SQLite: Check-in [8b962c94]

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview

Comment:	Remove an unnecessary parameter from sqlite3Utf8Read() resulting in a slight performance increase.
Downloads:	Tarball \| ZIP archive \| SQL archive
Timelines:	family \| ancestors \| descendants \| both \| trunk
Files:	files \| file ages \| folders
SHA1:	8b962c94a2f829f71efdd0ca770469b02feea3e1
User & Date:	drh 2012-09-17 18:56:32

Context

2012-09-17
19:12		Remove obsolete bits from the bitvector that defines disabled optimizations in the SQLITE_TESTCTRL_OPTIMIZATIONS verb of sqlite3_test_control(). (check-in: 4c21ee2d user: drh tags: trunk)
18:56		Remove an unnecessary parameter from sqlite3Utf8Read() resulting in a slight performance increase. (check-in: 8b962c94 user: drh tags: trunk)
17:16		Make sure the KeyInfo.aSortOrder array is always allocated so that we never have to test for KeyInfo.aSortOrder==0 in performance-critical loops. (check-in: 45793f0b user: drh tags: trunk)

Changes

Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/func.c.

/*
** For LIKE and GLOB matching on EBCDIC machines, assume that every
** character is exactly one byte in size.  Also, all characters are
** able to participate in upper-case-to-lower-case mappings in EBCDIC
** whereas only characters less than 0x80 do in ASCII.
*/
#if defined(SQLITE_EBCDIC)
# define sqlite3Utf8Read(A,C)  (*(A++))
# define GlogUpperToLower(A)   A = sqlite3UpperToLower[A]
#else
# define GlogUpperToLower(A)   if( !((A)&~0x7f) ){ A = sqlite3UpperToLower[A]; }
#endif

static const struct compareInfo globInfo = { '*', '?', '[', 0 };
/* The correct SQL-92 behavior is for the LIKE operator to ignore
................................................................................
  int seen;
  u8 matchOne = pInfo->matchOne;
  u8 matchAll = pInfo->matchAll;
  u8 matchSet = pInfo->matchSet;
  u8 noCase = pInfo->noCase; 
  int prevEscape = 0;     /* True if the previous character was 'escape' */

  while( (c = sqlite3Utf8Read(zPattern,&zPattern))!=0 ){
    if( !prevEscape && c==matchAll ){
      while( (c=sqlite3Utf8Read(zPattern,&zPattern)) == matchAll
               || c == matchOne ){
        if( c==matchOne && sqlite3Utf8Read(zString, &zString)==0 ){
          return 0;
        }
      }
      if( c==0 ){
        return 1;
      }else if( c==esc ){
        c = sqlite3Utf8Read(zPattern, &zPattern);
        if( c==0 ){
          return 0;
        }
      }else if( c==matchSet ){
        assert( esc==0 );         /* This is GLOB, not LIKE */
        assert( matchSet<0x80 );  /* '[' is a single-byte character */
        while( *zString && patternCompare(&zPattern[-1],zString,pInfo,esc)==0 ){
          SQLITE_SKIP_UTF8(zString);
        }
        return *zString!=0;
      }
      while( (c2 = sqlite3Utf8Read(zString,&zString))!=0 ){
        if( noCase ){
          GlogUpperToLower(c2);
          GlogUpperToLower(c);
          while( c2 != 0 && c2 != c ){
            c2 = sqlite3Utf8Read(zString, &zString);
            GlogUpperToLower(c2);
          }
        }else{
          while( c2 != 0 && c2 != c ){
            c2 = sqlite3Utf8Read(zString, &zString);
          }
        }
        if( c2==0 ) return 0;
        if( patternCompare(zPattern,zString,pInfo,esc) ) return 1;
      }
      return 0;
    }else if( !prevEscape && c==matchOne ){
      if( sqlite3Utf8Read(zString, &zString)==0 ){
        return 0;
      }
    }else if( c==matchSet ){
      u32 prior_c = 0;
      assert( esc==0 );    /* This only occurs for GLOB, not LIKE */
      seen = 0;
      invert = 0;
      c = sqlite3Utf8Read(zString, &zString);
      if( c==0 ) return 0;
      c2 = sqlite3Utf8Read(zPattern, &zPattern);
      if( c2=='^' ){
        invert = 1;
        c2 = sqlite3Utf8Read(zPattern, &zPattern);
      }
      if( c2==']' ){
        if( c==']' ) seen = 1;
        c2 = sqlite3Utf8Read(zPattern, &zPattern);
      }
      while( c2 && c2!=']' ){
        if( c2=='-' && zPattern[0]!=']' && zPattern[0]!=0 && prior_c>0 ){
          c2 = sqlite3Utf8Read(zPattern, &zPattern);
          if( c>=prior_c && c<=c2 ) seen = 1;
          prior_c = 0;
        }else{
          if( c==c2 ){
            seen = 1;
          }
          prior_c = c2;
        }
        c2 = sqlite3Utf8Read(zPattern, &zPattern);
      }
      if( c2==0 || (seen ^ invert)==0 ){
        return 0;
      }
    }else if( esc==c && !prevEscape ){
      prevEscape = 1;
    }else{
      c2 = sqlite3Utf8Read(zString, &zString);
      if( noCase ){
        GlogUpperToLower(c);
        GlogUpperToLower(c2);
      }
      if( c!=c2 ){
        return 0;
      }
................................................................................
    const unsigned char *zEsc = sqlite3_value_text(argv[2]);
    if( zEsc==0 ) return;
    if( sqlite3Utf8CharLen((char*)zEsc, -1)!=1 ){
      sqlite3_result_error(context, 
          "ESCAPE expression must be a single character", -1);
      return;
    }
    escape = sqlite3Utf8Read(zEsc, &zEsc);
  }
  if( zA && zB ){
    struct compareInfo *pInfo = sqlite3_user_data(context);
#ifdef SQLITE_TEST
    sqlite3_like_count++;
#endif

