#pragma ident "%Z%%M% %I% %E% SMI"
/*
** 2001 September 15
**
** 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.
**
** $Id: where.c,v 1.89.2.2 2004/07/19 19:30:50 drh Exp $
*/
#include "sqliteInt.h"
/*
** The query generator uses an array of instances of this structure to
** help it analyze the subexpressions of the WHERE clause. Each WHERE
** clause subexpression is separated from the others by an AND operator.
*/
struct ExprInfo {
Expr *p; /* Pointer to the subexpression */
** p->pLeft is not the column of any table */
** p->pRight is not the column of any table */
};
/*
** An instance of the following structure keeps track of a mapping
** between VDBE cursor numbers and bitmasks. The VDBE cursor numbers
** are small integers contained in SrcList_item.iCursor and Expr.iTable
** fields. For any given WHERE clause, we want to track which cursors
** are being used, so we assign a single bit in a 32-bit word to track
** that cursor. Then a 32-bit integer is able to show the set of all
** cursors being used.
*/
struct ExprMaskSet {
int n; /* Number of assigned cursor values */
};
/*
** Determine the number of elements in an array.
*/
#define ARRAYSIZE(X) (sizeof(X)/sizeof(X[0]))
/*
** This routine is used to divide the WHERE expression into subexpressions
** separated by the AND operator.
**
** aSlot[] is an array of subexpressions structures.
** There are nSlot spaces left in this array. This routine attempts to
** split pExpr into subexpressions and fills aSlot[] with those subexpressions.
** The return value is the number of slots filled.
*/
int cnt = 0;
return 1;
}
}else{
}
return cnt;
}
/*
** Initialize an expression mask set
*/
/*
** Return the bitmask for the given cursor. Assign a new bitmask
** if this is the first time the cursor has been seen.
*/
int i;
for(i=0; i<pMaskSet->n; i++){
}
pMaskSet->n++;
return 1<<i;
}
return 0;
}
/*
** Destroy an expression mask set
*/
/*
** This routine walks (recursively) an expression tree and generates
** a bitmask indicating which tables are used in that expression
** tree.
**
** In order for this routine to work, the calling function must have
** previously invoked sqliteExprResolveIds() on the expression. See
** the header comment on that routine for additional information.
** The sqliteExprResolveIds() routines looks for column names and
** sets their opcodes to TK_COLUMN and their Expr.iTable fields to
** the VDBE cursor number of the table.
*/
unsigned int mask = 0;
if( p==0 ) return 0;
return mask;
}
if( p->pRight ){
}
if( p->pLeft ){
}
if( p->pList ){
int i;
}
}
return mask;
}
/*
** Return TRUE if the given operator is one of the operators that is
** allowed for an indexable WHERE clause. The allowed operators are
** "=", "<", ">", "<=", ">=", and "IN".
*/
switch( op ){
case TK_LT:
case TK_LE:
case TK_GT:
case TK_GE:
case TK_EQ:
case TK_IN:
return 1;
default:
return 0;
}
}
/*
** The input to this routine is an ExprInfo structure with only the
** "p" field filled in. The job of this routine is to analyze the
** subexpression and populate all the other fields of the ExprInfo
** structure.
*/
}
}
}
}
/*
** pOrderBy is an ORDER BY clause from a SELECT statement. pTab is the
** left-most table in the FROM clause of that same SELECT statement and
** the table has a cursor number of "base".
**
** This routine attempts to find an index for pTab that generates the
** correct record sequence for the given ORDER BY clause. The return value
** is a pointer to an index that does the job. NULL is returned if the
** table has no index that will generate the correct sort order.
**
** If there are two or more indices that generate the correct sort order
** and pPreferredIdx is one of those indices, then return pPreferredIdx.
**
** nEqCol is the number of columns of pPreferredIdx that are used as
** equality constraints. Any index returned must have exactly this same
** set of columns. The ORDER BY clause only matches index columns beyond the
** the first nEqCol columns.
**
** All terms of the ORDER BY clause must be either ASC or DESC. The
** *pbRev value is set to 1 if the ORDER BY clause is all DESC and it is
** set to 0 if the ORDER BY clause is all ASC.
*/
int base, /* Cursor number for pTab */
int nEqCol, /* Number of index columns used with == constraints */
int *pbRev /* Set to 1 if ORDER BY is DESC */
){
int i, j;
int sortOrder;
Expr *p;
/* Indices can only be used if all ORDER BY terms are either
** DESC or ASC. Indices cannot be used on a mixture. */
return 0;
}
/* Do not sort by index if there is a COLLATE clause */
return 0;
}
/* Can not use an index sort on anything that is not a column in the
** left-most table of the FROM clause */
return 0;
