.$AutoVacuumMode.5 { # Open and close the db to make sure the page cache is empty. db close sqlite3 db test.db + execsql "PRAGMA mmap_size = 0" # Write the second-to-last 20 bytes of the blob via a blob handle. # set ::blob [db incrblob blobs v 1] seek $::blob -40 end @@ -198,10 +201,11 @@ do_test incrblob-2.$AutoVacuumMode.8 { # Open and close the db to make sure the page cache is empty. db close sqlite3 db test.db + execsql { PRAGMA mmap_size = 0 } execsql { SELECT i FROM blobs } } {45} do_test incrblob-2.$AutoVacuumMode.9 { @@ -503,11 +507,11 @@ db2 close } sqlite3_soft_heap_limit $cmdlinearg(soft-heap-limit) #----------------------------------------------------------------------- -# The following tests verify the behaviour of the incremental IO +# The following tests verify the behavior of the incremental IO # APIs in the following cases: # # 7.1 A row that containing an open blob is modified. # # 7.2 A CREATE TABLE requires that an overflow page that is part @@ -514,11 +518,11 @@ # of an open blob is moved. # # 7.3 An INCREMENTAL VACUUM moves an overflow page that is part # of an open blob. # -# In the first case above, correct behaviour is for all subsequent +# In the first case above, correct behavior is for all subsequent # read/write operations on the blob-handle to return SQLITE_ABORT. # More accurately, blob-handles are invalidated whenever the table # they belong to is written to. # # The second two cases have no external effect. They are testing Index: test/io.test ================================================================== --- test/io.test +++ test/io.test @@ -14,10 +14,11 @@ # more database pages than it has to, stuff like that). # set testdir [file dirname $argv0] source $testdir/tester.tcl +set ::testprefix io db close sqlite3_simulate_device sqlite3 db test.db -vfs devsym @@ -36,10 +37,14 @@ # fsync() calls on the journal file, no need to set nRec # field in the single journal header). # # io-5.* - Test that the default page size is selected and used # correctly. +# +# io-6.* - Test that the pager-cache is not being flushed unnecessarily +# after a transaction that uses the special atomic-write path +# is committed. # set ::nWrite 0 proc nWrite {db} { set bt [btree_from_db $db] @@ -205,11 +210,11 @@ # is executed. # # Changed 2010-03-27: The size of the database is now stored in # bytes 28..31 and so when a page is added to the database, page 1 # is immediately modified and the journal file immediately comes into -# existance. To fix this test, the BEGIN is changed into a a +# existence. To fix this test, the BEGIN is changed into a a # BEGIN IMMEDIATE and the INSERT is omitted. # do_test io-2.6.1 { execsql { BEGIN IMMEDIATE; @@ -562,8 +567,71 @@ CREATE TABLE abc(a, b, c); } expr {[file size test.db]/2} } $pgsize } + +#---------------------------------------------------------------------- +# +do_test io-6.1 { + db close + sqlite3_simulate_device -char atomic + forcedelete test.db + sqlite3 db test.db -vfs devsym + execsql { + PRAGMA mmap_size = 0; + PRAGMA page_size = 1024; + CREATE TABLE t1(x); + CREATE TABLE t2(x); + CREATE TABLE t3(x); + CREATE INDEX i3 ON t3(x); + INSERT INTO t3 VALUES(randomblob(100)); + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + INSERT INTO t3 SELECT randomblob(100) FROM t3; + } + + db_save_and_close +} {} + +foreach {tn sql} { + 1 { BEGIN; + INSERT INTO t1 VALUES('123'); + INSERT INTO t2 VALUES('456'); + COMMIT; + } + 2 { BEGIN; + INSERT INTO t1 VALUES('123'); + COMMIT; + } +} { + db_restore + sqlite3 db test.db -vfs devsym + execsql { + PRAGMA mmap_size = 0; + SELECT x FROM t3 ORDER BY rowid; + SELECT x FROM t3 ORDER BY x; + } + do_execsql_test 6.2.$tn.1 { PRAGMA integrity_check } {ok} + do_execsql_test 6.2.$tn.2 $sql + + # Corrupt the database file on disk. This should not matter for the + # purposes of the following "PRAGMA integrity_check", as the entire + # database should be cached in the pager-cache. If corruption is + # reported, it indicates that executing $sql caused the pager cache + # to be flushed. Which is a bug. + hexio_write test.db [expr 1024 * 5] [string repeat 00 2048] + do_execsql_test 6.2.$tn.3 { PRAGMA integrity_check } {ok} + db close +} sqlite3_simulate_device -char {} -sectorsize 0 finish_test + Index: test/loadext.test ================================================================== --- test/loadext.test +++ test/loadext.test @@ -137,24 +137,26 @@ forcedelete ${testextension}xx set rc [catch { sqlite3_load_extension db "${testextension}xx" } msg] list $rc $msg -} [list 1 [format $dlerror_nosuchfile ${testextension}xx]] +} /[list 1 [format $dlerror_nosuchfile ${testextension}xx.*]]/ # Try to load an extension for which the file is not a shared object # do_test loadext-2.2 { - set fd [open "${testextension}xx" w] + set fd [open "./notasharedlib.so" w] + puts $fd blah + close $fd + set fd [open "./notasharedlib.dll" w] puts $fd blah close $fd set rc [catch { - sqlite3_load_extension db "${testextension}xx" + sqlite3_load_extension db "./notasharedlib" } msg] - set expected_error_pattern [format $dlerror_notadll ${testextension}xx] - list $rc [string match $expected_error_pattern $msg] -} [list 1 1] + list $rc $msg +} /[list 1 [format $dlerror_notadll ./notasharedlib.*]]/ # Try to load an extension for which the file is present but the # entry point is not. # do_test loadext-2.3 { @@ -194,11 +196,11 @@ }] if {$::tcl_platform(os) eq "Darwin"} { regsub {0x[1234567890abcdefABCDEF]*} $res XXX res } set res -} [list 1 [format $dlerror_nosymbol $testextension sqlite3_extension_init]] +} /[list 1 [format $dlerror_nosymbol $testextension sqlite3_.*_init]]/ do_test loadext-3.3 { catchsql { SELECT load_extension($::testextension,'testloadext_init') } } {0 {{}}} Index: test/malloc.test ================================================================== --- test/malloc.test +++ test/malloc.test @@ -840,11 +840,11 @@ INSERT INTO t1 VALUES(3, 4); } -sqlbody { SELECT test_agg_errmsg16(), group_concat(a) FROM t1 } -# At one point, if an OOM occured immediately after obtaining a shared lock +# At one point, if an OOM occurred immediately after obtaining a shared lock # on the database file, the file remained locked. This test case ensures # that bug has been fixed.i if {[db eval {PRAGMA locking_mode}]!="exclusive"} { do_malloc_test 37 -tclprep { sqlite3 db2 test.db Index: test/malloc3.test ================================================================== --- test/malloc3.test +++ test/malloc3.test @@ -593,12 +593,12 @@ set rc [catch {db eval [lindex $v 2]} msg] ;# True error occurs set nac [sqlite3_get_autocommit $::DB] ;# New Auto-Commit if {$rc != 0 && $nac && !$ac} { # Before [db eval] the auto-commit flag was clear. Now it - # is set. Since an error occured we assume this was not a - # commit - therefore a rollback occured. Check that the + # is set. Since an error occurred we assume this was not a + # commit - therefore a rollback occurred. Check that the # rollback-hook was invoked. do_test malloc3-rollback_hook_count.$iterid { set ::rollback_hook_count } {1} } Index: test/malloc_common.tcl ================================================================== --- test/malloc_common.tcl +++ test/malloc_common.tcl @@ -262,11 +262,11 @@ proc faultsim_test_result_int {args} { upvar testrc testrc testresult testresult testnfail testnfail set t [list $testrc $testresult] set r $args if { ($testnfail==0 && $t != [lindex $r 0]) || [lsearch $r $t]<0 } { - error "nfail=$testnfail rc=$testrc result=$testresult" + error "nfail=$testnfail rc=$testrc result=$testresult list=$r" } } #-------------------------------------------------------------------------- # Usage do_one_faultsim_test NAME ?OPTIONS...? Index: test/memdb.test ================================================================== --- test/memdb.test +++ test/memdb.test @@ -363,11 +363,11 @@ } } {} ifcapable subquery&&vtab { do_test memdb-7.1 { - register_wholenumber_module db + load_static_extension db wholenumber execsql { CREATE TABLE t6(x); CREATE VIRTUAL TABLE nums USING wholenumber; INSERT INTO t6 SELECT value FROM nums WHERE value BETWEEN 1 AND 256; SELECT count(*) FROM (SELECT DISTINCT x FROM t6); ADDED test/mmap1.test Index: test/mmap1.test ================================================================== --- test/mmap1.test +++ test/mmap1.test @@ -0,0 +1,323 @@ +# 2013 March 20 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +ifcapable !mmap { + finish_test + return +} +source $testdir/lock_common.tcl +set testprefix mmap1 + +proc nRead {db} { + set bt [btree_from_db $db] + db_enter $db + array set stats [btree_pager_stats $bt] + db_leave $db + # puts [array get stats] + return $stats(read) +} + +proc register_rblob_code {dbname seed} { + return [subst -nocommands { + set ::rcnt $seed + proc rblob {n} { + set ::rcnt [expr (([set ::rcnt] << 3) + [set ::rcnt] + 456) & 0xFFFFFFFF] + set str [format %.8x [expr [set ::rcnt] ^ 0xbdf20da3]] + string range [string repeat [set str] [expr [set n]/4]] 1 [set n] + } + $dbname func rblob rblob + }] +} + +# For cases 1.1 and 1.4, the number of pages read using xRead() is 4 on +# unix and 9 on windows. The difference is that windows only ever maps +# an integer number of OS pages (i.e. creates mappings that are a multiple +# of 4KB in size). Whereas on unix any sized mapping may be created. +# +foreach {t mmap_size nRead c2init} { + 1.1 { PRAGMA mmap_size = 67108864 } /[49]/ {PRAGMA mmap_size = 0} + 1.2 { PRAGMA mmap_size = 53248 } 150 {PRAGMA mmap_size = 0} + 1.3 { PRAGMA mmap_size = 0 } 344 {PRAGMA mmap_size = 0} + 1.4 { PRAGMA mmap_size = 67108864 } /[49]/ {PRAGMA mmap_size = 67108864 } + 1.5 { PRAGMA mmap_size = 53248 } 150 {PRAGMA mmap_size = 67108864 } + 1.6 { PRAGMA mmap_size = 0 } 344 {PRAGMA mmap_size = 67108864 } +} { + + do_multiclient_test tn { + sql1 {PRAGMA page_size=1024} + sql1 $mmap_size + sql2 $c2init + + code2 [register_rblob_code db2 0] + + sql2 { + PRAGMA page_size=1024; + PRAGMA auto_vacuum = 1; + CREATE TABLE t1(a, b, UNIQUE(a, b)); + INSERT INTO t1 VALUES(rblob(500), rblob(500)); + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 2 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 4 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 8 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 16 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 32 + } + do_test $t.$tn.1 { + sql1 "SELECT count(*) FROM t1; PRAGMA integrity_check ; PRAGMA page_count" + } {32 ok 77} + + # Have connection 2 shrink the file. Check connection 1 can still read it. + sql2 { DELETE FROM t1 WHERE rowid%2; } + do_test $t.$tn.2 { + sql1 "SELECT count(*) FROM t1; PRAGMA integrity_check ; PRAGMA page_count" + } {16 ok 42} + + # Have connection 2 grow the file. Check connection 1 can still read it. + sql2 { INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1 } + do_test $t.$tn.3 { + sql1 "SELECT count(*) FROM t1; PRAGMA integrity_check ; PRAGMA page_count" + } {32 ok 79} + + # Have connection 2 grow the file again. Check connection 1 is still ok. + sql2 { INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1 } + do_test $t.$tn.4 { + sql1 "SELECT count(*) FROM t1; PRAGMA integrity_check ; PRAGMA page_count" + } {64 ok 149} + + # Check that the number of pages read by connection 1 indicates that the + # "PRAGMA mmap_size" command worked. + do_test $t.$tn.5 { nRead db } $nRead + } +} + +set ::rcnt 0 +proc rblob {n} { + set ::rcnt [expr (($::rcnt << 3) + $::rcnt + 456) & 0xFFFFFFFF] + set str [format %.8x [expr $::rcnt ^ 0xbdf20da3]] + string range [string repeat $str [expr $n/4]] 1 $n +} + +reset_db +db func rblob rblob + +do_execsql_test 2.1 { + PRAGMA auto_vacuum = 1; + PRAGMA mmap_size = 67108864; + PRAGMA journal_mode = wal; + CREATE TABLE t1(a, b, UNIQUE(a, b)); + INSERT INTO t1 VALUES(rblob(500), rblob(500)); + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 2 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 4 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 8 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 16 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 32 + PRAGMA wal_checkpoint; +} {67108864 wal 0 103 103} + +do_execsql_test 2.2 { + PRAGMA auto_vacuum; + SELECT count(*) FROM t1; +} {1 32} + +do_test 2.3 { + sqlite3 db2 test.db + db2 func rblob rblob + db2 eval { + DELETE FROM t1 WHERE (rowid%4); + PRAGMA wal_checkpoint; + } + db2 eval { + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 16 + SELECT count(*) FROM t1; + } +} {16} + +do_execsql_test 2.4 { + PRAGMA wal_checkpoint; +} {0 24 24} + +db2 close +reset_db +db func rblob rblob +do_execsql_test 3.1 { + PRAGMA auto_vacuum = 1; + + CREATE TABLE t1(a, b, UNIQUE(a, b)); + INSERT INTO t1 VALUES(rblob(500), rblob(500)); + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 2 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 4 + INSERT INTO t1 SELECT rblob(500), rblob(500) FROM t1; -- 8 + + CREATE TABLE t2(a, b, UNIQUE(a, b)); + INSERT INTO t2 SELECT * FROM t1; +} {} + +do_test 3.2 { + set nRow 0 + db eval {SELECT * FROM t2 ORDER BY a, b} { + if {$nRow==4} { db eval { DELETE FROM t1 } } + incr nRow + } + set nRow +} {8} + +#------------------------------------------------------------------------- +# Ensure that existing cursors using xFetch() pages see changes made +# to rows using the incrblob API. +# +reset_db +set aaa [string repeat a 400] +set bbb [string repeat b 400] +set ccc [string repeat c 400] +set ddd [string repeat d 400] +set eee [string repeat e 400] + +do_execsql_test 4.1 { + PRAGMA page_size = 1024; + CREATE TABLE t1(x); + INSERT INTO t1 VALUES($aaa); + INSERT INTO t1 VALUES($bbb); + INSERT INTO t1 VALUES($ccc); + INSERT INTO t1 VALUES($ddd); + SELECT * FROM t1; + BEGIN; +} [list $aaa $bbb $ccc $ddd] + +do_test 4.2 { + set ::STMT [sqlite3_prepare db "SELECT * FROM t1 ORDER BY rowid" -1 dummy] + sqlite3_step $::STMT + sqlite3_column_text $::STMT 0 +} $aaa + +do_test 4.3 { + foreach r {2 3 4} { + set fd [db incrblob t1 x $r] + puts -nonewline $fd $eee + close $fd + } + + set res [list] + while {"SQLITE_ROW" == [sqlite3_step $::STMT]} { + lappend res [sqlite3_column_text $::STMT 0] + } + set res +} [list $eee $eee $eee] + +do_test 4.4 { + sqlite3_finalize $::STMT +} SQLITE_OK + +do_execsql_test 4.5 { COMMIT } + +#------------------------------------------------------------------------- +# Ensure that existing cursors holding xFetch() references are not +# confused if those pages are moved to make way for the root page of a +# new table or index. +# +reset_db +do_execsql_test 5.1 { + PRAGMA auto_vacuum = 2; + PRAGMA page_size = 1024; + CREATE TABLE t1(x); + INSERT INTO t1 VALUES($aaa); + INSERT INTO t1 VALUES($bbb); + INSERT INTO t1 VALUES($ccc); + INSERT INTO t1 VALUES($ddd); + + PRAGMA auto_vacuum; + SELECT * FROM t1; +} [list 2 $aaa $bbb $ccc $ddd] + +do_test 5.2 { + set ::STMT [sqlite3_prepare db "SELECT * FROM t1 ORDER BY rowid" -1 dummy] + sqlite3_step $::STMT + sqlite3_column_text $::STMT 0 +} $aaa + +do_execsql_test 5.3 { + CREATE TABLE t2(x); + INSERT INTO t2 VALUES('tricked you!'); + INSERT INTO t2 VALUES('tricked you!'); +} + +do_test 5.4 { + sqlite3_step $::STMT + sqlite3_column_text $::STMT 0 +} $bbb + +do_test 5.5 { + sqlite3_finalize $::STMT +} SQLITE_OK + +#------------------------------------------------------------------------- +# Test various mmap_size settings. +# +foreach {tn1 mmap1 mmap2} { + 1 6144 167773 + 2 18432 140399 + 3 43008 401302 + 4 92160 253899 + 5 190464 2 + 6 387072 752431 + 7 780288 291143 + 8 1566720 594306 + 9 3139584 829137 + 10 6285312 793963 + 11 12576768 1015590 +} { + do_multiclient_test tn { + sql1 { + CREATE TABLE t1(a PRIMARY KEY); + CREATE TABLE t2(x); + INSERT INTO t2 VALUES(''); + } + + code1 [register_rblob_code db 0] + code2 [register_rblob_code db2 444] + + sql1 "PRAGMA mmap_size = $mmap1" + sql2 "PRAGMA mmap_size = $mmap2" + + do_test $tn1.$tn { + for {set i 1} {$i <= 100} {incr i} { + if {$i % 2} { + set c1 sql1 + set c2 sql2 + } else { + set c1 sql2 + set c2 sql1 + } + + $c1 { + INSERT INTO t1 VALUES( rblob(5000) ); + UPDATE t2 SET x = (SELECT md5sum(a) FROM t1); + } + + set res [$c2 { + SELECT count(*) FROM t1; + SELECT x == (SELECT md5sum(a) FROM t1) FROM t2; + PRAGMA integrity_check; + }] + if {$res != [list $i 1 ok]} { + do_test $tn1.$tn.$i { + set ::res + } [list $i 1 ok] + } + } + set res 1 + } {1} + } +} + + +finish_test ADDED test/mmap2.test Index: test/mmap2.test ================================================================== --- test/mmap2.test +++ test/mmap2.test @@ -0,0 +1,86 @@ +# 2013 March 20 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# This file tests the effect of the mmap() or mremap() system calls +# returning an error on the library. +# +# If either mmap() or mremap() fails, SQLite should log an error +# message, then continue accessing the database using read() and +# write() exclusively. +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +set testprefix mmap2 + +if {$::tcl_platform(platform)!="unix" || [test_syscall defaultvfs] != "unix"} { + finish_test + return +} +ifcapable !mmap { + finish_test + return +} + +db close +sqlite3_shutdown +test_sqlite3_log xLog +proc xLog {error_code msg} { + if {[string match os_unix.c* $msg]} { + lappend ::log $msg + } +} + +foreach syscall {mmap mremap} { + test_syscall uninstall + if {[catch {test_syscall install $syscall}]} continue + + for {set i 1} {$i < 20} {incr i} { + reset_db + + test_syscall fault $i 1 + test_syscall errno $syscall ENOMEM + set ::log "" + + do_execsql_test 1.