/*
** For LIKE and GLOB matching on EBCDIC machines, assume that every
** character is exactly one byte in size.  Also, all characters are
** able to participate in upper-case-to-lower-case mappings in EBCDIC
** whereas only characters less than 0x80 do in ASCII.
*/
#if defined(SQLITE_EBCDIC)
# define sqlite3Utf8Read(A)    (*((*A)++))
# define GlogUpperToLower(A)   A = sqlite3UpperToLower[A]
#else
# define GlogUpperToLower(A)   if( !((A)&~0x7f) ){ A = sqlite3UpperToLower[A]; }
#endif

static const struct compareInfo globInfo = { '*', '?', '[', 0 };
/* The correct SQL-92 behavior is for the LIKE operator to ignore
................................................................................
  int seen;
  u8 matchOne = pInfo->matchOne;
  u8 matchAll = pInfo->matchAll;
  u8 matchSet = pInfo->matchSet;
  u8 noCase = pInfo->noCase; 
  int prevEscape = 0;     /* True if the previous character was 'escape' */

  while( (c = sqlite3Utf8Read(&zPattern))!=0 ){
    if( c==matchAll && !prevEscape ){
      while( (c=sqlite3Utf8Read(&zPattern)) == matchAll
               || c == matchOne ){
        if( c==matchOne && sqlite3Utf8Read(&zString)==0 ){
          return 0;
        }
      }
      if( c==0 ){
        return 1;
      }else if( c==esc ){
        c = sqlite3Utf8Read(&zPattern);
        if( c==0 ){
          return 0;
        }
      }else if( c==matchSet ){
        assert( esc==0 );         /* This is GLOB, not LIKE */
        assert( matchSet<0x80 );  /* '[' is a single-byte character */
        while( *zString && patternCompare(&zPattern[-1],zString,pInfo,esc)==0 ){
          SQLITE_SKIP_UTF8(zString);
        }
        return *zString!=0;
      }
      while( (c2 = sqlite3Utf8Read(&zString))!=0 ){
        if( noCase ){
          GlogUpperToLower(c2);
          GlogUpperToLower(c);
          while( c2 != 0 && c2 != c ){
            c2 = sqlite3Utf8Read(&zString);
            GlogUpperToLower(c2);
          }
        }else{
          while( c2 != 0 && c2 != c ){
            c2 = sqlite3Utf8Read(&zString);
          }
        }
        if( c2==0 ) return 0;
        if( patternCompare(zPattern,zString,pInfo,esc) ) return 1;
      }
      return 0;
    }else if( c==matchOne && !prevEscape ){
      if( sqlite3Utf8Read(&zString)==0 ){
        return 0;
      }
    }else if( c==matchSet ){
      u32 prior_c = 0;
      assert( esc==0 );    /* This only occurs for GLOB, not LIKE */
      seen = 0;
      invert = 0;
      c = sqlite3Utf8Read(&zString);
      if( c==0 ) return 0;
      c2 = sqlite3Utf8Read(&zPattern);
      if( c2=='^' ){
        invert = 1;
        c2 = sqlite3Utf8Read(&zPattern);
      }
      if( c2==']' ){
        if( c==']' ) seen = 1;
        c2 = sqlite3Utf8Read(&zPattern);
      }
      while( c2 && c2!=']' ){
        if( c2=='-' && zPattern[0]!=']' && zPattern[0]!=0 && prior_c>0 ){
          c2 = sqlite3Utf8Read(&zPattern);
          if( c>=prior_c && c<=c2 ) seen = 1;
          prior_c = 0;
        }else{
          if( c==c2 ){
            seen = 1;
          }
          prior_c = c2;
        }
        c2 = sqlite3Utf8Read(&zPattern);
      }
      if( c2==0 || (seen ^ invert)==0 ){
        return 0;
      }
    }else if( esc==c && !prevEscape ){
      prevEscape = 1;
    }else{
      c2 = sqlite3Utf8Read(&zString);
      if( noCase ){
        GlogUpperToLower(c);
        GlogUpperToLower(c2);
      }
      if( c!=c2 ){
        return 0;
      }
................................................................................
    const unsigned char *zEsc = sqlite3_value_text(argv[2]);
    if( zEsc==0 ) return;
    if( sqlite3Utf8CharLen((char*)zEsc, -1)!=1 ){
      sqlite3_result_error(context, 
          "ESCAPE expression must be a single character", -1);
      return;
    }
    escape = sqlite3Utf8Read(&zEsc);
  }
  if( zA && zB ){
    struct compareInfo *pInfo = sqlite3_user_data(context);
#ifdef SQLITE_TEST
    sqlite3_like_count++;
#endif