}
}
/* If we get this far, it means the ORDER BY clause consists only of
** ascending columns in the left-most table of the FROM clause. Now
** check for a matching index.
*/
pMatch = 0;
for(i=j=0; i<nEqCol; i++){
}
if( i<nEqCol ) continue;
for(i=0; i+j<nExpr; i++){
}
if( i+j>=nExpr ){
if( pIdx==pPreferredIdx ) break;
}
}
}
return pMatch;
}
/*
** 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 did not originate
** 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. 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.
*/
*ppExpr = 0;
}
}
/*
** Generate the beginning of the loop used for WHERE clause processing.
** The return value is a pointer to an (opaque) structure that contains
** information needed to terminate the loop. Later, the calling routine
** should invoke sqliteWhereEnd() with the return value of this function
** in order to complete the WHERE clause processing.
**
** If an error occurs, this routine returns NULL.
**
** The basic idea is to do a nested loop, one loop for each table in
** the FROM clause of a select. (INSERT and UPDATE statements are the
** same as a SELECT with only a single table in the FROM clause.) For
** example, if the SQL is this:
**
** SELECT * FROM t1, t2, t3 WHERE ...;
**
** Then the code generated is conceptually like the following:
**
** foreach row1 in t1 do \ Code generated
** foreach row2 in t2 do |-- by sqliteWhereBegin()
** foreach row3 in t3 do /
** ...
** end \ Code generated
** end |-- by sqliteWhereEnd()
** end /
**
** There are Btree cursors associated with each table. t1 uses cursor
** number pTabList->a[0].iCursor. t2 uses the cursor pTabList->a[1].iCursor.
** And so forth. This routine generates code to open those VDBE cursors
** and sqliteWhereEnd() generates the code to close them.
**
** If the WHERE clause is empty, the foreach loops must each scan their
** entire tables. Thus a three-way join is an O(N^3) operation. But if
** the tables have indices and there are terms in the WHERE clause that
** refer to those indices, a complete table scan can be avoided and the
** code will run much faster. Most of the work of this routine is checking
** to see if there are indices that can be used to speed up the loop.
**
** Terms of the WHERE clause are also used to limit which rows actually
** make it to the "..." in the middle of the loop. After each "foreach",
** terms of the WHERE clause that use only terms in that loop and outer
** loops are evaluated and if false a jump is made around all subsequent
** inner loops (or around the "..." if the test occurs within the inner-
** most loop)
**
** OUTER JOINS
**
** An outer join of tables t1 and t2 is conceptally coded as follows:
**
** foreach row1 in t1 do
** flag = 0
** foreach row2 in t2 do
** start:
** ...
** flag = 1
** end
** if flag==0 then
** move the row2 cursor to a null row
** goto start
** fi
** end
**
** ORDER BY CLAUSE PROCESSING
**
** *ppOrderBy is a pointer to the ORDER BY clause of a SELECT statement,
** if there is one. If there is no ORDER BY clause or if this routine
** is called from an UPDATE or DELETE statement, then ppOrderBy is NULL.
**
** If an index can be used so that the natural output order of the table
** scan is correct for the ORDER BY clause, then that index is used and
** *ppOrderBy is set to NULL. This is an optimization that prevents an
** unnecessary sort of the result set if an index appropriate for the
** ORDER BY clause already exists.
**
** If the where clause loops cannot be arranged to provide the correct
** output order, then the *ppOrderBy is unchanged.
*/
int pushKey, /* If TRUE, leave the table key on the stack */
){
int i; /* Loop counter */
/* pushKey is only allowed if there is a single table (as in an INSERT or
** UPDATE statement)
*/
/* Split the WHERE clause into separate subexpressions where each
** subexpression is separated by an AND operator. If the aExpr[]
** array fills up, the last entry might point to an expression which
** contains additional unfactored AND operators.
*/
return 0;
}
/* Allocate and initialize the WhereInfo structure that will become the
** return value.
*/
if( sqlite_malloc_failed ){
return 0;
}
/* Special case: a WHERE clause that is constant. Evaluate the
** expression and either jump over all of the code or fall thru.
*/
pWhere = 0;
}
/* Analyze all of the subexpressions.
*/
for(i=0; i<nExpr; i++){
/* If we are executing a trigger body, remove all references to
** new.* and old.* tables from the prerequisite masks.
*/
int x;
}
}
}
}
/* Figure out what index to use (if any) for each nested loop.
