Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(248)

Side by Side Diff: third_party/sqlite/src/select.c

Issue 3108030: Move bundled copy of sqlite one level deeper to better separate it... (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/src/
Patch Set: Created 10 years, 4 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
« no previous file with comments | « third_party/sqlite/src/rowset.c ('k') | third_party/sqlite/src/shell.c » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
(Empty)
1 /*
2 ** 2001 September 15
3 **
4 ** The author disclaims copyright to this source code. In place of
5 ** a legal notice, here is a blessing:
6 **
7 ** May you do good and not evil.
8 ** May you find forgiveness for yourself and forgive others.
9 ** May you share freely, never taking more than you give.
10 **
11 *************************************************************************
12 ** This file contains C code routines that are called by the parser
13 ** to handle SELECT statements in SQLite.
14 **
15 ** $Id: select.c,v 1.526 2009/08/01 15:09:58 drh Exp $
16 */
17 #include "sqliteInt.h"
18
19
20 /*
21 ** Delete all the content of a Select structure but do not deallocate
22 ** the select structure itself.
23 */
24 static void clearSelect(sqlite3 *db, Select *p){
25 sqlite3ExprListDelete(db, p->pEList);
26 sqlite3SrcListDelete(db, p->pSrc);
27 sqlite3ExprDelete(db, p->pWhere);
28 sqlite3ExprListDelete(db, p->pGroupBy);
29 sqlite3ExprDelete(db, p->pHaving);
30 sqlite3ExprListDelete(db, p->pOrderBy);
31 sqlite3SelectDelete(db, p->pPrior);
32 sqlite3ExprDelete(db, p->pLimit);
33 sqlite3ExprDelete(db, p->pOffset);
34 }
35
36 /*
37 ** Initialize a SelectDest structure.
38 */
39 void sqlite3SelectDestInit(SelectDest *pDest, int eDest, int iParm){
40 pDest->eDest = (u8)eDest;
41 pDest->iParm = iParm;
42 pDest->affinity = 0;
43 pDest->iMem = 0;
44 pDest->nMem = 0;
45 }
46
47
48 /*
49 ** Allocate a new Select structure and return a pointer to that
50 ** structure.
51 */
52 Select *sqlite3SelectNew(
53 Parse *pParse, /* Parsing context */
54 ExprList *pEList, /* which columns to include in the result */
55 SrcList *pSrc, /* the FROM clause -- which tables to scan */
56 Expr *pWhere, /* the WHERE clause */
57 ExprList *pGroupBy, /* the GROUP BY clause */
58 Expr *pHaving, /* the HAVING clause */
59 ExprList *pOrderBy, /* the ORDER BY clause */
60 int isDistinct, /* true if the DISTINCT keyword is present */
61 Expr *pLimit, /* LIMIT value. NULL means not used */
62 Expr *pOffset /* OFFSET value. NULL means no offset */
63 ){
64 Select *pNew;
65 Select standin;
66 sqlite3 *db = pParse->db;
67 pNew = sqlite3DbMallocZero(db, sizeof(*pNew) );
68 assert( db->mallocFailed || !pOffset || pLimit ); /* OFFSET implies LIMIT */
69 if( pNew==0 ){
70 pNew = &standin;
71 memset(pNew, 0, sizeof(*pNew));
72 }
73 if( pEList==0 ){
74 pEList = sqlite3ExprListAppend(pParse, 0, sqlite3Expr(db,TK_ALL,0));
75 }
76 pNew->pEList = pEList;
77 pNew->pSrc = pSrc;
78 pNew->pWhere = pWhere;
79 pNew->pGroupBy = pGroupBy;
80 pNew->pHaving = pHaving;
81 pNew->pOrderBy = pOrderBy;
82 pNew->selFlags = isDistinct ? SF_Distinct : 0;
83 pNew->op = TK_SELECT;
84 pNew->pLimit = pLimit;
85 pNew->pOffset = pOffset;
86 assert( pOffset==0 || pLimit!=0 );
87 pNew->addrOpenEphm[0] = -1;
88 pNew->addrOpenEphm[1] = -1;
89 pNew->addrOpenEphm[2] = -1;
90 if( db->mallocFailed ) {
91 clearSelect(db, pNew);
92 if( pNew!=&standin ) sqlite3DbFree(db, pNew);
93 pNew = 0;
94 }
95 return pNew;
96 }
97
98 /*
99 ** Delete the given Select structure and all of its substructures.
100 */
101 void sqlite3SelectDelete(sqlite3 *db, Select *p){
102 if( p ){
103 clearSelect(db, p);
104 sqlite3DbFree(db, p);
105 }
106 }
107
108 /*
109 ** Given 1 to 3 identifiers preceeding the JOIN keyword, determine the
110 ** type of join. Return an integer constant that expresses that type
111 ** in terms of the following bit values:
112 **
113 ** JT_INNER
114 ** JT_CROSS
115 ** JT_OUTER
116 ** JT_NATURAL
117 ** JT_LEFT
118 ** JT_RIGHT
119 **
120 ** A full outer join is the combination of JT_LEFT and JT_RIGHT.
121 **
122 ** If an illegal or unsupported join type is seen, then still return
123 ** a join type, but put an error in the pParse structure.
124 */
125 int sqlite3JoinType(Parse *pParse, Token *pA, Token *pB, Token *pC){
126 int jointype = 0;
127 Token *apAll[3];
128 Token *p;
129 /* 0123456789 123456789 123456789 123 */
130 static const char zKeyText[] = "naturaleftouterightfullinnercross";
131 static const struct {
132 u8 i; /* Beginning of keyword text in zKeyText[] */
133 u8 nChar; /* Length of the keyword in characters */
134 u8 code; /* Join type mask */
135 } aKeyword[] = {
136 /* natural */ { 0, 7, JT_NATURAL },
137 /* left */ { 6, 4, JT_LEFT|JT_OUTER },
138 /* outer */ { 10, 5, JT_OUTER },
139 /* right */ { 14, 5, JT_RIGHT|JT_OUTER },
140 /* full */ { 19, 4, JT_LEFT|JT_RIGHT|JT_OUTER },
141 /* inner */ { 23, 5, JT_INNER },
142 /* cross */ { 28, 5, JT_INNER|JT_CROSS },
143 };
144 int i, j;
145 apAll[0] = pA;
146 apAll[1] = pB;
147 apAll[2] = pC;
148 for(i=0; i<3 && apAll[i]; i++){
149 p = apAll[i];
150 for(j=0; j<ArraySize(aKeyword); j++){
151 if( p->n==aKeyword[j].nChar
152 && sqlite3StrNICmp((char*)p->z, &zKeyText[aKeyword[j].i], p->n)==0 ){
153 jointype |= aKeyword[j].code;
154 break;
155 }
156 }
157 testcase( j==0 || j==1 || j==2 || j==3 || j==4 || j==5 || j==6 );
158 if( j>=ArraySize(aKeyword) ){
159 jointype |= JT_ERROR;
160 break;
161 }
162 }
163 if(
164 (jointype & (JT_INNER|JT_OUTER))==(JT_INNER|JT_OUTER) ||
165 (jointype & JT_ERROR)!=0
166 ){
167 const char *zSp = " ";
168 assert( pB!=0 );
169 if( pC==0 ){ zSp++; }
170 sqlite3ErrorMsg(pParse, "unknown or unsupported join type: "
171 "%T %T%s%T", pA, pB, zSp, pC);
172 jointype = JT_INNER;
173 }else if( (jointype & JT_OUTER)!=0
174 && (jointype & (JT_LEFT|JT_RIGHT))!=JT_LEFT ){
175 sqlite3ErrorMsg(pParse,
176 "RIGHT and FULL OUTER JOINs are not currently supported");
177 jointype = JT_INNER;
178 }
179 return jointype;
180 }
181
182 /*
183 ** Return the index of a column in a table. Return -1 if the column
184 ** is not contained in the table.
185 */
186 static int columnIndex(Table *pTab, const char *zCol){
187 int i;
188 for(i=0; i<pTab->nCol; i++){
189 if( sqlite3StrICmp(pTab->aCol[i].zName, zCol)==0 ) return i;
190 }
191 return -1;
192 }
193
194 /*
195 ** Create an expression node for an identifier with the name of zName
196 */
197 Expr *sqlite3CreateIdExpr(Parse *pParse, const char *zName){
198 return sqlite3Expr(pParse->db, TK_ID, zName);
199 }
200
201 /*
202 ** Add a term to the WHERE expression in *ppExpr that requires the
203 ** zCol column to be equal in the two tables pTab1 and pTab2.
204 */
205 static void addWhereTerm(
206 Parse *pParse, /* Parsing context */
207 const char *zCol, /* Name of the column */
208 const Table *pTab1, /* First table */
209 const char *zAlias1, /* Alias for first table. May be NULL */
210 const Table *pTab2, /* Second table */
211 const char *zAlias2, /* Alias for second table. May be NULL */
212 int iRightJoinTable, /* VDBE cursor for the right table */
213 Expr **ppExpr, /* Add the equality term to this expression */
214 int isOuterJoin /* True if dealing with an OUTER join */
215 ){
216 Expr *pE1a, *pE1b, *pE1c;
217 Expr *pE2a, *pE2b, *pE2c;
218 Expr *pE;
219
220 pE1a = sqlite3CreateIdExpr(pParse, zCol);
221 pE2a = sqlite3CreateIdExpr(pParse, zCol);
222 if( zAlias1==0 ){
223 zAlias1 = pTab1->zName;
224 }
225 pE1b = sqlite3CreateIdExpr(pParse, zAlias1);
226 if( zAlias2==0 ){
227 zAlias2 = pTab2->zName;
228 }
229 pE2b = sqlite3CreateIdExpr(pParse, zAlias2);
230 pE1c = sqlite3PExpr(pParse, TK_DOT, pE1b, pE1a, 0);
231 pE2c = sqlite3PExpr(pParse, TK_DOT, pE2b, pE2a, 0);
232 pE = sqlite3PExpr(pParse, TK_EQ, pE1c, pE2c, 0);
233 if( pE && isOuterJoin ){
234 ExprSetProperty(pE, EP_FromJoin);
235 assert( !ExprHasAnyProperty(pE, EP_TokenOnly|EP_Reduced) );
236 ExprSetIrreducible(pE);
237 pE->iRightJoinTable = (i16)iRightJoinTable;
238 }
239 *ppExpr = sqlite3ExprAnd(pParse->db,*ppExpr, pE);
240 }
241
242 /*
243 ** Set the EP_FromJoin property on all terms of the given expression.
244 ** And set the Expr.iRightJoinTable to iTable for every term in the
245 ** expression.
246 **
247 ** The EP_FromJoin property is used on terms of an expression to tell
248 ** the LEFT OUTER JOIN processing logic that this term is part of the
249 ** join restriction specified in the ON or USING clause and not a part
250 ** of the more general WHERE clause. These terms are moved over to the
251 ** WHERE clause during join processing but we need to remember that they
252 ** originated in the ON or USING clause.
253 **
254 ** The Expr.iRightJoinTable tells the WHERE clause processing that the
255 ** expression depends on table iRightJoinTable even if that table is not
256 ** explicitly mentioned in the expression. That information is needed
257 ** for cases like this:
258 **
259 ** SELECT * FROM t1 LEFT JOIN t2 ON t1.a=t2.b AND t1.x=5
260 **
261 ** The where clause needs to defer the handling of the t1.x=5
262 ** term until after the t2 loop of the join. In that way, a
263 ** NULL t2 row will be inserted whenever t1.x!=5. If we do not
264 ** defer the handling of t1.x=5, it will be processed immediately
265 ** after the t1 loop and rows with t1.x!=5 will never appear in
266 ** the output, which is incorrect.
267 */
268 static void setJoinExpr(Expr *p, int iTable){
269 while( p ){
270 ExprSetProperty(p, EP_FromJoin);
271 assert( !ExprHasAnyProperty(p, EP_TokenOnly|EP_Reduced) );
272 ExprSetIrreducible(p);
273 p->iRightJoinTable = (i16)iTable;
274 setJoinExpr(p->pLeft, iTable);
275 p = p->pRight;
276 }
277 }
278
279 /*
280 ** This routine processes the join information for a SELECT statement.
281 ** ON and USING clauses are converted into extra terms of the WHERE clause.
282 ** NATURAL joins also create extra WHERE clause terms.
283 **
284 ** The terms of a FROM clause are contained in the Select.pSrc structure.
285 ** The left most table is the first entry in Select.pSrc. The right-most
286 ** table is the last entry. The join operator is held in the entry to
287 ** the left. Thus entry 0 contains the join operator for the join between
288 ** entries 0 and 1. Any ON or USING clauses associated with the join are
289 ** also attached to the left entry.
290 **
291 ** This routine returns the number of errors encountered.
292 */
293 static int sqliteProcessJoin(Parse *pParse, Select *p){
294 SrcList *pSrc; /* All tables in the FROM clause */
295 int i, j; /* Loop counters */
296 struct SrcList_item *pLeft; /* Left table being joined */
297 struct SrcList_item *pRight; /* Right table being joined */
298
299 pSrc = p->pSrc;
300 pLeft = &pSrc->a[0];
301 pRight = &pLeft[1];
302 for(i=0; i<pSrc->nSrc-1; i++, pRight++, pLeft++){
303 Table *pLeftTab = pLeft->pTab;
304 Table *pRightTab = pRight->pTab;
305 int isOuter;
306
307 if( NEVER(pLeftTab==0 || pRightTab==0) ) continue;
308 isOuter = (pRight->jointype & JT_OUTER)!=0;
309
310 /* When the NATURAL keyword is present, add WHERE clause terms for
311 ** every column that the two tables have in common.
312 */
313 if( pRight->jointype & JT_NATURAL ){
314 if( pRight->pOn || pRight->pUsing ){
315 sqlite3ErrorMsg(pParse, "a NATURAL join may not have "
316 "an ON or USING clause", 0);
317 return 1;
318 }
319 for(j=0; j<pLeftTab->nCol; j++){
320 char *zName = pLeftTab->aCol[j].zName;
321 if( columnIndex(pRightTab, zName)>=0 ){
322 addWhereTerm(pParse, zName, pLeftTab, pLeft->zAlias,
323 pRightTab, pRight->zAlias,
324 pRight->iCursor, &p->pWhere, isOuter);
325
326 }
327 }
328 }
329
330 /* Disallow both ON and USING clauses in the same join
331 */
332 if( pRight->pOn && pRight->pUsing ){
333 sqlite3ErrorMsg(pParse, "cannot have both ON and USING "
334 "clauses in the same join");
335 return 1;
336 }
337
338 /* Add the ON clause to the end of the WHERE clause, connected by
339 ** an AND operator.
340 */
341 if( pRight->pOn ){
342 if( isOuter ) setJoinExpr(pRight->pOn, pRight->iCursor);
343 p->pWhere = sqlite3ExprAnd(pParse->db, p->pWhere, pRight->pOn);
344 pRight->pOn = 0;
345 }
346
347 /* Create extra terms on the WHERE clause for each column named
348 ** in the USING clause. Example: If the two tables to be joined are
349 ** A and B and the USING clause names X, Y, and Z, then add this
350 ** to the WHERE clause: A.X=B.X AND A.Y=B.Y AND A.Z=B.Z
351 ** Report an error if any column mentioned in the USING clause is
352 ** not contained in both tables to be joined.
353 */
354 if( pRight->pUsing ){
355 IdList *pList = pRight->pUsing;
356 for(j=0; j<pList->nId; j++){
357 char *zName = pList->a[j].zName;
358 if( columnIndex(pLeftTab, zName)<0 || columnIndex(pRightTab, zName)<0 ){
359 sqlite3ErrorMsg(pParse, "cannot join using column %s - column "
360 "not present in both tables", zName);
361 return 1;
362 }
363 addWhereTerm(pParse, zName, pLeftTab, pLeft->zAlias,
364 pRightTab, pRight->zAlias,
365 pRight->iCursor, &p->pWhere, isOuter);
366 }
367 }
368 }
369 return 0;
370 }
371
372 /*
373 ** Insert code into "v" that will push the record on the top of the
374 ** stack into the sorter.
375 */
376 static void pushOntoSorter(
377 Parse *pParse, /* Parser context */
378 ExprList *pOrderBy, /* The ORDER BY clause */
379 Select *pSelect, /* The whole SELECT statement */
380 int regData /* Register holding data to be sorted */
381 ){
382 Vdbe *v = pParse->pVdbe;
383 int nExpr = pOrderBy->nExpr;
384 int regBase = sqlite3GetTempRange(pParse, nExpr+2);
385 int regRecord = sqlite3GetTempReg(pParse);
386 sqlite3ExprCacheClear(pParse);
387 sqlite3ExprCodeExprList(pParse, pOrderBy, regBase, 0);
388 sqlite3VdbeAddOp2(v, OP_Sequence, pOrderBy->iECursor, regBase+nExpr);
389 sqlite3ExprCodeMove(pParse, regData, regBase+nExpr+1, 1);
390 sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nExpr + 2, regRecord);
391 sqlite3VdbeAddOp2(v, OP_IdxInsert, pOrderBy->iECursor, regRecord);
392 sqlite3ReleaseTempReg(pParse, regRecord);
393 sqlite3ReleaseTempRange(pParse, regBase, nExpr+2);
394 if( pSelect->iLimit ){
395 int addr1, addr2;
396 int iLimit;
397 if( pSelect->iOffset ){
398 iLimit = pSelect->iOffset+1;
399 }else{
400 iLimit = pSelect->iLimit;
401 }
402 addr1 = sqlite3VdbeAddOp1(v, OP_IfZero, iLimit);
403 sqlite3VdbeAddOp2(v, OP_AddImm, iLimit, -1);
404 addr2 = sqlite3VdbeAddOp0(v, OP_Goto);
405 sqlite3VdbeJumpHere(v, addr1);
406 sqlite3VdbeAddOp1(v, OP_Last, pOrderBy->iECursor);
407 sqlite3VdbeAddOp1(v, OP_Delete, pOrderBy->iECursor);
408 sqlite3VdbeJumpHere(v, addr2);
409 pSelect->iLimit = 0;
410 }
411 }
412
413 /*
414 ** Add code to implement the OFFSET
415 */
416 static void codeOffset(
417 Vdbe *v, /* Generate code into this VM */
418 Select *p, /* The SELECT statement being coded */
419 int iContinue /* Jump here to skip the current record */
420 ){
421 if( p->iOffset && iContinue!=0 ){
422 int addr;
423 sqlite3VdbeAddOp2(v, OP_AddImm, p->iOffset, -1);
424 addr = sqlite3VdbeAddOp1(v, OP_IfNeg, p->iOffset);
425 sqlite3VdbeAddOp2(v, OP_Goto, 0, iContinue);
426 VdbeComment((v, "skip OFFSET records"));
427 sqlite3VdbeJumpHere(v, addr);
428 }
429 }
430
431 /*
432 ** Add code that will check to make sure the N registers starting at iMem
433 ** form a distinct entry. iTab is a sorting index that holds previously
434 ** seen combinations of the N values. A new entry is made in iTab
435 ** if the current N values are new.
