/*
** 2015-06-06
**
** 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 contains C code that generates VDBE code used to process
** the WHERE clause of SQL statements.
**
** This file was split off from where.c on 2015-06-06 in order to reduce the
** size of where.c and make it easier to edit. This file contains the routines
** that actually generate the bulk of the WHERE loop code. The original where.c
** file retains the code that does query planning and analysis.
*/
#include "sqliteInt.h"
#include "whereInt.h"
#ifndef SQLITE_OMIT_EXPLAIN
/*
** Return the name of the i-th column of the pIdx index.
*/
static const char *explainIndexColumnName(Index *pIdx, int i){
i = pIdx->aiColumn[i];
if( i==XN_EXPR ) return "<expr>";
if( i==XN_ROWID ) return "rowid";
return pIdx->pTable->aCol[i].zCnName;
}
/*
** This routine is a helper for explainIndexRange() below
**
** pStr holds the text of an expression that we are building up one term
** at a time. This routine adds a new term to the end of the expression.
** Terms are separated by AND so add the "AND" text for second and subsequent
** terms only.
*/
static void explainAppendTerm(
StrAccum *pStr, /* The text expression being built */
Index *pIdx, /* Index to read column names from */
int nTerm, /* Number of terms */
int iTerm, /* Zero-based index of first term. */
int bAnd, /* Non-zero to append " AND " */
const char *zOp /* Name of the operator */
){
int i;
assert( nTerm>=1 );
if( bAnd ) sqlite3_str_append(pStr, " AND ", 5);
if( nTerm>1 ) sqlite3_str_append(pStr, "(", 1);
for(i=0; i<nTerm; i++){
if( i ) sqlite3_str_append(pStr, ",", 1);
sqlite3_str_appendall(pStr, explainIndexColumnName(pIdx, iTerm+i));
}
if( nTerm>1 ) sqlite3_str_append(pStr, ")", 1);
sqlite3_str_append(pStr, zOp, 1);
if( nTerm>1 ) sqlite3_str_append(pStr, "(", 1);
for(i=0; i<nTerm; i++){
if( i ) sqlite3_str_append(pStr, ",", 1);
sqlite3_str_append(pStr, "?", 1);
}
if( nTerm>1 ) sqlite3_str_append(pStr, ")", 1);
}
/*
** Argument pLevel describes a strategy for scanning table pTab. This
** function appends text to pStr that describes the subset of table
** rows scanned by the strategy in the form of an SQL expression.
**
** For example, if the query:
**
** SELECT * FROM t1 WHERE a=1 AND b>2;
**
** is run and there is an index on (a, b), then this function returns a
** string similar to:
**
** "a=? AND b>?"
*/
static void explainIndexRange(StrAccum *pStr, WhereLoop *pLoop){
Index *pIndex = pLoop->u.btree.pIndex;
u16 nEq = pLoop->u.btree.nEq;
u16 nSkip = pLoop->nSkip;
int i, j;
if( nEq==0 && (pLoop->wsFlags&(WHERE_BTM_LIMIT|WHERE_TOP_LIMIT))==0 ) return;
sqlite3_str_append(pStr, " (", 2);
for(i=0; i<nEq; i++){
const char *z = explainIndexColumnName(pIndex, i);
if( i ) sqlite3_str_append(pStr, " AND ", 5);
sqlite3_str_appendf(pStr, i>=nSkip ? "%s=?" : "ANY(%s)", z);
}
j = i;
if( pLoop->wsFlags&WHERE_BTM_LIMIT ){
explainAppendTerm(pStr, pIndex, pLoop->u.btree.nBtm, j, i, ">");
i = 1;
}
if( pLoop->wsFlags&WHERE_TOP_LIMIT ){
explainAppendTerm(pStr, pIndex, pLoop->u.btree.nTop, j, i, "<");
}
sqlite3_str_append(pStr, ")", 1);
}
/*
** This function is a no-op unless currently processing an EXPLAIN QUERY PLAN
** command, or if stmt_scanstatus_v2() stats are enabled, or if SQLITE_DEBUG
** was defined at compile-time. If it is not a no-op, a single OP_Explain
** opcode is added to the output to describe the table scan strategy in pLevel.
**
** If an OP_Explain opcode is added to the VM, its address is returned.
** Otherwise, if no OP_Explain is coded, zero is returned.
*/
int sqlite3WhereExplainOneScan(
Parse *pParse, /* Parse context */
SrcList *pTabList, /* Table list this loop refers to */
WhereLevel *pLevel, /* Scan to write OP_Explain opcode for */
u16 wctrlFlags /* Flags passed to sqlite3WhereBegin() */
){
int ret = 0;
#if !defined(SQLITE_DEBUG)
if( sqlite3ParseToplevel(pParse)->explain==2 || IS_STMT_SCANSTATUS(pParse->db) )
#endif
{
SrcItem *pItem = &pTabList->a[pLevel->iFrom];
Vdbe *v = pParse->pVdbe; /* VM being constructed */
sqlite3 *db = pParse->db; /* Database handle */
int isSearch; /* True for a SEARCH. False for SCAN. */
WhereLoop *pLoop; /* The controlling WhereLoop object */
u32 flags; /* Flags that describe this loop */
char *zMsg; /* Text to add to EQP output */
StrAccum str; /* EQP output string */
char zBuf[100]; /* Initial space for EQP output string */
pLoop = pLevel->pWLoop;
flags = pLoop->wsFlags;
if( (flags&WHERE_MULTI_OR) || (wctrlFlags&WHERE_OR_SUBCLAUSE) ) return 0;
isSearch = (flags&(WHERE_BTM_LIMIT|WHERE_TOP_LIMIT))!=0
|| ((flags&WHERE_VIRTUALTABLE)==0 && (pLoop->u.btree.nEq>0))
|| (wctrlFlags&(WHERE_ORDERBY_MIN|WHERE_ORDERBY_MAX));
sqlite3StrAccumInit(&str, db, zBuf, sizeof(zBuf), SQLITE_MAX_LENGTH);
str.printfFlags = SQLITE_PRINTF_INTERNAL;
sqlite3_str_appendf(&str, "%s %S", isSearch ? "SEARCH" : "SCAN", pItem);
if( (flags & (WHERE_IPK|WHERE_VIRTUALTABLE))==0 ){
const char *zFmt = 0;
Index *pIdx;
assert( pLoop->u.btree.pIndex!=0 );
pIdx = pLoop->u.btree.pIndex;
assert( !(flags&WHERE_AUTO_INDEX) || (flags&WHERE_IDX_ONLY) );
if( !HasRowid(pItem->pTab) && IsPrimaryKeyIndex(pIdx) ){
if( isSearch ){
zFmt = "PRIMARY KEY";
}
}else if( flags & WHERE_PARTIALIDX ){
zFmt = "AUTOMATIC PARTIAL COVERING INDEX";
}else if( flags & WHERE_AUTO_INDEX ){
zFmt = "AUTOMATIC COVERING INDEX";
}else if( flags & WHERE_IDX_ONLY ){
zFmt = "COVERING INDEX %s";
}else{
zFmt = "INDEX %s";
}
if( zFmt ){
sqlite3_str_append(&str, " USING ", 7);
sqlite3_str_appendf(&str, zFmt, pIdx->zName);
explainIndexRange(&str, pLoop);
}
}else if( (flags & WHERE_IPK)!=0 && (flags & WHERE_CONSTRAINT)!=0 ){
char cRangeOp;
#if 0 /* Better output, but breaks many tests */
const Table *pTab = pItem->pTab;
const char *zRowid = pTab->iPKey>=0 ? pTab->aCol[pTab->iPKey].zCnName:
"rowid";
#else
const char *zRowid = "rowid";
#endif
sqlite3_str_appendf(&str, " USING INTEGER PRIMARY KEY (%s", zRowid);
if( flags&(WHERE_COLUMN_EQ|WHERE_COLUMN_IN) ){
cRangeOp = '=';
}else if( (flags&WHERE_BOTH_LIMIT)==WHERE_BOTH_LIMIT ){
sqlite3_str_appendf(&str, ">? AND %s", zRowid);
cRangeOp = '<';
}else if( flags&WHERE_BTM_LIMIT ){
cRangeOp = '>';
}else{
assert( flags&WHERE_TOP_LIMIT);
cRangeOp = '<';
}
sqlite3_str_appendf(&str, "%c?)", cRangeOp);
}
#ifndef SQLITE_OMIT_VIRTUALTABLE
else if( (flags & WHERE_VIRTUALTABLE)!=0 ){
sqlite3_str_appendf(&str, " VIRTUAL TABLE INDEX %d:%s",
pLoop->u.vtab.idxNum, pLoop->u.vtab.idxStr);
}
#endif
if( pItem->fg.jointype & JT_LEFT ){
sqlite3_str_appendf(&str, " LEFT-JOIN");
}
#ifdef SQLITE_EXPLAIN_ESTIMATED_ROWS
if( pLoop->nOut>=10 ){
sqlite3_str_appendf(&str, " (~%llu rows)",
sqlite3LogEstToInt(pLoop->nOut));
}else{
sqlite3_str_append(&str, " (~1 row)", 9);
}
#endif
zMsg = sqlite3StrAccumFinish(&str);
sqlite3ExplainBreakpoint("",zMsg);
ret = sqlite3VdbeAddOp4(v, OP_Explain, sqlite3VdbeCurrentAddr(v),
pParse->addrExplain, 0, zMsg,P4_DYNAMIC);
}
return ret;
}
/*
** Add a single OP_Explain opcode that describes a Bloom filter.
**
** Or if not processing EXPLAIN QUERY PLAN and not in a SQLITE_DEBUG and/or
** SQLITE_ENABLE_STMT_SCANSTATUS build, then OP_Explain opcodes are not
** required and this routine is a no-op.
**
** If an OP_Explain opcode is added to the VM, its address is returned.
** Otherwise, if no OP_Explain is coded, zero is returned.
*/
int sqlite3WhereExplainBloomFilter(
const Parse *pParse, /* Parse context */
const WhereInfo *pWInfo, /* WHERE clause */
const WhereLevel *pLevel /* Bloom filter on this level */
){
int ret = 0;
SrcItem *pItem = &pWInfo->pTabList->a[pLevel->iFrom];
Vdbe *v = pParse->pVdbe; /* VM being constructed */
sqlite3 *db = pParse->db; /* Database handle */
char *zMsg; /* Text to add to EQP output */
int i; /* Loop counter */
WhereLoop *pLoop; /* The where loop */
StrAccum str; /* EQP output string */
char zBuf[100]; /* Initial space for EQP output string */
sqlite3StrAccumInit(&str, db, zBuf, sizeof(zBuf), SQLITE_MAX_LENGTH);
str.printfFlags = SQLITE_PRINTF_INTERNAL;
sqlite3_str_appendf(&str, "BLOOM FILTER ON %S (", pItem);
pLoop = pLevel->pWLoop;
if( pLoop->wsFlags & WHERE_IPK ){
const Table *pTab = pItem->pTab;
if( pTab->iPKey>=0 ){
sqlite3_str_appendf(&str, "%s=?", pTab->aCol[pTab->iPKey].zCnName);
}else{
sqlite3_str_appendf(&str, "rowid=?");
}
}else{
for(i=pLoop->nSkip; i<pLoop->u.btree.nEq; i++){
const char *z = explainIndexColumnName(pLoop->u.btree.pIndex, i);
if( i>pLoop->nSkip ) sqlite3_str_append(&str, " AND ", 5);
sqlite3_str_appendf(&str, "%s=?", z);
}
}
sqlite3_str_append(&str, ")", 1);
zMsg = sqlite3StrAccumFinish(&str);
ret = sqlite3VdbeAddOp4(v, OP_Explain, sqlite3VdbeCurrentAddr(v),
pParse->addrExplain, 0, zMsg,P4_DYNAMIC);
sqlite3VdbeScanStatus(v, sqlite3VdbeCurrentAddr(v)-1, 0, 0, 0, 0);
return ret;
}
#endif /* SQLITE_OMIT_EXPLAIN */
#ifdef SQLITE_ENABLE_STMT_SCANSTATUS
/*
** Configure the VM passed as the first argument with an
** sqlite3_stmt_scanstatus() entry corresponding to the scan used to
** implement level pLvl. Argument pSrclist is a pointer to the FROM
** clause that the scan reads data from.
**
** If argument addrExplain is not 0, it must be the address of an
** OP_Explain instruction that describes the same loop.
*/
void sqlite3WhereAddScanStatus(
Vdbe *v, /* Vdbe to add scanstatus entry to */
SrcList *pSrclist, /* FROM clause pLvl reads data from */
WhereLevel *pLvl, /* Level to add scanstatus() entry for */
int addrExplain /* Address of OP_Explain (or 0) */
){
if( IS_STMT_SCANSTATUS( sqlite3VdbeDb(v) ) ){
const char *zObj = 0;
WhereLoop *pLoop = pLvl->pWLoop;
int wsFlags = pLoop->wsFlags;
int viaCoroutine = 0;
if( (wsFlags & WHERE_VIRTUALTABLE)==0 && pLoop->u.btree.pIndex!=0 ){
zObj = pLoop->u.btree.pIndex->zName;
}else{
zObj = pSrclist->a[pLvl->iFrom].zName;
viaCoroutine = pSrclist->a[pLvl->iFrom].fg.viaCoroutine;
}
sqlite3VdbeScanStatus(
v, addrExplain, pLvl->addrBody, pLvl->addrVisit, pLoop->nOut, zObj
);
if( viaCoroutine==0 ){
if( (wsFlags & (WHERE_MULTI_OR|WHERE_AUTO_INDEX))==0 ){
sqlite3VdbeScanStatusRange(v, addrExplain, -1, pLvl->iTabCur);
}
if( wsFlags & WHERE_INDEXED ){
sqlite3VdbeScanStatusRange(v, addrExplain, -1, pLvl->iIdxCur);