Changes to src/sqliteInt.h.

int sqlite3FixExprList(DbFixer*, ExprList*);
int sqlite3FixTriggerStep(DbFixer*, TriggerStep*);
int sqlite3AtoF(const char *z, double*, int, u8);
int sqlite3GetInt32(const char *, int*);
int sqlite3Atoi(const char*);
int sqlite3Utf16ByteLen(const void *pData, int nChar);
int sqlite3Utf8CharLen(const char *pData, int nByte);
u32 sqlite3Utf8Read(const u8*, const u8**);

/*
** Routines to read and write variable-length integers.  These used to
** be defined locally, but now we use the varint routines in the util.c
** file.  Code should use the MACRO forms below, as the Varint32 versions
** are coded to assume the single byte case is already handled (which 
** the MACRO form does).

int sqlite3FixExprList(DbFixer*, ExprList*);
int sqlite3FixTriggerStep(DbFixer*, TriggerStep*);
int sqlite3AtoF(const char *z, double*, int, u8);
int sqlite3GetInt32(const char *, int*);
int sqlite3Atoi(const char*);
int sqlite3Utf16ByteLen(const void *pData, int nChar);
int sqlite3Utf8CharLen(const char *pData, int nByte);
u32 sqlite3Utf8Read(const u8**);

/*
** Routines to read and write variable-length integers.  These used to
** be defined locally, but now we use the varint routines in the util.c
** file.  Code should use the MACRO forms below, as the Varint32 versions
** are coded to assume the single byte case is already handled (which 
** the MACRO form does).

Changes to src/utf.c.

      c = (c<<6) + (0x3f & *(zIn++));                      \
    }                                                      \
    if( c<0x80                                             \
        || (c&0xFFFFF800)==0xD800                          \
        || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }        \
  }
u32 sqlite3Utf8Read(
  const unsigned char *zIn,       /* First byte of UTF-8 character */
  const unsigned char **pzNext    /* Write first byte past UTF-8 char here */
){
  unsigned int c;

  /* Same as READ_UTF8() above but without the zTerm parameter.
  ** For this routine, we assume the UTF8 string is always zero-terminated.
  */
  c = *(zIn++);
  if( c>=0xc0 ){
    c = sqlite3Utf8Trans1[c-0xc0];
    while( (*zIn & 0xc0)==0x80 ){
      c = (c<<6) + (0x3f & *(zIn++));
    }
    if( c<0x80
        || (c&0xFFFFF800)==0xD800
        || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }
  }
  *pzNext = zIn;
  return c;
}