436 **
437 ** A jump to addrRepeat is made and the N+1 values are popped from the
438 ** stack if the top N elements are not distinct.
439 */
440 static void codeDistinct(
441 Parse *pParse, /* Parsing and code generating context */
442 int iTab, /* A sorting index used to test for distinctness */
443 int addrRepeat, /* Jump to here if not distinct */
444 int N, /* Number of elements */
445 int iMem /* First element */
446 ){
447 Vdbe *v;
448 int r1;
449
450 v = pParse->pVdbe;
451 r1 = sqlite3GetTempReg(pParse);
452 sqlite3VdbeAddOp3(v, OP_MakeRecord, iMem, N, r1);
453 sqlite3VdbeAddOp3(v, OP_Found, iTab, addrRepeat, r1);
454 sqlite3VdbeAddOp2(v, OP_IdxInsert, iTab, r1);
455 sqlite3ReleaseTempReg(pParse, r1);
456 }
457
458 /*
459 ** Generate an error message when a SELECT is used within a subexpression
460 ** (example: "a IN (SELECT * FROM table)") but it has more than 1 result
461 ** column. We do this in a subroutine because the error occurs in multiple
462 ** places.
463 */
464 static int checkForMultiColumnSelectError(
465 Parse *pParse, /* Parse context. */
466 SelectDest *pDest, /* Destination of SELECT results */
467 int nExpr /* Number of result columns returned by SELECT */
468 ){
469 int eDest = pDest->eDest;
470 if( nExpr>1 && (eDest==SRT_Mem || eDest==SRT_Set) ){
471 sqlite3ErrorMsg(pParse, "only a single result allowed for "
472 "a SELECT that is part of an expression");
473 return 1;
474 }else{
475 return 0;
476 }
477 }
478
479 /*
480 ** This routine generates the code for the inside of the inner loop
481 ** of a SELECT.
482 **
483 ** If srcTab and nColumn are both zero, then the pEList expressions
484 ** are evaluated in order to get the data for this row. If nColumn>0
485 ** then data is pulled from srcTab and pEList is used only to get the
486 ** datatypes for each column.
487 */
488 static void selectInnerLoop(
489 Parse *pParse, /* The parser context */
490 Select *p, /* The complete select statement being coded */
491 ExprList *pEList, /* List of values being extracted */
492 int srcTab, /* Pull data from this table */
493 int nColumn, /* Number of columns in the source table */
494 ExprList *pOrderBy, /* If not NULL, sort results using this key */
495 int distinct, /* If >=0, make sure results are distinct */
496 SelectDest *pDest, /* How to dispose of the results */
497 int iContinue, /* Jump here to continue with next row */
498 int iBreak /* Jump here to break out of the inner loop */
499 ){
500 Vdbe *v = pParse->pVdbe;
501 int i;
502 int hasDistinct; /* True if the DISTINCT keyword is present */
503 int regResult; /* Start of memory holding result set */
504 int eDest = pDest->eDest; /* How to dispose of results */
505 int iParm = pDest->iParm; /* First argument to disposal method */
506 int nResultCol; /* Number of result columns */
507
508 assert( v );
509 if( NEVER(v==0) ) return;
510 assert( pEList!=0 );
511 hasDistinct = distinct>=0;
512 if( pOrderBy==0 && !hasDistinct ){
513 codeOffset(v, p, iContinue);
514 }
515
516 /* Pull the requested columns.
517 */
518 if( nColumn>0 ){
519 nResultCol = nColumn;
520 }else{
521 nResultCol = pEList->nExpr;
522 }
523 if( pDest->iMem==0 ){
524 pDest->iMem = pParse->nMem+1;
525 pDest->nMem = nResultCol;
526 pParse->nMem += nResultCol;
527 }else{
528 assert( pDest->nMem==nResultCol );
529 }
530 regResult = pDest->iMem;
531 if( nColumn>0 ){
532 for(i=0; i<nColumn; i++){
533 sqlite3VdbeAddOp3(v, OP_Column, srcTab, i, regResult+i);
534 }
535 }else if( eDest!=SRT_Exists ){
536 /* If the destination is an EXISTS(...) expression, the actual
537 ** values returned by the SELECT are not required.
538 */
539 sqlite3ExprCacheClear(pParse);
540 sqlite3ExprCodeExprList(pParse, pEList, regResult, eDest==SRT_Output);
541 }
542 nColumn = nResultCol;
543
544 /* If the DISTINCT keyword was present on the SELECT statement
545 ** and this row has been seen before, then do not make this row
546 ** part of the result.
547 */
548 if( hasDistinct ){
549 assert( pEList!=0 );
550 assert( pEList->nExpr==nColumn );
551 codeDistinct(pParse, distinct, iContinue, nColumn, regResult);
552 if( pOrderBy==0 ){
553 codeOffset(v, p, iContinue);
554 }
555 }
556
557 if( checkForMultiColumnSelectError(pParse, pDest, pEList->nExpr) ){
558 return;
559 }
560
561 switch( eDest ){
562 /* In this mode, write each query result to the key of the temporary
563 ** table iParm.
564 */
565 #ifndef SQLITE_OMIT_COMPOUND_SELECT
566 case SRT_Union: {
567 int r1;
568 r1 = sqlite3GetTempReg(pParse);
569 sqlite3VdbeAddOp3(v, OP_MakeRecord, regResult, nColumn, r1);
570 sqlite3VdbeAddOp2(v, OP_IdxInsert, iParm, r1);
571 sqlite3ReleaseTempReg(pParse, r1);
572 break;
573 }
574
575 /* Construct a record from the query result, but instead of
576 ** saving that record, use it as a key to delete elements from
577 ** the temporary table iParm.
578 */
579 case SRT_Except: {
580 sqlite3VdbeAddOp3(v, OP_IdxDelete, iParm, regResult, nColumn);
581 break;
582 }
583 #endif
584
585 /* Store the result as data using a unique key.
586 */
587 case SRT_Table:
588 case SRT_EphemTab: {
589 int r1 = sqlite3GetTempReg(pParse);
590 testcase( eDest==SRT_Table );
591 testcase( eDest==SRT_EphemTab );
592 sqlite3VdbeAddOp3(v, OP_MakeRecord, regResult, nColumn, r1);
593 if( pOrderBy ){
594 pushOntoSorter(pParse, pOrderBy, p, r1);
595 }else{
596 int r2 = sqlite3GetTempReg(pParse);
597 sqlite3VdbeAddOp2(v, OP_NewRowid, iParm, r2);
598 sqlite3VdbeAddOp3(v, OP_Insert, iParm, r1, r2);
599 sqlite3VdbeChangeP5(v, OPFLAG_APPEND);
600 sqlite3ReleaseTempReg(pParse, r2);
601 }
602 sqlite3ReleaseTempReg(pParse, r1);
603 break;
604 }
605
606 #ifndef SQLITE_OMIT_SUBQUERY
607 /* If we are creating a set for an "expr IN (SELECT ...)" construct,
608 ** then there should be a single item on the stack. Write this
609 ** item into the set table with bogus data.
610 */
611 case SRT_Set: {
612 assert( nColumn==1 );
613 p->affinity = sqlite3CompareAffinity(pEList->a[0].pExpr, pDest->affinity);
614 if( pOrderBy ){
615 /* At first glance you would think we could optimize out the
616 ** ORDER BY in this case since the order of entries in the set
617 ** does not matter. But there might be a LIMIT clause, in which
618 ** case the order does matter */
619 pushOntoSorter(pParse, pOrderBy, p, regResult);
620 }else{
621 int r1 = sqlite3GetTempReg(pParse);
622 sqlite3VdbeAddOp4(v, OP_MakeRecord, regResult, 1, r1, &p->affinity, 1);
623 sqlite3ExprCacheAffinityChange(pParse, regResult, 1);
624 sqlite3VdbeAddOp2(v, OP_IdxInsert, iParm, r1);
625 sqlite3ReleaseTempReg(pParse, r1);
626 }
627 break;
628 }
629
630 /* If any row exist in the result set, record that fact and abort.
631 */
632 case SRT_Exists: {
633 sqlite3VdbeAddOp2(v, OP_Integer, 1, iParm);
634 /* The LIMIT clause will terminate the loop for us */
635 break;
636 }
637
638 /* If this is a scalar select that is part of an expression, then
639 ** store the results in the appropriate memory cell and break out
640 ** of the scan loop.
641 */
642 case SRT_Mem: {
643 assert( nColumn==1 );
644 if( pOrderBy ){
645 pushOntoSorter(pParse, pOrderBy, p, regResult);
646 }else{
647 sqlite3ExprCodeMove(pParse, regResult, iParm, 1);
648 /* The LIMIT clause will jump out of the loop for us */
649 }
650 break;
651 }
652 #endif /* #ifndef SQLITE_OMIT_SUBQUERY */
653
654 /* Send the data to the callback function or to a subroutine. In the
655 ** case of a subroutine, the subroutine itself is responsible for
656 ** popping the data from the stack.
657 */
658 case SRT_Coroutine:
659 case SRT_Output: {
660 testcase( eDest==SRT_Coroutine );
661 testcase( eDest==SRT_Output );
662 if( pOrderBy ){
663 int r1 = sqlite3GetTempReg(pParse);
664 sqlite3VdbeAddOp3(v, OP_MakeRecord, regResult, nColumn, r1);
665 pushOntoSorter(pParse, pOrderBy, p, r1);
666 sqlite3ReleaseTempReg(pParse, r1);
667 }else if( eDest==SRT_Coroutine ){
668 sqlite3VdbeAddOp1(v, OP_Yield, pDest->iParm);
669 }else{
670 sqlite3VdbeAddOp2(v, OP_ResultRow, regResult, nColumn);
671 sqlite3ExprCacheAffinityChange(pParse, regResult, nColumn);
672 }
673 break;
674 }
675
676 #if !defined(SQLITE_OMIT_TRIGGER)
677 /* Discard the results. This is used for SELECT statements inside
678 ** the body of a TRIGGER. The purpose of such selects is to call
679 ** user-defined functions that have side effects. We do not care
680 ** about the actual results of the select.
681 */
682 default: {
683 assert( eDest==SRT_Discard );
684 break;
685 }
686 #endif
687 }
688
689 /* Jump to the end of the loop if the LIMIT is reached.
690 */
691 if( p->iLimit ){
692 assert( pOrderBy==0 ); /* If there is an ORDER BY, the call to
693 ** pushOntoSorter() would have cleared p->iLimit */
694 sqlite3VdbeAddOp2(v, OP_AddImm, p->iLimit, -1);
695 sqlite3VdbeAddOp2(v, OP_IfZero, p->iLimit, iBreak);
696 }
697 }
698
699 /*
700 ** Given an expression list, generate a KeyInfo structure that records
701 ** the collating sequence for each expression in that expression list.
702 **
703 ** If the ExprList is an ORDER BY or GROUP BY clause then the resulting
704 ** KeyInfo structure is appropriate for initializing a virtual index to
705 ** implement that clause. If the ExprList is the result set of a SELECT
706 ** then the KeyInfo structure is appropriate for initializing a virtual
707 ** index to implement a DISTINCT test.
708 **
709 ** Space to hold the KeyInfo structure is obtain from malloc. The calling
710 ** function is responsible for seeing that this structure is eventually
711 ** freed. Add the KeyInfo structure to the P4 field of an opcode using
712 ** P4_KEYINFO_HANDOFF is the usual way of dealing with this.
713 */
714 static KeyInfo *keyInfoFromExprList(Parse *pParse, ExprList *pList){
715 sqlite3 *db = pParse->db;
716 int nExpr;
717 KeyInfo *pInfo;
718 struct ExprList_item *pItem;
719 int i;
720
721 nExpr = pList->nExpr;
722 pInfo = sqlite3DbMallocZero(db, sizeof(*pInfo) + nExpr*(sizeof(CollSeq*)+1) );
723 if( pInfo ){
724 pInfo->aSortOrder = (u8*)&pInfo->aColl[nExpr];
725 pInfo->nField = (u16)nExpr;
726 pInfo->enc = ENC(db);
727 pInfo->db = db;
728 for(i=0, pItem=pList->a; i<nExpr; i++, pItem++){
729 CollSeq *pColl;
730 pColl = sqlite3ExprCollSeq(pParse, pItem->pExpr);
731 if( !pColl ){
732 pColl = db->pDfltColl;
733 }
734 pInfo->aColl[i] = pColl;
735 pInfo->aSortOrder[i] = pItem->sortOrder;
736 }
737 }
738 return pInfo;
739 }
740
741
742 /*
743 ** If the inner loop was generated using a non-null pOrderBy argument,
744 ** then the results were placed in a sorter. After the loop is terminated
745 ** we need to run the sorter and output the results. The following
746 ** routine generates the code needed to do that.
747 */
748 static void generateSortTail(
749 Parse *pParse, /* Parsing context */
750 Select *p, /* The SELECT statement */
751 Vdbe *v, /* Generate code into this VDBE */
752 int nColumn, /* Number of columns of data */
753 SelectDest *pDest /* Write the sorted results here */
754 ){
755 int addrBreak = sqlite3VdbeMakeLabel(v); /* Jump here to exit loop */
756 int addrContinue = sqlite3VdbeMakeLabel(v); /* Jump here for next cycle */
757 int addr;
758 int iTab;
759 int pseudoTab = 0;
760 ExprList *pOrderBy = p->pOrderBy;
761
762 int eDest = pDest->eDest;
763 int iParm = pDest->iParm;
764
765 int regRow;
766 int regRowid;
767
768 iTab = pOrderBy->iECursor;
769 regRow = sqlite3GetTempReg(pParse);
770 if( eDest==SRT_Output || eDest==SRT_Coroutine ){
771 pseudoTab = pParse->nTab++;
772 sqlite3VdbeAddOp3(v, OP_OpenPseudo, pseudoTab, regRow, nColumn);
773 regRowid = 0;
774 }else{
775 regRowid = sqlite3GetTempReg(pParse);
776 }
777 addr = 1 + sqlite3VdbeAddOp2(v, OP_Sort, iTab, addrBreak);
778 codeOffset(v, p, addrContinue);
779 sqlite3VdbeAddOp3(v, OP_Column, iTab, pOrderBy->nExpr + 1, regRow);
780 switch( eDest ){
781 case SRT_Table:
782 case SRT_EphemTab: {
783 testcase( eDest==SRT_Table );
784 testcase( eDest==SRT_EphemTab );
785 sqlite3VdbeAddOp2(v, OP_NewRowid, iParm, regRowid);
786 sqlite3VdbeAddOp3(v, OP_Insert, iParm, regRow, regRowid);
787 sqlite3VdbeChangeP5(v, OPFLAG_APPEND);
788 break;
789 }
790 #ifndef SQLITE_OMIT_SUBQUERY
791 case SRT_Set: {
792 assert( nColumn==1 );
793 sqlite3VdbeAddOp4(v, OP_MakeRecord, regRow, 1, regRowid, &p->affinity, 1);
794 sqlite3ExprCacheAffinityChange(pParse, regRow, 1);
795 sqlite3VdbeAddOp2(v, OP_IdxInsert, iParm, regRowid);
796 break;
797 }
798 case SRT_Mem: {
799 assert( nColumn==1 );
800 sqlite3ExprCodeMove(pParse, regRow, iParm, 1);
801 /* The LIMIT clause will terminate the loop for us */
802 break;
803 }
804 #endif
805 default: {
806 int i;
807 assert( eDest==SRT_Output || eDest==SRT_Coroutine );
808 testcase( eDest==SRT_Output );
809 testcase( eDest==SRT_Coroutine );
810 for(i=0; i<nColumn; i++){
811 assert( regRow!=pDest->iMem+i );
812 sqlite3VdbeAddOp3(v, OP_Column, pseudoTab, i, pDest->iMem+i);
813 if( i==0 ){
814 sqlite3VdbeChangeP5(v, OPFLAG_CLEARCACHE);
815 }
816 }
817 if( eDest==SRT_Output ){
818 sqlite3VdbeAddOp2(v, OP_ResultRow, pDest->iMem, nColumn);
819 sqlite3ExprCacheAffinityChange(pParse, pDest->iMem, nColumn);
820 }else{
821 sqlite3VdbeAddOp1(v, OP_Yield, pDest->iParm);
822 }
823 break;
824 }
825 }
826 sqlite3ReleaseTempReg(pParse, regRow);
827 sqlite3ReleaseTempReg(pParse, regRowid);
828
829 /* LIMIT has been implemented by the pushOntoSorter() routine.
830 */
831 assert( p->iLimit==0 );
832
833 /* The bottom of the loop
834 */
835 sqlite3VdbeResolveLabel(v, addrContinue);
836 sqlite3VdbeAddOp2(v, OP_Next, iTab, addr);
837 sqlite3VdbeResolveLabel(v, addrBreak);
838 if( eDest==SRT_Output || eDest==SRT_Coroutine ){
839 sqlite3VdbeAddOp2(v, OP_Close, pseudoTab, 0);
840 }
841 }
842
843 /*
844 ** Return a pointer to a string containing the 'declaration type' of the
845 ** expression pExpr. The string may be treated as static by the caller.
846 **
847 ** The declaration type is the exact datatype definition extracted from the
848 ** original CREATE TABLE statement if the expression is a column. The
849 ** declaration type for a ROWID field is INTEGER. Exactly when an expression
850 ** is considered a column can be complex in the presence of subqueries. The
851 ** result-set expression in all of the following SELECT statements is
852 ** considered a column by this function.
853 **
854 ** SELECT col FROM tbl;
855 ** SELECT (SELECT col FROM tbl;
856 ** SELECT (SELECT col FROM tbl);
857 ** SELECT abc FROM (SELECT col AS abc FROM tbl);
858 **
859 ** The declaration type for any expression other than a column is NULL.