}
}
}
}
#endif
/*
** Disable a term in the WHERE clause. Except, do not disable the term
** if it controls a LEFT OUTER JOIN and it did not originate in the ON
** or USING clause of that join.
**
** Consider the term t2.z='ok' in the following queries:
**
** (1) SELECT * FROM t1 LEFT JOIN t2 ON t1.a=t2.x WHERE t2.z='ok'
** (2) SELECT * FROM t1 LEFT JOIN t2 ON t1.a=t2.x AND t2.z='ok'
** (3) SELECT * FROM t1, t2 WHERE t1.a=t2.x AND t2.z='ok'
**
** The t2.z='ok' is disabled in the in (2) because it originates
** in the ON clause. The term is disabled in (3) because it is not part
** of a LEFT OUTER JOIN. In (1), the term is not disabled.
**
** Disabling a term causes that term to not be tested in the inner loop
** of the join. Disabling is an optimization. When terms are satisfied
** by indices, we disable them to prevent redundant tests in the inner
** loop. We would get the correct results if nothing were ever disabled,
** but joins might run a little slower. The trick is to disable as much
** as we can without disabling too much. If we disabled in (1), we'd get
** the wrong answer. See ticket #813.
**
** If all the children of a term are disabled, then that term is also
** automatically disabled. In this way, terms get disabled if derived
** virtual terms are tested first. For example:
**
** x GLOB 'abc*' AND x>='abc' AND x<'acd'
** \___________/ \______/ \_____/
** parent child1 child2
**
** Only the parent term was in the original WHERE clause. The child1
** and child2 terms were added by the LIKE optimization. If both of
** the virtual child terms are valid, then testing of the parent can be
** skipped.
**
** Usually the parent term is marked as TERM_CODED. But if the parent
** term was originally TERM_LIKE, then the parent gets TERM_LIKECOND instead.
** The TERM_LIKECOND marking indicates that the term should be coded inside
** a conditional such that is only evaluated on the second pass of a
** LIKE-optimization loop, when scanning BLOBs instead of strings.
*/
static void disableTerm(WhereLevel *pLevel, WhereTerm *pTerm){
int nLoop = 0;
assert( pTerm!=0 );
while( (pTerm->wtFlags & TERM_CODED)==0
&& (pLevel->iLeftJoin==0 || ExprHasProperty(pTerm->pExpr, EP_OuterON))
&& (pLevel->notReady & pTerm->prereqAll)==0
){
if( nLoop && (pTerm->wtFlags & TERM_LIKE)!=0 ){
pTerm->wtFlags |= TERM_LIKECOND;
}else{
pTerm->wtFlags |= TERM_CODED;
}
#ifdef WHERETRACE_ENABLED
if( (sqlite3WhereTrace & 0x4001)==0x4001 ){
sqlite3DebugPrintf("DISABLE-");
sqlite3WhereTermPrint(pTerm, (int)(pTerm - (pTerm->pWC->a)));
}
#endif
if( pTerm->iParent<0 ) break;
pTerm = &pTerm->pWC->a[pTerm->iParent];
assert( pTerm!=0 );
pTerm->nChild--;
if( pTerm->nChild!=0 ) break;
nLoop++;
}
}
/*
** Code an OP_Affinity opcode to apply the column affinity string zAff
** to the n registers starting at base.
**
** As an optimization, SQLITE_AFF_BLOB and SQLITE_AFF_NONE entries (which
** are no-ops) at the beginning and end of zAff are ignored. If all entries
** in zAff are SQLITE_AFF_BLOB or SQLITE_AFF_NONE, then no code gets generated.
**
** This routine makes its own copy of zAff so that the caller is free
** to modify zAff after this routine returns.
*/
static void codeApplyAffinity(Parse *pParse, int base, int n, char *zAff){
Vdbe *v = pParse->pVdbe;
if( zAff==0 ){
assert( pParse->db->mallocFailed );
return;
}
assert( v!=0 );
/* Adjust base and n to skip over SQLITE_AFF_BLOB and SQLITE_AFF_NONE
** entries at the beginning and end of the affinity string.
*/
assert( SQLITE_AFF_NONE<SQLITE_AFF_BLOB );
while( n>0 && zAff[0]<=SQLITE_AFF_BLOB ){
n--;
base++;
zAff++;
}
while( n>1 && zAff[n-1]<=SQLITE_AFF_BLOB ){
n--;
}
/* Code the OP_Affinity opcode if there is anything left to do. */
if( n>0 ){
sqlite3VdbeAddOp4(v, OP_Affinity, base, n, 0, zAff, n);
}
}
/*
** Expression pRight, which is the RHS of a comparison operation, is
** either a vector of n elements or, if n==1, a scalar expression.
** Before the comparison operation, affinity zAff is to be applied
** to the pRight values. This function modifies characters within the
** affinity string to SQLITE_AFF_BLOB if either:
**
** * the comparison will be performed with no affinity, or
** * the affinity change in zAff is guaranteed not to change the value.
*/
static void updateRangeAffinityStr(
Expr *pRight, /* RHS of comparison */
int n, /* Number of vector elements in comparison */
char *zAff /* Affinity string to modify */
){
int i;
for(i=0; i<n; i++){
Expr *p = sqlite3VectorFieldSubexpr(pRight, i);
if( sqlite3CompareAffinity(p, zAff[i])==SQLITE_AFF_BLOB
|| sqlite3ExprNeedsNoAffinityChange(p, zAff[i])
){
zAff[i] = SQLITE_AFF_BLOB;
}
}
}
/*
** pX is an expression of the form: (vector) IN (SELECT ...)
** In other words, it is a vector IN operator with a SELECT clause on the
** LHS. But not all terms in the vector are indexable and the terms might
** not be in the correct order for indexing.
**
** This routine makes a copy of the input pX expression and then adjusts
** the vector on the LHS with corresponding changes to the SELECT so that
** the vector contains only index terms and those terms are in the correct
** order. The modified IN expression is returned. The caller is responsible
** for deleting the returned expression.
**
** Example:
**
** CREATE TABLE t1(a,b,c,d,e,f);
** CREATE INDEX t1x1 ON t1(e,c);
** SELECT * FROM t1 WHERE (a,b,c,d,e) IN (SELECT v,w,x,y,z FROM t2)
** \_______________________________________/
** The pX expression
**
** Since only columns e and c can be used with the index, in that order,
** the modified IN expression that is returned will be:
**
** (e,c) IN (SELECT z,x FROM t2)
**
** The reduced pX is different from the original (obviously) and thus is
** only used for indexing, to improve performance. The original unaltered
** IN expression must also be run on each output row for correctness.
*/
static Expr *removeUnindexableInClauseTerms(
Parse *pParse, /* The parsing context */
int iEq, /* Look at loop terms starting here */
WhereLoop *pLoop, /* The current loop */
Expr *pX /* The IN expression to be reduced */
){
sqlite3 *db = pParse->db;
Select *pSelect; /* Pointer to the SELECT on the RHS */
Expr *pNew;
pNew = sqlite3ExprDup(db, pX, 0);
if( db->mallocFailed==0 ){
for(pSelect=pNew->x.pSelect; pSelect; pSelect=pSelect->pPrior){
ExprList *pOrigRhs; /* Original unmodified RHS */
ExprList *pOrigLhs = 0; /* Original unmodified LHS */
ExprList *pRhs = 0; /* New RHS after modifications */
ExprList *pLhs = 0; /* New LHS after mods */
int i; /* Loop counter */
assert( ExprUseXSelect(pNew) );
pOrigRhs = pSelect->pEList;
assert( pNew->pLeft!=0 );
assert( ExprUseXList(pNew->pLeft) );
if( pSelect==pNew->x.pSelect ){
pOrigLhs = pNew->pLeft->x.pList;
}
for(i=iEq; i<pLoop->nLTerm; i++){
if( pLoop->aLTerm[i]->pExpr==pX ){
int iField;
assert( (pLoop->aLTerm[i]->eOperator & (WO_OR|WO_AND))==0 );
iField = pLoop->aLTerm[i]->u.x.iField - 1;
if( pOrigRhs->a[iField].pExpr==0 ) continue; /* Duplicate PK column */
pRhs = sqlite3ExprListAppend(pParse, pRhs, pOrigRhs->a[iField].pExpr);
pOrigRhs->a[iField].pExpr = 0;
if( pOrigLhs ){
assert( pOrigLhs->a[iField].pExpr!=0 );
pLhs = sqlite3ExprListAppend(pParse,pLhs,pOrigLhs->a[iField].pExpr);
pOrigLhs->a[iField].pExpr = 0;
}
}
}
sqlite3ExprListDelete(db, pOrigRhs);
if( pOrigLhs ){
sqlite3ExprListDelete(db, pOrigLhs);
pNew->pLeft->x.pList = pLhs;
}
pSelect->pEList = pRhs;
if( pLhs && pLhs->nExpr==1 ){
/* Take care here not to generate a TK_VECTOR containing only a
** single value. Since the parser never creates such a vector, some
** of the subroutines do not handle this case. */
Expr *p = pLhs->a[0].pExpr;
pLhs->a[0].pExpr = 0;
sqlite3ExprDelete(db, pNew->pLeft);
pNew->pLeft = p;
}
if( pSelect->pOrderBy ){
/* If the SELECT statement has an ORDER BY clause, zero the
** iOrderByCol variables. These are set to non-zero when an
** ORDER BY term exactly matches one of the terms of the
** result-set. Since the result-set of the SELECT statement may
** have been modified or reordered, these variables are no longer
** set correctly. Since setting them is just an optimization,
** it's easiest just to zero them here. */
ExprList *pOrderBy = pSelect->pOrderBy;
for(i=0; i<pOrderBy->nExpr; i++){
pOrderBy->a[i].u.x.iOrderByCol = 0;
}
}
#if 0
printf("For indexing, change the IN expr:\n");
sqlite3TreeViewExpr(0, pX, 0);
printf("Into:\n");
sqlite3TreeViewExpr(0, pNew, 0);
#endif
}
}
return pNew;
}
/*
** Generate code for a single equality term of the WHERE clause. An equality
** term can be either X=expr or X IN (...). pTerm is the term to be
** coded.
**
** The current value for the constraint is left in a register, the index
** of which is returned. An attempt is made store the result in iTarget but
** this is only guaranteed for TK_ISNULL and TK_IN constraints. If the
** constraint is a TK_EQ or TK_IS, then the current value might be left in
** some other register and it is the caller's responsibility to compensate.
**
** For a constraint of the form X=expr, the expression is evaluated in
** straight-line code. For constraints of the form X IN (...)
** this routine sets up a loop that will iterate over all values of X.
*/
static int codeEqualityTerm(
Parse *pParse, /* The parsing context */
WhereTerm *pTerm, /* The term of the WHERE clause to be coded */
WhereLevel *pLevel, /* The level of the FROM clause we are working on */
int iEq, /* Index of the equality term within this level */
int bRev, /* True for reverse-order IN operations */
int iTarget /* Attempt to leave results in this register */
){
Expr *pX = pTerm->pExpr;
Vdbe *v = pParse->pVdbe;
int iReg; /* Register holding results */
assert( pLevel->pWLoop->aLTerm[iEq]==pTerm );
assert( iTarget>0 );
if( pX->op==TK_EQ || pX->op==TK_IS ){
iReg = sqlite3ExprCodeTarget(pParse, pX->pRight, iTarget);
}else if( pX->op==TK_ISNULL ){
iReg = iTarget;
sqlite3VdbeAddOp2(v, OP_Null, 0, iReg);
#ifndef SQLITE_OMIT_SUBQUERY
}else{
int eType = IN_INDEX_NOOP;
int iTab;
struct InLoop *pIn;
WhereLoop *pLoop = pLevel->pWLoop;
int i;
int nEq = 0;
int *aiMap = 0;
if( (pLoop->wsFlags & WHERE_VIRTUALTABLE)==0
&& pLoop->u.btree.pIndex!=0
&& pLoop->u.btree.pIndex->aSortOrder[iEq]
){
testcase( iEq==0 );
testcase( bRev );
bRev = !bRev;
}
assert( pX->op==TK_IN );
iReg = iTarget;
for(i=0; i<iEq; i++){
if( pLoop->aLTerm[i] && pLoop->aLTerm[i]->pExpr==pX ){
disableTerm(pLevel, pTerm);
return iTarget;
}
}
for(i=iEq;i<pLoop->nLTerm; i++){
assert( pLoop->aLTerm[i]!=0 );
if( pLoop->aLTerm[i]->pExpr==pX ) nEq++;
}
iTab = 0;
if( !ExprUseXSelect(pX) || pX->x.pSelect->pEList->nExpr==1 ){
eType = sqlite3FindInIndex(pParse, pX, IN_INDEX_LOOP, 0, 0, &iTab);
}else{
Expr *pExpr = pTerm->pExpr;
if( pExpr->iTable==0 || !ExprHasProperty(pExpr, EP_Subrtn) ){
sqlite3 *db = pParse->db;
pX = removeUnindexableInClauseTerms(pParse, iEq, pLoop, pX);
if( !db->mallocFailed ){
aiMap = (int*)sqlite3DbMallocZero(pParse->db, sizeof(int)*nEq);
eType = sqlite3FindInIndex(pParse, pX, IN_INDEX_LOOP, 0, aiMap,&iTab);
pExpr->iTable = iTab;
}
sqlite3ExprDelete(db, pX);
}else{
int n = sqlite3ExprVectorSize(pX->pLeft);
aiMap = (int*)sqlite3DbMallocZero(pParse->db, sizeof(int)*MAX(nEq,n));
eType = sqlite3FindInIndex(pParse, pX, IN_INDEX_LOOP, 0, aiMap, &iTab);
}
pX = pExpr;
}
if( eType==IN_INDEX_INDEX_DESC ){
testcase( bRev );
bRev = !bRev;
}
sqlite3VdbeAddOp2(v, bRev ? OP_Last : OP_Rewind, iTab, 0);
VdbeCoverageIf(v, bRev);
VdbeCoverageIf(v, !bRev);
assert( (pLoop->wsFlags & WHERE_MULTI_OR)==0 );
pLoop->wsFlags |= WHERE_IN_ABLE;
if( pLevel->u.in.nIn==0 ){
pLevel->addrNxt = sqlite3VdbeMakeLabel(pParse);
}
if( iEq>0 && (pLoop->wsFlags & WHERE_IN_SEEKSCAN)==0 ){
pLoop->wsFlags |= WHERE_IN_EARLYOUT;
}
i = pLevel->u.in.nIn;
pLevel->u.in.nIn += nEq;
pLevel->u.in.aInLoop =
sqlite3WhereRealloc(pTerm->pWC->pWInfo,
pLevel->u.in.aInLoop,
sizeof(pLevel->u.in.aInLoop[0])*pLevel->u.in.nIn);
pIn = pLevel->u.in.aInLoop;
if( pIn ){
int iMap = 0; /* Index in aiMap[] */
pIn += i;
for(i=iEq;i<pLoop->nLTerm; i++){
if( pLoop->aLTerm[i]->pExpr==pX ){
int iOut = iReg + i - iEq;
if( eType==IN_INDEX_ROWID ){
pIn->addrInTop = sqlite3VdbeAddOp2(v, OP_Rowid, iTab, iOut);
}else{
int iCol = aiMap ? aiMap[iMap++] : 0;
pIn->addrInTop = sqlite3VdbeAddOp3(v,OP_Column,iTab, iCol, iOut);
}
sqlite3VdbeAddOp1(v, OP_IsNull, iOut); VdbeCoverage(v);
if( i==iEq ){
pIn->iCur = iTab;
pIn->eEndLoopOp = bRev ? OP_Prev : OP_Next;
if( iEq>0 ){
pIn->iBase = iReg - i;
pIn->nPrefix = i;
}else{
pIn->nPrefix = 0;
}
}else{
pIn->eEndLoopOp = OP_Noop;
}
pIn++;
}
}
testcase( iEq>0
&& (pLoop->wsFlags & WHERE_IN_SEEKSCAN)==0
&& (pLoop->wsFlags & WHERE_VIRTUALTABLE)!=0 );
if( iEq>0
&& (pLoop->wsFlags & (WHERE_IN_SEEKSCAN|WHERE_VIRTUALTABLE))==0
){
sqlite3VdbeAddOp3(v, OP_SeekHit, pLevel->iIdxCur, 0, iEq);
}
}else{
pLevel->u.in.nIn = 0;
}
sqlite3DbFree(pParse->db, aiMap);
#endif
}
/* As an optimization, try to disable the WHERE clause term that is
** driving the index as it will always be true. The correct answer is
** obtained regardless, but we might get the answer with fewer CPU cycles
** by omitting the term.
**
** But do not disable the term unless we are certain that the term is
** not a transitive constraint. For an example of where that does not
** work, see https://sqlite.org/forum/forumpost/eb8613976a (2021-05-04)
*/
if( (pLevel->pWLoop->wsFlags & WHERE_TRANSCONS)==0
|| (pTerm->eOperator & WO_EQUIV)==0
){
disableTerm(pLevel, pTerm);
}
return iReg;