$syscall.$i.1 { + CREATE TABLE t1(a, b, UNIQUE(a, b)); + INSERT INTO t1 VALUES(randomblob(1000), randomblob(1000)); + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + INSERT INTO t1 SELECT randomblob(1000), randomblob(1000) FROM t1; + } + + set nFail [test_syscall fault 0 0] + + do_execsql_test 1.$syscall.$i.2 { + SELECT count(*) FROM t1; + PRAGMA integrity_check; + } {64 ok} + + do_test 1.$syscall.$i.3 { + expr {$nFail==0 || $nFail==1} + } {1} + + do_test 1.$syscall.$i.4.nFail=$nFail { + regexp ".*${syscall}.*" $::log + } [expr $nFail>0] + } +} + +db close +test_syscall uninstall +sqlite3_shutdown +test_sqlite3_log +sqlite3_initialize +finish_test Index: test/notify2.test ================================================================== --- test/notify2.test +++ test/notify2.test @@ -148,13 +148,13 @@ } } elseif {$rc} { # Hit some other kind of error. This is a malfunction. error $msg } else { - # No error occured. Check that any SELECT statements in the transaction + # No error occurred. Check that any SELECT statements in the transaction # returned "1". Otherwise, the invariant was false, indicating that - # some malfunction has occured. + # some malfunction has occurred. foreach r $msg { if {$r != 1} { puts "Invariant check failed: $msg" } } } } # Close the database connection and return 0. ADDED test/numcast.test Index: test/numcast.test ================================================================== --- test/numcast.test +++ test/numcast.test @@ -0,0 +1,46 @@ +# 2013 March 20 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# This file implements regression tests for SQLite library. +# This particular file does testing of casting strings into numeric +# values. +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl + +foreach enc {utf8 utf16le utf16be} { + do_test numcast-$enc.0 { + db close + sqlite3 db :memory: + db eval "PRAGMA encoding='$enc'" + set x [db eval {PRAGMA encoding}] + string map {- {}} [string tolower $x] + } $enc + foreach {idx str rval ival} { + 1 12345.0 12345.0 12345 + 2 12345.0e0 12345.0 12345 + 3 -12345.0e0 -12345.0 -12345 + 4 -12345.25 -12345.25 -12345 + 5 { -12345.0} -12345.0 -12345 + 6 { 876xyz} 876.0 876 + 7 { 456ķ89} 456.0 456 + 8 { Ġ 321.5} 0.0 0 + } { + do_test numcast-$enc.$idx.1 { + db eval {SELECT CAST($str AS real)} + } $rval + do_test numcast-$enc.$idx.2 { + db eval {SELECT CAST($str AS integer)} + } $ival + } +} + +finish_test ADDED test/orderby4.test Index: test/orderby4.test ================================================================== --- test/orderby4.test +++ test/orderby4.test @@ -0,0 +1,56 @@ +# 2013 March 26 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# This file implements regression tests for SQLite library. The +# focus of this file is testing that the optimizations that disable +# ORDER BY clauses work correctly on multi-value primary keys and +# unique indices when only some prefix of the terms in the key are +# used. See ticket http://www.sqlite.org/src/info/a179fe74659 +# + + +set testdir [file dirname $argv0] +source $testdir/tester.tcl +set ::testprefix orderby4 + +# Generate test data for a join. Verify that the join gets the +# correct answer. +# +do_execsql_test 1.1 { + CREATE TABLE t1(a, b, PRIMARY KEY(a,b)); + INSERT INTO t1 VALUES(1,1),(1,2); + CREATE TABLE t2(x, y, PRIMARY KEY(x,y)); + INSERT INTO t2 VALUES(3,3),(4,4); + SELECT a, x FROM t1, t2 ORDER BY 1, 2; +} {1 3 1 3 1 4 1 4} +do_execsql_test 1.2 { + SELECT a, x FROM t1 CROSS JOIN t2 ORDER BY 1, 2; +} {1 3 1 3 1 4 1 4} +do_execsql_test 1.3 { + SELECT a, x FROM t2 CROSS JOIN t1 ORDER BY 1, 2; +} {1 3 1 3 1 4 1 4} + +do_execsql_test 2.1 { + CREATE TABLE t3(a); + INSERT INTO t3 VALUES(1),(1); + CREATE INDEX t3a ON t3(a); + CREATE TABLE t4(x); + INSERT INTO t4 VALUES(3),(4); + CREATE INDEX t4x ON t4(x); + SELECT a, x FROM t3, t4 ORDER BY 1, 2; +} {1 3 1 3 1 4 1 4} +do_execsql_test 2.2 { + SELECT a, x FROM t3 CROSS JOIN t4 ORDER BY 1, 2; +} {1 3 1 3 1 4 1 4} +do_execsql_test 2.3 { + SELECT a, x FROM t4 CROSS JOIN t3 ORDER BY 1, 2; +} {1 3 1 3 1 4 1 4} + +finish_test Index: test/pager1.test ================================================================== --- test/pager1.test +++ test/pager1.test @@ -13,10 +13,11 @@ set testdir [file dirname $argv0] source $testdir/tester.tcl source $testdir/lock_common.tcl source $testdir/malloc_common.tcl source $testdir/wal_common.tcl +set testprefix pager1 # Do not use a codec for tests in this file, as the database file is # manipulated directly using tcl scripts (using the [hexio_write] command). # do_not_use_codec @@ -1380,10 +1381,11 @@ # Test that regardless of the value returned by xSectorSize(), the # minimum effective sector-size is 512 and the maximum 65536 bytes. # testvfs tv -default 1 foreach sectorsize { + 16 32 64 128 256 512 1024 2048 4096 8192 16384 32768 65536 131072 262144 } { tv sectorsize $sectorsize tv devchar {} @@ -1402,11 +1404,11 @@ CREATE TABLE t2(a, b); CREATE TABLE t3(a, b); COMMIT; } file size test.db-journal - } [expr $sectorsize > 65536 ? 65536 : $sectorsize] + } [expr $sectorsize > 65536 ? 65536 : ($sectorsize<32 ? 512 : $sectorsize)] do_test pager1-10.$sectorsize.2 { execsql { INSERT INTO t3 VALUES(a_string(300), a_string(300)); INSERT INTO t3 SELECT * FROM t3; /* 2 */ @@ -2241,11 +2243,10 @@ ROLLBACK TO abc; COMMIT; } db close } {} -breakpoint do_test pager1-25-2 { faultsim_delete_and_reopen execsql { SAVEPOINT abc; CREATE TABLE t1(a, b); @@ -2520,7 +2521,298 @@ do_test pager1-33.2 { file rename bak-journal test.db-journal execsql { SELECT * FROM t1 } } {one two} } + +#------------------------------------------------------------------------- +# Test that appending pages to the database file then moving those pages +# to the free-list before the transaction is committed does not cause +# an error. +# +foreach {tn pragma strsize} { + 1 { PRAGMA mmap_size = 0 } 2400 + 2 { } 2400 + 3 { PRAGMA mmap_size = 0 } 4400 + 4 { } 4400 +} { + reset_db + db func a_string a_string + db eval $pragma + do_execsql_test 34.$tn.1 { + CREATE TABLE t1(a, b); + INSERT INTO t1 VALUES(1, 2); + } + do_execsql_test 34.$tn.2 { + BEGIN; + INSERT INTO t1 VALUES(2, a_string($strsize)); + DELETE FROM t1 WHERE oid=2; + COMMIT; + PRAGMA integrity_check; + } {ok} +} + +#------------------------------------------------------------------------- +# +reset_db +do_test 35 { + sqlite3 db test.db + + execsql { + CREATE TABLE t1(x, y); + PRAGMA journal_mode = WAL; + INSERT INTO t1 VALUES(1, 2); + } + + execsql { + BEGIN; + CREATE TABLE t2(a, b); + } + + hexio_write test.db-shm [expr 16*1024] [string repeat 0055 8192] + catchsql ROLLBACK +} {0 {}} + +do_multiclient_test tn { + sql1 { + PRAGMA auto_vacuum = 0; + CREATE TABLE t1(x, y); + INSERT INTO t1 VALUES(1, 2); + } + + do_test 36.$tn.1 { + sql2 { PRAGMA max_page_count = 2 } + list [catch { sql2 { CREATE TABLE t2(x) } } msg] $msg + } {1 {database or disk is full}} + + sql1 { PRAGMA checkpoint_fullfsync = 1 } + sql1 { CREATE TABLE t2(x) } + + do_test 36.$tn.2 { + sql2 { INSERT INTO t2 VALUES('xyz') } + list [catch { sql2 { CREATE TABLE t3(x) } } msg] $msg + } {1 {database or disk is full}} +} + +forcedelete test1 test2 +foreach {tn uri} { + 1 {file:?mode=memory&cache=shared} + 2 {file:one?mode=memory&cache=shared} + 3 {file:test1?cache=shared} + 4 {file:test2?another=parameter&yet=anotherone} +} { + do_test 37.$tn { + catch { db close } + sqlite3_shutdown + sqlite3_config_uri 1 + sqlite3 db $uri + + db eval { + CREATE TABLE t1(x); + INSERT INTO t1 VALUES(1); + SELECT * FROM t1; + } + } {1} + + do_execsql_test 37.