860 */
861 static const char *columnType(
862 NameContext *pNC,
863 Expr *pExpr,
864 const char **pzOriginDb,
865 const char **pzOriginTab,
866 const char **pzOriginCol
867 ){
868 char const *zType = 0;
869 char const *zOriginDb = 0;
870 char const *zOriginTab = 0;
871 char const *zOriginCol = 0;
872 int j;
873 if( NEVER(pExpr==0) || pNC->pSrcList==0 ) return 0;
874
875 switch( pExpr->op ){
876 case TK_AGG_COLUMN:
877 case TK_COLUMN: {
878 /* The expression is a column. Locate the table the column is being
879 ** extracted from in NameContext.pSrcList. This table may be real
880 ** database table or a subquery.
881 */
882 Table *pTab = 0; /* Table structure column is extracted from */
883 Select *pS = 0; /* Select the column is extracted from */
884 int iCol = pExpr->iColumn; /* Index of column in pTab */
885 testcase( pExpr->op==TK_AGG_COLUMN );
886 testcase( pExpr->op==TK_COLUMN );
887 while( pNC && !pTab ){
888 SrcList *pTabList = pNC->pSrcList;
889 for(j=0;j<pTabList->nSrc && pTabList->a[j].iCursor!=pExpr->iTable;j++);
890 if( j<pTabList->nSrc ){
891 pTab = pTabList->a[j].pTab;
892 pS = pTabList->a[j].pSelect;
893 }else{
894 pNC = pNC->pNext;
895 }
896 }
897
898 if( pTab==0 ){
899 /* At one time, code such as "SELECT new.x" within a trigger would
900 ** cause this condition to run. Since then, we have restructured how
901 ** trigger code is generated and so this condition is no longer
902 ** possible. However, it can still be true for statements like
903 ** the following:
904 **
905 ** CREATE TABLE t1(col INTEGER);
906 ** SELECT (SELECT t1.col) FROM FROM t1;
907 **
908 ** when columnType() is called on the expression "t1.col" in the
909 ** sub-select. In this case, set the column type to NULL, even
910 ** though it should really be "INTEGER".
911 **
912 ** This is not a problem, as the column type of "t1.col" is never
913 ** used. When columnType() is called on the expression
914 ** "(SELECT t1.col)", the correct type is returned (see the TK_SELECT
915 ** branch below. */
916 break;
917 }
918
919 assert( pTab && pExpr->pTab==pTab );
920 if( pS ){
921 /* The "table" is actually a sub-select or a view in the FROM clause
922 ** of the SELECT statement. Return the declaration type and origin
923 ** data for the result-set column of the sub-select.
924 */
925 if( ALWAYS(iCol>=0 && iCol<pS->pEList->nExpr) ){
926 /* If iCol is less than zero, then the expression requests the
927 ** rowid of the sub-select or view. This expression is legal (see
928 ** test case misc2.2.2) - it always evaluates to NULL.
929 */
930 NameContext sNC;
931 Expr *p = pS->pEList->a[iCol].pExpr;
932 sNC.pSrcList = pS->pSrc;
933 sNC.pNext = pNC;
934 sNC.pParse = pNC->pParse;
935 zType = columnType(&sNC, p, &zOriginDb, &zOriginTab, &zOriginCol);
936 }
937 }else if( ALWAYS(pTab->pSchema) ){
938 /* A real table */
939 assert( !pS );
940 if( iCol<0 ) iCol = pTab->iPKey;
941 assert( iCol==-1 || (iCol>=0 && iCol<pTab->nCol) );
942 if( iCol<0 ){
943 zType = "INTEGER";
944 zOriginCol = "rowid";
945 }else{
946 zType = pTab->aCol[iCol].zType;
947 zOriginCol = pTab->aCol[iCol].zName;
948 }
949 zOriginTab = pTab->zName;
950 if( pNC->pParse ){
951 int iDb = sqlite3SchemaToIndex(pNC->pParse->db, pTab->pSchema);
952 zOriginDb = pNC->pParse->db->aDb[iDb].zName;
953 }
954 }
955 break;
956 }
957 #ifndef SQLITE_OMIT_SUBQUERY
958 case TK_SELECT: {
959 /* The expression is a sub-select. Return the declaration type and
960 ** origin info for the single column in the result set of the SELECT
961 ** statement.
962 */
963 NameContext sNC;
964 Select *pS = pExpr->x.pSelect;
965 Expr *p = pS->pEList->a[0].pExpr;
966 assert( ExprHasProperty(pExpr, EP_xIsSelect) );
967 sNC.pSrcList = pS->pSrc;
968 sNC.pNext = pNC;
969 sNC.pParse = pNC->pParse;
970 zType = columnType(&sNC, p, &zOriginDb, &zOriginTab, &zOriginCol);
971 break;
972 }
973 #endif
974 }
975
976 if( pzOriginDb ){
977 assert( pzOriginTab && pzOriginCol );
978 *pzOriginDb = zOriginDb;
979 *pzOriginTab = zOriginTab;
980 *pzOriginCol = zOriginCol;
981 }
982 return zType;
983 }
984
985 /*
986 ** Generate code that will tell the VDBE the declaration types of columns
987 ** in the result set.
988 */
989 static void generateColumnTypes(
990 Parse *pParse, /* Parser context */
991 SrcList *pTabList, /* List of tables */
992 ExprList *pEList /* Expressions defining the result set */
993 ){
994 #ifndef SQLITE_OMIT_DECLTYPE
995 Vdbe *v = pParse->pVdbe;
996 int i;
997 NameContext sNC;
998 sNC.pSrcList = pTabList;
999 sNC.pParse = pParse;
1000 for(i=0; i<pEList->nExpr; i++){
1001 Expr *p = pEList->a[i].pExpr;
1002 const char *zType;
1003 #ifdef SQLITE_ENABLE_COLUMN_METADATA
1004 const char *zOrigDb = 0;
1005 const char *zOrigTab = 0;
1006 const char *zOrigCol = 0;
1007 zType = columnType(&sNC, p, &zOrigDb, &zOrigTab, &zOrigCol);
1008
1009 /* The vdbe must make its own copy of the column-type and other
1010 ** column specific strings, in case the schema is reset before this
1011 ** virtual machine is deleted.
1012 */
1013 sqlite3VdbeSetColName(v, i, COLNAME_DATABASE, zOrigDb, SQLITE_TRANSIENT);
1014 sqlite3VdbeSetColName(v, i, COLNAME_TABLE, zOrigTab, SQLITE_TRANSIENT);
1015 sqlite3VdbeSetColName(v, i, COLNAME_COLUMN, zOrigCol, SQLITE_TRANSIENT);
1016 #else
1017 zType = columnType(&sNC, p, 0, 0, 0);
1018 #endif
1019 sqlite3VdbeSetColName(v, i, COLNAME_DECLTYPE, zType, SQLITE_TRANSIENT);
1020 }
1021 #endif /* SQLITE_OMIT_DECLTYPE */
1022 }
1023
1024 /*
1025 ** Generate code that will tell the VDBE the names of columns
1026 ** in the result set. This information is used to provide the
1027 ** azCol[] values in the callback.
1028 */
1029 static void generateColumnNames(
1030 Parse *pParse, /* Parser context */
1031 SrcList *pTabList, /* List of tables */
1032 ExprList *pEList /* Expressions defining the result set */
1033 ){
1034 Vdbe *v = pParse->pVdbe;
1035 int i, j;
1036 sqlite3 *db = pParse->db;
1037 int fullNames, shortNames;
1038
1039 #ifndef SQLITE_OMIT_EXPLAIN
1040 /* If this is an EXPLAIN, skip this step */
1041 if( pParse->explain ){
1042 return;
1043 }
1044 #endif
1045
1046 if( pParse->colNamesSet || NEVER(v==0) || db->mallocFailed ) return;
1047 pParse->colNamesSet = 1;
1048 fullNames = (db->flags & SQLITE_FullColNames)!=0;
1049 shortNames = (db->flags & SQLITE_ShortColNames)!=0;
1050 sqlite3VdbeSetNumCols(v, pEList->nExpr);
1051 for(i=0; i<pEList->nExpr; i++){
1052 Expr *p;
1053 p = pEList->a[i].pExpr;
1054 if( NEVER(p==0) ) continue;
1055 if( pEList->a[i].zName ){
1056 char *zName = pEList->a[i].zName;
1057 sqlite3VdbeSetColName(v, i, COLNAME_NAME, zName, SQLITE_TRANSIENT);
1058 }else if( (p->op==TK_COLUMN || p->op==TK_AGG_COLUMN) && pTabList ){
1059 Table *pTab;
1060 char *zCol;
1061 int iCol = p->iColumn;
1062 for(j=0; ALWAYS(j<pTabList->nSrc); j++){
1063 if( pTabList->a[j].iCursor==p->iTable ) break;
1064 }
1065 assert( j<pTabList->nSrc );
1066 pTab = pTabList->a[j].pTab;
1067 if( iCol<0 ) iCol = pTab->iPKey;
1068 assert( iCol==-1 || (iCol>=0 && iCol<pTab->nCol) );
1069 if( iCol<0 ){
1070 zCol = "rowid";
1071 }else{
1072 zCol = pTab->aCol[iCol].zName;
1073 }
1074 if( !shortNames && !fullNames ){
1075 sqlite3VdbeSetColName(v, i, COLNAME_NAME,
1076 sqlite3DbStrDup(db, pEList->a[i].zSpan), SQLITE_DYNAMIC);
1077 }else if( fullNames ){
1078 char *zName = 0;
1079 zName = sqlite3MPrintf(db, "%s.%s", pTab->zName, zCol);
1080 sqlite3VdbeSetColName(v, i, COLNAME_NAME, zName, SQLITE_DYNAMIC);
1081 }else{
1082 sqlite3VdbeSetColName(v, i, COLNAME_NAME, zCol, SQLITE_TRANSIENT);
1083 }
1084 }else{
1085 sqlite3VdbeSetColName(v, i, COLNAME_NAME,
1086 sqlite3DbStrDup(db, pEList->a[i].zSpan), SQLITE_DYNAMIC);
1087 }
1088 }
1089 generateColumnTypes(pParse, pTabList, pEList);
1090 }
1091
1092 #ifndef SQLITE_OMIT_COMPOUND_SELECT
1093 /*
1094 ** Name of the connection operator, used for error messages.
1095 */
1096 static const char *selectOpName(int id){
1097 char *z;
1098 switch( id ){
1099 case TK_ALL: z = "UNION ALL"; break;
1100 case TK_INTERSECT: z = "INTERSECT"; break;
1101 case TK_EXCEPT: z = "EXCEPT"; break;
1102 default: z = "UNION"; break;
1103 }
1104 return z;
1105 }
1106 #endif /* SQLITE_OMIT_COMPOUND_SELECT */
1107
1108 /*
1109 ** Given a an expression list (which is really the list of expressions
1110 ** that form the result set of a SELECT statement) compute appropriate
1111 ** column names for a table that would hold the expression list.
1112 **
1113 ** All column names will be unique.
1114 **
1115 ** Only the column names are computed. Column.zType, Column.zColl,
1116 ** and other fields of Column are zeroed.
1117 **
1118 ** Return SQLITE_OK on success. If a memory allocation error occurs,
1119 ** store NULL in *paCol and 0 in *pnCol and return SQLITE_NOMEM.
1120 */
1121 static int selectColumnsFromExprList(
1122 Parse *pParse, /* Parsing context */
1123 ExprList *pEList, /* Expr list from which to derive column names */
1124 int *pnCol, /* Write the number of columns here */
1125 Column **paCol /* Write the new column list here */
1126 ){
1127 sqlite3 *db = pParse->db; /* Database connection */
1128 int i, j; /* Loop counters */
1129 int cnt; /* Index added to make the name unique */
1130 Column *aCol, *pCol; /* For looping over result columns */
1131 int nCol; /* Number of columns in the result set */
1132 Expr *p; /* Expression for a single result column */
1133 char *zName; /* Column name */
1134 int nName; /* Size of name in zName[] */
1135
1136 *pnCol = nCol = pEList->nExpr;
1137 aCol = *paCol = sqlite3DbMallocZero(db, sizeof(aCol[0])*nCol);
1138 if( aCol==0 ) return SQLITE_NOMEM;
1139 for(i=0, pCol=aCol; i<nCol; i++, pCol++){
1140 /* Get an appropriate name for the column
1141 */
1142 p = pEList->a[i].pExpr;
1143 assert( p->pRight==0 || ExprHasProperty(p->pRight, EP_IntValue)
1144 || p->pRight->u.zToken==0 || p->pRight->u.zToken[0]!=0 );
1145 if( (zName = pEList->a[i].zName)!=0 ){
1146 /* If the column contains an "AS <name>" phrase, use <name> as the name */
1147 zName = sqlite3DbStrDup(db, zName);
1148 }else{
1149 Expr *pColExpr = p; /* The expression that is the result column name */
1150 Table *pTab; /* Table associated with this expression */
1151 while( pColExpr->op==TK_DOT ) pColExpr = pColExpr->pRight;
1152 if( pColExpr->op==TK_COLUMN && ALWAYS(pColExpr->pTab!=0) ){
1153 /* For columns use the column name name */
1154 int iCol = pColExpr->iColumn;
1155 pTab = pColExpr->pTab;
1156 if( iCol<0 ) iCol = pTab->iPKey;
1157 zName = sqlite3MPrintf(db, "%s",
1158 iCol>=0 ? pTab->aCol[iCol].zName : "rowid");
1159 }else if( pColExpr->op==TK_ID ){
1160 assert( !ExprHasProperty(pColExpr, EP_IntValue) );
1161 zName = sqlite3MPrintf(db, "%s", pColExpr->u.zToken);
1162 }else{
1163 /* Use the original text of the column expression as its name */
1164 zName = sqlite3MPrintf(db, "%s", pEList->a[i].zSpan);
1165 }
1166 }
1167 if( db->mallocFailed ){
1168 sqlite3DbFree(db, zName);
1169 break;
1170 }
1171
1172 /* Make sure the column name is unique. If the name is not unique,
1173 ** append a integer to the name so that it becomes unique.
1174 */
1175 nName = sqlite3Strlen30(zName);
1176 for(j=cnt=0; j<i; j++){
1177 if( sqlite3StrICmp(aCol[j].zName, zName)==0 ){
1178 char *zNewName;
1179 zName[nName] = 0;
1180 zNewName = sqlite3MPrintf(db, "%s:%d", zName, ++cnt);
1181 sqlite3DbFree(db, zName);
1182 zName = zNewName;
1183 j = -1;
1184 if( zName==0 ) break;
1185 }
1186 }
1187 pCol->zName = zName;
1188 }
1189 if( db->mallocFailed ){
1190 for(j=0; j<i; j++){
1191 sqlite3DbFree(db, aCol[j].zName);
1192 }
1193 sqlite3DbFree(db, aCol);
1194 *paCol = 0;
1195 *pnCol = 0;
1196 return SQLITE_NOMEM;
1197 }
1198 return SQLITE_OK;
1199 }
1200
1201 /*
1202 ** Add type and collation information to a column list based on
1203 ** a SELECT statement.
1204 **
1205 ** The column list presumably came from selectColumnNamesFromExprList().
1206 ** The column list has only names, not types or collations. This
1207 ** routine goes through and adds the types and collations.
1208 **
1209 ** This routine requires that all identifiers in the SELECT
1210 ** statement be resolved.
1211 */
1212 static void selectAddColumnTypeAndCollation(
1213 Parse *pParse, /* Parsing contexts */
1214 int nCol, /* Number of columns */
1215 Column *aCol, /* List of columns */
1216 Select *pSelect /* SELECT used to determine types and collations */
1217 ){
1218 sqlite3 *db = pParse->db;
1219 NameContext sNC;
1220 Column *pCol;
1221 CollSeq *pColl;
1222 int i;
1223 Expr *p;
1224 struct ExprList_item *a;
1225
1226 assert( pSelect!=0 );
1227 assert( (pSelect->selFlags & SF_Resolved)!=0 );
1228 assert( nCol==pSelect->pEList->nExpr || db->mallocFailed );
1229 if( db->mallocFailed ) return;
1230 memset(&sNC, 0, sizeof(sNC));
1231 sNC.pSrcList = pSelect->pSrc;
1232 a = pSelect->pEList->a;
1233 for(i=0, pCol=aCol; i<nCol; i++, pCol++){
1234 p = a[i].pExpr;
1235 pCol->zType = sqlite3DbStrDup(db, columnType(&sNC, p, 0, 0, 0));
1236 pCol->affinity = sqlite3ExprAffinity(p);
1237 if( pCol->affinity==0 ) pCol->affinity = SQLITE_AFF_NONE;
1238 pColl = sqlite3ExprCollSeq(pParse, p);
1239 if( pColl ){
1240 pCol->zColl = sqlite3DbStrDup(db, pColl->zName);
1241 }
1242 }
1243 }
1244
1245 /*
1246 ** Given a SELECT statement, generate a Table structure that describes
1247 ** the result set of that SELECT.
1248 */
1249 Table *sqlite3ResultSetOfSelect(Parse *pParse, Select *pSelect){
1250 Table *pTab;
1251 sqlite3 *db = pParse->db;
1252 int savedFlags;
1253
1254 savedFlags = db->flags;
1255 db->flags &= ~SQLITE_FullColNames;
1256 db->flags |= SQLITE_ShortColNames;
1257 sqlite3SelectPrep(pParse, pSelect, 0);
1258 if( pParse->nErr ) return 0;
1259 while( pSelect->pPrior ) pSelect = pSelect->pPrior;
1260 db->flags = savedFlags;
1261 pTab = sqlite3DbMallocZero(db, sizeof(Table) );
1262 if( pTab==0 ){
1263 return 0;
1264 }
1265 /* The sqlite3ResultSetOfSelect() is only used n contexts where lookaside
1266 ** is disabled, so we might as well hard-code pTab->dbMem to NULL. */
1267 assert( db->lookaside.bEnabled==0 );
1268 pTab->dbMem = 0;
1269 pTab->nRef = 1;
1270 pTab->zName = 0;
1271 selectColumnsFromExprList(pParse, pSelect->pEList, &pTab->nCol, &pTab->aCol);
1272 selectAddColumnTypeAndCollation(pParse, pTab->nCol, pTab->aCol, pSelect);
1273 pTab->iPKey = -1;
1274 if( db->mallocFailed ){
1275 sqlite3DeleteTable(pTab);
1276 return 0;
1277 }
1278 return pTab;
1279 }
1280
1281 /*
1282 ** Get a VDBE for the given parser context. Create a new one if necessary.