}
/*
** Generate code that will evaluate all == and IN constraints for an
** index scan.
**
** For example, consider table t1(a,b,c,d,e,f) with index i1(a,b,c).
** Suppose the WHERE clause is this: a==5 AND b IN (1,2,3) AND c>5 AND c<10
** The index has as many as three equality constraints, but in this
** example, the third "c" value is an inequality. So only two
** constraints are coded. This routine will generate code to evaluate
** a==5 and b IN (1,2,3). The current values for a and b will be stored
** in consecutive registers and the index of the first register is returned.
**
** In the example above nEq==2. But this subroutine works for any value
** of nEq including 0. If nEq==0, this routine is nearly a no-op.
** The only thing it does is allocate the pLevel->iMem memory cell and
** compute the affinity string.
**
** The nExtraReg parameter is 0 or 1. It is 0 if all WHERE clause constraints
** are == or IN and are covered by the nEq. nExtraReg is 1 if there is
** an inequality constraint (such as the "c>=5 AND c<10" in the example) that
** occurs after the nEq quality constraints.
**
** This routine allocates a range of nEq+nExtraReg memory cells and returns
** the index of the first memory cell in that range. The code that
** calls this routine will use that memory range to store keys for
** start and termination conditions of the loop.
** key value of the loop. If one or more IN operators appear, then
** this routine allocates an additional nEq memory cells for internal
** use.
**
** Before returning, *pzAff is set to point to a buffer containing a
** copy of the column affinity string of the index allocated using
** sqlite3DbMalloc(). Except, entries in the copy of the string associated
** with equality constraints that use BLOB or NONE affinity are set to
** SQLITE_AFF_BLOB. This is to deal with SQL such as the following:
**
** CREATE TABLE t1(a TEXT PRIMARY KEY, b);
** SELECT ... FROM t1 AS t2, t1 WHERE t1.a = t2.b;
**
** In the example above, the index on t1(a) has TEXT affinity. But since
** the right hand side of the equality constraint (t2.b) has BLOB/NONE affinity,
** no conversion should be attempted before using a t2.b value as part of
** a key to search the index. Hence the first byte in the returned affinity
** string in this example would be set to SQLITE_AFF_BLOB.
*/
static int codeAllEqualityTerms(
Parse *pParse, /* Parsing context */
WhereLevel *pLevel, /* Which nested loop of the FROM we are coding */
int bRev, /* Reverse the order of IN operators */
int nExtraReg, /* Number of extra registers to allocate */
char **pzAff /* OUT: Set to point to affinity string */
){
u16 nEq; /* The number of == or IN constraints to code */
u16 nSkip; /* Number of left-most columns to skip */
Vdbe *v = pParse->pVdbe; /* The vm under construction */
Index *pIdx; /* The index being used for this loop */
WhereTerm *pTerm; /* A single constraint term */
WhereLoop *pLoop; /* The WhereLoop object */
int j; /* Loop counter */
int regBase; /* Base register */
int nReg; /* Number of registers to allocate */
char *zAff; /* Affinity string to return */
/* This module is only called on query plans that use an index. */
pLoop = pLevel->pWLoop;
assert( (pLoop->wsFlags & WHERE_VIRTUALTABLE)==0 );
nEq = pLoop->u.btree.nEq;
nSkip = pLoop->nSkip;
pIdx = pLoop->u.btree.pIndex;
assert( pIdx!=0 );
/* Figure out how many memory cells we will need then allocate them.
*/
regBase = pParse->nMem + 1;
nReg = pLoop->u.btree.nEq + nExtraReg;
pParse->nMem += nReg;
zAff = sqlite3DbStrDup(pParse->db,sqlite3IndexAffinityStr(pParse->db,pIdx));
assert( zAff!=0 || pParse->db->mallocFailed );
if( nSkip ){
int iIdxCur = pLevel->iIdxCur;
sqlite3VdbeAddOp3(v, OP_Null, 0, regBase, regBase+nSkip-1);
sqlite3VdbeAddOp1(v, (bRev?OP_Last:OP_Rewind), iIdxCur);
VdbeCoverageIf(v, bRev==0);
VdbeCoverageIf(v, bRev!=0);
VdbeComment((v, "begin skip-scan on %s", pIdx->zName));
j = sqlite3VdbeAddOp0(v, OP_Goto);
assert( pLevel->addrSkip==0 );
pLevel->addrSkip = sqlite3VdbeAddOp4Int(v, (bRev?OP_SeekLT:OP_SeekGT),
iIdxCur, 0, regBase, nSkip);
VdbeCoverageIf(v, bRev==0);
VdbeCoverageIf(v, bRev!=0);
sqlite3VdbeJumpHere(v, j);
for(j=0; j<nSkip; j++){
sqlite3VdbeAddOp3(v, OP_Column, iIdxCur, j, regBase+j);
testcase( pIdx->aiColumn[j]==XN_EXPR );
VdbeComment((v, "%s", explainIndexColumnName(pIdx, j)));
}
}
/* Evaluate the equality constraints
*/
assert( zAff==0 || (int)strlen(zAff)>=nEq );
for(j=nSkip; j<nEq; j++){
int r1;
pTerm = pLoop->aLTerm[j];
assert( pTerm!=0 );
/* The following testcase is true for indices with redundant columns.
** Ex: CREATE INDEX i1 ON t1(a,b,a); SELECT * FROM t1 WHERE a=0 AND b=0; */
testcase( (pTerm->wtFlags & TERM_CODED)!=0 );
testcase( pTerm->wtFlags & TERM_VIRTUAL );
r1 = codeEqualityTerm(pParse, pTerm, pLevel, j, bRev, regBase+j);
if( r1!=regBase+j ){
if( nReg==1 ){
sqlite3ReleaseTempReg(pParse, regBase);
regBase = r1;
}else{
sqlite3VdbeAddOp2(v, OP_Copy, r1, regBase+j);
}
}
}
for(j=nSkip; j<nEq; j++){
pTerm = pLoop->aLTerm[j];
if( pTerm->eOperator & WO_IN ){
if( pTerm->pExpr->flags & EP_xIsSelect ){
/* No affinity ever needs to be (or should be) applied to a value
** from the RHS of an "? IN (SELECT ...)" expression. The
** sqlite3FindInIndex() routine has already ensured that the
** affinity of the comparison has been applied to the value. */
if( zAff ) zAff[j] = SQLITE_AFF_BLOB;
}
}else if( (pTerm->eOperator & WO_ISNULL)==0 ){
Expr *pRight = pTerm->pExpr->pRight;
if( (pTerm->wtFlags & TERM_IS)==0 && sqlite3ExprCanBeNull(pRight) ){
sqlite3VdbeAddOp2(v, OP_IsNull, regBase+j, pLevel->addrBrk);
VdbeCoverage(v);
}
if( pParse->nErr==0 ){
assert( pParse->db->mallocFailed==0 );
if( sqlite3CompareAffinity(pRight, zAff[j])==SQLITE_AFF_BLOB ){
zAff[j] = SQLITE_AFF_BLOB;
}
if( sqlite3ExprNeedsNoAffinityChange(pRight, zAff[j]) ){
zAff[j] = SQLITE_AFF_BLOB;
}
}
}
}
*pzAff = zAff;
return regBase;
}
#ifndef SQLITE_LIKE_DOESNT_MATCH_BLOBS
/*
** If the most recently coded instruction is a constant range constraint
** (a string literal) that originated from the LIKE optimization, then
** set P3 and P5 on the OP_String opcode so that the string will be cast
** to a BLOB at appropriate times.
**
** The LIKE optimization trys to evaluate "x LIKE 'abc%'" as a range
** expression: "x>='ABC' AND x<'abd'". But this requires that the range
** scan loop run twice, once for strings and a second time for BLOBs.
** The OP_String opcodes on the second pass convert the upper and lower
** bound string constants to blobs. This routine makes the necessary changes
** to the OP_String opcodes for that to happen.
**
** Except, of course, if SQLITE_LIKE_DOESNT_MATCH_BLOBS is defined, then
** only the one pass through the string space is required, so this routine
** becomes a no-op.
*/
static void whereLikeOptimizationStringFixup(
Vdbe *v, /* prepared statement under construction */
WhereLevel *pLevel, /* The loop that contains the LIKE operator */
WhereTerm *pTerm /* The upper or lower bound just coded */
){
if( pTerm->wtFlags & TERM_LIKEOPT ){
VdbeOp *pOp;
assert( pLevel->iLikeRepCntr>0 );
pOp = sqlite3VdbeGetLastOp(v);
assert( pOp!=0 );
assert( pOp->opcode==OP_String8
|| pTerm->pWC->pWInfo->pParse->db->mallocFailed );
pOp->p3 = (int)(pLevel->iLikeRepCntr>>1); /* Register holding counter */
pOp->p5 = (u8)(pLevel->iLikeRepCntr&1); /* ASC or DESC */
}
}
#else
# define whereLikeOptimizationStringFixup(A,B,C)
#endif
#ifdef SQLITE_ENABLE_CURSOR_HINTS
/*
** Information is passed from codeCursorHint() down to individual nodes of
** the expression tree (by sqlite3WalkExpr()) using an instance of this
** structure.
*/
struct CCurHint {
int iTabCur; /* Cursor for the main table */
int iIdxCur; /* Cursor for the index, if pIdx!=0. Unused otherwise */
Index *pIdx; /* The index used to access the table */
};
/*
** This function is called for every node of an expression that is a candidate
** for a cursor hint on an index cursor. For TK_COLUMN nodes that reference
** the table CCurHint.iTabCur, verify that the same column can be
** accessed through the index. If it cannot, then set pWalker->eCode to 1.
*/
static int codeCursorHintCheckExpr(Walker *pWalker, Expr *pExpr){
struct CCurHint *pHint = pWalker->u.pCCurHint;
assert( pHint->pIdx!=0 );
if( pExpr->op==TK_COLUMN
&& pExpr->iTable==pHint->iTabCur
&& sqlite3TableColumnToIndex(pHint->pIdx, pExpr->iColumn)<0
){
pWalker->eCode = 1;
}
return WRC_Continue;
}
/*
** Test whether or not expression pExpr, which was part of a WHERE clause,
** should be included in the cursor-hint for a table that is on the rhs
** of a LEFT JOIN. Set Walker.eCode to non-zero before returning if the
** expression is not suitable.
**
** An expression is unsuitable if it might evaluate to non NULL even if
** a TK_COLUMN node that does affect the value of the expression is set
** to NULL. For example:
**
** col IS NULL
** col IS NOT NULL
** coalesce(col, 1)
** CASE WHEN col THEN 0 ELSE 1 END
*/
static int codeCursorHintIsOrFunction(Walker *pWalker, Expr *pExpr){
if( pExpr->op==TK_IS
|| pExpr->op==TK_ISNULL || pExpr->op==TK_ISNOT
|| pExpr->op==TK_NOTNULL || pExpr->op==TK_CASE
){
pWalker->eCode = 1;
}else if( pExpr->op==TK_FUNCTION ){
int d1;
char d2[4];
if( 0==sqlite3IsLikeFunction(pWalker->pParse->db, pExpr, &d1, d2) ){
pWalker->eCode = 1;
}
}
return WRC_Continue;
}
/*
** This function is called on every node of an expression tree used as an
** argument to the OP_CursorHint instruction. If the node is a TK_COLUMN
** that accesses any table other than the one identified by
** CCurHint.iTabCur, then do the following:
**
** 1) allocate a register and code an OP_Column instruction to read
** the specified column into the new register, and
**
** 2) transform the expression node to a TK_REGISTER node that reads
** from the newly populated register.
**
** Also, if the node is a TK_COLUMN that does access the table idenified
** by pCCurHint.iTabCur, and an index is being used (which we will
** know because CCurHint.pIdx!=0) then transform the TK_COLUMN into
** an access of the index rather than the original table.
*/
static int codeCursorHintFixExpr(Walker *pWalker, Expr *pExpr){
int rc = WRC_Continue;
struct CCurHint *pHint = pWalker->u.pCCurHint;
if( pExpr->op==TK_COLUMN ){
if( pExpr->iTable!=pHint->iTabCur ){
int reg = ++pWalker->pParse->nMem; /* Register for column value */
sqlite3ExprCode(pWalker->pParse, pExpr, reg);
pExpr->op = TK_REGISTER;
pExpr->iTable = reg;
}else if( pHint->pIdx!=0 ){
pExpr->iTable = pHint->iIdxCur;
pExpr->iColumn = sqlite3TableColumnToIndex(pHint->pIdx, pExpr->iColumn);
assert( pExpr->iColumn>=0 );
}
}else if( pExpr->op==TK_AGG_FUNCTION ){
/* An aggregate function in the WHERE clause of a query means this must
** be a correlated sub-query, and expression pExpr is an aggregate from
** the parent context. Do not walk the function arguments in this case.
**
** todo: It should be possible to replace this node with a TK_REGISTER
** expression, as the result of the expression must be stored in a
** register at this point. The same holds for TK_AGG_COLUMN nodes. */
rc = WRC_Prune;
}
return rc;
}
/*
** Insert an OP_CursorHint instruction if it is appropriate to do so.
*/
static void codeCursorHint(
SrcItem *pTabItem, /* FROM clause item */
WhereInfo *pWInfo, /* The where clause */
WhereLevel *pLevel, /* Which loop to provide hints for */
WhereTerm *pEndRange /* Hint this end-of-scan boundary term if not NULL */
){
Parse *pParse = pWInfo->pParse;
sqlite3 *db = pParse->db;
Vdbe *v = pParse->pVdbe;
Expr *pExpr = 0;
WhereLoop *pLoop = pLevel->pWLoop;
int iCur;
WhereClause *pWC;
WhereTerm *pTerm;
int i, j;
struct CCurHint sHint;
Walker sWalker;
if( OptimizationDisabled(db, SQLITE_CursorHints) ) return;
iCur = pLevel->iTabCur;
assert( iCur==pWInfo->pTabList->a[pLevel->iFrom].iCursor );
sHint.iTabCur = iCur;
sHint.iIdxCur = pLevel->iIdxCur;
sHint.pIdx = pLoop->u.btree.pIndex;
memset(&sWalker, 0, sizeof(sWalker));
sWalker.pParse = pParse;
sWalker.u.pCCurHint = &sHint;
pWC = &pWInfo->sWC;
for(i=0; i<pWC->nBase; i++){
pTerm = &pWC->a[i];
if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue;
if( pTerm->prereqAll & pLevel->notReady ) continue;