$tn.2 { + VACUUM; + SELECT * FROM t1; + } {1} + + db close + sqlite3_shutdown + sqlite3_config_uri 0 +} + +do_test 38.1 { + catch { db close } + forcedelete test.db + set fd [open test.db w] + puts $fd "hello world" + close $fd + sqlite3 db test.db + catchsql { CREATE TABLE t1(x) } +} {1 {file is encrypted or is not a database}} +do_test 38.2 { + catch { db close } + forcedelete test.db +} {} + +do_test 39.1 { + sqlite3 db test.db + execsql { + PRAGMA auto_vacuum = 1; + CREATE TABLE t1(x); + INSERT INTO t1 VALUES('xxx'); + INSERT INTO t1 VALUES('two'); + INSERT INTO t1 VALUES(randomblob(400)); + INSERT INTO t1 VALUES(randomblob(400)); + INSERT INTO t1 VALUES(randomblob(400)); + INSERT INTO t1 VALUES(randomblob(400)); + BEGIN; + UPDATE t1 SET x = 'one' WHERE rowid=1; + } + set ::stmt [sqlite3_prepare db "SELECT * FROM t1 ORDER BY rowid" -1 dummy] + sqlite3_step $::stmt + sqlite3_column_text $::stmt 0 +} {one} +do_test 39.2 { + execsql { CREATE TABLE t2(x) } + sqlite3_step $::stmt + sqlite3_column_text $::stmt 0 +} {two} +do_test 39.3 { + sqlite3_finalize $::stmt + execsql COMMIT +} {} + +do_execsql_test 39.4 { + PRAGMA auto_vacuum = 2; + CREATE TABLE t3(x); + CREATE TABLE t4(x); + + DROP TABLE t2; + DROP TABLE t3; + DROP TABLE t4; +} +do_test 39.5 { + db close + sqlite3 db test.db + execsql { + PRAGMA cache_size = 1; + PRAGMA incremental_vacuum; + PRAGMA integrity_check; + } +} {ok} + +do_test 40.1 { + reset_db + execsql { + PRAGMA auto_vacuum = 1; + CREATE TABLE t1(x PRIMARY KEY); + INSERT INTO t1 VALUES(randomblob(1200)); + PRAGMA page_count; + } +} {6} +do_test 40.2 { + execsql { + INSERT INTO t1 VALUES(randomblob(1200)); + INSERT INTO t1 VALUES(randomblob(1200)); + INSERT INTO t1 VALUES(randomblob(1200)); + } +} {} +do_test 40.3 { + db close + sqlite3 db test.db + execsql { + PRAGMA cache_size = 1; + CREATE TABLE t2(x); + PRAGMA integrity_check; + } +} {ok} + +do_test 41.1 { + reset_db + execsql { + CREATE TABLE t1(x PRIMARY KEY); + INSERT INTO t1 VALUES(randomblob(200)); + INSERT INTO t1 SELECT randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200) FROM t1; + } +} {} +do_test 41.2 { + testvfs tv -default 1 + tv sectorsize 16384; + tv devchar [list] + db close + sqlite3 db test.db + execsql { + PRAGMA cache_size = 1; + DELETE FROM t1 WHERE rowid%4; + PRAGMA integrity_check; + } +} {ok} +db close +tv delete + +set pending_prev [sqlite3_test_control_pending_byte 0x1000000] +do_test 42.1 { + reset_db + execsql { + CREATE TABLE t1(x, y); + INSERT INTO t1 VALUES(randomblob(200), randomblob(200)); + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + } + db close + sqlite3_test_control_pending_byte 0x0010000 + sqlite3 db test.db + db eval { PRAGMA mmap_size = 0 } + catchsql { SELECT sum(length(y)) FROM t1 } +} {1 {database disk image is malformed}} +do_test 42.2 { + reset_db + execsql { + CREATE TABLE t1(x, y); + INSERT INTO t1 VALUES(randomblob(200), randomblob(200)); + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + } + db close + + testvfs tv -default 1 + tv sectorsize 16384; + tv devchar [list] + sqlite3 db test.db -vfs tv + execsql { UPDATE t1 SET x = randomblob(200) } +} {} +db close +tv delete +sqlite3_test_control_pending_byte $pending_prev + +do_test 43.1 { + reset_db + execsql { + CREATE TABLE t1(x, y); + INSERT INTO t1 VALUES(1, 2); + CREATE TABLE t2(x, y); + INSERT INTO t2 VALUES(1, 2); + CREATE TABLE t3(x, y); + INSERT INTO t3 VALUES(1, 2); + } + db close + sqlite3 db test.db + + db eval { PRAGMA mmap_size = 0 } + db eval { SELECT * FROM t1 } + sqlite3_db_status db CACHE_MISS 0 +} {0 2 0} + +do_test 43.2 { + db eval { SELECT * FROM t2 } + sqlite3_db_status db CACHE_MISS 1 +} {0 3 0} + +do_test 43.3 { + db eval { SELECT * FROM t3 } + sqlite3_db_status db CACHE_MISS 0 +} {0 1 0} finish_test + Index: test/pager2.test ================================================================== --- test/pager2.test +++ test/pager2.test @@ -116,11 +116,10 @@ db close tv delete #------------------------------------------------------------------------- -# # pager2-2.1: Test a ROLLBACK with journal_mode=off. # pager2-2.2: Test shrinking the database (auto-vacuum) with # journal_mode=off # do_test pager2-2.1 { @@ -145,7 +144,25 @@ DELETE FROM t1; PRAGMA incremental_vacuum; } file size test.db } {3072} + +#------------------------------------------------------------------------- +# Test that shared in-memory databases seem to work. +# +db close +do_test pager2-3.1 { + forcedelete test.db + sqlite3_shutdown + sqlite3_config_uri 1 + + sqlite3 db1 {file:test.db?mode=memory&cache=shared} + sqlite3 db2 {file:test.db?mode=memory&cache=shared} + sqlite3 db3 test.db + + db1 eval { CREATE TABLE t1(a, b) } + db2 eval { INSERT INTO t1 VALUES(1, 2) } + list [catch { db3 eval { INSERT INTO t1 VALUES(3, 4) } } msg] $msg +} {1 {no such table: t1}} finish_test Index: test/pagerfault.test ================================================================== --- test/pagerfault.test +++ test/pagerfault.test @@ -1244,7 +1244,306 @@ catch { close $::channel } catchsql { ROLLBACK } faultsim_integrity_check } + +#------------------------------------------------------------------------- +# +do_test pagerfault-28-pre { + faultsim_delete_and_reopen + db func a_string a_string + execsql { + PRAGMA page_size = 512; + + PRAGMA journal_mode = wal; + PRAGMA wal_autocheckpoint = 0; + PRAGMA cache_size = 100000; + + BEGIN; + CREATE TABLE t2(a UNIQUE, b UNIQUE); + INSERT INTO t2 VALUES( a_string(800), a_string(800) ); + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + COMMIT; + CREATE TABLE t1(a PRIMARY KEY, b); + } + expr {[file size test.db-shm] >= 96*1024} +} {1} +faultsim_save_and_close + +do_faultsim_test pagerfault-28a -faults oom* -prep { + faultsim_restore_and_reopen + execsql { PRAGMA mmap_size=0 } + + sqlite3 db2 test.db + db2 eval { SELECT count(*) FROM t2 } + + db func a_string a_string + execsql { + BEGIN; + INSERT INTO t1 VALUES(a_string(2000), a_string(2000)); + INSERT INTO t1 VALUES(a_string(2000), a_string(2000)); + } + set ::STMT [sqlite3_prepare db "SELECT * FROM t1 ORDER BY a" -1 DUMMY] + sqlite3_step $::STMT +} -body { + execsql { ROLLBACK } +} -test { + db2 close + sqlite3_finalize $::STMT + catchsql { ROLLBACK } + faultsim_integrity_check +} + +faultsim_restore_and_reopen +sqlite3 db2 test.db +db2 eval {SELECT count(*) FROM t2} +db close + +do_faultsim_test pagerfault-28b -faults oom* -prep { + sqlite3 db test.db +} -body { + execsql { SELECT count(*) FROM t2 } +} -test { + faultsim_test_result {0 2048} + db close +} + +db2 close + +#------------------------------------------------------------------------- +# Try this: +# +# 1) Put the pager in ERROR state (error during rollback) +# +# 2) Next time the connection is used inject errors into all xWrite() and +# xUnlock() calls. This causes the hot-journal rollback to fail and +# the pager to declare its locking state UNKNOWN. +# +# 3) Same again. +# +# 4a) Stop injecting errors. Allow the rollback to succeed. Check that +# the database is Ok. Or, +# +# 4b) Close and reopen the db. Check that the db is Ok. +# +proc custom_injectinstall {} { + testvfs custom -default true + custom filter {xWrite xUnlock} +} +proc custom_injectuninstall {} { + catch {db close} + catch {db2 close} + custom delete +} +proc custom_injectstart {iFail} { + custom ioerr $iFail 1 +} +proc custom_injectstop {} { + custom ioerr +} +set ::FAULTSIM(custom) [list \ + -injectinstall custom_injectinstall \ + -injectstart custom_injectstart \ + -injectstop custom_injectstop \ + -injecterrlist {{1 {disk I/O error}}} \ + -injectuninstall custom_injectuninstall \ +] + +do_test pagerfault-29-pre { + faultsim_delete_and_reopen + db func a_string a_string + execsql { + PRAGMA page_size = 1024; + PRAGMA cache_size = 5; + + BEGIN; + CREATE TABLE t2(a UNIQUE, b UNIQUE); + INSERT INTO t2 VALUES( a_string(800), a_string(800) ); + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + INSERT INTO t2 SELECT a_string(800), a_string(800) FROM t2; + COMMIT; + } + expr {[file size test.db] >= 50*1024} +} {1} +faultsim_save_and_close +foreach {tn tt} { + 29 { catchsql ROLLBACK } + 30 { db close ; sqlite3 db test.db } +} { + do_faultsim_test pagerfault-$tn -faults custom -prep { + faultsim_restore_and_reopen + db func a_string a_string + execsql { + PRAGMA cache_size = 5; + BEGIN; + UPDATE t2 SET a = a_string(799); + } + } -body { + catchsql ROLLBACK + catchsql ROLLBACK + catchsql ROLLBACK + } -test { + eval $::tt + if {"ok" != [db one {PRAGMA integrity_check}]} { + error "integrity check failed" + } + } +} + +do_test pagerfault-31-pre { + sqlite3_shutdown + sqlite3_config_uri 1 +} {SQLITE_OK} +do_faultsim_test pagerfault-31 -faults oom* -body { + sqlite3 db {file:one?mode=memory&cache=shared} + db eval { + CREATE TABLE t1(x); + INSERT INTO t1 VALUES(1); + SELECT * FROM t1; + } +} -test { + faultsim_test_result {0 1} {1 {}} + catch { db close } +} +sqlite3_shutdown +sqlite3_config_uri 0 + +do_test pagerfault-32-pre { + reset_db + execsql { + CREATE TABLE t1(x); + INSERT INTO t1 VALUES('one'); + } +} {} +faultsim_save_and_close + +do_faultsim_test pagerfault-32 -prep { + faultsim_restore_and_reopen + db eval { SELECT * FROM t1; } +} -body { + execsql { SELECT * FROM t1; } +} -test { + faultsim_test_result {0 one} +} +sqlite3_shutdown +sqlite3_config_uri 0 + +do_faultsim_test pagerfault-33a -prep { + sqlite3 db :memory: + execsql { + CREATE TABLE t1(a, b); + INSERT INTO t1 VALUES(1, 2); + } +} -body { + execsql { VACUUM } +} -test { + faultsim_test_result {0 {}} +} +do_faultsim_test pagerfault-33b -prep { + sqlite3 db "" + execsql { + CREATE TABLE