1283 ** If an error occurs, return NULL and leave a message in pParse.
1284 */
1285 Vdbe *sqlite3GetVdbe(Parse *pParse){
1286 Vdbe *v = pParse->pVdbe;
1287 if( v==0 ){
1288 v = pParse->pVdbe = sqlite3VdbeCreate(pParse->db);
1289 #ifndef SQLITE_OMIT_TRACE
1290 if( v ){
1291 sqlite3VdbeAddOp0(v, OP_Trace);
1292 }
1293 #endif
1294 }
1295 return v;
1296 }
1297
1298
1299 /*
1300 ** Compute the iLimit and iOffset fields of the SELECT based on the
1301 ** pLimit and pOffset expressions. pLimit and pOffset hold the expressions
1302 ** that appear in the original SQL statement after the LIMIT and OFFSET
1303 ** keywords. Or NULL if those keywords are omitted. iLimit and iOffset
1304 ** are the integer memory register numbers for counters used to compute
1305 ** the limit and offset. If there is no limit and/or offset, then
1306 ** iLimit and iOffset are negative.
1307 **
1308 ** This routine changes the values of iLimit and iOffset only if
1309 ** a limit or offset is defined by pLimit and pOffset. iLimit and
1310 ** iOffset should have been preset to appropriate default values
1311 ** (usually but not always -1) prior to calling this routine.
1312 ** Only if pLimit!=0 or pOffset!=0 do the limit registers get
1313 ** redefined. The UNION ALL operator uses this property to force
1314 ** the reuse of the same limit and offset registers across multiple
1315 ** SELECT statements.
1316 */
1317 static void computeLimitRegisters(Parse *pParse, Select *p, int iBreak){
1318 Vdbe *v = 0;
1319 int iLimit = 0;
1320 int iOffset;
1321 int addr1;
1322 if( p->iLimit ) return;
1323
1324 /*
1325 ** "LIMIT -1" always shows all rows. There is some
1326 ** contraversy about what the correct behavior should be.
1327 ** The current implementation interprets "LIMIT 0" to mean
1328 ** no rows.
1329 */
1330 sqlite3ExprCacheClear(pParse);
1331 assert( p->pOffset==0 || p->pLimit!=0 );
1332 if( p->pLimit ){
1333 p->iLimit = iLimit = ++pParse->nMem;
1334 v = sqlite3GetVdbe(pParse);
1335 if( NEVER(v==0) ) return; /* VDBE should have already been allocated */
1336 sqlite3ExprCode(pParse, p->pLimit, iLimit);
1337 sqlite3VdbeAddOp1(v, OP_MustBeInt, iLimit);
1338 VdbeComment((v, "LIMIT counter"));
1339 sqlite3VdbeAddOp2(v, OP_IfZero, iLimit, iBreak);
1340 if( p->pOffset ){
1341 p->iOffset = iOffset = ++pParse->nMem;
1342 pParse->nMem++; /* Allocate an extra register for limit+offset */
1343 sqlite3ExprCode(pParse, p->pOffset, iOffset);
1344 sqlite3VdbeAddOp1(v, OP_MustBeInt, iOffset);
1345 VdbeComment((v, "OFFSET counter"));
1346 addr1 = sqlite3VdbeAddOp1(v, OP_IfPos, iOffset);
1347 sqlite3VdbeAddOp2(v, OP_Integer, 0, iOffset);
1348 sqlite3VdbeJumpHere(v, addr1);
1349 sqlite3VdbeAddOp3(v, OP_Add, iLimit, iOffset, iOffset+1);
1350 VdbeComment((v, "LIMIT+OFFSET"));
1351 addr1 = sqlite3VdbeAddOp1(v, OP_IfPos, iLimit);
1352 sqlite3VdbeAddOp2(v, OP_Integer, -1, iOffset+1);
1353 sqlite3VdbeJumpHere(v, addr1);
1354 }
1355 }
1356 }
1357
1358 #ifndef SQLITE_OMIT_COMPOUND_SELECT
1359 /*
1360 ** Return the appropriate collating sequence for the iCol-th column of
1361 ** the result set for the compound-select statement "p". Return NULL if
1362 ** the column has no default collating sequence.
1363 **
1364 ** The collating sequence for the compound select is taken from the
1365 ** left-most term of the select that has a collating sequence.
1366 */
1367 static CollSeq *multiSelectCollSeq(Parse *pParse, Select *p, int iCol){
1368 CollSeq *pRet;
1369 if( p->pPrior ){
1370 pRet = multiSelectCollSeq(pParse, p->pPrior, iCol);
1371 }else{
1372 pRet = 0;
1373 }
1374 assert( iCol>=0 );
1375 if( pRet==0 && iCol<p->pEList->nExpr ){
1376 pRet = sqlite3ExprCollSeq(pParse, p->pEList->a[iCol].pExpr);
1377 }
1378 return pRet;
1379 }
1380 #endif /* SQLITE_OMIT_COMPOUND_SELECT */
1381
1382 /* Forward reference */
1383 static int multiSelectOrderBy(
1384 Parse *pParse, /* Parsing context */
1385 Select *p, /* The right-most of SELECTs to be coded */
1386 SelectDest *pDest /* What to do with query results */
1387 );
1388
1389
1390 #ifndef SQLITE_OMIT_COMPOUND_SELECT
1391 /*
1392 ** This routine is called to process a compound query form from
1393 ** two or more separate queries using UNION, UNION ALL, EXCEPT, or
1394 ** INTERSECT
1395 **
1396 ** "p" points to the right-most of the two queries. the query on the
1397 ** left is p->pPrior. The left query could also be a compound query
1398 ** in which case this routine will be called recursively.
1399 **
1400 ** The results of the total query are to be written into a destination
1401 ** of type eDest with parameter iParm.
1402 **
1403 ** Example 1: Consider a three-way compound SQL statement.
1404 **
1405 ** SELECT a FROM t1 UNION SELECT b FROM t2 UNION SELECT c FROM t3
1406 **
1407 ** This statement is parsed up as follows:
1408 **
1409 ** SELECT c FROM t3
1410 ** |
1411 ** `-----> SELECT b FROM t2
1412 ** |
1413 ** `------> SELECT a FROM t1
1414 **
1415 ** The arrows in the diagram above represent the Select.pPrior pointer.
1416 ** So if this routine is called with p equal to the t3 query, then
1417 ** pPrior will be the t2 query. p->op will be TK_UNION in this case.
1418 **
1419 ** Notice that because of the way SQLite parses compound SELECTs, the
1420 ** individual selects always group from left to right.
1421 */
1422 static int multiSelect(
1423 Parse *pParse, /* Parsing context */
1424 Select *p, /* The right-most of SELECTs to be coded */
1425 SelectDest *pDest /* What to do with query results */
1426 ){
1427 int rc = SQLITE_OK; /* Success code from a subroutine */
1428 Select *pPrior; /* Another SELECT immediately to our left */
1429 Vdbe *v; /* Generate code to this VDBE */
1430 SelectDest dest; /* Alternative data destination */
1431 Select *pDelete = 0; /* Chain of simple selects to delete */
1432 sqlite3 *db; /* Database connection */
1433
1434 /* Make sure there is no ORDER BY or LIMIT clause on prior SELECTs. Only
1435 ** the last (right-most) SELECT in the series may have an ORDER BY or LIMIT.
1436 */
1437 assert( p && p->pPrior ); /* Calling function guarantees this much */
1438 db = pParse->db;
1439 pPrior = p->pPrior;
1440 assert( pPrior->pRightmost!=pPrior );
1441 assert( pPrior->pRightmost==p->pRightmost );
1442 dest = *pDest;
1443 if( pPrior->pOrderBy ){
1444 sqlite3ErrorMsg(pParse,"ORDER BY clause should come after %s not before",
1445 selectOpName(p->op));
1446 rc = 1;
1447 goto multi_select_end;
1448 }
1449 if( pPrior->pLimit ){
1450 sqlite3ErrorMsg(pParse,"LIMIT clause should come after %s not before",
1451 selectOpName(p->op));
1452 rc = 1;
1453 goto multi_select_end;
1454 }
1455
1456 v = sqlite3GetVdbe(pParse);
1457 assert( v!=0 ); /* The VDBE already created by calling function */
1458
1459 /* Create the destination temporary table if necessary
1460 */
1461 if( dest.eDest==SRT_EphemTab ){
1462 assert( p->pEList );
1463 sqlite3VdbeAddOp2(v, OP_OpenEphemeral, dest.iParm, p->pEList->nExpr);
1464 dest.eDest = SRT_Table;
1465 }
1466
1467 /* Make sure all SELECTs in the statement have the same number of elements
1468 ** in their result sets.
1469 */
1470 assert( p->pEList && pPrior->pEList );
1471 if( p->pEList->nExpr!=pPrior->pEList->nExpr ){
1472 sqlite3ErrorMsg(pParse, "SELECTs to the left and right of %s"
1473 " do not have the same number of result columns", selectOpName(p->op));
1474 rc = 1;
1475 goto multi_select_end;
1476 }
1477
1478 /* Compound SELECTs that have an ORDER BY clause are handled separately.
1479 */
1480 if( p->pOrderBy ){
1481 return multiSelectOrderBy(pParse, p, pDest);
1482 }
1483
1484 /* Generate code for the left and right SELECT statements.
1485 */
1486 switch( p->op ){
1487 case TK_ALL: {
1488 int addr = 0;
1489 assert( !pPrior->pLimit );
1490 pPrior->pLimit = p->pLimit;
1491 pPrior->pOffset = p->pOffset;
1492 rc = sqlite3Select(pParse, pPrior, &dest);
1493 p->pLimit = 0;
1494 p->pOffset = 0;
1495 if( rc ){
1496 goto multi_select_end;
1497 }
1498 p->pPrior = 0;
1499 p->iLimit = pPrior->iLimit;
1500 p->iOffset = pPrior->iOffset;
1501 if( p->iLimit ){
1502 addr = sqlite3VdbeAddOp1(v, OP_IfZero, p->iLimit);
1503 VdbeComment((v, "Jump ahead if LIMIT reached"));
1504 }
1505 rc = sqlite3Select(pParse, p, &dest);
1506 testcase( rc!=SQLITE_OK );
1507 pDelete = p->pPrior;
1508 p->pPrior = pPrior;
1509 if( addr ){
1510 sqlite3VdbeJumpHere(v, addr);
1511 }
1512 break;
1513 }
1514 case TK_EXCEPT:
1515 case TK_UNION: {
1516 int unionTab; /* Cursor number of the temporary table holding result */
1517 u8 op = 0; /* One of the SRT_ operations to apply to self */
1518 int priorOp; /* The SRT_ operation to apply to prior selects */
1519 Expr *pLimit, *pOffset; /* Saved values of p->nLimit and p->nOffset */
1520 int addr;
1521 SelectDest uniondest;
1522
1523 testcase( p->op==TK_EXCEPT );
1524 testcase( p->op==TK_UNION );
1525 priorOp = SRT_Union;
1526 if( dest.eDest==priorOp && ALWAYS(!p->pLimit &&!p->pOffset) ){
1527 /* We can reuse a temporary table generated by a SELECT to our
1528 ** right.
1529 */
1530 assert( p->pRightmost!=p ); /* Can only happen for leftward elements
1531 ** of a 3-way or more compound */
1532 assert( p->pLimit==0 ); /* Not allowed on leftward elements */
1533 assert( p->pOffset==0 ); /* Not allowed on leftward elements */
1534 unionTab = dest.iParm;
1535 }else{
1536 /* We will need to create our own temporary table to hold the
1537 ** intermediate results.
1538 */
1539 unionTab = pParse->nTab++;
1540 assert( p->pOrderBy==0 );
1541 addr = sqlite3VdbeAddOp2(v, OP_OpenEphemeral, unionTab, 0);
1542 assert( p->addrOpenEphm[0] == -1 );
1543 p->addrOpenEphm[0] = addr;
1544 p->pRightmost->selFlags |= SF_UsesEphemeral;
1545 assert( p->pEList );
1546 }
1547
1548 /* Code the SELECT statements to our left
1549 */
1550 assert( !pPrior->pOrderBy );
1551 sqlite3SelectDestInit(&uniondest, priorOp, unionTab);
1552 rc = sqlite3Select(pParse, pPrior, &uniondest);
1553 if( rc ){
1554 goto multi_select_end;
1555 }
1556
1557 /* Code the current SELECT statement
1558 */
1559 if( p->op==TK_EXCEPT ){
1560 op = SRT_Except;
1561 }else{
1562 assert( p->op==TK_UNION );
1563 op = SRT_Union;
1564 }
1565 p->pPrior = 0;
1566 pLimit = p->pLimit;
1567 p->pLimit = 0;
1568 pOffset = p->pOffset;
1569 p->pOffset = 0;
1570 uniondest.eDest = op;
1571 rc = sqlite3Select(pParse, p, &uniondest);
1572 testcase( rc!=SQLITE_OK );
1573 /* Query flattening in sqlite3Select() might refill p->pOrderBy.
1574 ** Be sure to delete p->pOrderBy, therefore, to avoid a memory leak. */
1575 sqlite3ExprListDelete(db, p->pOrderBy);
1576 pDelete = p->pPrior;
1577 p->pPrior = pPrior;
1578 p->pOrderBy = 0;
1579 sqlite3ExprDelete(db, p->pLimit);
1580 p->pLimit = pLimit;
1581 p->pOffset = pOffset;
1582 p->iLimit = 0;
1583 p->iOffset = 0;
1584
1585 /* Convert the data in the temporary table into whatever form
1586 ** it is that we currently need.
1587 */
1588 assert( unionTab==dest.iParm || dest.eDest!=priorOp );
1589 if( dest.eDest!=priorOp ){
1590 int iCont, iBreak, iStart;
1591 assert( p->pEList );
1592 if( dest.eDest==SRT_Output ){
1593 Select *pFirst = p;
1594 while( pFirst->pPrior ) pFirst = pFirst->pPrior;
1595 generateColumnNames(pParse, 0, pFirst->pEList);
1596 }
1597 iBreak = sqlite3VdbeMakeLabel(v);
1598 iCont = sqlite3VdbeMakeLabel(v);
1599 computeLimitRegisters(pParse, p, iBreak);
1600 sqlite3VdbeAddOp2(v, OP_Rewind, unionTab, iBreak);
1601 iStart = sqlite3VdbeCurrentAddr(v);
1602 selectInnerLoop(pParse, p, p->pEList, unionTab, p->pEList->nExpr,
1603 0, -1, &dest, iCont, iBreak);
1604 sqlite3VdbeResolveLabel(v, iCont);
1605 sqlite3VdbeAddOp2(v, OP_Next, unionTab, iStart);
1606 sqlite3VdbeResolveLabel(v, iBreak);
1607 sqlite3VdbeAddOp2(v, OP_Close, unionTab, 0);
1608 }
1609 break;
1610 }
1611 default: assert( p->op==TK_INTERSECT ); {
1612 int tab1, tab2;
1613 int iCont, iBreak, iStart;
1614 Expr *pLimit, *pOffset;
1615 int addr;
1616 SelectDest intersectdest;
1617 int r1;
1618
1619 /* INTERSECT is different from the others since it requires
1620 ** two temporary tables. Hence it has its own case. Begin
1621 ** by allocating the tables we will need.
1622 */
1623 tab1 = pParse->nTab++;
1624 tab2 = pParse->nTab++;
1625 assert( p->pOrderBy==0 );
1626
1627 addr = sqlite3VdbeAddOp2(v, OP_OpenEphemeral, tab1, 0);
1628 assert( p->addrOpenEphm[0] == -1 );
1629 p->addrOpenEphm[0] = addr;
1630 p->pRightmost->selFlags |= SF_UsesEphemeral;
1631 assert( p->pEList );
1632
1633 /* Code the SELECTs to our left into temporary table "tab1".
1634 */
1635 sqlite3SelectDestInit(&intersectdest, SRT_Union, tab1);
1636 rc = sqlite3Select(pParse, pPrior, &intersectdest);
1637 if( rc ){
1638 goto multi_select_end;
1639 }
1640
1641 /* Code the current SELECT into temporary table "tab2"
1642 */
1643 addr = sqlite3VdbeAddOp2(v, OP_OpenEphemeral, tab2, 0);
1644 assert( p->addrOpenEphm[1] == -1 );
1645 p->addrOpenEphm[1] = addr;
1646 p->pPrior = 0;
1647 pLimit = p->pLimit;
1648 p->pLimit = 0;
1649 pOffset = p->pOffset;
1650 p->pOffset = 0;
1651 intersectdest.iParm = tab2;
1652 rc = sqlite3Select(pParse, p, &intersectdest);
1653 testcase( rc!=SQLITE_OK );
1654 pDelete = p->pPrior;
1655 p->pPrior = pPrior;
1656 sqlite3ExprDelete(db, p->pLimit);
1657 p->pLimit = pLimit;
1658 p->pOffset = pOffset;
1659
1660 /* Generate code to take the intersection of the two temporary
1661 ** tables.
1662 */
1663 assert( p->pEList );
1664 if( dest.eDest==SRT_Output ){
1665 Select *pFirst = p;
1666 while( pFirst->pPrior ) pFirst = pFirst->pPrior;
1667 generateColumnNames(pParse, 0, pFirst->pEList);
1668 }
1669 iBreak = sqlite3VdbeMakeLabel(v);
1670 iCont = sqlite3VdbeMakeLabel(v);
1671 computeLimitRegisters(pParse, p, iBreak);
1672 sqlite3VdbeAddOp2(v, OP_Rewind, tab1, iBreak);
1673 r1 = sqlite3GetTempReg(pParse);
1674 iStart = sqlite3VdbeAddOp2(v, OP_RowKey, tab1, r1);
1675 sqlite3VdbeAddOp3(v, OP_NotFound, tab2, iCont, r1);
1676 sqlite3ReleaseTempReg(pParse, r1);
1677 selectInnerLoop(pParse, p, p->pEList, tab1, p->pEList->nExpr,
1678 0, -1, &dest, iCont, iBreak);
1679 sqlite3VdbeResolveLabel(v, iCont);
1680 sqlite3VdbeAddOp2(v, OP_Next, tab1, iStart);
1681 sqlite3VdbeResolveLabel(v, iBreak);
1682 sqlite3VdbeAddOp2(v, OP_Close, tab2, 0);
1683 sqlite3VdbeAddOp2(v, OP_Close, tab1, 0);
1684 break;
1685 }
1686 }
1687
1688 /* Compute collating sequences used by
1689 ** temporary tables needed to implement the compound select.