** Make pWInfo->a[i].pIdx point to the index to use for the i-th nested
** loop where i==0 is the outer loop and i==pTabList->nSrc-1 is the inner
** loop.
**
** If terms exist that use the ROWID of any table, then set the
** iDirectEq[], iDirectLt[], or iDirectGt[] elements for that table
** to the index of the term containing the ROWID. We always prefer
** to use a ROWID which can directly access a table rather than an
** index which requires reading an index first to get the rowid then
** doing a second read of the actual database table.
**
** Actually, if there are more than 32 tables in the join, only the
** first 32 tables are candidates for indices. This is (again) due
** to the limit of 32 bits in an integer bitmask.
*/
loopMask = 0;
int j;
int bestScore = 0;
/* Check to see if there is an expression that uses only the
** ROWID field of this table. For terms of the form ROWID==expr
** set iDirectEq[i] to the index of the term. For terms of the
** form ROWID<expr or ROWID<=expr set iDirectLt[i] to the term index.
** For terms like ROWID>expr or ROWID>=expr set iDirectGt[i].
**
** (Added:) Treat ROWID IN expr like ROWID=expr.
*/
iDirectEq[i] = -1;
iDirectLt[i] = -1;
iDirectGt[i] = -1;
for(j=0; j<nExpr; j++){
case TK_IN:
case TK_LE:
case TK_GE:
}
}
case TK_LE:
case TK_GE:
}
}
}
if( iDirectEq[i]>=0 ){
continue;
}
/* Do a search for usable indices. Leave pBestIdx pointing to
** the "best" index. pBestIdx is left set to NULL if no indices
** are usable.
**
** The best index is determined as follows. For each of the
** left-most terms that is fixed by an equality operator, add
** 8 to the score. The right-most term of the index may be
** constrained by an inequality. Add 1 if for an "x<..." constraint
** and add 2 for an "x>..." constraint. Chose the index that
** gives the best score.
**
** This scoring system is designed so that the score can later be
** used to determine how the index is used. If the score&7 is 0
** then all constraints are equalities. If score&1 is not 0 then
** there is an inequality used as a termination key. (ex: "x<...")
** If score&2 is not 0 then there is an inequality used as the
** start key. (ex: "x>..."). A score or 4 is the special case
** of an IN operator constraint. (ex: "x IN ...").
**
** The IN operator (as in "<expr> IN (...)") is treated the same as
** an equality comparison except that it can only be used on the
** left-most column of an index and other terms of the WHERE clause
** cannot be used in conjunction with the IN operator to help satisfy
** other columns of the index.
*/
for(j=0; j<nExpr; j++){
int k;
case TK_IN: {
if( k==0 ) inMask |= 1;
break;
}
case TK_EQ: {
eqMask |= 1<<k;
break;
}
case TK_LE:
case TK_LT: {
ltMask |= 1<<k;
break;
}
case TK_GE:
case TK_GT: {
gtMask |= 1<<k;
break;
}
default: {
/* CANT_HAPPEN */
assert( 0 );
break;
}
}
break;
}
}
}
int k;
case TK_EQ: {
eqMask |= 1<<k;
break;
}
case TK_LE:
case TK_LT: {
gtMask |= 1<<k;
break;
}
case TK_GE:
case TK_GT: {
ltMask |= 1<<k;
break;
}
default: {
/* CANT_HAPPEN */
assert( 0 );
break;
}
}
break;
}
}
}
}
/* The following loop ends with nEq set to the number of columns
** on the left of the index with == constraints.
*/
if( (m & eqMask)!=m ) break;
}
m = 1<<nEq;
}
}
if( pBestIdx ){
}
}
/* Check to see if the ORDER BY clause is or can be satisfied by the
** use of an index on the first table.
*/
int bRev = 0;
/* If there is already an IN index on the left-most table,
** it will not give the correct sort order.
** So, pretend that no suitable index is found.
*/
pSortIdx = 0;
/* If the left-most column is accessed using its ROWID, then do
** not try to sort by index.
*/
pSortIdx = 0;
}else{
}
if( pIdx==0 ){
}
*ppOrderBy = 0;
}
}
/* Open all tables in the pTabList and all indices used by those tables.
*/
}
}
/* Generate the code to do the search
*/
loopMask = 0;
int j, k;
/* 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.
*/
sqliteVdbeAddOp(v, OP_String, 0, 0);
}
/* Case 1: 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.