/* Any terms specified as part of the ON(...) clause for any LEFT
** JOIN for which the current table is not the rhs are omitted
** from the cursor-hint.
**
** If this table is the rhs of a LEFT JOIN, "IS" or "IS NULL" terms
** that were specified as part of the WHERE clause must be excluded.
** This is to address the following:
**
** SELECT ... t1 LEFT JOIN t2 ON (t1.a=t2.b) WHERE t2.c IS NULL;
**
** Say there is a single row in t2 that matches (t1.a=t2.b), but its
** t2.c values is not NULL. If the (t2.c IS NULL) constraint is
** pushed down to the cursor, this row is filtered out, causing
** SQLite to synthesize a row of NULL values. Which does match the
** WHERE clause, and so the query returns a row. Which is incorrect.
**
** For the same reason, WHERE terms such as:
**
** WHERE 1 = (t2.c IS NULL)
**
** are also excluded. See codeCursorHintIsOrFunction() for details.
*/
if( pTabItem->fg.jointype & JT_LEFT ){
Expr *pExpr = pTerm->pExpr;
if( !ExprHasProperty(pExpr, EP_OuterON)
|| pExpr->w.iJoin!=pTabItem->iCursor
){
sWalker.eCode = 0;
sWalker.xExprCallback = codeCursorHintIsOrFunction;
sqlite3WalkExpr(&sWalker, pTerm->pExpr);
if( sWalker.eCode ) continue;
}
}else{
if( ExprHasProperty(pTerm->pExpr, EP_OuterON) ) continue;
}
/* All terms in pWLoop->aLTerm[] except pEndRange are used to initialize
** the cursor. These terms are not needed as hints for a pure range
** scan (that has no == terms) so omit them. */
if( pLoop->u.btree.nEq==0 && pTerm!=pEndRange ){
for(j=0; j<pLoop->nLTerm && pLoop->aLTerm[j]!=pTerm; j++){}
if( j<pLoop->nLTerm ) continue;
}
/* No subqueries or non-deterministic functions allowed */
if( sqlite3ExprContainsSubquery(pTerm->pExpr) ) continue;
/* For an index scan, make sure referenced columns are actually in
** the index. */
if( sHint.pIdx!=0 ){
sWalker.eCode = 0;
sWalker.xExprCallback = codeCursorHintCheckExpr;
sqlite3WalkExpr(&sWalker, pTerm->pExpr);
if( sWalker.eCode ) continue;
}
/* If we survive all prior tests, that means this term is worth hinting */
pExpr = sqlite3ExprAnd(pParse, pExpr, sqlite3ExprDup(db, pTerm->pExpr, 0));
}
if( pExpr!=0 ){
sWalker.xExprCallback = codeCursorHintFixExpr;
sqlite3WalkExpr(&sWalker, pExpr);
sqlite3VdbeAddOp4(v, OP_CursorHint,
(sHint.pIdx ? sHint.iIdxCur : sHint.iTabCur), 0, 0,
(const char*)pExpr, P4_EXPR);
}
}
#else
# define codeCursorHint(A,B,C,D) /* No-op */
#endif /* SQLITE_ENABLE_CURSOR_HINTS */
/*
** Cursor iCur is open on an intkey b-tree (a table). Register iRowid contains
** a rowid value just read from cursor iIdxCur, open on index pIdx. This
** function generates code to do a deferred seek of cursor iCur to the
** rowid stored in register iRowid.
**
** Normally, this is just:
**
** OP_DeferredSeek $iCur $iRowid
**
** Which causes a seek on $iCur to the row with rowid $iRowid.
**
** However, if the scan currently being coded is a branch of an OR-loop and
** the statement currently being coded is a SELECT, then additional information
** is added that might allow OP_Column to omit the seek and instead do its
** lookup on the index, thus avoiding an expensive seek operation. To
** enable this optimization, the P3 of OP_DeferredSeek is set to iIdxCur
** and P4 is set to an array of integers containing one entry for each column
** in the table. For each table column, if the column is the i'th
** column of the index, then the corresponding array entry is set to (i+1).
** If the column does not appear in the index at all, the array entry is set
** to 0. The OP_Column opcode can check this array to see if the column it
** wants is in the index and if it is, it will substitute the index cursor
** and column number and continue with those new values, rather than seeking
** the table cursor.
*/
static void codeDeferredSeek(
WhereInfo *pWInfo, /* Where clause context */
Index *pIdx, /* Index scan is using */
int iCur, /* Cursor for IPK b-tree */
int iIdxCur /* Index cursor */
){
Parse *pParse = pWInfo->pParse; /* Parse context */
Vdbe *v = pParse->pVdbe; /* Vdbe to generate code within */
assert( iIdxCur>0 );
assert( pIdx->aiColumn[pIdx->nColumn-1]==-1 );
pWInfo->bDeferredSeek = 1;
sqlite3VdbeAddOp3(v, OP_DeferredSeek, iIdxCur, 0, iCur);
if( (pWInfo->wctrlFlags & (WHERE_OR_SUBCLAUSE|WHERE_RIGHT_JOIN))
&& DbMaskAllZero(sqlite3ParseToplevel(pParse)->writeMask)
){
int i;
Table *pTab = pIdx->pTable;
u32 *ai = (u32*)sqlite3DbMallocZero(pParse->db, sizeof(u32)*(pTab->nCol+1));
if( ai ){
ai[0] = pTab->nCol;
for(i=0; i<pIdx->nColumn-1; i++){
int x1, x2;
assert( pIdx->aiColumn[i]<pTab->nCol );
x1 = pIdx->aiColumn[i];
x2 = sqlite3TableColumnToStorage(pTab, x1);
testcase( x1!=x2 );
if( x1>=0 ) ai[x2+1] = i+1;
}
sqlite3VdbeChangeP4(v, -1, (char*)ai, P4_INTARRAY);
}
}
}
/*
** If the expression passed as the second argument is a vector, generate
** code to write the first nReg elements of the vector into an array
** of registers starting with iReg.
**
** If the expression is not a vector, then nReg must be passed 1. In
** this case, generate code to evaluate the expression and leave the
** result in register iReg.
*/
static void codeExprOrVector(Parse *pParse, Expr *p, int iReg, int nReg){
assert( nReg>0 );
if( p && sqlite3ExprIsVector(p) ){
#ifndef SQLITE_OMIT_SUBQUERY
if( ExprUseXSelect(p) ){
Vdbe *v = pParse->pVdbe;
int iSelect;
assert( p->op==TK_SELECT );
iSelect = sqlite3CodeSubselect(pParse, p);
sqlite3VdbeAddOp3(v, OP_Copy, iSelect, iReg, nReg-1);
}else
#endif
{
int i;
const ExprList *pList;
assert( ExprUseXList(p) );
pList = p->x.pList;
assert( nReg<=pList->nExpr );
for(i=0; i<nReg; i++){
sqlite3ExprCode(pParse, pList->a[i].pExpr, iReg+i);
}
}
}else{
assert( nReg==1 || pParse->nErr );
sqlite3ExprCode(pParse, p, iReg);
}
}
/*
** The pTruth expression is always true because it is the WHERE clause
** a partial index that is driving a query loop. Look through all of the
** WHERE clause terms on the query, and if any of those terms must be
** true because pTruth is true, then mark those WHERE clause terms as
** coded.
*/
static void whereApplyPartialIndexConstraints(
Expr *pTruth,
int iTabCur,
WhereClause *pWC
){
int i;
WhereTerm *pTerm;
while( pTruth->op==TK_AND ){
whereApplyPartialIndexConstraints(pTruth->pLeft, iTabCur, pWC);
pTruth = pTruth->pRight;
}
for(i=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){
Expr *pExpr;
if( pTerm->wtFlags & TERM_CODED ) continue;
pExpr = pTerm->pExpr;
if( sqlite3ExprCompare(0, pExpr, pTruth, iTabCur)==0 ){
pTerm->wtFlags |= TERM_CODED;
}
}
}
/*
** This routine is called right after An OP_Filter has been generated and
** before the corresponding index search has been performed. This routine
** checks to see if there are additional Bloom filters in inner loops that
** can be checked prior to doing the index lookup. If there are available
** inner-loop Bloom filters, then evaluate those filters now, before the
** index lookup. The idea is that a Bloom filter check is way faster than
** an index lookup, and the Bloom filter might return false, meaning that
** the index lookup can be skipped.
**
** We know that an inner loop uses a Bloom filter because it has the
** WhereLevel.regFilter set. If an inner-loop Bloom filter is checked,
** then clear the WhereLevel.regFilter value to prevent the Bloom filter
** from being checked a second time when the inner loop is evaluated.
*/
static SQLITE_NOINLINE void filterPullDown(
Parse *pParse, /* Parsing context */
WhereInfo *pWInfo, /* Complete information about the WHERE clause */
int iLevel, /* Which level of pWInfo->a[] should be coded */
int addrNxt, /* Jump here to bypass inner loops */
Bitmask notReady /* Loops that are not ready */
){
while( ++iLevel < pWInfo->nLevel ){
WhereLevel *pLevel = &pWInfo->a[iLevel];
WhereLoop *pLoop = pLevel->pWLoop;
if( pLevel->regFilter==0 ) continue;
if( pLevel->pWLoop->nSkip ) continue;
/* ,--- Because sqlite3ConstructBloomFilter() has will not have set
** vvvvv--' pLevel->regFilter if this were true. */
if( NEVER(pLoop->prereq & notReady) ) continue;
assert( pLevel->addrBrk==0 );
pLevel->addrBrk = addrNxt;
if( pLoop->wsFlags & WHERE_IPK ){
WhereTerm *pTerm = pLoop->aLTerm[0];
int regRowid;
assert( pTerm!=0 );
assert( pTerm->pExpr!=0 );
testcase( pTerm->wtFlags & TERM_VIRTUAL );
regRowid = sqlite3GetTempReg(pParse);
regRowid = codeEqualityTerm(pParse, pTerm, pLevel, 0, 0, regRowid);
sqlite3VdbeAddOp2(pParse->pVdbe, OP_MustBeInt, regRowid, addrNxt);
VdbeCoverage(pParse->pVdbe);
sqlite3VdbeAddOp4Int(pParse->pVdbe, OP_Filter, pLevel->regFilter,
addrNxt, regRowid, 1);
VdbeCoverage(pParse->pVdbe);
}else{
u16 nEq = pLoop->u.btree.nEq;
int r1;
char *zStartAff;
assert( pLoop->wsFlags & WHERE_INDEXED );
assert( (pLoop->wsFlags & WHERE_COLUMN_IN)==0 );
r1 = codeAllEqualityTerms(pParse,pLevel,0,0,&zStartAff);
codeApplyAffinity(pParse, r1, nEq, zStartAff);
sqlite3DbFree(pParse->db, zStartAff);
sqlite3VdbeAddOp4Int(pParse->pVdbe, OP_Filter, pLevel->regFilter,
addrNxt, r1, nEq);
VdbeCoverage(pParse->pVdbe);
}
pLevel->regFilter = 0;
pLevel->addrBrk = 0;
}
}
/*
** Generate code for the start of the iLevel-th loop in the WHERE clause
** implementation described by pWInfo.
*/
Bitmask sqlite3WhereCodeOneLoopStart(
Parse *pParse, /* Parsing context */
Vdbe *v, /* Prepared statement under construction */
WhereInfo *pWInfo, /* Complete information about the WHERE clause */
int iLevel, /* Which level of pWInfo->a[] should be coded */
WhereLevel *pLevel, /* The current level pointer */
Bitmask notReady /* Which tables are currently available */
){
int j, k; /* Loop counters */
int iCur; /* The VDBE cursor for the table */
int addrNxt; /* Where to jump to continue with the next IN case */
int bRev; /* True if we need to scan in reverse order */
WhereLoop *pLoop; /* The WhereLoop object being coded */
WhereClause *pWC; /* Decomposition of the entire WHERE clause */
WhereTerm *pTerm; /* A WHERE clause term */
sqlite3 *db; /* Database connection */
SrcItem *pTabItem; /* FROM clause term being coded */
int addrBrk; /* Jump here to break out of the loop */
int addrHalt; /* addrBrk for the outermost 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 */
Index *pIdx = 0; /* Index used by loop (if any) */
int iLoop; /* Iteration of constraint generator loop */
pWC = &pWInfo->sWC;
db = pParse->db;
pLoop = pLevel->pWLoop;
pTabItem = &pWInfo->pTabList->a[pLevel->iFrom];
iCur = pTabItem->iCursor;
pLevel->notReady = notReady & ~sqlite3WhereGetMask(&pWInfo->sMaskSet, iCur);
bRev = (pWInfo->revMask>>iLevel)&1;
VdbeModuleComment((v, "Begin WHERE-loop%d: %s",iLevel,pTabItem->pTab->zName));
#if WHERETRACE_ENABLED /* 0x4001 */
if( sqlite3WhereTrace & 0x1 ){
sqlite3DebugPrintf("Coding level %d of %d: notReady=%llx iFrom=%d\n",
iLevel, pWInfo->nLevel, (u64)notReady, pLevel->iFrom);
if( sqlite3WhereTrace & 0x1000 ){
sqlite3WhereLoopPrint(pLoop, pWC);
}
}
if( (sqlite3WhereTrace & 0x4001)==0x4001 ){
if( iLevel==0 ){
sqlite3DebugPrintf("WHERE clause being coded:\n");
sqlite3TreeViewExpr(0, pWInfo->pWhere, 0);
}
sqlite3DebugPrintf("All WHERE-clause terms before coding:\n");
sqlite3WhereClausePrint(pWC);
}
#endif
/* 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.
**
** When there is an IN operator, we also have a "addrNxt" label that
** means to continue with the next IN value combination. When
** there are no IN operators in the constraints, the "addrNxt" label
** is the same as "addrBrk".
*/
addrBrk = pLevel->addrBrk = pLevel->addrNxt = sqlite3VdbeMakeLabel(pParse);
addrCont = pLevel->addrCont = sqlite3VdbeMakeLabel(pParse);
/* If this is the right table of a LEFT OUTER JOIN, allocate and
** initialize a memory cell that records if this table matches any
** row of the left table of the join.
*/
assert( (pWInfo->wctrlFlags & (WHERE_OR_SUBCLAUSE|WHERE_RIGHT_JOIN))
|| pLevel->iFrom>0 || (pTabItem[0].fg.jointype & JT_LEFT)==0
);
if( pLevel->iFrom>0 && (pTabItem[0].fg.jointype & JT_LEFT)!=0 ){
pLevel->iLeftJoin = ++pParse->nMem;
sqlite3VdbeAddOp2(v, OP_Integer, 0, pLevel->iLeftJoin);
VdbeComment((v, "init LEFT JOIN no-match flag"));
}
/* Compute a safe address to jump to if we discover that the table for
** this loop is empty and can never contribute content. */
for(j=iLevel; j>0; j--){
if( pWInfo->a[j].iLeftJoin ) break;
if( pWInfo->a[j].pRJ ) break;
}
addrHalt = pWInfo->a[j].addrBrk;
/* Special case of a FROM clause subquery implemented as a co-routine */
if( pTabItem->fg.viaCoroutine ){
int regYield = pTabItem->regReturn;
sqlite3VdbeAddOp3(v, OP_InitCoroutine, regYield, 0, pTabItem->addrFillSub);
pLevel->p2 = sqlite3VdbeAddOp2(v, OP_Yield, regYield, addrBrk);
VdbeCoverage(v);
VdbeComment((v, "next row of %s", pTabItem->pTab->zName));
pLevel->op = OP_Goto;