1690 ** Attach the KeyInfo structure to all temporary tables.
1691 **
1692 ** This section is run by the right-most SELECT statement only.
1693 ** SELECT statements to the left always skip this part. The right-most
1694 ** SELECT might also skip this part if it has no ORDER BY clause and
1695 ** no temp tables are required.
1696 */
1697 if( p->selFlags & SF_UsesEphemeral ){
1698 int i; /* Loop counter */
1699 KeyInfo *pKeyInfo; /* Collating sequence for the result set */
1700 Select *pLoop; /* For looping through SELECT statements */
1701 CollSeq **apColl; /* For looping through pKeyInfo->aColl[] */
1702 int nCol; /* Number of columns in result set */
1703
1704 assert( p->pRightmost==p );
1705 nCol = p->pEList->nExpr;
1706 pKeyInfo = sqlite3DbMallocZero(db,
1707 sizeof(*pKeyInfo)+nCol*(sizeof(CollSeq*) + 1));
1708 if( !pKeyInfo ){
1709 rc = SQLITE_NOMEM;
1710 goto multi_select_end;
1711 }
1712
1713 pKeyInfo->enc = ENC(db);
1714 pKeyInfo->nField = (u16)nCol;
1715
1716 for(i=0, apColl=pKeyInfo->aColl; i<nCol; i++, apColl++){
1717 *apColl = multiSelectCollSeq(pParse, p, i);
1718 if( 0==*apColl ){
1719 *apColl = db->pDfltColl;
1720 }
1721 }
1722
1723 for(pLoop=p; pLoop; pLoop=pLoop->pPrior){
1724 for(i=0; i<2; i++){
1725 int addr = pLoop->addrOpenEphm[i];
1726 if( addr<0 ){
1727 /* If [0] is unused then [1] is also unused. So we can
1728 ** always safely abort as soon as the first unused slot is found */
1729 assert( pLoop->addrOpenEphm[1]<0 );
1730 break;
1731 }
1732 sqlite3VdbeChangeP2(v, addr, nCol);
1733 sqlite3VdbeChangeP4(v, addr, (char*)pKeyInfo, P4_KEYINFO);
1734 pLoop->addrOpenEphm[i] = -1;
1735 }
1736 }
1737 sqlite3DbFree(db, pKeyInfo);
1738 }
1739
1740 multi_select_end:
1741 pDest->iMem = dest.iMem;
1742 pDest->nMem = dest.nMem;
1743 sqlite3SelectDelete(db, pDelete);
1744 return rc;
1745 }
1746 #endif /* SQLITE_OMIT_COMPOUND_SELECT */
1747
1748 /*
1749 ** Code an output subroutine for a coroutine implementation of a
1750 ** SELECT statment.
1751 **
1752 ** The data to be output is contained in pIn->iMem. There are
1753 ** pIn->nMem columns to be output. pDest is where the output should
1754 ** be sent.
1755 **
1756 ** regReturn is the number of the register holding the subroutine
1757 ** return address.
1758 **
1759 ** If regPrev>0 then it is a the first register in a vector that
1760 ** records the previous output. mem[regPrev] is a flag that is false
1761 ** if there has been no previous output. If regPrev>0 then code is
1762 ** generated to suppress duplicates. pKeyInfo is used for comparing
1763 ** keys.
1764 **
1765 ** If the LIMIT found in p->iLimit is reached, jump immediately to
1766 ** iBreak.
1767 */
1768 static int generateOutputSubroutine(
1769 Parse *pParse, /* Parsing context */
1770 Select *p, /* The SELECT statement */
1771 SelectDest *pIn, /* Coroutine supplying data */
1772 SelectDest *pDest, /* Where to send the data */
1773 int regReturn, /* The return address register */
1774 int regPrev, /* Previous result register. No uniqueness if 0 */
1775 KeyInfo *pKeyInfo, /* For comparing with previous entry */
1776 int p4type, /* The p4 type for pKeyInfo */
1777 int iBreak /* Jump here if we hit the LIMIT */
1778 ){
1779 Vdbe *v = pParse->pVdbe;
1780 int iContinue;
1781 int addr;
1782
1783 addr = sqlite3VdbeCurrentAddr(v);
1784 iContinue = sqlite3VdbeMakeLabel(v);
1785
1786 /* Suppress duplicates for UNION, EXCEPT, and INTERSECT
1787 */
1788 if( regPrev ){
1789 int j1, j2;
1790 j1 = sqlite3VdbeAddOp1(v, OP_IfNot, regPrev);
1791 j2 = sqlite3VdbeAddOp4(v, OP_Compare, pIn->iMem, regPrev+1, pIn->nMem,
1792 (char*)pKeyInfo, p4type);
1793 sqlite3VdbeAddOp3(v, OP_Jump, j2+2, iContinue, j2+2);
1794 sqlite3VdbeJumpHere(v, j1);
1795 sqlite3ExprCodeCopy(pParse, pIn->iMem, regPrev+1, pIn->nMem);
1796 sqlite3VdbeAddOp2(v, OP_Integer, 1, regPrev);
1797 }
1798 if( pParse->db->mallocFailed ) return 0;
1799
1800 /* Suppress the the first OFFSET entries if there is an OFFSET clause
1801 */
1802 codeOffset(v, p, iContinue);
1803
1804 switch( pDest->eDest ){
1805 /* Store the result as data using a unique key.
1806 */
1807 case SRT_Table:
1808 case SRT_EphemTab: {
1809 int r1 = sqlite3GetTempReg(pParse);
1810 int r2 = sqlite3GetTempReg(pParse);
1811 testcase( pDest->eDest==SRT_Table );
1812 testcase( pDest->eDest==SRT_EphemTab );
1813 sqlite3VdbeAddOp3(v, OP_MakeRecord, pIn->iMem, pIn->nMem, r1);
1814 sqlite3VdbeAddOp2(v, OP_NewRowid, pDest->iParm, r2);
1815 sqlite3VdbeAddOp3(v, OP_Insert, pDest->iParm, r1, r2);
1816 sqlite3VdbeChangeP5(v, OPFLAG_APPEND);
1817 sqlite3ReleaseTempReg(pParse, r2);
1818 sqlite3ReleaseTempReg(pParse, r1);
1819 break;
1820 }
1821
1822 #ifndef SQLITE_OMIT_SUBQUERY
1823 /* If we are creating a set for an "expr IN (SELECT ...)" construct,
1824 ** then there should be a single item on the stack. Write this
1825 ** item into the set table with bogus data.
1826 */
1827 case SRT_Set: {
1828 int r1;
1829 assert( pIn->nMem==1 );
1830 p->affinity =
1831 sqlite3CompareAffinity(p->pEList->a[0].pExpr, pDest->affinity);
1832 r1 = sqlite3GetTempReg(pParse);
1833 sqlite3VdbeAddOp4(v, OP_MakeRecord, pIn->iMem, 1, r1, &p->affinity, 1);
1834 sqlite3ExprCacheAffinityChange(pParse, pIn->iMem, 1);
1835 sqlite3VdbeAddOp2(v, OP_IdxInsert, pDest->iParm, r1);
1836 sqlite3ReleaseTempReg(pParse, r1);
1837 break;
1838 }
1839
1840 #if 0 /* Never occurs on an ORDER BY query */
1841 /* If any row exist in the result set, record that fact and abort.
1842 */
1843 case SRT_Exists: {
1844 sqlite3VdbeAddOp2(v, OP_Integer, 1, pDest->iParm);
1845 /* The LIMIT clause will terminate the loop for us */
1846 break;
1847 }
1848 #endif
1849
1850 /* If this is a scalar select that is part of an expression, then
1851 ** store the results in the appropriate memory cell and break out
1852 ** of the scan loop.
1853 */
1854 case SRT_Mem: {
1855 assert( pIn->nMem==1 );
1856 sqlite3ExprCodeMove(pParse, pIn->iMem, pDest->iParm, 1);
1857 /* The LIMIT clause will jump out of the loop for us */
1858 break;
1859 }
1860 #endif /* #ifndef SQLITE_OMIT_SUBQUERY */
1861
1862 /* The results are stored in a sequence of registers
1863 ** starting at pDest->iMem. Then the co-routine yields.
1864 */
1865 case SRT_Coroutine: {
1866 if( pDest->iMem==0 ){
1867 pDest->iMem = sqlite3GetTempRange(pParse, pIn->nMem);
1868 pDest->nMem = pIn->nMem;
1869 }
1870 sqlite3ExprCodeMove(pParse, pIn->iMem, pDest->iMem, pDest->nMem);
1871 sqlite3VdbeAddOp1(v, OP_Yield, pDest->iParm);
1872 break;
1873 }
1874
1875 /* If none of the above, then the result destination must be
1876 ** SRT_Output. This routine is never called with any other
1877 ** destination other than the ones handled above or SRT_Output.
1878 **
1879 ** For SRT_Output, results are stored in a sequence of registers.
1880 ** Then the OP_ResultRow opcode is used to cause sqlite3_step() to
1881 ** return the next row of result.
1882 */
1883 default: {
1884 assert( pDest->eDest==SRT_Output );
1885 sqlite3VdbeAddOp2(v, OP_ResultRow, pIn->iMem, pIn->nMem);
1886 sqlite3ExprCacheAffinityChange(pParse, pIn->iMem, pIn->nMem);
1887 break;
1888 }
1889 }
1890
1891 /* Jump to the end of the loop if the LIMIT is reached.
1892 */
1893 if( p->iLimit ){
1894 sqlite3VdbeAddOp2(v, OP_AddImm, p->iLimit, -1);
1895 sqlite3VdbeAddOp2(v, OP_IfZero, p->iLimit, iBreak);
1896 }
1897
1898 /* Generate the subroutine return
1899 */
1900 sqlite3VdbeResolveLabel(v, iContinue);
1901 sqlite3VdbeAddOp1(v, OP_Return, regReturn);
1902
1903 return addr;
1904 }
1905
1906 /*
1907 ** Alternative compound select code generator for cases when there
1908 ** is an ORDER BY clause.
1909 **
1910 ** We assume a query of the following form:
1911 **
1912 ** <selectA> <operator> <selectB> ORDER BY <orderbylist>
1913 **
1914 ** <operator> is one of UNION ALL, UNION, EXCEPT, or INTERSECT. The idea
1915 ** is to code both <selectA> and <selectB> with the ORDER BY clause as
1916 ** co-routines. Then run the co-routines in parallel and merge the results
1917 ** into the output. In addition to the two coroutines (called selectA and
1918 ** selectB) there are 7 subroutines:
1919 **
1920 ** outA: Move the output of the selectA coroutine into the output
1921 ** of the compound query.
1922 **
1923 ** outB: Move the output of the selectB coroutine into the output
1924 ** of the compound query. (Only generated for UNION and
1925 ** UNION ALL. EXCEPT and INSERTSECT never output a row that
1926 ** appears only in B.)
1927 **
1928 ** AltB: Called when there is data from both coroutines and A<B.
1929 **
1930 ** AeqB: Called when there is data from both coroutines and A==B.
1931 **
1932 ** AgtB: Called when there is data from both coroutines and A>B.
1933 **
1934 ** EofA: Called when data is exhausted from selectA.
1935 **
1936 ** EofB: Called when data is exhausted from selectB.
1937 **
1938 ** The implementation of the latter five subroutines depend on which
1939 ** <operator> is used:
1940 **
1941 **
1942 ** UNION ALL UNION EXCEPT INTERSECT
1943 ** ------------- ----------------- -------------- -----------------
1944 ** AltB: outA, nextA outA, nextA outA, nextA nextA
1945 **
1946 ** AeqB: outA, nextA nextA nextA outA, nextA
1947 **
1948 ** AgtB: outB, nextB outB, nextB nextB nextB
1949 **
1950 ** EofA: outB, nextB outB, nextB halt halt
1951 **
1952 ** EofB: outA, nextA outA, nextA outA, nextA halt
1953 **
1954 ** In the AltB, AeqB, and AgtB subroutines, an EOF on A following nextA
1955 ** causes an immediate jump to EofA and an EOF on B following nextB causes
1956 ** an immediate jump to EofB. Within EofA and EofB, and EOF on entry or
1957 ** following nextX causes a jump to the end of the select processing.
1958 **
1959 ** Duplicate removal in the UNION, EXCEPT, and INTERSECT cases is handled
1960 ** within the output subroutine. The regPrev register set holds the previously
1961 ** output value. A comparison is made against this value and the output
1962 ** is skipped if the next results would be the same as the previous.
1963 **
1964 ** The implementation plan is to implement the two coroutines and seven
1965 ** subroutines first, then put the control logic at the bottom. Like this:
1966 **
1967 ** goto Init
1968 ** coA: coroutine for left query (A)
1969 ** coB: coroutine for right query (B)
1970 ** outA: output one row of A
1971 ** outB: output one row of B (UNION and UNION ALL only)
1972 ** EofA: ...
1973 ** EofB: ...
1974 ** AltB: ...
1975 ** AeqB: ...
1976 ** AgtB: ...
1977 ** Init: initialize coroutine registers
1978 ** yield coA
1979 ** if eof(A) goto EofA
1980 ** yield coB
1981 ** if eof(B) goto EofB
1982 ** Cmpr: Compare A, B
1983 ** Jump AltB, AeqB, AgtB
1984 ** End: ...
1985 **
1986 ** We call AltB, AeqB, AgtB, EofA, and EofB "subroutines" but they are not
1987 ** actually called using Gosub and they do not Return. EofA and EofB loop
1988 ** until all data is exhausted then jump to the "end" labe. AltB, AeqB,
1989 ** and AgtB jump to either L2 or to one of EofA or EofB.
1990 */
1991 #ifndef SQLITE_OMIT_COMPOUND_SELECT
1992 static int multiSelectOrderBy(
1993 Parse *pParse, /* Parsing context */
1994 Select *p, /* The right-most of SELECTs to be coded */
1995 SelectDest *pDest /* What to do with query results */
1996 ){
1997 int i, j; /* Loop counters */
1998 Select *pPrior; /* Another SELECT immediately to our left */
1999 Vdbe *v; /* Generate code to this VDBE */
2000 SelectDest destA; /* Destination for coroutine A */
2001 SelectDest destB; /* Destination for coroutine B */
2002 int regAddrA; /* Address register for select-A coroutine */
2003 int regEofA; /* Flag to indicate when select-A is complete */
2004 int regAddrB; /* Address register for select-B coroutine */
2005 int regEofB; /* Flag to indicate when select-B is complete */
2006 int addrSelectA; /* Address of the select-A coroutine */
2007 int addrSelectB; /* Address of the select-B coroutine */
2008 int regOutA; /* Address register for the output-A subroutine */
2009 int regOutB; /* Address register for the output-B subroutine */
2010 int addrOutA; /* Address of the output-A subroutine */
2011 int addrOutB = 0; /* Address of the output-B subroutine */
2012 int addrEofA; /* Address of the select-A-exhausted subroutine */
2013 int addrEofB; /* Address of the select-B-exhausted subroutine */
2014 int addrAltB; /* Address of the A<B subroutine */
2015 int addrAeqB; /* Address of the A==B subroutine */
2016 int addrAgtB; /* Address of the A>B subroutine */
2017 int regLimitA; /* Limit register for select-A */
2018 int regLimitB; /* Limit register for select-A */
2019 int regPrev; /* A range of registers to hold previous output */
2020 int savedLimit; /* Saved value of p->iLimit */
2021 int savedOffset; /* Saved value of p->iOffset */
2022 int labelCmpr; /* Label for the start of the merge algorithm */
2023 int labelEnd; /* Label for the end of the overall SELECT stmt */
2024 int j1; /* Jump instructions that get retargetted */
2025 int op; /* One of TK_ALL, TK_UNION, TK_EXCEPT, TK_INTERSECT */
2026 KeyInfo *pKeyDup = 0; /* Comparison information for duplicate removal */
2027 KeyInfo *pKeyMerge; /* Comparison information for merging rows */
2028 sqlite3 *db; /* Database connection */
2029 ExprList *pOrderBy; /* The ORDER BY clause */
2030 int nOrderBy; /* Number of terms in the ORDER BY clause */
2031 int *aPermute; /* Mapping from ORDER BY terms to result set columns */
2032
2033 assert( p->pOrderBy!=0 );
2034 assert( pKeyDup==0 ); /* "Managed" code needs this. Ticket #3382. */
2035 db = pParse->db;
2036 v = pParse->pVdbe;
2037 assert( v!=0 ); /* Already thrown the error if VDBE alloc failed */
2038 labelEnd = sqlite3VdbeMakeLabel(v);
2039 labelCmpr = sqlite3VdbeMakeLabel(v);
2040
2041
2042 /* Patch up the ORDER BY clause
2043 */
2044 op = p->op;
2045 pPrior = p->pPrior;
2046 assert( pPrior->pOrderBy==0 );
2047 pOrderBy = p->pOrderBy;
2048 assert( pOrderBy );
2049 nOrderBy = pOrderBy->nExpr;
2050
2051 /* For operators other than UNION ALL we have to make sure that
2052 ** the ORDER BY clause covers every term of the result set. Add
2053 ** terms to the ORDER BY clause as necessary.
2054 */
2055 if( op!=TK_ALL ){
2056 for(i=1; db->mallocFailed==0 && i<=p->pEList->nExpr; i++){
2057 struct ExprList_item *pItem;
2058 for(j=0, pItem=pOrderBy->a; j<nOrderBy; j++, pItem++){
2059 assert( pItem->iCol>0 );
2060 if( pItem->iCol==i ) break;
2061 }
2062 if( j==nOrderBy ){
2063 Expr *pNew = sqlite3Expr(db, TK_INTEGER, 0);
2064 if( pNew==0 ) return SQLITE_NOMEM;
2065 pNew->flags |= EP_IntValue;
2066 pNew->u.iValue = i;
2067 pOrderBy = sqlite3ExprListAppend(pParse, pOrderBy, pNew);
2068 pOrderBy->a[nOrderBy++].iCol = (u16)i;
2069 }
2070 }
2071 }
2072
2073 /* Compute the comparison permutation and keyinfo that is used with
2074 ** the permutation used to determine if the next
2075 ** row of results comes from selectA or selectB. Also add explicit
2076 ** collations to the ORDER BY clause terms so that when the subqueries
2077 ** to the right and the left are evaluated, they use the correct
2078 ** collation.