t1(a, b); + INSERT INTO t1 VALUES(1, 2); + } +} -body { + execsql { VACUUM } +} -test { + faultsim_test_result {0 {}} +} + +do_test pagerfault-34-pre { + reset_db + execsql { + CREATE TABLE t1(x PRIMARY KEY); + } +} {} +faultsim_save_and_close +do_faultsim_test pagerfault-34 -prep { + faultsim_restore_and_reopen + execsql { + BEGIN; + INSERT INTO t1 VALUES( randomblob(4000) ); + DELETE FROM t1; + } +} -body { + execsql COMMIT +} -test { + faultsim_test_result {0 {}} +} + +do_test pagerfault-35-pre { + faultsim_delete_and_reopen + execsql { + CREATE TABLE t1(x PRIMARY KEY, y); + INSERT INTO t1 VALUES(randomblob(200), randomblob(200)); + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + } + faultsim_save_and_close +} {} +testvfs tv -default 1 +tv sectorsize 8192; +tv devchar [list] +do_faultsim_test pagerfault-35 -prep { + faultsim_restore_and_reopen +} -body { + execsql { UPDATE t1 SET x=randomblob(200) } +} -test { + faultsim_test_result {0 {}} +} +catch {db close} +tv delete + +sqlite3_shutdown +sqlite3_config_uri 1 +do_test pagerfault-36-pre { + faultsim_delete_and_reopen + execsql { + CREATE TABLE t1(x PRIMARY KEY, y); + INSERT INTO t1 VALUES(randomblob(200), randomblob(200)); + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + INSERT INTO t1 SELECT randomblob(200), randomblob(200) FROM t1; + } + faultsim_save_and_close +} {} +do_faultsim_test pagerfault-36 -prep { + faultsim_restore + sqlite3 db file:test.db?cache=shared + sqlite3 db2 file:test.db?cache=shared + db2 eval { + BEGIN; + SELECT count(*) FROM sqlite_master; + } + db eval { + PRAGMA cache_size = 1; + BEGIN; + UPDATE t1 SET x = randomblob(200); + } +} -body { + execsql ROLLBACK db +} -test { + catch { db eval {UPDATE t1 SET x = randomblob(200)} } + faultsim_test_result {0 {}} + catch { db close } + catch { db2 close } +} + +sqlite3_shutdown +sqlite3_config_uri 0 + finish_test Index: test/pageropt.test ================================================================== --- test/pageropt.test +++ test/pageropt.test @@ -85,16 +85,21 @@ } [list 0 0 0 $blobcontent] # But if the other thread modifies the database, then the cache # must refill. # +ifcapable mmap { + set x [expr {[permutation]=="mmap" ? 1 : 6}] +} else { + set x 6 +} do_test pageropt-1.5 { db2 eval {CREATE TABLE t2(y)} pagercount_sql { SELECT hex(x) FROM t1 } -} [list 6 0 0 $blobcontent] +} [list $x 0 0 $blobcontent] do_test pageropt-1.6 { pagercount_sql { SELECT hex(x) FROM t1 } } [list 0 0 0 $blobcontent] Index: test/permutations.test ================================================================== --- test/permutations.test +++ test/permutations.test @@ -135,10 +135,18 @@ This test suite is the same as the "quick" tests, except that some files that test malloc and IO errors are omitted. } -files [ test_set $allquicktests -exclude *malloc* *ioerr* *fault* ] + +test_suite "mmap" -prefix "mm-" -description { + Similar to veryquick. Except with memory mapping disabled. +} -presql { + pragma mmap_size = 268435456; +} -files [ + test_set $allquicktests -exclude *malloc* *ioerr* *fault* -include malloc.test +] test_suite "valgrind" -prefix "" -description { Run the "veryquick" test suite with a couple of multi-process tests (that fail under valgrind) omitted. } -files [ @@ -176,10 +184,11 @@ fts3aa.test fts3ab.test fts3ac.test fts3ad.test fts3ae.test fts3af.test fts3ag.test fts3ah.test fts3ai.test fts3aj.test fts3ak.test fts3al.test fts3am.test fts3an.test fts3ao.test fts3atoken.test fts3b.test fts3c.test fts3cov.test fts3d.test fts3defer.test fts3defer2.test fts3e.test fts3expr.test fts3expr2.test + fts3expr3.test fts3near.test fts3query.test fts3shared.test fts3snippet.test fts3sort.test fts3fault.test fts3malloc.test fts3matchinfo.test fts3aux1.test fts3comp1.test fts3auto.test fts4aa.test fts4content.test Index: test/pragma.test ================================================================== --- test/pragma.test +++ test/pragma.test @@ -934,10 +934,20 @@ } } return "unknown" } +# Application_ID +# +do_test pragma-8.3.1 { + execsql { + PRAGMA application_id; + } +} {0} +do_test pragma-8.3.2 { + execsql {PRAGMA Application_ID(12345); PRAGMA application_id;} +} {12345} # Test temp_store and temp_store_directory pragmas # ifcapable pager_pragmas { do_test pragma-9.1 { Index: test/regexp1.test ================================================================== --- test/regexp1.test +++ test/regexp1.test @@ -14,11 +14,11 @@ set testdir [file dirname $argv0] source $testdir/tester.tcl do_test regexp1-1.1 { - sqlite3_add_regexp_func db + load_static_extension db regexp db eval { CREATE TABLE t1(x INTEGER PRIMARY KEY, y TEXT); INSERT INTO t1 VALUES(1, 'For since by man came death,'); INSERT INTO t1 VALUES(2, 'by man came also the resurrection of the dead.'); INSERT INTO t1 VALUES(3, 'For as in Adam all die,'); ADDED test/resolver01.test Index: test/resolver01.test ================================================================== --- test/resolver01.test +++ test/resolver01.test @@ -0,0 +1,39 @@ +# 2013-04-13 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# This file tests features of the name resolver (the component that +# figures out what identifiers in the SQL statement refer to) that +# were fixed by ticket [2500cdb9be] +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl + +do_test resolver01-1.1 { + catchsql { + CREATE TABLE t1(x, y); INSERT INTO t1 VALUES(11,22); + CREATE TABLE t2(y, z); INSERT INTO t2 VALUES(33,44); + SELECT 1 AS y FROM t1, t2 ORDER BY y; + } +} {0 1} +do_test resolver01-1.2 { + catchsql { + SELECT 2 AS y FROM t1, t2 ORDER BY y COLLATE nocase; + } +} {0 2} +do_test resolver01-1.3 { + catchsql { + SELECT 3 AS y FROM t1, t2 ORDER BY +y; + } +} {0 3} + + +finish_test Index: test/selectA.test ================================================================== --- test/selectA.test +++ test/selectA.test @@ -279,17 +279,17 @@ } } {{} C c {} U u 5200000.0 X x -23 Y y mad Z z 1 a a 9.9 b B hello d D abc e e hare m M} do_test selectA-2.35 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 - ORDER BY b COLLATE NOCASE,a,c + ORDER BY y COLLATE NOCASE,x,z } } {1 a a 9.9 b B {} C c hello d D abc e e hare m M {} U u 5200000.0 X x -23 Y y mad Z z} do_test selectA-2.36 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 - ORDER BY b COLLATE NOCASE DESC,a,c + ORDER BY y COLLATE NOCASE DESC,x,z } } {mad Z z -23 Y y 5200000.0 X x {} U u hare m M abc e e hello d D {} C c 9.9 b B 1 a a} do_test selectA-2.37 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 @@ -309,11 +309,11 @@ } } {mad Z z -23 Y y 5200000.0 X x {} U u hare m M abc e e hello d D {} C c 9.9 b B 1 a a} do_test selectA-2.40 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 - ORDER BY c COLLATE BINARY DESC,a,b + ORDER BY z COLLATE BINARY DESC,x,y } } {mad Z z -23 Y y 5200000.0 X x {} U u abc e e {} C c 1 a a hare m M hello d D 9.9 b B} do_test selectA-2.41 { execsql { SELECT a,b,c FROM t1 EXCEPT SELECT a,b,c FROM t1 WHERE b>='d' @@ -600,11 +600,11 @@ } } {{} C c {} U u 5200000.0 X x -23 Y y mad Z z 1 a a 9.9 b B hello d D abc e e hare m M} do_test selectA-2.86 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t3 - ORDER BY b COLLATE NOCASE,a,c + ORDER BY y COLLATE NOCASE,x,z } } {1 a a 9.9 b B {} C c hello d D abc e e hare m M {} U u 5200000.0 X x -23 Y y mad Z z} do_test selectA-2.87 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t3 @@ -630,11 +630,11 @@ } } {mad Z z -23 Y y 5200000.0 X x {} U u hare m M abc e e hello d D {} C c 9.9 b B 1 a a} do_test selectA-2.91 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t3 - ORDER BY c COLLATE BINARY DESC,a,b + ORDER BY z COLLATE BINARY DESC,x,y } } {mad Z z -23 Y y 5200000.0 X x {} U u abc e e {} C c 1 a a hare m M hello d D 9.9 b B} do_test selectA-2.92 { execsql { SELECT x,y,z FROM t2 @@ -891,17 +891,17 @@ } } {{} C c {} U u 5200000.0 X x -23 Y y mad Z z 1 a a 9.9 b B hello d D abc e e hare m M} do_test selectA-3.35 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 - ORDER BY b COLLATE NOCASE,a,c + ORDER BY y COLLATE NOCASE,x,z } } {1 a a 9.9 b B {} C c hello d D abc e e hare m M {} U u 5200000.0 X x -23 Y y mad Z z} do_test selectA-3.36 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 - ORDER BY b COLLATE NOCASE DESC,a,c + ORDER BY y COLLATE NOCASE DESC,x,z } } {mad Z z -23 Y y 5200000.0 X x {} U u hare m M abc e e hello d D {} C c 9.9 b B 1 a a} do_test selectA-3.37 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 @@ -921,11 +921,11 @@ } } {mad Z z -23 Y y 5200000.0 X x {} U u hare m M abc e e hello d D {} C c 9.9 b B 1 a a} do_test selectA-3.40 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t1 - ORDER BY c COLLATE BINARY DESC,a,b + ORDER