}else
#ifndef SQLITE_OMIT_VIRTUALTABLE
if( (pLoop->wsFlags & WHERE_VIRTUALTABLE)!=0 ){
/* Case 1: The table is a virtual-table. Use the VFilter and VNext
** to access the data.
*/
int iReg; /* P3 Value for OP_VFilter */
int addrNotFound;
int nConstraint = pLoop->nLTerm;
iReg = sqlite3GetTempRange(pParse, nConstraint+2);
addrNotFound = pLevel->addrBrk;
for(j=0; j<nConstraint; j++){
int iTarget = iReg+j+2;
pTerm = pLoop->aLTerm[j];
if( NEVER(pTerm==0) ) continue;
if( pTerm->eOperator & WO_IN ){
if( SMASKBIT32(j) & pLoop->u.vtab.mHandleIn ){
int iTab = pParse->nTab++;
int iCache = ++pParse->nMem;
sqlite3CodeRhsOfIN(pParse, pTerm->pExpr, iTab);
sqlite3VdbeAddOp3(v, OP_VInitIn, iTab, iTarget, iCache);
}else{
codeEqualityTerm(pParse, pTerm, pLevel, j, bRev, iTarget);
addrNotFound = pLevel->addrNxt;
}
}else{
Expr *pRight = pTerm->pExpr->pRight;
codeExprOrVector(pParse, pRight, iTarget, 1);
if( pTerm->eMatchOp==SQLITE_INDEX_CONSTRAINT_OFFSET
&& pLoop->u.vtab.bOmitOffset
){
assert( pTerm->eOperator==WO_AUX );
assert( pWInfo->pSelect!=0 );
assert( pWInfo->pSelect->iOffset>0 );
sqlite3VdbeAddOp2(v, OP_Integer, 0, pWInfo->pSelect->iOffset);
VdbeComment((v,"Zero OFFSET counter"));
}
}
}
sqlite3VdbeAddOp2(v, OP_Integer, pLoop->u.vtab.idxNum, iReg);
sqlite3VdbeAddOp2(v, OP_Integer, nConstraint, iReg+1);
sqlite3VdbeAddOp4(v, OP_VFilter, iCur, addrNotFound, iReg,
pLoop->u.vtab.idxStr,
pLoop->u.vtab.needFree ? P4_DYNAMIC : P4_STATIC);
VdbeCoverage(v);
pLoop->u.vtab.needFree = 0;
/* An OOM inside of AddOp4(OP_VFilter) instruction above might have freed
** the u.vtab.idxStr. NULL it out to prevent a use-after-free */
if( db->mallocFailed ) pLoop->u.vtab.idxStr = 0;
pLevel->p1 = iCur;
pLevel->op = pWInfo->eOnePass ? OP_Noop : OP_VNext;
pLevel->p2 = sqlite3VdbeCurrentAddr(v);
assert( (pLoop->wsFlags & WHERE_MULTI_OR)==0 );
for(j=0; j<nConstraint; j++){
pTerm = pLoop->aLTerm[j];
if( j<16 && (pLoop->u.vtab.omitMask>>j)&1 ){
disableTerm(pLevel, pTerm);
continue;
}
if( (pTerm->eOperator & WO_IN)!=0
&& (SMASKBIT32(j) & pLoop->u.vtab.mHandleIn)==0
&& !db->mallocFailed
){
Expr *pCompare; /* The comparison operator */
Expr *pRight; /* RHS of the comparison */
VdbeOp *pOp; /* Opcode to access the value of the IN constraint */
int iIn; /* IN loop corresponding to the j-th constraint */
/* Reload the constraint value into reg[iReg+j+2]. The same value
** was loaded into the same register prior to the OP_VFilter, but
** the xFilter implementation might have changed the datatype or
** encoding of the value in the register, so it *must* be reloaded.
*/
for(iIn=0; ALWAYS(iIn<pLevel->u.in.nIn); iIn++){
pOp = sqlite3VdbeGetOp(v, pLevel->u.in.aInLoop[iIn].addrInTop);
if( (pOp->opcode==OP_Column && pOp->p3==iReg+j+2)
|| (pOp->opcode==OP_Rowid && pOp->p2==iReg+j+2)
){
testcase( pOp->opcode==OP_Rowid );
sqlite3VdbeAddOp3(v, pOp->opcode, pOp->p1, pOp->p2, pOp->p3);
break;
}
}
/* Generate code that will continue to the next row if
** the IN constraint is not satisfied
*/
pCompare = sqlite3PExpr(pParse, TK_EQ, 0, 0);
if( !db->mallocFailed ){
int iFld = pTerm->u.x.iField;
Expr *pLeft = pTerm->pExpr->pLeft;
assert( pLeft!=0 );
if( iFld>0 ){
assert( pLeft->op==TK_VECTOR );
assert( ExprUseXList(pLeft) );
assert( iFld<=pLeft->x.pList->nExpr );
pCompare->pLeft = pLeft->x.pList->a[iFld-1].pExpr;
}else{
pCompare->pLeft = pLeft;
}
pCompare->pRight = pRight = sqlite3Expr(db, TK_REGISTER, 0);
if( pRight ){
pRight->iTable = iReg+j+2;
sqlite3ExprIfFalse(
pParse, pCompare, pLevel->addrCont, SQLITE_JUMPIFNULL
);
}
pCompare->pLeft = 0;
}
sqlite3ExprDelete(db, pCompare);
}
}
/* These registers need to be preserved in case there is an IN operator
** loop. So we could deallocate the registers here (and potentially
** reuse them later) if (pLoop->wsFlags & WHERE_IN_ABLE)==0. But it seems
** simpler and safer to simply not reuse the registers.
**
** sqlite3ReleaseTempRange(pParse, iReg, nConstraint+2);
*/
}else
#endif /* SQLITE_OMIT_VIRTUALTABLE */
if( (pLoop->wsFlags & WHERE_IPK)!=0
&& (pLoop->wsFlags & (WHERE_COLUMN_IN|WHERE_COLUMN_EQ))!=0
){
/* Case 2: We can directly reference a single row using an
** equality comparison against the ROWID field. Or
** we reference multiple rows using a "rowid IN (...)"
** construct.
*/
assert( pLoop->u.btree.nEq==1 );
pTerm = pLoop->aLTerm[0];
assert( pTerm!=0 );
assert( pTerm->pExpr!=0 );
testcase( pTerm->wtFlags & TERM_VIRTUAL );
iReleaseReg = ++pParse->nMem;
iRowidReg = codeEqualityTerm(pParse, pTerm, pLevel, 0, bRev, iReleaseReg);
if( iRowidReg!=iReleaseReg ) sqlite3ReleaseTempReg(pParse, iReleaseReg);
addrNxt = pLevel->addrNxt;
if( pLevel->regFilter ){
sqlite3VdbeAddOp2(v, OP_MustBeInt, iRowidReg, addrNxt);
VdbeCoverage(v);
sqlite3VdbeAddOp4Int(v, OP_Filter, pLevel->regFilter, addrNxt,
iRowidReg, 1);
VdbeCoverage(v);
filterPullDown(pParse, pWInfo, iLevel, addrNxt, notReady);
}
sqlite3VdbeAddOp3(v, OP_SeekRowid, iCur, addrNxt, iRowidReg);
VdbeCoverage(v);
pLevel->op = OP_Noop;
}else if( (pLoop->wsFlags & WHERE_IPK)!=0
&& (pLoop->wsFlags & WHERE_COLUMN_RANGE)!=0
){
/* Case 3: We have an inequality comparison against the ROWID field.
*/
int testOp = OP_Noop;
int start;
int memEndValue = 0;
WhereTerm *pStart, *pEnd;
j = 0;
pStart = pEnd = 0;
if( pLoop->wsFlags & WHERE_BTM_LIMIT ) pStart = pLoop->aLTerm[j++];
if( pLoop->wsFlags & WHERE_TOP_LIMIT ) pEnd = pLoop->aLTerm[j++];
assert( pStart!=0 || pEnd!=0 );
if( bRev ){
pTerm = pStart;
pStart = pEnd;
pEnd = pTerm;
}
codeCursorHint(pTabItem, pWInfo, pLevel, pEnd);
if( pStart ){
Expr *pX; /* The expression that defines the start bound */
int r1, rTemp; /* Registers for holding the start boundary */
int op; /* Cursor seek operation */
/* The following constant maps TK_xx codes into corresponding
** seek opcodes. It depends on a particular ordering of TK_xx
*/
const u8 aMoveOp[] = {
/* TK_GT */ OP_SeekGT,
/* TK_LE */ OP_SeekLE,
/* TK_LT */ OP_SeekLT,
/* TK_GE */ OP_SeekGE
};
assert( TK_LE==TK_GT+1 ); /* Make sure the ordering.. */
assert( TK_LT==TK_GT+2 ); /* ... of the TK_xx values... */
assert( TK_GE==TK_GT+3 ); /* ... is correcct. */
assert( (pStart->wtFlags & TERM_VNULL)==0 );
testcase( pStart->wtFlags & TERM_VIRTUAL );
pX = pStart->pExpr;
assert( pX!=0 );
testcase( pStart->leftCursor!=iCur ); /* transitive constraints */
if( sqlite3ExprIsVector(pX->pRight) ){
r1 = rTemp = sqlite3GetTempReg(pParse);
codeExprOrVector(pParse, pX->pRight, r1, 1);
testcase( pX->op==TK_GT );
testcase( pX->op==TK_GE );
testcase( pX->op==TK_LT );
testcase( pX->op==TK_LE );
op = aMoveOp[((pX->op - TK_GT - 1) & 0x3) | 0x1];
assert( pX->op!=TK_GT || op==OP_SeekGE );
assert( pX->op!=TK_GE || op==OP_SeekGE );
assert( pX->op!=TK_LT || op==OP_SeekLE );
assert( pX->op!=TK_LE || op==OP_SeekLE );
}else{
r1 = sqlite3ExprCodeTemp(pParse, pX->pRight, &rTemp);
disableTerm(pLevel, pStart);
op = aMoveOp[(pX->op - TK_GT)];
}
sqlite3VdbeAddOp3(v, op, iCur, addrBrk, r1);
VdbeComment((v, "pk"));
VdbeCoverageIf(v, pX->op==TK_GT);
VdbeCoverageIf(v, pX->op==TK_LE);
VdbeCoverageIf(v, pX->op==TK_LT);
VdbeCoverageIf(v, pX->op==TK_GE);
sqlite3ReleaseTempReg(pParse, rTemp);
}else{
sqlite3VdbeAddOp2(v, bRev ? OP_Last : OP_Rewind, iCur, addrHalt);
VdbeCoverageIf(v, bRev==0);
VdbeCoverageIf(v, bRev!=0);
}
if( pEnd ){
Expr *pX;
pX = pEnd->pExpr;
assert( pX!=0 );
assert( (pEnd->wtFlags & TERM_VNULL)==0 );
testcase( pEnd->leftCursor!=iCur ); /* Transitive constraints */
testcase( pEnd->wtFlags & TERM_VIRTUAL );
memEndValue = ++pParse->nMem;
codeExprOrVector(pParse, pX->pRight, memEndValue, 1);
if( 0==sqlite3ExprIsVector(pX->pRight)
&& (pX->op==TK_LT || pX->op==TK_GT)
){
testOp = bRev ? OP_Le : OP_Ge;
}else{
testOp = bRev ? OP_Lt : OP_Gt;
}
if( 0==sqlite3ExprIsVector(pX->pRight) ){
disableTerm(pLevel, pEnd);
}
}
start = sqlite3VdbeCurrentAddr(v);
pLevel->op = bRev ? OP_Prev : OP_Next;
pLevel->p1 = iCur;
pLevel->p2 = start;
assert( pLevel->p5==0 );
if( testOp!=OP_Noop ){
iRowidReg = ++pParse->nMem;
sqlite3VdbeAddOp2(v, OP_Rowid, iCur, iRowidReg);
sqlite3VdbeAddOp3(v, testOp, memEndValue, addrBrk, iRowidReg);
VdbeCoverageIf(v, testOp==OP_Le);
VdbeCoverageIf(v, testOp==OP_Lt);
VdbeCoverageIf(v, testOp==OP_Ge);
VdbeCoverageIf(v, testOp==OP_Gt);
sqlite3VdbeChangeP5(v, SQLITE_AFF_NUMERIC | SQLITE_JUMPIFNULL);
}
}else if( pLoop->wsFlags & WHERE_INDEXED ){
/* Case 4: A scan using an index.
**
** The WHERE clause may contain zero or more equality
** terms ("==" or "IN" operators) that refer to the N
** left-most columns of the index. It may also contain
** inequality constraints (>, <, >= or <=) on the indexed
** column that immediately follows the N equalities. Only
** the right-most column can be an inequality - the rest must
** use the "==" and "IN" operators. For example, if the
** index is on (x,y,z), then the following clauses are all
** optimized:
**
** x=5
** x=5 AND y=10
** x=5 AND y<10
** x=5 AND y>5 AND y<10
** x=5 AND y=5 AND z<=10
**
** The z<10 term of the following cannot be used, only
** the x=5 term:
**
** x=5 AND z<10
**
** N may be zero if there are inequality constraints.
** If there are no inequality constraints, then N is at
** least one.
**
** This case is also used when there are no WHERE clause
** constraints but an index is selected anyway, in order
** to force the output order to conform to an ORDER BY.
*/
static const u8 aStartOp[] = {
0,
0,
OP_Rewind, /* 2: (!start_constraints && startEq && !bRev) */
OP_Last, /* 3: (!start_constraints && startEq && bRev) */
OP_SeekGT, /* 4: (start_constraints && !startEq && !bRev) */
OP_SeekLT, /* 5: (start_constraints && !startEq && bRev) */
OP_SeekGE, /* 6: (start_constraints && startEq && !bRev) */
OP_SeekLE /* 7: (start_constraints && startEq && bRev) */
};
static const u8 aEndOp[] = {
OP_IdxGE, /* 0: (end_constraints && !bRev && !endEq) */
OP_IdxGT, /* 1: (end_constraints && !bRev && endEq) */
OP_IdxLE, /* 2: (end_constraints && bRev && !endEq) */
OP_IdxLT, /* 3: (end_constraints && bRev && endEq) */
};
u16 nEq = pLoop->u.btree.nEq; /* Number of == or IN terms */
u16 nBtm = pLoop->u.btree.nBtm; /* Length of BTM vector */
u16 nTop = pLoop->u.btree.nTop; /* Length of TOP vector */
int regBase; /* Base register holding constraint values */
WhereTerm *pRangeStart = 0; /* Inequality constraint at range start */
WhereTerm *pRangeEnd = 0; /* Inequality constraint at range end */
int startEq; /* True if range start uses ==, >= or <= */
int endEq; /* True if range end uses ==, >= or <= */
int start_constraints; /* Start of range is constrained */
int nConstraint; /* Number of constraint terms */
int iIdxCur; /* The VDBE cursor for the index */
int nExtraReg = 0; /* Number of extra registers needed */
int op; /* Instruction opcode */
char *zStartAff; /* Affinity for start of range constraint */
char *zEndAff = 0; /* Affinity for end of range constraint */
u8 bSeekPastNull = 0; /* True to seek past initial nulls */
u8 bStopAtNull = 0; /* Add condition to terminate at NULLs */
int omitTable; /* True if we use the index only */
int regBignull = 0; /* big-null flag register */
int addrSeekScan = 0; /* Opcode of the OP_SeekScan, if any */
pIdx = pLoop->u.btree.pIndex;
iIdxCur = pLevel->iIdxCur;
assert( nEq>=pLoop->nSkip );