2079 */
2080 aPermute = sqlite3DbMallocRaw(db, sizeof(int)*nOrderBy);
2081 if( aPermute ){
2082 struct ExprList_item *pItem;
2083 for(i=0, pItem=pOrderBy->a; i<nOrderBy; i++, pItem++){
2084 assert( pItem->iCol>0 && pItem->iCol<=p->pEList->nExpr );
2085 aPermute[i] = pItem->iCol - 1;
2086 }
2087 pKeyMerge =
2088 sqlite3DbMallocRaw(db, sizeof(*pKeyMerge)+nOrderBy*(sizeof(CollSeq*)+1));
2089 if( pKeyMerge ){
2090 pKeyMerge->aSortOrder = (u8*)&pKeyMerge->aColl[nOrderBy];
2091 pKeyMerge->nField = (u16)nOrderBy;
2092 pKeyMerge->enc = ENC(db);
2093 for(i=0; i<nOrderBy; i++){
2094 CollSeq *pColl;
2095 Expr *pTerm = pOrderBy->a[i].pExpr;
2096 if( pTerm->flags & EP_ExpCollate ){
2097 pColl = pTerm->pColl;
2098 }else{
2099 pColl = multiSelectCollSeq(pParse, p, aPermute[i]);
2100 pTerm->flags |= EP_ExpCollate;
2101 pTerm->pColl = pColl;
2102 }
2103 pKeyMerge->aColl[i] = pColl;
2104 pKeyMerge->aSortOrder[i] = pOrderBy->a[i].sortOrder;
2105 }
2106 }
2107 }else{
2108 pKeyMerge = 0;
2109 }
2110
2111 /* Reattach the ORDER BY clause to the query.
2112 */
2113 p->pOrderBy = pOrderBy;
2114 pPrior->pOrderBy = sqlite3ExprListDup(pParse->db, pOrderBy, 0);
2115
2116 /* Allocate a range of temporary registers and the KeyInfo needed
2117 ** for the logic that removes duplicate result rows when the
2118 ** operator is UNION, EXCEPT, or INTERSECT (but not UNION ALL).
2119 */
2120 if( op==TK_ALL ){
2121 regPrev = 0;
2122 }else{
2123 int nExpr = p->pEList->nExpr;
2124 assert( nOrderBy>=nExpr || db->mallocFailed );
2125 regPrev = sqlite3GetTempRange(pParse, nExpr+1);
2126 sqlite3VdbeAddOp2(v, OP_Integer, 0, regPrev);
2127 pKeyDup = sqlite3DbMallocZero(db,
2128 sizeof(*pKeyDup) + nExpr*(sizeof(CollSeq*)+1) );
2129 if( pKeyDup ){
2130 pKeyDup->aSortOrder = (u8*)&pKeyDup->aColl[nExpr];
2131 pKeyDup->nField = (u16)nExpr;
2132 pKeyDup->enc = ENC(db);
2133 for(i=0; i<nExpr; i++){
2134 pKeyDup->aColl[i] = multiSelectCollSeq(pParse, p, i);
2135 pKeyDup->aSortOrder[i] = 0;
2136 }
2137 }
2138 }
2139
2140 /* Separate the left and the right query from one another
2141 */
2142 p->pPrior = 0;
2143 pPrior->pRightmost = 0;
2144 sqlite3ResolveOrderGroupBy(pParse, p, p->pOrderBy, "ORDER");
2145 if( pPrior->pPrior==0 ){
2146 sqlite3ResolveOrderGroupBy(pParse, pPrior, pPrior->pOrderBy, "ORDER");
2147 }
2148
2149 /* Compute the limit registers */
2150 computeLimitRegisters(pParse, p, labelEnd);
2151 if( p->iLimit && op==TK_ALL ){
2152 regLimitA = ++pParse->nMem;
2153 regLimitB = ++pParse->nMem;
2154 sqlite3VdbeAddOp2(v, OP_Copy, p->iOffset ? p->iOffset+1 : p->iLimit,
2155 regLimitA);
2156 sqlite3VdbeAddOp2(v, OP_Copy, regLimitA, regLimitB);
2157 }else{
2158 regLimitA = regLimitB = 0;
2159 }
2160 sqlite3ExprDelete(db, p->pLimit);
2161 p->pLimit = 0;
2162 sqlite3ExprDelete(db, p->pOffset);
2163 p->pOffset = 0;
2164
2165 regAddrA = ++pParse->nMem;
2166 regEofA = ++pParse->nMem;
2167 regAddrB = ++pParse->nMem;
2168 regEofB = ++pParse->nMem;
2169 regOutA = ++pParse->nMem;
2170 regOutB = ++pParse->nMem;
2171 sqlite3SelectDestInit(&destA, SRT_Coroutine, regAddrA);
2172 sqlite3SelectDestInit(&destB, SRT_Coroutine, regAddrB);
2173
2174 /* Jump past the various subroutines and coroutines to the main
2175 ** merge loop
2176 */
2177 j1 = sqlite3VdbeAddOp0(v, OP_Goto);
2178 addrSelectA = sqlite3VdbeCurrentAddr(v);
2179
2180
2181 /* Generate a coroutine to evaluate the SELECT statement to the
2182 ** left of the compound operator - the "A" select.
2183 */
2184 VdbeNoopComment((v, "Begin coroutine for left SELECT"));
2185 pPrior->iLimit = regLimitA;
2186 sqlite3Select(pParse, pPrior, &destA);
2187 sqlite3VdbeAddOp2(v, OP_Integer, 1, regEofA);
2188 sqlite3VdbeAddOp1(v, OP_Yield, regAddrA);
2189 VdbeNoopComment((v, "End coroutine for left SELECT"));
2190
2191 /* Generate a coroutine to evaluate the SELECT statement on
2192 ** the right - the "B" select
2193 */
2194 addrSelectB = sqlite3VdbeCurrentAddr(v);
2195 VdbeNoopComment((v, "Begin coroutine for right SELECT"));
2196 savedLimit = p->iLimit;
2197 savedOffset = p->iOffset;
2198 p->iLimit = regLimitB;
2199 p->iOffset = 0;
2200 sqlite3Select(pParse, p, &destB);
2201 p->iLimit = savedLimit;
2202 p->iOffset = savedOffset;
2203 sqlite3VdbeAddOp2(v, OP_Integer, 1, regEofB);
2204 sqlite3VdbeAddOp1(v, OP_Yield, regAddrB);
2205 VdbeNoopComment((v, "End coroutine for right SELECT"));
2206
2207 /* Generate a subroutine that outputs the current row of the A
2208 ** select as the next output row of the compound select.
2209 */
2210 VdbeNoopComment((v, "Output routine for A"));
2211 addrOutA = generateOutputSubroutine(pParse,
2212 p, &destA, pDest, regOutA,
2213 regPrev, pKeyDup, P4_KEYINFO_HANDOFF, labelEnd);
2214
2215 /* Generate a subroutine that outputs the current row of the B
2216 ** select as the next output row of the compound select.
2217 */
2218 if( op==TK_ALL || op==TK_UNION ){
2219 VdbeNoopComment((v, "Output routine for B"));
2220 addrOutB = generateOutputSubroutine(pParse,
2221 p, &destB, pDest, regOutB,
2222 regPrev, pKeyDup, P4_KEYINFO_STATIC, labelEnd);
2223 }
2224
2225 /* Generate a subroutine to run when the results from select A
2226 ** are exhausted and only data in select B remains.
2227 */
2228 VdbeNoopComment((v, "eof-A subroutine"));
2229 if( op==TK_EXCEPT || op==TK_INTERSECT ){
2230 addrEofA = sqlite3VdbeAddOp2(v, OP_Goto, 0, labelEnd);
2231 }else{
2232 addrEofA = sqlite3VdbeAddOp2(v, OP_If, regEofB, labelEnd);
2233 sqlite3VdbeAddOp2(v, OP_Gosub, regOutB, addrOutB);
2234 sqlite3VdbeAddOp1(v, OP_Yield, regAddrB);
2235 sqlite3VdbeAddOp2(v, OP_Goto, 0, addrEofA);
2236 }
2237
2238 /* Generate a subroutine to run when the results from select B
2239 ** are exhausted and only data in select A remains.
2240 */
2241 if( op==TK_INTERSECT ){
2242 addrEofB = addrEofA;
2243 }else{
2244 VdbeNoopComment((v, "eof-B subroutine"));
2245 addrEofB = sqlite3VdbeAddOp2(v, OP_If, regEofA, labelEnd);
2246 sqlite3VdbeAddOp2(v, OP_Gosub, regOutA, addrOutA);
2247 sqlite3VdbeAddOp1(v, OP_Yield, regAddrA);
2248 sqlite3VdbeAddOp2(v, OP_Goto, 0, addrEofB);
2249 }
2250
2251 /* Generate code to handle the case of A<B
2252 */
2253 VdbeNoopComment((v, "A-lt-B subroutine"));
2254 addrAltB = sqlite3VdbeAddOp2(v, OP_Gosub, regOutA, addrOutA);
2255 sqlite3VdbeAddOp1(v, OP_Yield, regAddrA);
2256 sqlite3VdbeAddOp2(v, OP_If, regEofA, addrEofA);
2257 sqlite3VdbeAddOp2(v, OP_Goto, 0, labelCmpr);
2258
2259 /* Generate code to handle the case of A==B
2260 */
2261 if( op==TK_ALL ){
2262 addrAeqB = addrAltB;
2263 }else if( op==TK_INTERSECT ){
2264 addrAeqB = addrAltB;
2265 addrAltB++;
2266 }else{
2267 VdbeNoopComment((v, "A-eq-B subroutine"));
2268 addrAeqB =
2269 sqlite3VdbeAddOp1(v, OP_Yield, regAddrA);
2270 sqlite3VdbeAddOp2(v, OP_If, regEofA, addrEofA);
2271 sqlite3VdbeAddOp2(v, OP_Goto, 0, labelCmpr);
2272 }
2273
2274 /* Generate code to handle the case of A>B
2275 */
2276 VdbeNoopComment((v, "A-gt-B subroutine"));
2277 addrAgtB = sqlite3VdbeCurrentAddr(v);
2278 if( op==TK_ALL || op==TK_UNION ){
2279 sqlite3VdbeAddOp2(v, OP_Gosub, regOutB, addrOutB);
2280 }
2281 sqlite3VdbeAddOp1(v, OP_Yield, regAddrB);
2282 sqlite3VdbeAddOp2(v, OP_If, regEofB, addrEofB);
2283 sqlite3VdbeAddOp2(v, OP_Goto, 0, labelCmpr);
2284
2285 /* This code runs once to initialize everything.
2286 */
2287 sqlite3VdbeJumpHere(v, j1);
2288 sqlite3VdbeAddOp2(v, OP_Integer, 0, regEofA);
2289 sqlite3VdbeAddOp2(v, OP_Integer, 0, regEofB);
2290 sqlite3VdbeAddOp2(v, OP_Gosub, regAddrA, addrSelectA);
2291 sqlite3VdbeAddOp2(v, OP_Gosub, regAddrB, addrSelectB);
2292 sqlite3VdbeAddOp2(v, OP_If, regEofA, addrEofA);
2293 sqlite3VdbeAddOp2(v, OP_If, regEofB, addrEofB);
2294
2295 /* Implement the main merge loop
2296 */
2297 sqlite3VdbeResolveLabel(v, labelCmpr);
2298 sqlite3VdbeAddOp4(v, OP_Permutation, 0, 0, 0, (char*)aPermute, P4_INTARRAY);
2299 sqlite3VdbeAddOp4(v, OP_Compare, destA.iMem, destB.iMem, nOrderBy,
2300 (char*)pKeyMerge, P4_KEYINFO_HANDOFF);
2301 sqlite3VdbeAddOp3(v, OP_Jump, addrAltB, addrAeqB, addrAgtB);
2302
2303 /* Release temporary registers
2304 */
2305 if( regPrev ){
2306 sqlite3ReleaseTempRange(pParse, regPrev, nOrderBy+1);
2307 }
2308
2309 /* Jump to the this point in order to terminate the query.
2310 */
2311 sqlite3VdbeResolveLabel(v, labelEnd);
2312
2313 /* Set the number of output columns
2314 */
2315 if( pDest->eDest==SRT_Output ){
2316 Select *pFirst = pPrior;
2317 while( pFirst->pPrior ) pFirst = pFirst->pPrior;
2318 generateColumnNames(pParse, 0, pFirst->pEList);
2319 }
2320
2321 /* Reassembly the compound query so that it will be freed correctly
2322 ** by the calling function */
2323 if( p->pPrior ){
2324 sqlite3SelectDelete(db, p->pPrior);
2325 }
2326 p->pPrior = pPrior;
2327
2328 /*** TBD: Insert subroutine calls to close cursors on incomplete
2329 **** subqueries ****/
2330 return SQLITE_OK;
2331 }
2332 #endif
2333
2334 #if !defined(SQLITE_OMIT_SUBQUERY) || !defined(SQLITE_OMIT_VIEW)
2335 /* Forward Declarations */
2336 static void substExprList(sqlite3*, ExprList*, int, ExprList*);
2337 static void substSelect(sqlite3*, Select *, int, ExprList *);
2338
2339 /*
2340 ** Scan through the expression pExpr. Replace every reference to
2341 ** a column in table number iTable with a copy of the iColumn-th
2342 ** entry in pEList. (But leave references to the ROWID column
2343 ** unchanged.)
2344 **
2345 ** This routine is part of the flattening procedure. A subquery
2346 ** whose result set is defined by pEList appears as entry in the
2347 ** FROM clause of a SELECT such that the VDBE cursor assigned to that
2348 ** FORM clause entry is iTable. This routine make the necessary
2349 ** changes to pExpr so that it refers directly to the source table
2350 ** of the subquery rather the result set of the subquery.
2351 */
2352 static Expr *substExpr(
2353 sqlite3 *db, /* Report malloc errors to this connection */
2354 Expr *pExpr, /* Expr in which substitution occurs */
2355 int iTable, /* Table to be substituted */
2356 ExprList *pEList /* Substitute expressions */
2357 ){
2358 if( pExpr==0 ) return 0;
2359 if( pExpr->op==TK_COLUMN && pExpr->iTable==iTable ){
2360 if( pExpr->iColumn<0 ){
2361 pExpr->op = TK_NULL;
2362 }else{
2363 Expr *pNew;
2364 assert( pEList!=0 && pExpr->iColumn<pEList->nExpr );
2365 assert( pExpr->pLeft==0 && pExpr->pRight==0 );
2366 pNew = sqlite3ExprDup(db, pEList->a[pExpr->iColumn].pExpr, 0);
2367 if( pNew && pExpr->pColl ){
2368 pNew->pColl = pExpr->pColl;
2369 }
2370 sqlite3ExprDelete(db, pExpr);
2371 pExpr = pNew;
2372 }
2373 }else{
2374 pExpr->pLeft = substExpr(db, pExpr->pLeft, iTable, pEList);
2375 pExpr->pRight = substExpr(db, pExpr->pRight, iTable, pEList);
2376 if( ExprHasProperty(pExpr, EP_xIsSelect) ){
2377 substSelect(db, pExpr->x.pSelect, iTable, pEList);
2378 }else{
2379 substExprList(db, pExpr->x.pList, iTable, pEList);
2380 }
2381 }
2382 return pExpr;
2383 }
2384 static void substExprList(
2385 sqlite3 *db, /* Report malloc errors here */
2386 ExprList *pList, /* List to scan and in which to make substitutes */
2387 int iTable, /* Table to be substituted */
2388 ExprList *pEList /* Substitute values */
2389 ){
2390 int i;
2391 if( pList==0 ) return;
2392 for(i=0; i<pList->nExpr; i++){
2393 pList->a[i].pExpr = substExpr(db, pList->a[i].pExpr, iTable, pEList);
2394 }
2395 }
2396 static void substSelect(
2397 sqlite3 *db, /* Report malloc errors here */
2398 Select *p, /* SELECT statement in which to make substitutions */
2399 int iTable, /* Table to be replaced */
2400 ExprList *pEList /* Substitute values */
2401 ){
2402 SrcList *pSrc;
2403 struct SrcList_item *pItem;
2404 int i;
2405 if( !p ) return;
2406 substExprList(db, p->pEList, iTable, pEList);
2407 substExprList(db, p->pGroupBy, iTable, pEList);
2408 substExprList(db, p->pOrderBy, iTable, pEList);
2409 p->pHaving = substExpr(db, p->pHaving, iTable, pEList);
2410 p->pWhere = substExpr(db, p->pWhere, iTable, pEList);
2411 substSelect(db, p->pPrior, iTable, pEList);
2412 pSrc = p->pSrc;
2413 assert( pSrc ); /* Even for (SELECT 1) we have: pSrc!=0 but pSrc->nSrc==0 */
2414 if( ALWAYS(pSrc) ){
2415 for(i=pSrc->nSrc, pItem=pSrc->a; i>0; i--, pItem++){
2416 substSelect(db, pItem->pSelect, iTable, pEList);
2417 }
2418 }
2419 }
2420 #endif /* !defined(SQLITE_OMIT_SUBQUERY) || !defined(SQLITE_OMIT_VIEW) */
2421
2422 #if !defined(SQLITE_OMIT_SUBQUERY) || !defined(SQLITE_OMIT_VIEW)
2423 /*
2424 ** This routine attempts to flatten subqueries in order to speed
2425 ** execution. It returns 1 if it makes changes and 0 if no flattening
2426 ** occurs.
2427 **
2428 ** To understand the concept of flattening, consider the following
2429 ** query:
2430 **
2431 ** SELECT a FROM (SELECT x+y AS a FROM t1 WHERE z<100) WHERE a>5
2432 **
2433 ** The default way of implementing this query is to execute the
2434 ** subquery first and store the results in a temporary table, then
2435 ** run the outer query on that temporary table. This requires two
2436 ** passes over the data. Furthermore, because the temporary table
2437 ** has no indices, the WHERE clause on the outer query cannot be
2438 ** optimized.