BY z COLLATE BINARY DESC,x,y } } {mad Z z -23 Y y 5200000.0 X x {} U u abc e e {} C c 1 a a hare m M hello d D 9.9 b B} do_test selectA-3.41 { execsql { SELECT a,b,c FROM t1 EXCEPT SELECT a,b,c FROM t1 WHERE b>='d' @@ -1212,11 +1212,11 @@ } } {{} C c {} U u 5200000.0 X x -23 Y y mad Z z 1 a a 9.9 b B hello d D abc e e hare m M} do_test selectA-3.86 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t3 - ORDER BY b COLLATE NOCASE,a,c + ORDER BY y COLLATE NOCASE,x,z } } {1 a a 9.9 b B {} C c hello d D abc e e hare m M {} U u 5200000.0 X x -23 Y y mad Z z} do_test selectA-3.87 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t3 @@ -1242,11 +1242,11 @@ } } {mad Z z -23 Y y 5200000.0 X x {} U u hare m M abc e e hello d D {} C c 9.9 b B 1 a a} do_test selectA-3.91 { execsql { SELECT x,y,z FROM t2 UNION SELECT a,b,c FROM t3 - ORDER BY c COLLATE BINARY DESC,a,b + ORDER BY z COLLATE BINARY DESC,x,y } } {mad Z z -23 Y y 5200000.0 X x {} U u abc e e {} C c 1 a a hare m M hello d D 9.9 b B} do_test selectA-3.92 { execsql { SELECT x,y,z FROM t2 Index: test/selectD.test ================================================================== --- test/selectD.test +++ test/selectD.test @@ -149,7 +149,26 @@ FROM (t1 LEFT JOIN t2 USING(a)) JOIN (t3 LEFT JOIN t4 USING(a)) ON t1.a=t3.a-111; } } {111 x1 111 x2 222 x3 {}} } + +# The following test was added on 2013-04-24 in order to verify that +# the datatypes and affinities of sub-sub-queries are set prior to computing +# the datatypes and affinities of the parent sub-queries because the +# latter computation depends on the former. +# +do_execsql_test selectD-4.1 { + CREATE TABLE t41(a INTEGER PRIMARY KEY, b INTEGER); + CREATE TABLE t42(d INTEGER PRIMARY KEY, e INTEGER); + CREATE TABLE t43(f INTEGER PRIMARY KEY, g INTEGER); + EXPLAIN QUERY PLAN + SELECT * + FROM t41 + LEFT JOIN (SELECT count(*) AS cnt, x1.d + FROM (t42 INNER JOIN t43 ON d=g) AS x1 + WHERE x1.d>5 + GROUP BY x1.d) AS x2 + ON t41.b=x2.d; +} {/.*SEARCH SUBQUERY 1 AS x2 USING AUTOMATIC.*/} finish_test ADDED test/selectE.test Index: test/selectE.test ================================================================== --- test/selectE.test +++ test/selectE.test @@ -0,0 +1,95 @@ +# 2013-05-07 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# This file implements regression tests for compound SELECT statements +# that have ORDER BY clauses with collating sequences that differ +# from the collating sequence used for comparison in the compound. +# +# Ticket 6709574d2a8d8b9be3a9cb1afbf4ff2de48ea4e7: +# drh added on 2013-05-06 15:21:16: +# +# In the code shown below (which is intended to be run from the +# sqlite3.exe command-line tool) the three SELECT statements should all +# generate the same answer. But the third one does not. It is as if the +# COLLATE clause on the ORDER BY somehow got pulled into the EXCEPT +# operator. Note that the ".print" commands are instructions to the +# sqlite3.exe shell program to output delimiter lines so that you can more +# easily tell where the output of one query ends and the next query +# begins. +# +# CREATE TABLE t1(a); +# INSERT INTO t1 VALUES('abc'),('def'); +# CREATE TABLE t2(a); +# INSERT INTO t2 VALUES('DEF'); +# +# SELECT a FROM t1 EXCEPT SELECT a FROM t2 ORDER BY a; +# .print ----- +# SELECT a FROM (SELECT a FROM t1 EXCEPT SELECT a FROM t2) +# ORDER BY a COLLATE nocase; +# .print ----- +# SELECT a FROM t1 EXCEPT SELECT a FROM t2 ORDER BY a COLLATE nocase; +# +# Bisecting shows that this problem was introduced in SQLite version 3.6.0 +# by check-in [8bbfa97837a74ef] on 2008-06-15. +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl + +do_test selectE-1.0 { + db eval { + CREATE TABLE t1(a); + INSERT INTO t1 VALUES('abc'),('def'),('ghi'); + CREATE TABLE t2(a); + INSERT INTO t2 VALUES('DEF'),('abc'); + CREATE TABLE t3(a); + INSERT INTO t3 VALUES('def'),('jkl'); + + SELECT a FROM t1 EXCEPT SELECT a FROM t2 + ORDER BY a COLLATE nocase; + } +} {def ghi} +do_test selectE-1.1 { + db eval { + SELECT a FROM t2 EXCEPT SELECT a FROM t3 + ORDER BY a COLLATE nocase; + } +} {abc DEF} +do_test selectE-1.2 { + db eval { + SELECT a FROM t2 EXCEPT SELECT a FROM t3 + ORDER BY a COLLATE binary; + } +} {DEF abc} +do_test selectE-1.3 { + db eval { + SELECT a FROM t2 EXCEPT SELECT a FROM t3 + ORDER BY a; + } +} {DEF abc} + +do_test selectE-2.1 { + db eval { + DELETE FROM t2; + DELETE FROM t3; + INSERT INTO t2 VALUES('ABC'),('def'),('GHI'),('jkl'); + INSERT INTO t3 SELECT lower(a) FROM t2; + SELECT a COLLATE nocase FROM t2 EXCEPT SELECT a FROM t3 + ORDER BY 1 + } +} {} +do_test selectE-2.2 { + db eval { + SELECT a COLLATE nocase FROM t2 EXCEPT SELECT a FROM t3 + ORDER BY 1 COLLATE binary + } +} {} + +finish_test Index: test/speed1p.test ================================================================== --- test/speed1p.test +++ test/speed1p.test @@ -22,10 +22,12 @@ #sqlite3_config_pagecache 1024 11000 set testdir [file dirname $argv0] source $testdir/tester.tcl speed_trial_init speed1 +sqlite3_memdebug_vfs_oom_test 0 + # Set a uniform random seed expr srand(0) # The number_name procedure below converts its argment (an integer) # into a string which is the English-language name for that number. @@ -75,11 +77,10 @@ } execsql { SELECT name FROM sqlite_master ORDER BY 1; } } {i2a i2b t1 t2} - # 50000 INSERTs on an unindexed table # set list {} for {set i 1} {$i<=50000} {incr i} { Index: test/spellfix.test ================================================================== --- test/spellfix.test +++ test/spellfix.test @@ -14,11 +14,11 @@ source $testdir/tester.tcl set testprefix spellfix ifcapable !vtab { finish_test ; return } -register_spellfix_module db +load_static_extension db spellfix nextchar set vocab { rabbi rabbit rabbits rabble rabid rabies raccoon raccoons race raced racer racers races racetrack racial racially racing rack racked racket racketeer racketeering racketeers rackets racking racks radar radars radial radially @@ -82,10 +82,30 @@ SELECT word, matchlen FROM t1 WHERE word MATCH $word ORDER BY score, word LIMIT 5 } $res } +# Tests of the next_char function. +# +do_test 1.10 { + db eval { + CREATE TABLE vocab(w TEXT PRIMARY KEY); + INSERT INTO vocab SELECT word FROM t1; + } +} {} +do_execsql_test 1.11 { + SELECT next_char('re','vocab','w'); +} {a} +do_execsql_test 1.12 { + SELECT next_char('r','vocab','w'); +} {ae} +do_execsql_test 1.13 { + SELECT next_char('','vocab','w'); +} {r} +do_test 1.14 { + catchsql {SELECT next_char('','xyzzy','a')} +} {1 {no such table: xyzzy}} do_execsql_test 2.1 { CREATE VIRTUAL TABLE t2 USING spellfix1; INSERT INTO t2 (word, soundslike) VALUES('school', 'skuul'); INSERT INTO t2 (word, soundslike) VALUES('psalm', 'sarm'); Index: test/subquery.test ================================================================== --- test/subquery.test +++ test/subquery.test @@ -419,11 +419,11 @@ #------------------------------------------------------------------ # These tests - subquery-4.* - use the TCL statement cache to try # and expose bugs to do with re-using statements that have been # passed to sqlite3_reset(). # -# One problem was that VDBE memory cells were not being initialised +# One problem was that VDBE memory cells were not being initialized # to NULL on the second and subsequent executions. # do_test subquery-4.1.1 { execsql { SELECT (SELECT a FROM t1); Index: test/syscall.test ================================================================== --- test/syscall.test +++ test/syscall.test @@ -58,11 +58,11 @@ # foreach s { open close access getcwd stat fstat ftruncate fcntl read pread write pwrite fchmod fallocate pread64 pwrite64 unlink openDirectory mkdir rmdir - statvfs fchown umask + statvfs fchown umask mmap munmap mremap } { if {[test_syscall exists $s]} {lappend syscall_list $s} } do_test 3.1 { lsort [test_syscall list] } [lsort $syscall_list] Index: test/sysfault.test ================================================================== --- test/sysfault.test +++ test/sysfault.test @@ -241,7 +241,37 @@ } } -test { faultsim_test_result {0 20000} } -finish_test +#------------------------------------------------------------------------- +# Test errors in mmap(). +# +proc vfsfault_install {} { + test_syscall reset + test_syscall install {mmap} +} + +faultsim_delete_and_reopen +execsql { + CREATE TABLE t1(a, b); + INSERT INTO t1 VALUES(1, 2); +} +faultsim_save_and_close + +do_faultsim_test 4 -faults vfsfault-* -prep { + faultsim_restore_and_reopen + file_control_chunksize_test db main 8192 + execsql { + PRAGMA mmap_size = 1000000; + } +} -body { + test_syscall errno mmap EACCES + + execsql { + SELECT * FROM t1; + } +} -test { + faultsim_test_result {0 {1 2}} {1 {disk I/O error}} +} +finish_test Index: test/temptable.test ================================================================== --- test/temptable.test +++ test/temptable.test @@ -148,11 +148,11 @@ } {1 {no such index: i2}} # Check for correct name collision processing. A name collision can # occur when process A creates a temporary table T then process B # creates a permanent table also named T. The temp table in process A -# hides the existance of the permanent table. +# hides the existence of the permanent table. # do_test temptable-4.1 { execsql { CREATE TEMP TABLE t2(x,y); INSERT INTO t2 VALUES(10,20); Index: test/tester.tcl ================================================================== --- test/tester.tcl +++ test/tester.tcl @@ -121,11 +121,11 @@ uplevel #0 $::G(perm:dbconfig) } set res } else { # This command is not opening a new database connection. Pass the - # arguments through to the C implemenation as the are. + # arguments through to the C implementation as the are. # uplevel 1 sqlite_orig $args } } } @@ -460,10 +460,11 @@ if {0==[info exists ::SLAVE]} { set TC(errors) 0 set TC(count) 0 set TC(fail_list) [list] set TC(omit_list) [list] + set TC(warn_list) [list] proc set_test_counter {counter args} { if {[llength $args]} { set ::TC($counter) [lindex $args 0] } @@ -493,10 +494,22 @@ if {$nFail>=$::cmdlinearg(maxerror)} { puts "*** Giving up..." finalize_testing } } + +# Remember a warning message to be displayed at the conclusion of all testing +# +proc warning {msg {append 1}} { + puts "Warning: $msg" + set warnList [set_test_counter warn_list] + if {$append} { + lappend warnList $msg + } + set_test_counter warn_list $warnList +} + # Increment the number of tests run # proc incr_ntest {} { set_test_counter count [expr [set_test_counter count] + 1] @@ -782,10 +795,13 @@ puts "$nErr errors out of $nTest tests" if {$nErr>0} { puts "Failures on these tests: [set_test_counter fail_list]" } + foreach warning [set_test_counter warn_list] { + puts "Warning: $warning" + } run_thread_tests 1 if {[llength $omitList]>0} { puts "Omitted test cases:" set prec {} foreach {rec} [lsort $omitList] { @@ -1035,11 +1051,11 @@ # # The return value is a list of two elements. The first element is a # boolean, indicating whether or not the process actually crashed or # reported some other error. The second element in the returned list is the # error message. This is "child process exited abnormally" if the crash -# occured. +# occurred. # # crashsql -delay CRASHDELAY -file CRASHFILE ?-blocksize BLOCKSIZE? $sql # proc crashsql {args} { @@ -1315,11 +1331,11 @@ set stats(state) } 0 } } - # If an IO error occured, then the checksum of the database should + # If an IO error occurred, then the checksum of the database should # be the same as before the script that caused the IO error was run. # if {$::go && $::sqlite_io_error_hardhit && $::ioerropts(-cksum)} { do_test $testname.$n.6 { catch {db close} Index: test/tkt-2d1a5c67d.test ================================================================== --- test/tkt-2d1a5c67d.test +++ test/tkt-2d1a5c67d.test @@ -44,11 +44,11 @@ } db close forcedelete test.db test.db-wal sqlite3 db test.db -register_wholenumber_module db +load_static_extension db wholenumber db eval { PRAGMA journal_mode=WAL; CREATE TABLE t1(a,b); CREATE INDEX t1b ON t1(b); CREATE TABLE t2(x,y); ADDED test/tkt-6bfb98dfc0.test Index: test/tkt-6bfb98dfc0.test ================================================================== --- test/tkt-6bfb98dfc0.test +++ test/tkt-6bfb98dfc0.test @@ -0,0 +1,61 @@ +# 2013 March 27 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# This file implements regression tests for SQLite library. Specifically, +# it tests that ticket [6bfb98dfc0] +# +# The final INSERT in the script below reports that the database is +# corrupt (SQLITE_CORRUPT) and aborts even though the database is not +# corrupt. +# +# PRAGMA page_size=512; +# CREATE TABLE t1(x INTEGER PRIMARY KEY, y); +# INSERT INTO t1 VALUES(1,randomblob(400)); +# INSERT INTO t1 VALUES(2,randomblob(400)); +# INSERT INTO t1 SELECT x+2, randomblob(400) FROM t1; +# INSERT INTO t1 SELECT x+4, randomblob(400) FROM t1; +# INSERT INTO t1 SELECT x+8, randomblob(400) FROM t1; +# INSERT INTO t1 SELECT x+16, randomblob(400) FROM t1; +# INSERT INTO t1 SELECT x+32, randomblob(400) FROM t1; +# INSERT INTO t1 SELECT x+64, randomblob(400) FROM t1 WHERE x<10; +# CREATE TRIGGER r1 AFTER INSERT ON t1 WHEN new.x=74 BEGIN +# DELETE FROM t1; +# INSERT INTO t1 VALUES(75, randomblob(400)); +# INSERT INTO t1 VALUES(76, randomblob(400)); +# END; +# INSERT INTO t1 VALUES(74, randomblob(400)); +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl + +do_test tkt-6bfb98dfc0.100 { + db eval { + PRAGMA page_size=512; + CREATE TABLE t1(x INTEGER PRIMARY KEY, y); + INSERT INTO t1 VALUES(1,randomblob(400)); + INSERT INTO t1 VALUES(2,randomblob(400)); + INSERT INTO t1 SELECT x+2, randomblob(400) FROM t1; + INSERT INTO t1 SELECT x+4, randomblob(400) FROM t1; + INSERT INTO t1 SELECT x+8, randomblob(400) FROM t1; + INSERT INTO t1 SELECT x+16, randomblob(400) FROM t1; + INSERT INTO t1 SELECT x+32, randomblob(400) FROM t1; + INSERT INTO t1 SELECT x+64, randomblob(400) FROM t1 WHERE x<10; + CREATE TRIGGER r1 AFTER INSERT ON t1 WHEN new.x=74 BEGIN + DELETE FROM t1; + INSERT INTO t1 VALUES(75, randomblob(400)); + INSERT INTO t1 VALUES(76, randomblob(400)); + END; + INSERT INTO t1 VALUES(74, randomblob(400)); + SELECT x, length(y) FROM t1 ORDER BY x; + } +} {75 400 76 400} + +finish_test Index: test/tkt2409.test ================================================================== --- test/tkt2409.test +++ test/tkt2409.test @@ -105,11 +105,11 @@ # Check the integrity of the cache. # integrity_check tkt2409-1.3 # Check that the transaction was rolled back. Because the INSERT -# statement in which the "I/O error" occured did not open a statement +# statement in which the "I/O error" occurred did not open a statement # transaction, SQLite had no choice but to roll back the transaction. # do_test tkt2409-1.4 { unread_lock_db catchsql { ROLLBACK } @@ -173,11 +173,11 @@ # Check the integrity of the cache. # integrity_check tkt2409-3.3 # Check that the transaction was rolled back. Because the INSERT -# statement in which the "I/O error" occured did not open a statement +# statement in which the "I/O error" occurred did not open a statement # transaction, SQLite had no choice but to roll back the transaction. # do_test tkt2409-3.4 { unread_lock_db catchsql { ROLLBACK } Index: test/tkt2822.test ================================================================== --- test/tkt2822.test +++ test/tkt2822.test @@ -206,16 +206,19 @@ } } {1 9} # In "ORDER BY +b" the term is now an expression rather than # a label. It therefore matches by rule (3) instead of rule (2). +# +# 2013-04-13: This is busted. Changed to conform to PostgreSQL and +# MySQL and Oracle behavior. # do_test tkt2822-5.5 { execsql { SELECT a AS b FROM t3 ORDER BY +b; } -} {9 1} +} {1 9} # Tests for rule 2 in compound queries # do_test tkt2822-6.1 { execsql { Index: test/wal.test ================================================================== --- test/wal.test +++ test/wal.test @@ -725,39 +725,42 @@ do_test wal-11.9 { db close list [expr [file size test.db]/1024] [log_deleted test.db-wal] } {37 1} sqlite3_wal db test.db +set nWal 39 +if {[permutation]!="mmap"} {set nWal 37} +ifcapable !mmap {set nWal 37} do_test wal-11.10 { execsql { PRAGMA cache_size = 10; BEGIN; INSERT INTO t1 SELECT blob(900) FROM t1; -- 32 SELECT count(*) FROM t1; } list [expr [file size test.db]/1024] [file size test.db-wal] -} [list 37 [wal_file_size 37 1024]] +} [list 37 [wal_file_size $nWal 1024]] do_test wal-11.11 { execsql { SELECT count(*) FROM t1; ROLLBACK; SELECT count(*) FROM t1; } } {32 16} do_test wal-11.12 { list [expr [file size test.db]/1024] [file size test.db-wal] -} [list 37 [wal_file_size 37 1024]] +} [list 37 [wal_file_size $nWal 1024]] do_test wal-11.13 { execsql { INSERT INTO t1 VALUES( blob(900) ); SELECT count(*) FROM t1; PRAGMA integrity_check; } } {17 ok} do_test wal-11.14 { list [expr [file size test.db]/1024] [file size test.db-wal] -} [list 37 [wal_file_size 37 1024]] +} [list 37 [wal_file_size $nWal 1024]] #------------------------------------------------------------------------- # This block of tests, wal-12.