/* Find any inequality constraint terms for the start and end
** of the range.
*/
j = nEq;
if( pLoop->wsFlags & WHERE_BTM_LIMIT ){
pRangeStart = pLoop->aLTerm[j++];
nExtraReg = MAX(nExtraReg, pLoop->u.btree.nBtm);
/* Like optimization range constraints always occur in pairs */
assert( (pRangeStart->wtFlags & TERM_LIKEOPT)==0 ||
(pLoop->wsFlags & WHERE_TOP_LIMIT)!=0 );
}
if( pLoop->wsFlags & WHERE_TOP_LIMIT ){
pRangeEnd = pLoop->aLTerm[j++];
nExtraReg = MAX(nExtraReg, pLoop->u.btree.nTop);
#ifndef SQLITE_LIKE_DOESNT_MATCH_BLOBS
if( (pRangeEnd->wtFlags & TERM_LIKEOPT)!=0 ){
assert( pRangeStart!=0 ); /* LIKE opt constraints */
assert( pRangeStart->wtFlags & TERM_LIKEOPT ); /* occur in pairs */
pLevel->iLikeRepCntr = (u32)++pParse->nMem;
sqlite3VdbeAddOp2(v, OP_Integer, 1, (int)pLevel->iLikeRepCntr);
VdbeComment((v, "LIKE loop counter"));
pLevel->addrLikeRep = sqlite3VdbeCurrentAddr(v);
/* iLikeRepCntr actually stores 2x the counter register number. The
** bottom bit indicates whether the search order is ASC or DESC. */
testcase( bRev );
testcase( pIdx->aSortOrder[nEq]==SQLITE_SO_DESC );
assert( (bRev & ~1)==0 );
pLevel->iLikeRepCntr <<=1;
pLevel->iLikeRepCntr |= bRev ^ (pIdx->aSortOrder[nEq]==SQLITE_SO_DESC);
}
#endif
if( pRangeStart==0 ){
j = pIdx->aiColumn[nEq];
if( (j>=0 && pIdx->pTable->aCol[j].notNull==0) || j==XN_EXPR ){
bSeekPastNull = 1;
}
}
}
assert( pRangeEnd==0 || (pRangeEnd->wtFlags & TERM_VNULL)==0 );
/* If the WHERE_BIGNULL_SORT flag is set, then index column nEq uses
** a non-default "big-null" sort (either ASC NULLS LAST or DESC NULLS
** FIRST). In both cases separate ordered scans are made of those
** index entries for which the column is null and for those for which
** it is not. For an ASC sort, the non-NULL entries are scanned first.
** For DESC, NULL entries are scanned first.
*/
if( (pLoop->wsFlags & (WHERE_TOP_LIMIT|WHERE_BTM_LIMIT))==0
&& (pLoop->wsFlags & WHERE_BIGNULL_SORT)!=0
){
assert( bSeekPastNull==0 && nExtraReg==0 && nBtm==0 && nTop==0 );
assert( pRangeEnd==0 && pRangeStart==0 );
testcase( pLoop->nSkip>0 );
nExtraReg = 1;
bSeekPastNull = 1;
pLevel->regBignull = regBignull = ++pParse->nMem;
if( pLevel->iLeftJoin ){
sqlite3VdbeAddOp2(v, OP_Integer, 0, regBignull);
}
pLevel->addrBignull = sqlite3VdbeMakeLabel(pParse);
}
/* If we are doing a reverse order scan on an ascending index, or
** a forward order scan on a descending index, interchange the
** start and end terms (pRangeStart and pRangeEnd).
*/
if( (nEq<pIdx->nColumn && bRev==(pIdx->aSortOrder[nEq]==SQLITE_SO_ASC)) ){
SWAP(WhereTerm *, pRangeEnd, pRangeStart);
SWAP(u8, bSeekPastNull, bStopAtNull);
SWAP(u8, nBtm, nTop);
}
if( iLevel>0 && (pLoop->wsFlags & WHERE_IN_SEEKSCAN)!=0 ){
/* In case OP_SeekScan is used, ensure that the index cursor does not
** point to a valid row for the first iteration of this loop. */
sqlite3VdbeAddOp1(v, OP_NullRow, iIdxCur);
}
/* Generate code to evaluate all constraint terms using == or IN
** and store the values of those terms in an array of registers
** starting at regBase.
*/
codeCursorHint(pTabItem, pWInfo, pLevel, pRangeEnd);
regBase = codeAllEqualityTerms(pParse,pLevel,bRev,nExtraReg,&zStartAff);
assert( zStartAff==0 || sqlite3Strlen30(zStartAff)>=nEq );
if( zStartAff && nTop ){
zEndAff = sqlite3DbStrDup(db, &zStartAff[nEq]);
}
addrNxt = (regBignull ? pLevel->addrBignull : pLevel->addrNxt);
testcase( pRangeStart && (pRangeStart->eOperator & WO_LE)!=0 );
testcase( pRangeStart && (pRangeStart->eOperator & WO_GE)!=0 );
testcase( pRangeEnd && (pRangeEnd->eOperator & WO_LE)!=0 );
testcase( pRangeEnd && (pRangeEnd->eOperator & WO_GE)!=0 );
startEq = !pRangeStart || pRangeStart->eOperator & (WO_LE|WO_GE);
endEq = !pRangeEnd || pRangeEnd->eOperator & (WO_LE|WO_GE);
start_constraints = pRangeStart || nEq>0;
/* Seek the index cursor to the start of the range. */
nConstraint = nEq;
if( pRangeStart ){
Expr *pRight = pRangeStart->pExpr->pRight;
codeExprOrVector(pParse, pRight, regBase+nEq, nBtm);
whereLikeOptimizationStringFixup(v, pLevel, pRangeStart);
if( (pRangeStart->wtFlags & TERM_VNULL)==0
&& sqlite3ExprCanBeNull(pRight)
){
sqlite3VdbeAddOp2(v, OP_IsNull, regBase+nEq, addrNxt);
VdbeCoverage(v);
}
if( zStartAff ){
updateRangeAffinityStr(pRight, nBtm, &zStartAff[nEq]);
}
nConstraint += nBtm;
testcase( pRangeStart->wtFlags & TERM_VIRTUAL );
if( sqlite3ExprIsVector(pRight)==0 ){
disableTerm(pLevel, pRangeStart);
}else{
startEq = 1;
}
bSeekPastNull = 0;
}else if( bSeekPastNull ){
startEq = 0;
sqlite3VdbeAddOp2(v, OP_Null, 0, regBase+nEq);
start_constraints = 1;
nConstraint++;
}else if( regBignull ){
sqlite3VdbeAddOp2(v, OP_Null, 0, regBase+nEq);
start_constraints = 1;
nConstraint++;
}
codeApplyAffinity(pParse, regBase, nConstraint - bSeekPastNull, zStartAff);
if( pLoop->nSkip>0 && nConstraint==pLoop->nSkip ){
/* The skip-scan logic inside the call to codeAllEqualityConstraints()
** above has already left the cursor sitting on the correct row,
** so no further seeking is needed */
}else{
if( regBignull ){
sqlite3VdbeAddOp2(v, OP_Integer, 1, regBignull);
VdbeComment((v, "NULL-scan pass ctr"));
}
if( pLevel->regFilter ){
sqlite3VdbeAddOp4Int(v, OP_Filter, pLevel->regFilter, addrNxt,
regBase, nEq);
VdbeCoverage(v);
filterPullDown(pParse, pWInfo, iLevel, addrNxt, notReady);
}
op = aStartOp[(start_constraints<<2) + (startEq<<1) + bRev];
assert( op!=0 );
if( (pLoop->wsFlags & WHERE_IN_SEEKSCAN)!=0 && op==OP_SeekGE ){
assert( regBignull==0 );
/* TUNING: The OP_SeekScan opcode seeks to reduce the number
** of expensive seek operations by replacing a single seek with
** 1 or more step operations. The question is, how many steps
** should we try before giving up and going with a seek. The cost
** of a seek is proportional to the logarithm of the of the number
** of entries in the tree, so basing the number of steps to try
** on the estimated number of rows in the btree seems like a good
** guess. */
addrSeekScan = sqlite3VdbeAddOp1(v, OP_SeekScan,
(pIdx->aiRowLogEst[0]+9)/10);
if( pRangeStart || pRangeEnd ){
sqlite3VdbeChangeP5(v, 1);
sqlite3VdbeChangeP2(v, addrSeekScan, sqlite3VdbeCurrentAddr(v)+1);
addrSeekScan = 0;
}
VdbeCoverage(v);
}
sqlite3VdbeAddOp4Int(v, op, iIdxCur, addrNxt, regBase, nConstraint);
VdbeCoverage(v);
VdbeCoverageIf(v, op==OP_Rewind); testcase( op==OP_Rewind );
VdbeCoverageIf(v, op==OP_Last); testcase( op==OP_Last );
VdbeCoverageIf(v, op==OP_SeekGT); testcase( op==OP_SeekGT );
VdbeCoverageIf(v, op==OP_SeekGE); testcase( op==OP_SeekGE );
VdbeCoverageIf(v, op==OP_SeekLE); testcase( op==OP_SeekLE );
VdbeCoverageIf(v, op==OP_SeekLT); testcase( op==OP_SeekLT );
assert( bSeekPastNull==0 || bStopAtNull==0 );
if( regBignull ){
assert( bSeekPastNull==1 || bStopAtNull==1 );
assert( bSeekPastNull==!bStopAtNull );
assert( bStopAtNull==startEq );
sqlite3VdbeAddOp2(v, OP_Goto, 0, sqlite3VdbeCurrentAddr(v)+2);
op = aStartOp[(nConstraint>1)*4 + 2 + bRev];
sqlite3VdbeAddOp4Int(v, op, iIdxCur, addrNxt, regBase,
nConstraint-startEq);
VdbeCoverage(v);
VdbeCoverageIf(v, op==OP_Rewind); testcase( op==OP_Rewind );
VdbeCoverageIf(v, op==OP_Last); testcase( op==OP_Last );
VdbeCoverageIf(v, op==OP_SeekGE); testcase( op==OP_SeekGE );
VdbeCoverageIf(v, op==OP_SeekLE); testcase( op==OP_SeekLE );
assert( op==OP_Rewind || op==OP_Last || op==OP_SeekGE || op==OP_SeekLE);
}
}
/* Load the value for the inequality constraint at the end of the
** range (if any).
*/
nConstraint = nEq;
assert( pLevel->p2==0 );
if( pRangeEnd ){
Expr *pRight = pRangeEnd->pExpr->pRight;
assert( addrSeekScan==0 );
codeExprOrVector(pParse, pRight, regBase+nEq, nTop);
whereLikeOptimizationStringFixup(v, pLevel, pRangeEnd);
if( (pRangeEnd->wtFlags & TERM_VNULL)==0
&& sqlite3ExprCanBeNull(pRight)
){
sqlite3VdbeAddOp2(v, OP_IsNull, regBase+nEq, addrNxt);
VdbeCoverage(v);
}
if( zEndAff ){
updateRangeAffinityStr(pRight, nTop, zEndAff);
codeApplyAffinity(pParse, regBase+nEq, nTop, zEndAff);
}else{
assert( pParse->db->mallocFailed );
}
nConstraint += nTop;
testcase( pRangeEnd->wtFlags & TERM_VIRTUAL );
if( sqlite3ExprIsVector(pRight)==0 ){
disableTerm(pLevel, pRangeEnd);
}else{
endEq = 1;
}
}else if( bStopAtNull ){
if( regBignull==0 ){
sqlite3VdbeAddOp2(v, OP_Null, 0, regBase+nEq);
endEq = 0;
}
nConstraint++;
}
if( zStartAff ) sqlite3DbNNFreeNN(db, zStartAff);
if( zEndAff ) sqlite3DbNNFreeNN(db, zEndAff);
/* Top of the loop body */
pLevel->p2 = sqlite3VdbeCurrentAddr(v);
/* Check if the index cursor is past the end of the range. */
if( nConstraint ){
if( regBignull ){
/* Except, skip the end-of-range check while doing the NULL-scan */
sqlite3VdbeAddOp2(v, OP_IfNot, regBignull, sqlite3VdbeCurrentAddr(v)+3);
VdbeComment((v, "If NULL-scan 2nd pass"));
VdbeCoverage(v);
}
op = aEndOp[bRev*2 + endEq];
sqlite3VdbeAddOp4Int(v, op, iIdxCur, addrNxt, regBase, nConstraint);
testcase( op==OP_IdxGT ); VdbeCoverageIf(v, op==OP_IdxGT );
testcase( op==OP_IdxGE ); VdbeCoverageIf(v, op==OP_IdxGE );
testcase( op==OP_IdxLT ); VdbeCoverageIf(v, op==OP_IdxLT );
testcase( op==OP_IdxLE ); VdbeCoverageIf(v, op==OP_IdxLE );
if( addrSeekScan ) sqlite3VdbeJumpHere(v, addrSeekScan);
}
if( regBignull ){
/* During a NULL-scan, check to see if we have reached the end of
** the NULLs */
assert( bSeekPastNull==!bStopAtNull );
assert( bSeekPastNull+bStopAtNull==1 );
assert( nConstraint+bSeekPastNull>0 );
sqlite3VdbeAddOp2(v, OP_If, regBignull, sqlite3VdbeCurrentAddr(v)+2);
VdbeComment((v, "If NULL-scan 1st pass"));
VdbeCoverage(v);
op = aEndOp[bRev*2 + bSeekPastNull];
sqlite3VdbeAddOp4Int(v, op, iIdxCur, addrNxt, regBase,
nConstraint+bSeekPastNull);
testcase( op==OP_IdxGT ); VdbeCoverageIf(v, op==OP_IdxGT );
testcase( op==OP_IdxGE ); VdbeCoverageIf(v, op==OP_IdxGE );
testcase( op==OP_IdxLT ); VdbeCoverageIf(v, op==OP_IdxLT );
testcase( op==OP_IdxLE ); VdbeCoverageIf(v, op==OP_IdxLE );
}
if( (pLoop->wsFlags & WHERE_IN_EARLYOUT)!=0 ){
sqlite3VdbeAddOp3(v, OP_SeekHit, iIdxCur, nEq, nEq);
}
/* Seek the table cursor, if required */
omitTable = (pLoop->wsFlags & WHERE_IDX_ONLY)!=0
&& (pWInfo->wctrlFlags & (WHERE_OR_SUBCLAUSE|WHERE_RIGHT_JOIN))==0;
if( omitTable ){
/* pIdx is a covering index. No need to access the main table. */
}else if( HasRowid(pIdx->pTable) ){
codeDeferredSeek(pWInfo, pIdx, iCur, iIdxCur);
}else if( iCur!=iIdxCur ){
Index *pPk = sqlite3PrimaryKeyIndex(pIdx->pTable);
iRowidReg = sqlite3GetTempRange(pParse, pPk->nKeyCol);
for(j=0; j<pPk->nKeyCol; j++){
k = sqlite3TableColumnToIndex(pIdx, pPk->aiColumn[j]);
sqlite3VdbeAddOp3(v, OP_Column, iIdxCur, k, iRowidReg+j);
}
sqlite3VdbeAddOp4Int(v, OP_NotFound, iCur, addrCont,
iRowidReg, pPk->nKeyCol); VdbeCoverage(v);