2439 **
2440 ** This routine attempts to rewrite queries such as the above into
2441 ** a single flat select, like this:
2442 **
2443 ** SELECT x+y AS a FROM t1 WHERE z<100 AND a>5
2444 **
2445 ** The code generated for this simpification gives the same result
2446 ** but only has to scan the data once. And because indices might
2447 ** exist on the table t1, a complete scan of the data might be
2448 ** avoided.
2449 **
2450 ** Flattening is only attempted if all of the following are true:
2451 **
2452 ** (1) The subquery and the outer query do not both use aggregates.
2453 **
2454 ** (2) The subquery is not an aggregate or the outer query is not a join.
2455 **
2456 ** (3) The subquery is not the right operand of a left outer join
2457 ** (Originally ticket #306. Strenghtened by ticket #3300)
2458 **
2459 ** (4) The subquery is not DISTINCT or the outer query is not a join.
2460 **
2461 ** (5) The subquery is not DISTINCT or the outer query does not use
2462 ** aggregates.
2463 **
2464 ** (6) The subquery does not use aggregates or the outer query is not
2465 ** DISTINCT.
2466 **
2467 ** (7) The subquery has a FROM clause.
2468 **
2469 ** (8) The subquery does not use LIMIT or the outer query is not a join.
2470 **
2471 ** (9) The subquery does not use LIMIT or the outer query does not use
2472 ** aggregates.
2473 **
2474 ** (10) The subquery does not use aggregates or the outer query does not
2475 ** use LIMIT.
2476 **
2477 ** (11) The subquery and the outer query do not both have ORDER BY clauses.
2478 **
2479 ** (12) Not implemented. Subsumed into restriction (3). Was previously
2480 ** a separate restriction deriving from ticket #350.
2481 **
2482 ** (13) The subquery and outer query do not both use LIMIT
2483 **
2484 ** (14) The subquery does not use OFFSET
2485 **
2486 ** (15) The outer query is not part of a compound select or the
2487 ** subquery does not have both an ORDER BY and a LIMIT clause.
2488 ** (See ticket #2339)
2489 **
2490 ** (16) The outer query is not an aggregate or the subquery does
2491 ** not contain ORDER BY. (Ticket #2942) This used to not matter
2492 ** until we introduced the group_concat() function.
2493 **
2494 ** (17) The sub-query is not a compound select, or it is a UNION ALL
2495 ** compound clause made up entirely of non-aggregate queries, and
2496 ** the parent query:
2497 **
2498 ** * is not itself part of a compound select,
2499 ** * is not an aggregate or DISTINCT query, and
2500 ** * has no other tables or sub-selects in the FROM clause.
2501 **
2502 ** The parent and sub-query may contain WHERE clauses. Subject to
2503 ** rules (11), (13) and (14), they may also contain ORDER BY,
2504 ** LIMIT and OFFSET clauses.
2505 **
2506 ** (18) If the sub-query is a compound select, then all terms of the
2507 ** ORDER by clause of the parent must be simple references to
2508 ** columns of the sub-query.
2509 **
2510 ** (19) The subquery does not use LIMIT or the outer query does not
2511 ** have a WHERE clause.
2512 **
2513 ** (20) If the sub-query is a compound select, then it must not use
2514 ** an ORDER BY clause. Ticket #3773. We could relax this constraint
2515 ** somewhat by saying that the terms of the ORDER BY clause must
2516 ** appear as unmodified result columns in the outer query. But
2517 ** have other optimizations in mind to deal with that case.
2518 **
2519 ** In this routine, the "p" parameter is a pointer to the outer query.
2520 ** The subquery is p->pSrc->a[iFrom]. isAgg is true if the outer query
2521 ** uses aggregates and subqueryIsAgg is true if the subquery uses aggregates.
2522 **
2523 ** If flattening is not attempted, this routine is a no-op and returns 0.
2524 ** If flattening is attempted this routine returns 1.
2525 **
2526 ** All of the expression analysis must occur on both the outer query and
2527 ** the subquery before this routine runs.
2528 */
2529 static int flattenSubquery(
2530 Parse *pParse, /* Parsing context */
2531 Select *p, /* The parent or outer SELECT statement */
2532 int iFrom, /* Index in p->pSrc->a[] of the inner subquery */
2533 int isAgg, /* True if outer SELECT uses aggregate functions */
2534 int subqueryIsAgg /* True if the subquery uses aggregate functions */
2535 ){
2536 const char *zSavedAuthContext = pParse->zAuthContext;
2537 Select *pParent;
2538 Select *pSub; /* The inner query or "subquery" */
2539 Select *pSub1; /* Pointer to the rightmost select in sub-query */
2540 SrcList *pSrc; /* The FROM clause of the outer query */
2541 SrcList *pSubSrc; /* The FROM clause of the subquery */
2542 ExprList *pList; /* The result set of the outer query */
2543 int iParent; /* VDBE cursor number of the pSub result set temp table */
2544 int i; /* Loop counter */
2545 Expr *pWhere; /* The WHERE clause */
2546 struct SrcList_item *pSubitem; /* The subquery */
2547 sqlite3 *db = pParse->db;
2548
2549 /* Check to see if flattening is permitted. Return 0 if not.
2550 */
2551 assert( p!=0 );
2552 assert( p->pPrior==0 ); /* Unable to flatten compound queries */
2553 pSrc = p->pSrc;
2554 assert( pSrc && iFrom>=0 && iFrom<pSrc->nSrc );
2555 pSubitem = &pSrc->a[iFrom];
2556 iParent = pSubitem->iCursor;
2557 pSub = pSubitem->pSelect;
2558 assert( pSub!=0 );
2559 if( isAgg && subqueryIsAgg ) return 0; /* Restriction (1) */
2560 if( subqueryIsAgg && pSrc->nSrc>1 ) return 0; /* Restriction (2) */
2561 pSubSrc = pSub->pSrc;
2562 assert( pSubSrc );
2563 /* Prior to version 3.1.2, when LIMIT and OFFSET had to be simple constants,
2564 ** not arbitrary expresssions, we allowed some combining of LIMIT and OFFSET
2565 ** because they could be computed at compile-time. But when LIMIT and OFFSET
2566 ** became arbitrary expressions, we were forced to add restrictions (13)
2567 ** and (14). */
2568 if( pSub->pLimit && p->pLimit ) return 0; /* Restriction (13) */
2569 if( pSub->pOffset ) return 0; /* Restriction (14) */
2570 if( p->pRightmost && pSub->pLimit && pSub->pOrderBy ){
2571 return 0; /* Restriction (15) */
2572 }
2573 if( pSubSrc->nSrc==0 ) return 0; /* Restriction (7) */
2574 if( ((pSub->selFlags & SF_Distinct)!=0 || pSub->pLimit)
2575 && (pSrc->nSrc>1 || isAgg) ){ /* Restrictions (4)(5)(8)(9) */
2576 return 0;
2577 }
2578 if( (p->selFlags & SF_Distinct)!=0 && subqueryIsAgg ){
2579 return 0; /* Restriction (6) */
2580 }
2581 if( p->pOrderBy && pSub->pOrderBy ){
2582 return 0; /* Restriction (11) */
2583 }
2584 if( isAgg && pSub->pOrderBy ) return 0; /* Restriction (16) */
2585 if( pSub->pLimit && p->pWhere ) return 0; /* Restriction (19) */
2586
2587 /* OBSOLETE COMMENT 1:
2588 ** Restriction 3: If the subquery is a join, make sure the subquery is
2589 ** not used as the right operand of an outer join. Examples of why this
2590 ** is not allowed:
2591 **
2592 ** t1 LEFT OUTER JOIN (t2 JOIN t3)
2593 **
2594 ** If we flatten the above, we would get
2595 **
2596 ** (t1 LEFT OUTER JOIN t2) JOIN t3
2597 **
2598 ** which is not at all the same thing.
2599 **
2600 ** OBSOLETE COMMENT 2:
2601 ** Restriction 12: If the subquery is the right operand of a left outer
2602 ** join, make sure the subquery has no WHERE clause.
2603 ** An examples of why this is not allowed:
2604 **
2605 ** t1 LEFT OUTER JOIN (SELECT * FROM t2 WHERE t2.x>0)
2606 **
2607 ** If we flatten the above, we would get
2608 **
2609 ** (t1 LEFT OUTER JOIN t2) WHERE t2.x>0
2610 **
2611 ** But the t2.x>0 test will always fail on a NULL row of t2, which
2612 ** effectively converts the OUTER JOIN into an INNER JOIN.
2613 **
2614 ** THIS OVERRIDES OBSOLETE COMMENTS 1 AND 2 ABOVE:
2615 ** Ticket #3300 shows that flattening the right term of a LEFT JOIN
2616 ** is fraught with danger. Best to avoid the whole thing. If the
2617 ** subquery is the right term of a LEFT JOIN, then do not flatten.
2618 */
2619 if( (pSubitem->jointype & JT_OUTER)!=0 ){
2620 return 0;
2621 }
2622
2623 /* Restriction 17: If the sub-query is a compound SELECT, then it must
2624 ** use only the UNION ALL operator. And none of the simple select queries
2625 ** that make up the compound SELECT are allowed to be aggregate or distinct
2626 ** queries.
2627 */
2628 if( pSub->pPrior ){
2629 if( pSub->pOrderBy ){
2630 return 0; /* Restriction 20 */
2631 }
2632 if( isAgg || (p->selFlags & SF_Distinct)!=0 || pSrc->nSrc!=1 ){
2633 return 0;
2634 }
2635 for(pSub1=pSub; pSub1; pSub1=pSub1->pPrior){
2636 testcase( (pSub1->selFlags & (SF_Distinct|SF_Aggregate))==SF_Distinct );
2637 testcase( (pSub1->selFlags & (SF_Distinct|SF_Aggregate))==SF_Aggregate );
2638 if( (pSub1->selFlags & (SF_Distinct|SF_Aggregate))!=0
2639 || (pSub1->pPrior && pSub1->op!=TK_ALL)
2640 || NEVER(pSub1->pSrc==0) || pSub1->pSrc->nSrc!=1
2641 ){
2642 return 0;
2643 }
2644 }
2645
2646 /* Restriction 18. */
2647 if( p->pOrderBy ){
2648 int ii;
2649 for(ii=0; ii<p->pOrderBy->nExpr; ii++){
2650 if( p->pOrderBy->a[ii].iCol==0 ) return 0;
2651 }
2652 }
2653 }
2654
2655 /***** If we reach this point, flattening is permitted. *****/
2656
2657 /* Authorize the subquery */
2658 pParse->zAuthContext = pSubitem->zName;
2659 sqlite3AuthCheck(pParse, SQLITE_SELECT, 0, 0, 0);
2660 pParse->zAuthContext = zSavedAuthContext;
2661
2662 /* If the sub-query is a compound SELECT statement, then (by restrictions
2663 ** 17 and 18 above) it must be a UNION ALL and the parent query must
2664 ** be of the form:
2665 **
2666 ** SELECT <expr-list> FROM (<sub-query>) <where-clause>
2667 **
2668 ** followed by any ORDER BY, LIMIT and/or OFFSET clauses. This block
2669 ** creates N-1 copies of the parent query without any ORDER BY, LIMIT or
2670 ** OFFSET clauses and joins them to the left-hand-side of the original
2671 ** using UNION ALL operators. In this case N is the number of simple
2672 ** select statements in the compound sub-query.
2673 **
2674 ** Example:
2675 **
2676 ** SELECT a+1 FROM (
2677 ** SELECT x FROM tab
2678 ** UNION ALL
2679 ** SELECT y FROM tab
2680 ** UNION ALL
2681 ** SELECT abs(z*2) FROM tab2
2682 ** ) WHERE a!=5 ORDER BY 1
2683 **
2684 ** Transformed into:
2685 **
2686 ** SELECT x+1 FROM tab WHERE x+1!=5
2687 ** UNION ALL
2688 ** SELECT y+1 FROM tab WHERE y+1!=5
2689 ** UNION ALL
2690 ** SELECT abs(z*2)+1 FROM tab2 WHERE abs(z*2)+1!=5
2691 ** ORDER BY 1
2692 **
2693 ** We call this the "compound-subquery flattening".
2694 */
2695 for(pSub=pSub->pPrior; pSub; pSub=pSub->pPrior){
2696 Select *pNew;
2697 ExprList *pOrderBy = p->pOrderBy;
2698 Expr *pLimit = p->pLimit;
2699 Select *pPrior = p->pPrior;
2700 p->pOrderBy = 0;
2701 p->pSrc = 0;
2702 p->pPrior = 0;
2703 p->pLimit = 0;
2704 pNew = sqlite3SelectDup(db, p, 0);
2705 p->pLimit = pLimit;
2706 p->pOrderBy = pOrderBy;
2707 p->pSrc = pSrc;
2708 p->op = TK_ALL;
2709 p->pRightmost = 0;
2710 if( pNew==0 ){
2711 pNew = pPrior;
2712 }else{
2713 pNew->pPrior = pPrior;
2714 pNew->pRightmost = 0;
2715 }
2716 p->pPrior = pNew;
2717 if( db->mallocFailed ) return 1;
2718 }
2719
2720 /* Begin flattening the iFrom-th entry of the FROM clause
2721 ** in the outer query.
2722 */
2723 pSub = pSub1 = pSubitem->pSelect;
2724
2725 /* Delete the transient table structure associated with the
2726 ** subquery
2727 */
2728 sqlite3DbFree(db, pSubitem->zDatabase);
2729 sqlite3DbFree(db, pSubitem->zName);
2730 sqlite3DbFree(db, pSubitem->zAlias);
2731 pSubitem->zDatabase = 0;
2732 pSubitem->zName = 0;
2733 pSubitem->zAlias = 0;
2734 pSubitem->pSelect = 0;
2735
2736 /* Defer deleting the Table object associated with the
2737 ** subquery until code generation is
2738 ** complete, since there may still exist Expr.pTab entries that
2739 ** refer to the subquery even after flattening. Ticket #3346.
2740 **
2741 ** pSubitem->pTab is always non-NULL by test restrictions and tests above.
2742 */
2743 if( ALWAYS(pSubitem->pTab!=0) ){
2744 Table *pTabToDel = pSubitem->pTab;
2745 if( pTabToDel->nRef==1 ){
2746 Parse *pToplevel = sqlite3ParseToplevel(pParse);
2747 pTabToDel->pNextZombie = pToplevel->pZombieTab;
2748 pToplevel->pZombieTab = pTabToDel;
2749 }else{
2750 pTabToDel->nRef--;
2751 }
2752 pSubitem->pTab = 0;
2753 }
2754
2755 /* The following loop runs once for each term in a compound-subquery
2756 ** flattening (as described above). If we are doing a different kind
2757 ** of flattening - a flattening other than a compound-subquery flattening -
2758 ** then this loop only runs once.
2759 **
2760 ** This loop moves all of the FROM elements of the subquery into the
2761 ** the FROM clause of the outer query. Before doing this, remember
2762 ** the cursor number for the original outer query FROM element in
2763 ** iParent. The iParent cursor will never be used. Subsequent code
2764 ** will scan expressions looking for iParent references and replace
2765 ** those references with expressions that resolve to the subquery FROM
2766 ** elements we are now copying in.
2767 */
2768 for(pParent=p; pParent; pParent=pParent->pPrior, pSub=pSub->pPrior){
2769 int nSubSrc;
2770 u8 jointype = 0;
2771 pSubSrc = pSub->pSrc; /* FROM clause of subquery */
2772 nSubSrc = pSubSrc->nSrc; /* Number of terms in subquery FROM clause */
2773 pSrc = pParent->pSrc; /* FROM clause of the outer query */
2774
2775 if( pSrc ){
2776 assert( pParent==p ); /* First time through the loop */
2777 jointype = pSubitem->jointype;
2778 }else{
2779 assert( pParent!=p ); /* 2nd and subsequent times through the loop */
2780 pSrc = pParent->pSrc = sqlite3SrcListAppend(db, 0, 0, 0);
2781 if( pSrc==0 ){
2782 assert( db->mallocFailed );
2783 break;
2784 }
2785 }
2786
2787 /* The subquery uses a single slot of the FROM clause of the outer
2788 ** query. If the subquery has more than one element in its FROM clause,
2789 ** then expand the outer query to make space for it to hold all elements
2790 ** of the subquery.
2791 **
2792 ** Example:
2793 **
2794 ** SELECT * FROM tabA, (SELECT * FROM sub1, sub2), tabB;
2795 **
2796 ** The outer query has 3 slots in its FROM clause. One slot of the
2797 ** outer query (the middle slot) is used by the subquery. The next
2798 ** block of code will expand the out query to 4 slots. The middle
2799 ** slot is expanded to two slots in order to make space for the
2800 ** two elements in the FROM clause of the subquery.
2801 */
2802 if( nSubSrc>1 ){
2803 pParent->pSrc = pSrc = sqlite3SrcListEnlarge(db, pSrc, nSubSrc-1,iFrom+1);
2804 if( db->mallocFailed ){
2805 break;
2806 }
2807 }
2808
2809 /* Transfer the FROM clause terms from the subquery into the
2810 ** outer query.
2811 */
2812 for(i=0; i<nSubSrc; i++){
2813 sqlite3IdListDelete(db, pSrc->a[i+iFrom].pUsing);
2814 pSrc->a[i+iFrom] = pSubSrc->a[i];
2815 memset(&pSubSrc->a[i], 0, sizeof(pSubSrc->a[i]));
2816 }
2817 pSrc->a[iFrom].jointype = jointype;
2818
2819 /* Now begin substituting subquery result set expressions for
2820 ** references to the iParent in the outer query.
2821 **
2822 ** Example:
2823 **
2824 ** SELECT a+5, b*10 FROM (SELECT x*3 AS a, y+10 AS b FROM t1) WHERE a>b;
2825 ** \ \_____________ subquery __________/ /
2826 ** \_____________________ outer query ______________________________/
2827 **
2828 ** We look at every expression in the outer query and every place we see
2829 ** "a" we substitute "x*3" and every place we see "b" we substitute "y+10".