/*
................................................................................
  }
  z = zOut;

  if( pMem->enc==SQLITE_UTF8 ){
    if( desiredEnc==SQLITE_UTF16LE ){
      /* UTF-8 -> UTF-16 Little-endian */
      while( zIn<zTerm ){
        /* c = sqlite3Utf8Read(zIn, zTerm, (const u8**)&zIn); */
        READ_UTF8(zIn, zTerm, c);
        WRITE_UTF16LE(z, c);
      }
    }else{
      assert( desiredEnc==SQLITE_UTF16BE );
      /* UTF-8 -> UTF-16 Big-endian */
      while( zIn<zTerm ){
        /* c = sqlite3Utf8Read(zIn, zTerm, (const u8**)&zIn); */
        READ_UTF8(zIn, zTerm, c);
        WRITE_UTF16BE(z, c);
      }
    }
    pMem->n = (int)(z - zOut);
    *z++ = 0;
  }else{
................................................................................
*/
int sqlite3Utf8To8(unsigned char *zIn){
  unsigned char *zOut = zIn;
  unsigned char *zStart = zIn;
  u32 c;

  while( zIn[0] && zOut<=zIn ){
    c = sqlite3Utf8Read(zIn, (const u8**)&zIn);
    if( c!=0xfffd ){
      WRITE_UTF8(zOut, c);
    }
  }
  *zOut = 0;
  return (int)(zOut - zStart);
}
................................................................................
  for(i=0; i<0x00110000; i++){
    z = zBuf;
    WRITE_UTF8(z, i);
    n = (int)(z-zBuf);
    assert( n>0 && n<=4 );
    z[0] = 0;
    z = zBuf;
    c = sqlite3Utf8Read(z, (const u8**)&z);
    t = i;
    if( i>=0xD800 && i<=0xDFFF ) t = 0xFFFD;
    if( (i&0xFFFFFFFE)==0xFFFE ) t = 0xFFFD;
    assert( c==t );
    assert( (z-zBuf)==n );
  }
  for(i=0; i<0x00110000; i++){

      c = (c<<6) + (0x3f & *(zIn++));                      \
    }                                                      \
    if( c<0x80                                             \
        || (c&0xFFFFF800)==0xD800                          \
        || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }        \
  }
u32 sqlite3Utf8Read(
  const unsigned char **pz    /* Pointer to string from which to read char */

){
  unsigned int c;

  /* Same as READ_UTF8() above but without the zTerm parameter.
  ** For this routine, we assume the UTF8 string is always zero-terminated.
  */
  c = *((*pz)++);
  if( c>=0xc0 ){
    c = sqlite3Utf8Trans1[c-0xc0];
    while( (*(*pz) & 0xc0)==0x80 ){
      c = (c<<6) + (0x3f & *((*pz)++));
    }
    if( c<0x80
        || (c&0xFFFFF800)==0xD800
        || (c&0xFFFFFFFE)==0xFFFE ){  c = 0xFFFD; }
  }

  return c;
}




/*
................................................................................
  }
  z = zOut;

  if( pMem->enc==SQLITE_UTF8 ){
    if( desiredEnc==SQLITE_UTF16LE ){
      /* UTF-8 -> UTF-16 Little-endian */
      while( zIn<zTerm ){

        READ_UTF8(zIn, zTerm, c);
        WRITE_UTF16LE(z, c);
      }
    }else{
      assert( desiredEnc==SQLITE_UTF16BE );
      /* UTF-8 -> UTF-16 Big-endian */
      while( zIn<zTerm ){

        READ_UTF8(zIn, zTerm, c);
        WRITE_UTF16BE(z, c);
      }
    }
    pMem->n = (int)(z - zOut);
    *z++ = 0;
  }else{
................................................................................
*/
int sqlite3Utf8To8(unsigned char *zIn){
  unsigned char *zOut = zIn;
  unsigned char *zStart = zIn;
  u32 c;

  while( zIn[0] && zOut<=zIn ){
    c = sqlite3Utf8Read((const u8**)&zIn);
    if( c!=0xfffd ){
      WRITE_UTF8(zOut, c);
    }
  }
  *zOut = 0;
  return (int)(zOut - zStart);
}
................................................................................
  for(i=0; i<0x00110000; i++){
    z = zBuf;
    WRITE_UTF8(z, i);
    n = (int)(z-zBuf);
    assert( n>0 && n<=4 );
    z[0] = 0;
    z = zBuf;
    c = sqlite3Utf8Read((const u8**)&z);
    t = i;
    if( i>=0xD800 && i<=0xDFFF ) t = 0xFFFD;
    if( (i&0xFFFFFFFE)==0xFFFE ) t = 0xFFFD;
    assert( c==t );
    assert( (z-zBuf)==n );
  }
  for(i=0; i<0x00110000; i++){