*/
k = iDirectEq[i];
}else{
}
}else{
}
haveKey = 0;
/* Case 2: There is an index and all terms of the WHERE clause that
** refer to the index use the "==" or "IN" operators.
*/
int start;
int testOp;
for(j=0; j<nColumn; j++){
for(k=0; k<nExpr; k++){
if( pX==0 ) continue;
){
break;
}
}else{
}
break;
}
}
){
break;
}
}
}
sqliteAddIdxKeyType(v, pIdx);
}else{
sqliteVdbeAddOp(v, OP_Dup, 0, 0);
sqliteVdbeAddOp(v, OP_IncrKey, 0, 0);
}
/* Scan in reverse order */
sqliteVdbeAddOp(v, OP_IncrKey, 0, 0);
}else{
/* Scan in the forward order */
}
haveKey = 1;
}else{
haveKey = 0;
}
/* Case 3: We have an inequality comparison against the ROWID field.
*/
int start;
if( iDirectGt[i]>=0 ){
k = iDirectGt[i];
}else{
}
}else{
}
if( iDirectLt[i]>=0 ){
k = iDirectLt[i];
}else{
}
/* sqliteVdbeAddOp(v, OP_MustBeInt, 0, sqliteVdbeCurrentAddr(v)+1); */
}else{
}
}
start = sqliteVdbeCurrentAddr(v);
}
haveKey = 0;
}else if( pIdx==0 ){
/* Case 4: There is no usable index. We must do a complete
** scan of the entire database table.
*/
int start;
start = sqliteVdbeCurrentAddr(v);
haveKey = 0;
}else{
/* Case 5: The WHERE clause term that refers to the right-most
** column of the index is an inequality. For example, if
** the index is on (x,y,z) and the WHERE clause is of the
** form "x=5 AND y<10" then this case is used. Only the
** right-most column can be an inequality - the rest must
** use the "==" operator.
**
** 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.
*/
int start;
int testOp;
/* Evaluate the equality constraints
*/
for(j=0; j<nEqColumn; j++){
for(k=0; k<nExpr; k++){
if( aExpr[k].p==0 ) continue;
){
break;
}
){
break;
}
}
}
/* Duplicate the equality term values because they will all be
** used twice: once to make the termination key and once to make the
** start key.
*/
for(j=0; j<nEqColumn; j++){
}
/* Labels for the beginning and end of the loop
*/
/* Generate the termination key. This is the key value that
** will end the search. There is no termination key if there
** are no equality terms and no "X<..." term.
**
** 2002-Dec-04: On a reverse-order scan, the so-called "termination"
** key computed here really ends up being the start key.
*/
if( (score & 1)!=0 ){
for(k=0; k<nExpr; k++){
if( pExpr==0 ) continue;
){
break;
}
){
break;
}
}
}else{
leFlag = 1;
}
sqliteAddIdxKeyType(v, pIdx);
if( leFlag ){
sqliteVdbeAddOp(v, OP_IncrKey, 0, 0);
}
}else{
}
}
/* Generate the start key. This is the key that defines the lower
** bound on the search. There is no start key if there are no
** equality terms and if there is no "X>..." term. In
** that case, generate a "Rewind" instruction in place of the
** start key search.
**
** 2002-Dec-04: In the case of a reverse-order search, the so-called
** "start" key really ends up being used as the termination key.
*/
if( (score & 2)!=0 ){
for(k=0; k<nExpr; k++){
if( pExpr==0 ) continue;
){
break;
}
){
break;
}
}
}else{
geFlag = 1;
}
sqliteAddIdxKeyType(v, pIdx);
if( !geFlag ){
sqliteVdbeAddOp(v, OP_IncrKey, 0, 0);
}
}else{
}
}else{
}
/* Generate the the top of the loop. If there is a termination
** key we have to test for that key and abort at the top of the
** loop.
*/
start = sqliteVdbeCurrentAddr(v);
}
haveKey = 1;
}else{
haveKey = 0;
}
/* Record the instruction used to terminate the loop.
*/
}
/* Insert code to test every subexpression that can be completely
** computed using the current set of tables.
*/
for(j=0; j<nExpr; j++){
if( aExpr[j].p==0 ) continue;
continue;
}
if( haveKey ){
haveKey = 0;
}
aExpr[j].p = 0;
}
/* 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.
*/
for(j=0; j<nExpr; j++){
if( aExpr[j].p==0 ) continue;
if( haveKey ){
/* Cannot happen. "haveKey" can only be true if pushKey is true
** an pushKey can only be true for DELETE and UPDATE and there are
** no outer joins with DELETE and UPDATE.
*/
haveKey = 0;
}
aExpr[j].p = 0;
}
}
}
}
return pWInfo;
}
/*
** Generate the end of the WHERE loop. See comments on
** sqliteWhereBegin() for additional information.
*/
int i;
}
}
int addr;
}
}
}
}
}
#if 0 /* Never reuse a cursor */
}
#endif
return;
}