*, tests the fix for a problem that # could occur if a log that is a prefix of an older log is written @@ -1510,11 +1513,12 @@ execsql { SELECT * FROM t1 } } {1 2 3 4} set nPage [expr 2+$AUTOVACUUM] do_test wal-23.4 { set ::log -} [list SQLITE_OK "Recovered $nPage frames from WAL file $walfile"] +} [list SQLITE_NOTICE_RECOVER_WAL \ + "recovered $nPage frames from WAL file $walfile"] ifcapable autovacuum { # This block tests that if the size of a database is reduced by a # transaction (because of an incremental or auto-vacuum), that no Index: test/wal5.test ================================================================== --- test/wal5.test +++ test/wal5.test @@ -233,11 +233,20 @@ do_test 2.3.$tn.3 { sql2 { BEGIN; SELECT * FROM t1 } } {1 2} do_test 2.3.$tn.4 { sql1 { INSERT INTO t1 VALUES(3, 4) } } {} do_test 2.3.$tn.5 { sql1 { INSERT INTO t2 VALUES(3, 4) } } {} do_test 2.3.$tn.6 { file_page_counts } {1 4 1 4} do_test 2.3.$tn.7 { code1 { do_wal_checkpoint db -mode full } } {1 4 3} - do_test 2.3.$tn.8 { file_page_counts } {1 4 2 4} + + # The checkpoint above only writes page 1 of the db file. The other + # page (page 2) is locked by the read-transaction opened by the + # [sql2] commmand above. So normally, the db is 1 page in size here. + # However, in mmap() mode, the db is pre-allocated to 2 pages at the + # start of the checkpoint, even though page 2 cannot be written. + set nDb 2 + if {[permutation]!="mmap"} {set nDb 1} + ifcapable !mmap {set nDb 1} + do_test 2.3.$tn.8 { file_page_counts } [list $nDb 4 2 4] } # Check that checkpoints block on the correct locks. And respond correctly # if they cannot obtain those locks. There are three locks that a checkpoint # may block on (in the following order): @@ -341,6 +350,5 @@ } } finish_test - Index: test/walfault.test ================================================================== --- test/walfault.test +++ test/walfault.test @@ -546,8 +546,46 @@ faultsim_test_result {0 {0 9 9}} faultsim_integrity_check set nRow [db eval {SELECT count(*) FROM abc}] if {!(($nRow==2 && $testrc) || $nRow==3)} { error "Bad db content" } } -finish_test + +#------------------------------------------------------------------------- +# Test fault-handling when switching out of exclusive-locking mode. +# +do_test walfault-14-pre { + faultsim_delete_and_reopen + execsql { + PRAGMA auto_vacuum = 0; + PRAGMA journal_mode = WAL; + BEGIN; + CREATE TABLE abc(a PRIMARY KEY); + INSERT INTO abc VALUES(randomblob(1500)); + INSERT INTO abc VALUES(randomblob(1500)); + COMMIT; + } + faultsim_save_and_close +} {} +do_faultsim_test walfault-14 -prep { + faultsim_restore_and_reopen + breakpoint + execsql { + SELECT count(*) FROM abc; + PRAGMA locking_mode = exclusive; + BEGIN; + INSERT INTO abc VALUES(randomblob(1500)); + COMMIT; + } +} -body { + db eval { + PRAGMA locking_mode = normal; + BEGIN; + INSERT INTO abc VALUES(randomblob(1500)); + COMMIT; + } +} -test { + faultsim_integrity_check + set nRow [db eval {SELECT count(*) FROM abc}] + if {$nRow!=3 && $nRow!=4} { error "Bad db content" } +} finish_test Index: test/where8.test ================================================================== --- test/where8.test +++ test/where8.test @@ -293,18 +293,32 @@ do_test where8-3.21 { execsql_status { SELECT a, d FROM t1, (t2) WHERE (a=d OR b=e) AND a<5 ORDER BY a } +} {1 1 2 2 3 3 4 2 4 4 0 0} +do_test where8-3.21.1 { + execsql_status { + SELECT a, d FROM t1, ((t2)) AS t3 WHERE (a=d OR b=e) AND a<5 ORDER BY a + } +} {1 1 2 2 3 3 4 2 4 4 0 0} +do_test where8-3.21.2 { + execsql_status { + SELECT a, d FROM t1, ((SELECT * FROM t2)) AS t3 WHERE (a=d OR b=e) AND a<5 ORDER BY a + } } {1 1 2 2 3 3 4 2 4 4 0 0} do_test where8-3.22 { execsql_status { SELECT a, d FROM ((((((t1))), (((t2)))))) WHERE (a=d OR b=e) AND a<5 ORDER BY a } } {1 1 2 2 3 3 4 2 4 4 0 0} - +do_test where8-3.23 { + execsql_status { + SELECT * FROM ((SELECT * FROM t2)) AS t3; + } +} {1 {} I 2 four IV 3 {} IX 4 sixteen XVI 5 {} XXV 6 thirtysix XXXVI 7 fortynine XLIX 8 sixtyeight LXIV 9 eightyone LXXXIX 10 {} C 9 0} #----------------------------------------------------------------------- # The following tests - where8-4.* - verify that adding or removing # indexes does not change the results returned by various queries. # Index: test/where9.test ================================================================== --- test/where9.test +++ test/where9.test @@ -890,7 +890,28 @@ FROM (t81) LEFT JOIN (main.t82) ON y=b JOIN t83 WHERE c==p OR d==p ORDER BY +a; } } {2 3 4 5 {} {} 5 55 3 4 5 6 2 4 5 55} + +# Fix for ticket [f2369304e47167e3e644e2f1fe9736063391d7b7] +# Incorrect results when OR is used in the ON clause of a LEFT JOIN +# +do_test where9-9.1 { + db eval { + CREATE TABLE t91(x); INSERT INTO t91 VALUES(1); + CREATE TABLE t92(y INTEGER PRIMARY KEY,a,b); + INSERT INTO t92 VALUES(1,2,3); + SELECT 1 FROM t91 LEFT JOIN t92 ON a=2 OR b=3; + SELECT 2 FROM t91 LEFT JOIN t92 ON a=2 AND b=3; + SELECT 3 FROM t91 LEFT JOIN t92 ON (a=2 OR b=3) AND y IS NULL; + SELECT 4 FROM t91 LEFT JOIN t92 ON (a=2 AND b=3) AND y IS NULL; + CREATE TEMP TABLE x9 AS SELECT * FROM t91 LEFT JOIN t92 ON a=2 OR b=3; + SELECT 5 FROM x9 WHERE y IS NULL; + SELECT 6 FROM t91 LEFT JOIN t92 ON a=2 OR b=3 WHERE y IS NULL; + SELECT 7 FROM t91 LEFT JOIN t92 ON a=2 AND b=3 WHERE y IS NULL; + SELECT 8 FROM t91 LEFT JOIN t92 ON a=22 OR b=33 WHERE y IS NULL; + SELECT 9 FROM t91 LEFT JOIN t92 ON a=22 AND b=33 WHERE y IS NULL; + } +} {1 2 3 4 8 9} finish_test Index: test/win32lock.test ================================================================== --- test/win32lock.test +++ test/win32lock.test @@ -25,10 +25,11 @@ test_sqlite3_log xLog proc xLog {error_code msg} { lappend ::log $msg } sqlite3 db test.db +db eval {PRAGMA mmap_size=0} do_test win32lock-1.1 { db eval { PRAGMA cache_size=10; CREATE TABLE t1(x,y); Index: test/zerodamage.test ================================================================== --- test/zerodamage.test +++ test/zerodamage.test @@ -57,11 +57,11 @@ set sz [file size $file] if {$sz>$::max_journal_size} {set ::max_journal_size $sz} } tv filter xDelete tv script xDeleteCallback - register_wholenumber_module db + load_static_extension db wholenumber db eval { PRAGMA page_size=1024; PRAGMA journal_mode=DELETE; PRAGMA cache_size=5; CREATE VIRTUAL TABLE nums USING wholenumber; Index: tool/build-shell.sh ================================================================== --- tool/build-shell.sh +++ tool/build-shell.sh @@ -13,14 +13,10 @@ -DSQLITE_THREADSAFE=0 \ -DSQLITE_ENABLE_VFSTRACE \ -DSQLITE_ENABLE_STAT3 \ -DSQLITE_ENABLE_FTS4 \ -DSQLITE_ENABLE_RTREE \ - -DSQLITE_ENABLE_REGEXP \ - -DSQLITE_ENABLE_SPELLFIX -DSQLITE_CORE=1 \ -DHAVE_READLINE \ -DHAVE_USLEEP=1 \ ../sqlite/src/shell.c \ - ../sqlite/src/test_regexp.c \ - ../sqlite/src/test_spellfix.c \ ../sqlite/src/test_vfstrace.c \ sqlite3.c -ldl -lreadline -lncurses Index: tool/mksqlite3c.tcl ================================================================== --- tool/mksqlite3c.tcl +++ tool/mksqlite3c.tcl @@ -21,11 +21,11 @@ # # The amalgamated SQLite code will be written into sqlite3.c # # Begin by reading the "sqlite3.h" header file. Extract the version number -# from in this file. The versioon number is needed to generate the header +# from in this file. The version number is needed to generate the header # comment of the amalgamation. # if {[lsearch $argv --nostatic]>=0} { set addstatic 0 } else { @@ -309,10 +309,11 @@ fts3_expr.c fts3_hash.c fts3_porter.c fts3_tokenizer.c fts3_tokenizer1.c + fts3_tokenize_vtab.c fts3_write.c fts3_snippet.c fts3_unicode.c fts3_unicode2.c Index: tool/mksqlite3h.tcl ================================================================== --- tool/mksqlite3h.tcl +++ tool/mksqlite3h.tcl @@ -66,13 +66,18 @@ set declpattern {^ *[a-zA-Z][a-zA-Z_0-9 ]+ \**sqlite3_[_a-zA-Z0-9]+\(} # Force the output to use unix line endings, even on Windows. fconfigure stdout -translation lf -# Process the src/sqlite.h.in ext/rtree/sqlite3rtree.h files. +set filelist [subst { + $TOP/src/sqlite.h.in + $TOP/ext/rtree/sqlite3rtree.h +}] + +# Process the source files. # -foreach file [list $TOP/src/sqlite.h.in $TOP/ext/rtree/sqlite3rtree.h] { +foreach file $filelist { set in [open $file] while {![eof $in]} { set line [gets $in] Index: tool/showdb.c ================================================================== --- tool/showdb.c +++ tool/showdb.c @@ -174,11 +174,11 @@ print_decode_line(aData, 48, 4, "Default page cache size"); print_decode_line(aData, 52, 4, "Largest auto-vac root page"); print_decode_line(aData, 56, 4, "Text encoding"); print_decode_line(aData, 60, 4, "User version"); print_decode_line(aData, 64, 4, "Incremental-vacuum mode"); - print_decode_line(aData, 68, 4, "meta[7]"); + print_decode_line(aData, 68, 4, "Application ID"); print_decode_line(aData, 72, 4, "meta[8]"); print_decode_line(aData, 76, 4, "meta[9]"); print_decode_line(aData, 80, 4, "meta[10]"); print_decode_line(aData, 84, 4, "meta[11]"); print_decode_line(aData, 88, 4, "meta[12]");