}
if( pLevel->iLeftJoin==0 ){
/* If a partial index is driving the loop, try to eliminate WHERE clause
** terms from the query that must be true due to the WHERE clause of
** the partial index.
**
** 2019-11-02 ticket 623eff57e76d45f6: This optimization does not work
** for a LEFT JOIN.
*/
if( pIdx->pPartIdxWhere ){
whereApplyPartialIndexConstraints(pIdx->pPartIdxWhere, iCur, pWC);
}
}else{
testcase( pIdx->pPartIdxWhere );
/* The following assert() is not a requirement, merely an observation:
** The OR-optimization doesn't work for the right hand table of
** a LEFT JOIN: */
assert( (pWInfo->wctrlFlags & (WHERE_OR_SUBCLAUSE|WHERE_RIGHT_JOIN))==0 );
}
/* Record the instruction used to terminate the loop. */
if( pLoop->wsFlags & WHERE_ONEROW ){
pLevel->op = OP_Noop;
}else if( bRev ){
pLevel->op = OP_Prev;
}else{
pLevel->op = OP_Next;
}
pLevel->p1 = iIdxCur;
pLevel->p3 = (pLoop->wsFlags&WHERE_UNQ_WANTED)!=0 ? 1:0;
if( (pLoop->wsFlags & WHERE_CONSTRAINT)==0 ){
pLevel->p5 = SQLITE_STMTSTATUS_FULLSCAN_STEP;
}else{
assert( pLevel->p5==0 );
}
if( omitTable ) pIdx = 0;
}else
#ifndef SQLITE_OMIT_OR_OPTIMIZATION
if( pLoop->wsFlags & WHERE_MULTI_OR ){
/* Case 5: Two or more separately indexed terms connected by OR
**
** Example:
**
** CREATE TABLE t1(a,b,c,d);
** CREATE INDEX i1 ON t1(a);
** CREATE INDEX i2 ON t1(b);
** CREATE INDEX i3 ON t1(c);
**
** SELECT * FROM t1 WHERE a=5 OR b=7 OR (c=11 AND d=13)
**
** In the example, there are three indexed terms connected by OR.
** The top of the loop looks like this:
**
** Null 1 # Zero the rowset in reg 1
**
** Then, for each indexed term, the following. The arguments to
** RowSetTest are such that the rowid of the current row is inserted
** into the RowSet. If it is already present, control skips the
** Gosub opcode and jumps straight to the code generated by WhereEnd().
**
** sqlite3WhereBegin(<term>)
** RowSetTest # Insert rowid into rowset
** Gosub 2 A
** sqlite3WhereEnd()
**
** Following the above, code to terminate the loop. Label A, the target
** of the Gosub above, jumps to the instruction right after the Goto.
**
** Null 1 # Zero the rowset in reg 1
** Goto B # The loop is finished.
**
** A: <loop body> # Return data, whatever.
**
** Return 2 # Jump back to the Gosub
**
** B: <after the loop>
**
** Added 2014-05-26: If the table is a WITHOUT ROWID table, then
** use an ephemeral index instead of a RowSet to record the primary
** keys of the rows we have already seen.
**
*/
WhereClause *pOrWc; /* The OR-clause broken out into subterms */
SrcList *pOrTab; /* Shortened table list or OR-clause generation */
Index *pCov = 0; /* Potential covering index (or NULL) */
int iCovCur = pParse->nTab++; /* Cursor used for index scans (if any) */
int regReturn = ++pParse->nMem; /* Register used with OP_Gosub */
int regRowset = 0; /* Register for RowSet object */
int regRowid = 0; /* Register holding rowid */
int iLoopBody = sqlite3VdbeMakeLabel(pParse);/* Start of loop body */
int iRetInit; /* Address of regReturn init */
int untestedTerms = 0; /* Some terms not completely tested */
int ii; /* Loop counter */
Expr *pAndExpr = 0; /* An ".. AND (...)" expression */
Table *pTab = pTabItem->pTab;
pTerm = pLoop->aLTerm[0];
assert( pTerm!=0 );
assert( pTerm->eOperator & WO_OR );
assert( (pTerm->wtFlags & TERM_ORINFO)!=0 );
pOrWc = &pTerm->u.pOrInfo->wc;
pLevel->op = OP_Return;
pLevel->p1 = regReturn;
/* Set up a new SrcList in pOrTab containing the table being scanned
** by this loop in the a[0] slot and all notReady tables in a[1..] slots.
** This becomes the SrcList in the recursive call to sqlite3WhereBegin().
*/
if( pWInfo->nLevel>1 ){
int nNotReady; /* The number of notReady tables */
SrcItem *origSrc; /* Original list of tables */
nNotReady = pWInfo->nLevel - iLevel - 1;
pOrTab = sqlite3DbMallocRawNN(db,
sizeof(*pOrTab)+ nNotReady*sizeof(pOrTab->a[0]));
if( pOrTab==0 ) return notReady;
pOrTab->nAlloc = (u8)(nNotReady + 1);
pOrTab->nSrc = pOrTab->nAlloc;
memcpy(pOrTab->a, pTabItem, sizeof(*pTabItem));
origSrc = pWInfo->pTabList->a;
for(k=1; k<=nNotReady; k++){
memcpy(&pOrTab->a[k], &origSrc[pLevel[k].iFrom], sizeof(pOrTab->a[k]));
}
}else{
pOrTab = pWInfo->pTabList;
}
/* Initialize the rowset register to contain NULL. An SQL NULL is
** equivalent to an empty rowset. Or, create an ephemeral index
** capable of holding primary keys in the case of a WITHOUT ROWID.
**
** Also initialize regReturn to contain the address of the instruction
** immediately following the OP_Return at the bottom of the loop. This
** is required in a few obscure LEFT JOIN cases where control jumps
** over the top of the loop into the body of it. In this case the
** correct response for the end-of-loop code (the OP_Return) is to
** fall through to the next instruction, just as an OP_Next does if
** called on an uninitialized cursor.
*/
if( (pWInfo->wctrlFlags & WHERE_DUPLICATES_OK)==0 ){
if( HasRowid(pTab) ){
regRowset = ++pParse->nMem;
sqlite3VdbeAddOp2(v, OP_Null, 0, regRowset);
}else{
Index *pPk = sqlite3PrimaryKeyIndex(pTab);
regRowset = pParse->nTab++;
sqlite3VdbeAddOp2(v, OP_OpenEphemeral, regRowset, pPk->nKeyCol);
sqlite3VdbeSetP4KeyInfo(pParse, pPk);
}
regRowid = ++pParse->nMem;
}
iRetInit = sqlite3VdbeAddOp2(v, OP_Integer, 0, regReturn);
/* If the original WHERE clause is z of the form: (x1 OR x2 OR ...) AND y
** Then for every term xN, evaluate as the subexpression: xN AND y
** That way, terms in y that are factored into the disjunction will
** be picked up by the recursive calls to sqlite3WhereBegin() below.
**
** 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
**
** 2022-02-04: Do not push down slices of a row-value comparison.
** In other words, "w" or "y" may not be a slice of a vector. Otherwise,
** the initialization of the right-hand operand of the vector comparison
** might not occur, or might occur only in an OR branch that is not
** taken. dbsqlfuzz 80a9fade844b4fb43564efc972bcb2c68270f5d1.
**
** 2022-03-03: Do not push down expressions that involve subqueries.
** The subquery might get coded as a subroutine. Any table-references
** in the subquery might be resolved to index-references for the index on
** the OR branch in which the subroutine is coded. But if the subroutine
** is invoked from a different OR branch that uses a different index, such
** index-references will not work. tag-20220303a
** https://sqlite.org/forum/forumpost/36937b197273d403
*/
if( pWC->nTerm>1 ){
int iTerm;
for(iTerm=0; iTerm<pWC->nTerm; iTerm++){
Expr *pExpr = pWC->a[iTerm].pExpr;
if( &pWC->a[iTerm] == pTerm ) continue;
testcase( pWC->a[iTerm].wtFlags & TERM_VIRTUAL );
testcase( pWC->a[iTerm].wtFlags & TERM_CODED );
testcase( pWC->a[iTerm].wtFlags & TERM_SLICE );
if( (pWC->a[iTerm].wtFlags & (TERM_VIRTUAL|TERM_CODED|TERM_SLICE))!=0 ){
continue;
}
if( (pWC->a[iTerm].eOperator & WO_ALL)==0 ) continue;
if( ExprHasProperty(pExpr, EP_Subquery) ) continue; /* tag-20220303a */
pExpr = sqlite3ExprDup(db, pExpr, 0);
pAndExpr = sqlite3ExprAnd(pParse, pAndExpr, pExpr);
}
if( pAndExpr ){
/* The extra 0x10000 bit on the opcode is masked off and does not
** become part of the new Expr.op. However, it does make the
** op==TK_AND comparison inside of sqlite3PExpr() false, and this
** prevents sqlite3PExpr() from applying the AND short-circuit
** optimization, which we do not want here. */
pAndExpr = sqlite3PExpr(pParse, TK_AND|0x10000, 0, pAndExpr);
}
}
/* Run a separate WHERE clause for each term of the OR clause. After
** eliminating duplicates from other WHERE clauses, the action for each
** sub-WHERE clause is to to invoke the main loop body as a subroutine.
*/
ExplainQueryPlan((pParse, 1, "MULTI-INDEX OR"));
for(ii=0; ii<pOrWc->nTerm; 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; /* Current OR clause term */
Expr *pDelete; /* Local copy of OR clause term */
int jmp1 = 0; /* Address of jump operation */
testcase( (pTabItem[0].fg.jointype & JT_LEFT)!=0
&& !ExprHasProperty(pOrExpr, EP_OuterON)
); /* See TH3 vtab25.400 and ticket 614b25314c766238 */
pDelete = pOrExpr = sqlite3ExprDup(db, pOrExpr, 0);
if( db->mallocFailed ){
sqlite3ExprDelete(db, pDelete);
continue;
}
if( pAndExpr ){
pAndExpr->pLeft = pOrExpr;
pOrExpr = pAndExpr;
}
/* Loop through table entries that match term pOrTerm. */
ExplainQueryPlan((pParse, 1, "INDEX %d", ii+1));
WHERETRACE(0xffffffff, ("Subplan for OR-clause:\n"));
pSubWInfo = sqlite3WhereBegin(pParse, pOrTab, pOrExpr, 0, 0, 0,
WHERE_OR_SUBCLAUSE, iCovCur);
assert( pSubWInfo || pParse->nErr );
if( pSubWInfo ){
WhereLoop *pSubLoop;
int addrExplain = sqlite3WhereExplainOneScan(
pParse, pOrTab, &pSubWInfo->a[0], 0
);
sqlite3WhereAddScanStatus(v, pOrTab, &pSubWInfo->a[0], addrExplain);
/* This is the sub-WHERE clause body. First skip over
** duplicate rows from prior sub-WHERE clauses, and record the
** rowid (or PRIMARY KEY) for the current row so that the same
** row will be skipped in subsequent sub-WHERE clauses.
*/
if( (pWInfo->wctrlFlags & WHERE_DUPLICATES_OK)==0 ){
int iSet = ((ii==pOrWc->nTerm-1)?-1:ii);
if( HasRowid(pTab) ){
sqlite3ExprCodeGetColumnOfTable(v, pTab, iCur, -1, regRowid);
jmp1 = sqlite3VdbeAddOp4Int(v, OP_RowSetTest, regRowset, 0,
regRowid, iSet);
VdbeCoverage(v);
}else{
Index *pPk = sqlite3PrimaryKeyIndex(pTab);
int nPk = pPk->nKeyCol;
int iPk;
int r;
/* Read the PK into an array of temp registers. */
r = sqlite3GetTempRange(pParse, nPk);
for(iPk=0; iPk<nPk; iPk++){
int iCol = pPk->aiColumn[iPk];
sqlite3ExprCodeGetColumnOfTable(v, pTab, iCur, iCol,r+iPk);
}
/* Check if the temp table already contains this key. If so,
** the row has already been included in the result set and
** can be ignored (by jumping past the Gosub below). Otherwise,
** insert the key into the temp table and proceed with processing
** the row.
**
** Use some of the same optimizations as OP_RowSetTest: If iSet
** is zero, assume that the key cannot already be present in
** the temp table. And if iSet is -1, assume that there is no
** need to insert the key into the temp table, as it will never
** be tested for. */
if( iSet ){
jmp1 = sqlite3VdbeAddOp4Int(v, OP_Found, regRowset, 0, r, nPk);
VdbeCoverage(v);
}
if( iSet>=0 ){
sqlite3VdbeAddOp3(v, OP_MakeRecord, r, nPk, regRowid);
sqlite3VdbeAddOp4Int(v, OP_IdxInsert, regRowset, regRowid,
r, nPk);
if( iSet ) sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT);
}
/* Release the array of temp registers */
sqlite3ReleaseTempRange(pParse, r, nPk);
}
}
/* Invoke the main loop body as a subroutine */
sqlite3VdbeAddOp2(v, OP_Gosub, regReturn, iLoopBody);
/* Jump here (skipping the main loop body subroutine) if the
** current sub-WHERE row is a duplicate from prior sub-WHEREs. */
if( jmp1 ) sqlite3VdbeJumpHere(v, jmp1);
/* The pSubWInfo->untestedTerms flag means that this OR term
** contained one or more AND term from a notReady table. The
** terms from the notReady table could not be tested and will
** need to be tested later.
*/
if( pSubWInfo->untestedTerms ) untestedTerms = 1;
/* If all of the OR-connected terms are optimized using the same
** index, and the index is opened using the same cursor number
** by each call to sqlite3WhereBegin() made by this loop, it may
** be possible to use that index as a covering index.
**
** If the call to sqlite3WhereBegin() above resulted in a scan that
** uses an index, and this is either the first OR-connected term
** processed or the index is the same as that used by all previous
** terms, set pCov to the candidate covering index. Otherwise, set
** pCov to NULL to indicate that no candidate covering index will
** be available.
*/
pSubLoop = pSubWInfo->a[0].pWLoop;
assert( (pSubLoop->wsFlags & WHERE_AUTO_INDEX)==0 );
if( (pSubLoop->wsFlags & WHERE_INDEXED)!=0
&& (ii==0 || pSubLoop->u.btree.pIndex==pCov)
&& (HasRowid(pTab) || !IsPrimaryKeyIndex(pSubLoop->u.btree.pIndex))
){
assert( pSubWInfo->a[0].iIdxCur==iCovCur );
pCov = pSubLoop->u.btree.pIndex;
}else{
pCov = 0;
}
if( sqlite3WhereUsesDeferredSeek(pSubWInfo) ){
pWInfo->bDeferredSeek = 1;
}
/* Finish the loop through table entries that match term pOrTerm. */
sqlite3WhereEnd(pSubWInfo);
ExplainQueryPlanPop(pParse);
}
sqlite3ExprDelete(db, pDelete);
}
}
ExplainQueryPlanPop(pParse);
assert( pLevel->pWLoop==pLoop );
assert( (pLoop->wsFlags & WHERE_MULTI_OR)!=0 );
assert( (pLoop->wsFlags & WHERE_IN_ABLE)==0 );
pLevel->u.pCoveringIdx = pCov;
if( pCov ) pLevel->iIdxCur = iCovCur;
if( pAndExpr ){
pAndExpr->pLeft = 0;
sqlite3ExprDelete(db, pAndExpr);
}
sqlite3VdbeChangeP1(v, iRetInit, sqlite3VdbeCurrentAddr(v));
sqlite3VdbeGoto(v, pLevel->addrBrk);
sqlite3VdbeResolveLabel(v, iLoopBody);
/* Set the P2 operand of the OP_Return opcode that will end the current
** loop to point to this spot, which is the top of the next containing
** loop. The byte-code formatter will use that P2 value as a hint to
** indent everything in between the this point and the final OP_Return.
** See tag-20220407a in vdbe.c and shell.c */
assert( pLevel->op==OP_Return );
pLevel->p2 = sqlite3VdbeCurrentAddr(v);
if( pWInfo->nLevel>1 ){ sqlite3DbFreeNN(db, pOrTab); }
if( !untestedTerms ) disableTerm(pLevel, pTerm);
}else
#endif /* SQLITE_OMIT_OR_OPTIMIZATION */
{
/* Case 6: There is no usable index. We must do a complete
** scan of the entire table.
*/
static const u8 aStep[] = { OP_Next, OP_Prev };
static const u8 aStart[] = { OP_Rewind, OP_Last };
assert( bRev==0 || bRev==1 );
if( pTabItem->fg.isRecursive ){
/* Tables marked isRecursive have only a single row that is stored in
** a pseudo-cursor. No need to Rewind or Next such cursors. */
pLevel->op = OP_Noop;
}else{
codeCursorHint(pTabItem, pWInfo, pLevel, 0);
pLevel->op = aStep[bRev];
pLevel->p1 = iCur;
pLevel->p2 = 1 + sqlite3VdbeAddOp2(v, aStart[bRev], iCur, addrHalt);
VdbeCoverageIf(v, bRev==0);
VdbeCoverageIf(v, bRev!=0);
pLevel->p5 = SQLITE_STMTSTATUS_FULLSCAN_STEP;
}
}
#ifdef SQLITE_ENABLE_STMT_SCANSTATUS
pLevel->addrVisit = sqlite3VdbeCurrentAddr(v);
#endif
/* Insert code to test every subexpression that can be completely
** computed using the current set of tables.
**
** This loop may run between one and three times, depending on the
** constraints to be generated. The value of stack variable iLoop
** determines the constraints coded by each iteration, as follows:
**
** iLoop==1: Code only expressions that are entirely covered by pIdx.
** iLoop==2: Code remaining expressions that do not contain correlated
** sub-queries.
** iLoop==3: Code all remaining expressions.
**
** An effort is made to skip unnecessary iterations of the loop.
*/
iLoop = (pIdx ? 1 : 2);
do{
int iNext = 0; /* Next value for iLoop */
for(pTerm=pWC->a, j=pWC->nTerm; j>0; j--, pTerm++){
Expr *pE;
int skipLikeAddr = 0;
testcase( pTerm->wtFlags & TERM_VIRTUAL );
testcase( pTerm->wtFlags & TERM_CODED );
if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue;
if( (pTerm->prereqAll & pLevel->notReady)!=0 ){
testcase( pWInfo->untestedTerms==0
&& (pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE)!=0 );
pWInfo->untestedTerms = 1;
continue;
}
pE = pTerm->pExpr;
assert( pE!=0 );
if( pTabItem->fg.jointype & (JT_LEFT|JT_LTORJ|JT_RIGHT) ){
if( !ExprHasProperty(pE,EP_OuterON|EP_InnerON) ){
/* Defer processing WHERE clause constraints until after outer
** join processing. tag-20220513a */
continue;
}else if( (pTabItem->fg.jointype & JT_LEFT)==JT_LEFT
&& !ExprHasProperty(pE,EP_OuterON) ){
continue;
}else{
Bitmask m = sqlite3WhereGetMask(&pWInfo->sMaskSet, pE->w.iJoin);
if( m & pLevel->notReady ){
/* An ON clause that is not ripe */
continue;
}
}
}
if( iLoop==1 && !sqlite3ExprCoveredByIndex(pE, pLevel->iTabCur, pIdx) ){
iNext = 2;
continue;
}
if( iLoop<3 && (pTerm->wtFlags & TERM_VARSELECT) ){
if( iNext==0 ) iNext = 3;
continue;
}
if( (pTerm->wtFlags & TERM_LIKECOND)!=0 ){
/* If the TERM_LIKECOND flag is set, that means that the range search
** is sufficient to guarantee that the LIKE operator is true, so we
** can skip the call to the like(A,B) function. But this only works
** for strings. So do not skip the call to the function on the pass
** that compares BLOBs. */
#ifdef SQLITE_LIKE_DOESNT_MATCH_BLOBS
continue;
#else
u32 x = pLevel->iLikeRepCntr;
if( x>0 ){
skipLikeAddr = sqlite3VdbeAddOp1(v, (x&1)?OP_IfNot:OP_If,(int)(x>>1));
VdbeCoverageIf(v, (x&1)==1);
VdbeCoverageIf(v, (x&1)==0);
}
#endif
}
#ifdef WHERETRACE_ENABLED /* 0xffffffff */
if( sqlite3WhereTrace ){
VdbeNoopComment((v, "WhereTerm[%d] (%p) priority=%d",
pWC->nTerm-j, pTerm, iLoop));
}
if( sqlite3WhereTrace & 0x4000 ){
sqlite3DebugPrintf("Coding auxiliary constraint:\n");
sqlite3WhereTermPrint(pTerm, pWC->nTerm-j);
}
#endif
sqlite3ExprIfFalse(pParse, pE, addrCont, SQLITE_JUMPIFNULL);
if( skipLikeAddr ) sqlite3VdbeJumpHere(v, skipLikeAddr);
pTerm->wtFlags |= TERM_CODED;
}
iLoop = iNext;
}while( iLoop>0 );
/* 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->nBase; j>0; j--, pTerm++){
Expr *pE, sEAlt;
WhereTerm *pAlt;
if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue;
if( (pTerm->eOperator & (WO_EQ|WO_IS))==0 ) continue;
if( (pTerm->eOperator & WO_EQUIV)==0 ) continue;
if( pTerm->leftCursor!=iCur ) continue;
if( pTabItem->fg.jointype & (JT_LEFT|JT_LTORJ|JT_RIGHT) ) continue;
pE = pTerm->pExpr;
#ifdef WHERETRACE_ENABLED /* 0x4001 */
if( (sqlite3WhereTrace & 0x4001)==0x4001 ){
sqlite3DebugPrintf("Coding transitive constraint:\n");
sqlite3WhereTermPrint(pTerm, pWC->nTerm-j);
}
#endif
assert( !ExprHasProperty(pE, EP_OuterON) );
assert( (pTerm->prereqRight & pLevel->notReady)!=0 );
assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 );
pAlt = sqlite3WhereFindTerm(pWC, iCur, pTerm->u.x.leftColumn, notReady,
WO_EQ|WO_IN|WO_IS, 0);
if( pAlt==0 ) continue;
if( pAlt->wtFlags & (TERM_CODED) ) continue;
if( (pAlt->eOperator & WO_IN)
&& ExprUseXSelect(pAlt->pExpr)
&& (pAlt->pExpr->x.pSelect->pEList->nExpr>1)
){
continue;
}
testcase( pAlt->eOperator & WO_EQ );
testcase( pAlt->eOperator & WO_IS );
testcase( pAlt->eOperator & WO_IN );
VdbeModuleComment((v, "begin transitive constraint"));
sEAlt = *pAlt->pExpr;
sEAlt.pLeft = pE->pLeft;
sqlite3ExprIfFalse(pParse, &sEAlt, addrCont, SQLITE_JUMPIFNULL);
pAlt->wtFlags |= TERM_CODED;
}
/* For a RIGHT OUTER JOIN, record the fact that the current row has
** been matched at least once.
*/
if( pLevel->pRJ ){
Table *pTab;
int nPk;
int r;
int jmp1 = 0;
WhereRightJoin *pRJ = pLevel->pRJ;
/* pTab is the right-hand table of the RIGHT JOIN. Generate code that
** will record that the current row of that table has been matched at
** least once. This is accomplished by storing the PK for the row in
** both the iMatch index and the regBloom Bloom filter.
*/
pTab = pWInfo->pTabList->a[pLevel->iFrom].pTab;
if( HasRowid(pTab) ){
r = sqlite3GetTempRange(pParse, 2);
sqlite3ExprCodeGetColumnOfTable(v, pTab, pLevel->iTabCur, -1, r+1);
nPk = 1;
}else{
int iPk;
Index *pPk = sqlite3PrimaryKeyIndex(pTab);
nPk = pPk->nKeyCol;
r = sqlite3GetTempRange(pParse, nPk+1);
for(iPk=0; iPk<nPk; iPk++){
int iCol = pPk->aiColumn[iPk];
sqlite3ExprCodeGetColumnOfTable(v, pTab, iCur, iCol,r+1+iPk);
}
}
jmp1 = sqlite3VdbeAddOp4Int(v, OP_Found, pRJ->iMatch, 0, r+1, nPk);
VdbeCoverage(v);
VdbeComment((v, "match against %s", pTab->zName));
sqlite3VdbeAddOp3(v, OP_MakeRecord, r+1, nPk, r);
sqlite3VdbeAddOp4Int(v, OP_IdxInsert, pRJ->iMatch, r, r+1, nPk);
sqlite3VdbeAddOp4Int(v, OP_FilterAdd, pRJ->regBloom, 0, r+1, nPk);
sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT);
sqlite3VdbeJumpHere(v, jmp1);
sqlite3ReleaseTempRange(pParse, r, nPk+1);
}
/* 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 ){
pLevel->addrFirst = sqlite3VdbeCurrentAddr(v);
sqlite3VdbeAddOp2(v, OP_Integer, 1, pLevel->iLeftJoin);
VdbeComment((v, "record LEFT JOIN hit"));
if( pLevel->pRJ==0 ){
goto code_outer_join_constraints; /* WHERE clause constraints */
}
}
if( pLevel->pRJ ){
/* Create a subroutine used to process all interior loops and code
** of the RIGHT JOIN. During normal operation, the subroutine will
** be in-line with the rest of the code. But at the end, a separate
** loop will run that invokes this subroutine for unmatched rows
** of pTab, with all tables to left begin set to NULL.
*/
WhereRightJoin *pRJ = pLevel->pRJ;
sqlite3VdbeAddOp2(v, OP_BeginSubrtn, 0, pRJ->regReturn);
pRJ->addrSubrtn = sqlite3VdbeCurrentAddr(v);
assert( pParse->withinRJSubrtn < 255 );
pParse->withinRJSubrtn++;
/* WHERE clause constraints must be deferred until after outer join
** row elimination has completed, since WHERE clause constraints apply
** to the results of the OUTER JOIN. The following loop generates the
** appropriate WHERE clause constraint checks. tag-20220513a.
*/
code_outer_join_constraints:
for(pTerm=pWC->a, j=0; j<pWC->nBase; j++, pTerm++){
testcase( pTerm->wtFlags & TERM_VIRTUAL );
testcase( pTerm->wtFlags & TERM_CODED );
if( pTerm->wtFlags & (TERM_VIRTUAL|TERM_CODED) ) continue;
if( (pTerm->prereqAll & pLevel->notReady)!=0 ){
assert( pWInfo->untestedTerms );
continue;
}
if( pTabItem->fg.jointype & JT_LTORJ ) continue;
assert( pTerm->pExpr );
sqlite3ExprIfFalse(pParse, pTerm->pExpr, addrCont, SQLITE_JUMPIFNULL);
pTerm->wtFlags |= TERM_CODED;
}
}
#if WHERETRACE_ENABLED /* 0x4001 */
if( sqlite3WhereTrace & 0x4000 ){
sqlite3DebugPrintf("All WHERE-clause terms after coding level %d:\n",
iLevel);
sqlite3WhereClausePrint(pWC);
}
if( sqlite3WhereTrace & 0x1 ){
sqlite3DebugPrintf("End Coding level %d: notReady=%llx\n",
iLevel, (u64)pLevel->notReady);
}
#endif
return pLevel->notReady;
}
/*
** Generate the code for the loop that finds all non-matched terms
** for a RIGHT JOIN.
*/
SQLITE_NOINLINE void sqlite3WhereRightJoinLoop(
WhereInfo *pWInfo,
int iLevel,
WhereLevel *pLevel
){
Parse *pParse = pWInfo->pParse;
Vdbe *v = pParse->pVdbe;
WhereRightJoin *pRJ = pLevel->pRJ;
Expr *pSubWhere = 0;
WhereClause *pWC = &pWInfo->sWC;
WhereInfo *pSubWInfo;
WhereLoop *pLoop = pLevel->pWLoop;
SrcItem *pTabItem = &pWInfo->pTabList->a[pLevel->iFrom];
SrcList sFrom;
Bitmask mAll = 0;
int k;
ExplainQueryPlan((pParse, 1, "RIGHT-JOIN %s", pTabItem->pTab->zName));
sqlite3VdbeNoJumpsOutsideSubrtn(v, pRJ->addrSubrtn, pRJ->endSubrtn,
pRJ->regReturn);
for(k=0; k<iLevel; k++){
int iIdxCur;
mAll |= pWInfo->a[k].pWLoop->maskSelf;
sqlite3VdbeAddOp1(v, OP_NullRow, pWInfo->a[k].iTabCur);
iIdxCur = pWInfo->a[k].iIdxCur;
if( iIdxCur ){
sqlite3VdbeAddOp1(v, OP_NullRow, iIdxCur);
}
}
if( (pTabItem->fg.jointype & JT_LTORJ)==0 ){
mAll |= pLoop->maskSelf;
for(k=0; k<pWC->nTerm; k++){
WhereTerm *pTerm = &pWC->a[k];
if( (pTerm->wtFlags & (TERM_VIRTUAL|TERM_SLICE))!=0
&& pTerm->eOperator!=WO_ROWVAL
){
break;
}
if( pTerm->prereqAll & ~mAll ) continue;
if( ExprHasProperty(pTerm->pExpr, EP_OuterON|EP_InnerON) ) continue;
pSubWhere = sqlite3ExprAnd(pParse, pSubWhere,
sqlite3ExprDup(pParse->db, pTerm->pExpr, 0));
}
}
sFrom.nSrc = 1;
sFrom.nAlloc = 1;
memcpy(&sFrom.a[0], pTabItem, sizeof(SrcItem));
sFrom.a[0].fg.jointype = 0;
assert( pParse->withinRJSubrtn < 100 );
pParse->withinRJSubrtn++;
pSubWInfo = sqlite3WhereBegin(pParse, &sFrom, pSubWhere, 0, 0, 0,
WHERE_RIGHT_JOIN, 0);
if( pSubWInfo ){
int iCur = pLevel->iTabCur;
int r = ++pParse->nMem;
int nPk;
int jmp;
int addrCont = sqlite3WhereContinueLabel(pSubWInfo);
Table *pTab = pTabItem->pTab;
if( HasRowid(pTab) ){
sqlite3ExprCodeGetColumnOfTable(v, pTab, iCur, -1, r);
nPk = 1;
}else{
int iPk;
Index *pPk = sqlite3PrimaryKeyIndex(pTab);
nPk = pPk->nKeyCol;
pParse->nMem += nPk - 1;
for(iPk=0; iPk<nPk; iPk++){
int iCol = pPk->aiColumn[iPk];
sqlite3ExprCodeGetColumnOfTable(v, pTab, iCur, iCol,r+iPk);
}
}
jmp = sqlite3VdbeAddOp4Int(v, OP_Filter, pRJ->regBloom, 0, r, nPk);
VdbeCoverage(v);
sqlite3VdbeAddOp4Int(v, OP_Found, pRJ->iMatch, addrCont, r, nPk);
VdbeCoverage(v);
sqlite3VdbeJumpHere(v, jmp);
sqlite3VdbeAddOp2(v, OP_Gosub, pRJ->regReturn, pRJ->addrSubrtn);
sqlite3WhereEnd(pSubWInfo);
}
sqlite3ExprDelete(pParse->db, pSubWhere);
ExplainQueryPlanPop(pParse);
assert( pParse->withinRJSubrtn>0 );
pParse->withinRJSubrtn--;
}