2830 */
2831 pList = pParent->pEList;
2832 for(i=0; i<pList->nExpr; i++){
2833 if( pList->a[i].zName==0 ){
2834 const char *zSpan = pList->a[i].zSpan;
2835 if( ALWAYS(zSpan) ){
2836 pList->a[i].zName = sqlite3DbStrDup(db, zSpan);
2837 }
2838 }
2839 }
2840 substExprList(db, pParent->pEList, iParent, pSub->pEList);
2841 if( isAgg ){
2842 substExprList(db, pParent->pGroupBy, iParent, pSub->pEList);
2843 pParent->pHaving = substExpr(db, pParent->pHaving, iParent, pSub->pEList);
2844 }
2845 if( pSub->pOrderBy ){
2846 assert( pParent->pOrderBy==0 );
2847 pParent->pOrderBy = pSub->pOrderBy;
2848 pSub->pOrderBy = 0;
2849 }else if( pParent->pOrderBy ){
2850 substExprList(db, pParent->pOrderBy, iParent, pSub->pEList);
2851 }
2852 if( pSub->pWhere ){
2853 pWhere = sqlite3ExprDup(db, pSub->pWhere, 0);
2854 }else{
2855 pWhere = 0;
2856 }
2857 if( subqueryIsAgg ){
2858 assert( pParent->pHaving==0 );
2859 pParent->pHaving = pParent->pWhere;
2860 pParent->pWhere = pWhere;
2861 pParent->pHaving = substExpr(db, pParent->pHaving, iParent, pSub->pEList);
2862 pParent->pHaving = sqlite3ExprAnd(db, pParent->pHaving,
2863 sqlite3ExprDup(db, pSub->pHaving, 0));
2864 assert( pParent->pGroupBy==0 );
2865 pParent->pGroupBy = sqlite3ExprListDup(db, pSub->pGroupBy, 0);
2866 }else{
2867 pParent->pWhere = substExpr(db, pParent->pWhere, iParent, pSub->pEList);
2868 pParent->pWhere = sqlite3ExprAnd(db, pParent->pWhere, pWhere);
2869 }
2870
2871 /* The flattened query is distinct if either the inner or the
2872 ** outer query is distinct.
2873 */
2874 pParent->selFlags |= pSub->selFlags & SF_Distinct;
2875
2876 /*
2877 ** SELECT ... FROM (SELECT ... LIMIT a OFFSET b) LIMIT x OFFSET y;
2878 **
2879 ** One is tempted to try to add a and b to combine the limits. But this
2880 ** does not work if either limit is negative.
2881 */
2882 if( pSub->pLimit ){
2883 pParent->pLimit = pSub->pLimit;
2884 pSub->pLimit = 0;
2885 }
2886 }
2887
2888 /* Finially, delete what is left of the subquery and return
2889 ** success.
2890 */
2891 sqlite3SelectDelete(db, pSub1);
2892
2893 return 1;
2894 }
2895 #endif /* !defined(SQLITE_OMIT_SUBQUERY) || !defined(SQLITE_OMIT_VIEW) */
2896
2897 /*
2898 ** Analyze the SELECT statement passed as an argument to see if it
2899 ** is a min() or max() query. Return WHERE_ORDERBY_MIN or WHERE_ORDERBY_MAX if
2900 ** it is, or 0 otherwise. At present, a query is considered to be
2901 ** a min()/max() query if:
2902 **
2903 ** 1. There is a single object in the FROM clause.
2904 **
2905 ** 2. There is a single expression in the result set, and it is
2906 ** either min(x) or max(x), where x is a column reference.
2907 */
2908 static u8 minMaxQuery(Select *p){
2909 Expr *pExpr;
2910 ExprList *pEList = p->pEList;
2911
2912 if( pEList->nExpr!=1 ) return WHERE_ORDERBY_NORMAL;
2913 pExpr = pEList->a[0].pExpr;
2914 if( pExpr->op!=TK_AGG_FUNCTION ) return 0;
2915 if( NEVER(ExprHasProperty(pExpr, EP_xIsSelect)) ) return 0;
2916 pEList = pExpr->x.pList;
2917 if( pEList==0 || pEList->nExpr!=1 ) return 0;
2918 if( pEList->a[0].pExpr->op!=TK_AGG_COLUMN ) return WHERE_ORDERBY_NORMAL;
2919 assert( !ExprHasProperty(pExpr, EP_IntValue) );
2920 if( sqlite3StrICmp(pExpr->u.zToken,"min")==0 ){
2921 return WHERE_ORDERBY_MIN;
2922 }else if( sqlite3StrICmp(pExpr->u.zToken,"max")==0 ){
2923 return WHERE_ORDERBY_MAX;
2924 }
2925 return WHERE_ORDERBY_NORMAL;
2926 }
2927
2928 /*
2929 ** The select statement passed as the first argument is an aggregate query.
2930 ** The second argment is the associated aggregate-info object. This
2931 ** function tests if the SELECT is of the form:
2932 **
2933 ** SELECT count(*) FROM <tbl>
2934 **
2935 ** where table is a database table, not a sub-select or view. If the query
2936 ** does match this pattern, then a pointer to the Table object representing
2937 ** <tbl> is returned. Otherwise, 0 is returned.
2938 */
2939 static Table *isSimpleCount(Select *p, AggInfo *pAggInfo){
2940 Table *pTab;
2941 Expr *pExpr;
2942
2943 assert( !p->pGroupBy );
2944
2945 if( p->pWhere || p->pEList->nExpr!=1
2946 || p->pSrc->nSrc!=1 || p->pSrc->a[0].pSelect
2947 ){
2948 return 0;
2949 }
2950 pTab = p->pSrc->a[0].pTab;
2951 pExpr = p->pEList->a[0].pExpr;
2952 assert( pTab && !pTab->pSelect && pExpr );
2953
2954 if( IsVirtual(pTab) ) return 0;
2955 if( pExpr->op!=TK_AGG_FUNCTION ) return 0;
2956 if( (pAggInfo->aFunc[0].pFunc->flags&SQLITE_FUNC_COUNT)==0 ) return 0;
2957 if( pExpr->flags&EP_Distinct ) return 0;
2958
2959 return pTab;
2960 }
2961
2962 /*
2963 ** If the source-list item passed as an argument was augmented with an
2964 ** INDEXED BY clause, then try to locate the specified index. If there
2965 ** was such a clause and the named index cannot be found, return
2966 ** SQLITE_ERROR and leave an error in pParse. Otherwise, populate
2967 ** pFrom->pIndex and return SQLITE_OK.
2968 */
2969 int sqlite3IndexedByLookup(Parse *pParse, struct SrcList_item *pFrom){
2970 if( pFrom->pTab && pFrom->zIndex ){
2971 Table *pTab = pFrom->pTab;
2972 char *zIndex = pFrom->zIndex;
2973 Index *pIdx;
2974 for(pIdx=pTab->pIndex;
2975 pIdx && sqlite3StrICmp(pIdx->zName, zIndex);
2976 pIdx=pIdx->pNext
2977 );
2978 if( !pIdx ){
2979 sqlite3ErrorMsg(pParse, "no such index: %s", zIndex, 0);
2980 return SQLITE_ERROR;
2981 }
2982 pFrom->pIndex = pIdx;
2983 }
2984 return SQLITE_OK;
2985 }
2986
2987 /*
2988 ** This routine is a Walker callback for "expanding" a SELECT statement.
2989 ** "Expanding" means to do the following:
2990 **
2991 ** (1) Make sure VDBE cursor numbers have been assigned to every
2992 ** element of the FROM clause.
2993 **
2994 ** (2) Fill in the pTabList->a[].pTab fields in the SrcList that
2995 ** defines FROM clause. When views appear in the FROM clause,
2996 ** fill pTabList->a[].pSelect with a copy of the SELECT statement
2997 ** that implements the view. A copy is made of the view's SELECT
2998 ** statement so that we can freely modify or delete that statement
2999 ** without worrying about messing up the presistent representation
3000 ** of the view.
3001 **
3002 ** (3) Add terms to the WHERE clause to accomodate the NATURAL keyword
3003 ** on joins and the ON and USING clause of joins.
3004 **
3005 ** (4) Scan the list of columns in the result set (pEList) looking
3006 ** for instances of the "*" operator or the TABLE.* operator.
3007 ** If found, expand each "*" to be every column in every table
3008 ** and TABLE.* to be every column in TABLE.
3009 **
3010 */
3011 static int selectExpander(Walker *pWalker, Select *p){
3012 Parse *pParse = pWalker->pParse;
3013 int i, j, k;
3014 SrcList *pTabList;
3015 ExprList *pEList;
3016 struct SrcList_item *pFrom;
3017 sqlite3 *db = pParse->db;
3018
3019 if( db->mallocFailed ){
3020 return WRC_Abort;
3021 }
3022 if( NEVER(p->pSrc==0) || (p->selFlags & SF_Expanded)!=0 ){
3023 return WRC_Prune;
3024 }
3025 p->selFlags |= SF_Expanded;
3026 pTabList = p->pSrc;
3027 pEList = p->pEList;
3028
3029 /* Make sure cursor numbers have been assigned to all entries in
3030 ** the FROM clause of the SELECT statement.
3031 */
3032 sqlite3SrcListAssignCursors(pParse, pTabList);
3033
3034 /* Look up every table named in the FROM clause of the select. If
3035 ** an entry of the FROM clause is a subquery instead of a table or view,
3036 ** then create a transient table structure to describe the subquery.
3037 */
3038 for(i=0, pFrom=pTabList->a; i<pTabList->nSrc; i++, pFrom++){
3039 Table *pTab;
3040 if( pFrom->pTab!=0 ){
3041 /* This statement has already been prepared. There is no need
3042 ** to go further. */
3043 assert( i==0 );
3044 return WRC_Prune;
3045 }
3046 if( pFrom->zName==0 ){
3047 #ifndef SQLITE_OMIT_SUBQUERY
3048 Select *pSel = pFrom->pSelect;
3049 /* A sub-query in the FROM clause of a SELECT */
3050 assert( pSel!=0 );
3051 assert( pFrom->pTab==0 );
3052 sqlite3WalkSelect(pWalker, pSel);
3053 pFrom->pTab = pTab = sqlite3DbMallocZero(db, sizeof(Table));
3054 if( pTab==0 ) return WRC_Abort;
3055 pTab->dbMem = db->lookaside.bEnabled ? db : 0;
3056 pTab->nRef = 1;
3057 pTab->zName = sqlite3MPrintf(db, "sqlite_subquery_%p_", (void*)pTab);
3058 while( pSel->pPrior ){ pSel = pSel->pPrior; }
3059 selectColumnsFromExprList(pParse, pSel->pEList, &pTab->nCol, &pTab->aCol);
3060 pTab->iPKey = -1;
3061 pTab->tabFlags |= TF_Ephemeral;
3062 #endif
3063 }else{
3064 /* An ordinary table or view name in the FROM clause */
3065 assert( pFrom->pTab==0 );
3066 pFrom->pTab = pTab =
3067 sqlite3LocateTable(pParse,0,pFrom->zName,pFrom->zDatabase);
3068 if( pTab==0 ) return WRC_Abort;
3069 pTab->nRef++;
3070 #if !defined(SQLITE_OMIT_VIEW) || !defined (SQLITE_OMIT_VIRTUALTABLE)
3071 if( pTab->pSelect || IsVirtual(pTab) ){
3072 /* We reach here if the named table is a really a view */
3073 if( sqlite3ViewGetColumnNames(pParse, pTab) ) return WRC_Abort;
3074 assert( pFrom->pSelect==0 );
3075 pFrom->pSelect = sqlite3SelectDup(db, pTab->pSelect, 0);
3076 sqlite3WalkSelect(pWalker, pFrom->pSelect);
3077 }
3078 #endif
3079 }
3080
3081 /* Locate the index named by the INDEXED BY clause, if any. */
3082 if( sqlite3IndexedByLookup(pParse, pFrom) ){
3083 return WRC_Abort;
3084 }
3085 }
3086
3087 /* Process NATURAL keywords, and ON and USING clauses of joins.
3088 */
3089 if( db->mallocFailed || sqliteProcessJoin(pParse, p) ){
3090 return WRC_Abort;
3091 }
3092
3093 /* For every "*" that occurs in the column list, insert the names of
3094 ** all columns in all tables. And for every TABLE.* insert the names
3095 ** of all columns in TABLE. The parser inserted a special expression
3096 ** with the TK_ALL operator for each "*" that it found in the column list.
3097 ** The following code just has to locate the TK_ALL expressions and expand
3098 ** each one to the list of all columns in all tables.
3099 **
3100 ** The first loop just checks to see if there are any "*" operators
3101 ** that need expanding.
3102 */
3103 for(k=0; k<pEList->nExpr; k++){
3104 Expr *pE = pEList->a[k].pExpr;
3105 if( pE->op==TK_ALL ) break;
3106 assert( pE->op!=TK_DOT || pE->pRight!=0 );
3107 assert( pE->op!=TK_DOT || (pE->pLeft!=0 && pE->pLeft->op==TK_ID) );
3108 if( pE->op==TK_DOT && pE->pRight->op==TK_ALL ) break;
3109 }
3110 if( k<pEList->nExpr ){
3111 /*
3112 ** If we get here it means the result set contains one or more "*"
3113 ** operators that need to be expanded. Loop through each expression
3114 ** in the result set and expand them one by one.
3115 */
3116 struct ExprList_item *a = pEList->a;
3117 ExprList *pNew = 0;
3118 int flags = pParse->db->flags;
3119 int longNames = (flags & SQLITE_FullColNames)!=0
3120 && (flags & SQLITE_ShortColNames)==0;
3121
3122 for(k=0; k<pEList->nExpr; k++){
3123 Expr *pE = a[k].pExpr;
3124 assert( pE->op!=TK_DOT || pE->pRight!=0 );
3125 if( pE->op!=TK_ALL && (pE->op!=TK_DOT || pE->pRight->op!=TK_ALL) ){
3126 /* This particular expression does not need to be expanded.
3127 */
3128 pNew = sqlite3ExprListAppend(pParse, pNew, a[k].pExpr);
3129 if( pNew ){
3130 pNew->a[pNew->nExpr-1].zName = a[k].zName;
3131 pNew->a[pNew->nExpr-1].zSpan = a[k].zSpan;
3132 a[k].zName = 0;
3133 a[k].zSpan = 0;
3134 }
3135 a[k].pExpr = 0;
3136 }else{
3137 /* This expression is a "*" or a "TABLE.*" and needs to be
3138 ** expanded. */
3139 int tableSeen = 0; /* Set to 1 when TABLE matches */
3140 char *zTName; /* text of name of TABLE */
3141 if( pE->op==TK_DOT ){
3142 assert( pE->pLeft!=0 );
3143 assert( !ExprHasProperty(pE->pLeft, EP_IntValue) );
3144 zTName = pE->pLeft->u.zToken;
3145 }else{
3146 zTName = 0;
3147 }
3148 for(i=0, pFrom=pTabList->a; i<pTabList->nSrc; i++, pFrom++){
3149 Table *pTab = pFrom->pTab;
3150 char *zTabName = pFrom->zAlias;
3151 if( zTabName==0 ){
3152 zTabName = pTab->zName;
3153 }
3154 if( db->mallocFailed ) break;
3155 if( zTName && sqlite3StrICmp(zTName, zTabName)!=0 ){
3156 continue;
3157 }
3158 tableSeen = 1;
3159 for(j=0; j<pTab->nCol; j++){
3160 Expr *pExpr, *pRight;
3161 char *zName = pTab->aCol[j].zName;
3162 char *zColname; /* The computed column name */
3163 char *zToFree; /* Malloced string that needs to be freed */
3164 Token sColname; /* Computed column name as a token */
3165
3166 /* If a column is marked as 'hidden' (currently only possible
3167 ** for virtual tables), do not include it in the expanded
3168 ** result-set list.
3169 */
3170 if( IsHiddenColumn(&pTab->aCol[j]) ){
3171 assert(IsVirtual(pTab));
3172 continue;
3173 }
3174
3175 if( i>0 && zTName==0 ){
3176 struct SrcList_item *pLeft = &pTabList->a[i-1];
3177 if( (pLeft[1].jointype & JT_NATURAL)!=0 &&
3178 columnIndex(pLeft->pTab, zName)>=0 ){
3179 /* In a NATURAL join, omit the join columns from the
3180 ** table on the right */
3181 continue;
3182 }
3183 if( sqlite3IdListIndex(pLeft[1].pUsing, zName)>=0 ){
3184 /* In a join with a USING clause, omit columns in the
3185 ** using clause from the table on the right. */
3186 continue;
3187 }
3188 }
3189 pRight = sqlite3Expr(db, TK_ID, zName);
3190 zColname = zName;
3191 zToFree = 0;
3192 if( longNames || pTabList->nSrc>1 ){
3193 Expr *pLeft;
3194 pLeft = sqlite3Expr(db, TK_ID, zTabName);
3195 pExpr = sqlite3PExpr(pParse, TK_DOT, pLeft, pRight, 0);
3196 if( longNames ){
3197 zColname = sqlite3MPrintf(db, "%s.%s", zTabName, zName);
3198 zToFree = zColname;
3199 }
3200 }else{
3201 pExpr = pRight;
3202 }
3203 pNew = sqlite3ExprListAppend(pParse, pNew, pExpr);
3204 sColname.z = zColname;
3205 sColname.n = sqlite3Strlen30(zColname);
3206 sqlite3ExprListSetName(pParse, pNew, &sColname, 0);
3207 sqlite3DbFree(db, zToFree);
3208 }
3209 }
3210 if( !tableSeen ){
3211 if( zTName ){
3212 sqlite3ErrorMsg(pParse, "no such table: %s", zTName);
3213 }else{
3214 sqlite3ErrorMsg(pParse, "no tables specified");
3215 }
3216 }
3217 }
3218 }
3219 sqlite3ExprListDelete(db, pEList);
3220 p->pEList = pNew;
3221 }
3222 #if SQLITE_MAX_COLUMN
3223 if( p->pEList && p->pEList->nExpr>db->aLimit[SQLITE_LIMIT_COLUMN] ){
3224 sqlite3ErrorMsg(pParse, "too many columns in result set");
3225 }
3226 #endif
3227 return WRC_Continue;
3228 }
3229
3230 /*
3231 ** No-op routine for the parse-tree walker.
3232 **
3233 ** When this routine is the Walker.xExprCallback then expression trees
3234 ** are walked without any actions being taken at each node. Presumably,
3235 ** when this routine is used for Walker.xExprCallback then
3236 ** Walker.xSelectCallback is set to do something useful for every
3237 ** subquery in the parser tree.
3238 */
3239 static int exprWalkNoop(Walker *NotUsed, Expr *NotUsed2){
3240 UNUSED_PARAMETER2(NotUsed, NotUsed2);
3241 return WRC_Continue;
3242 }
3243
3244 /*
3245 ** This routine "expands" a SELECT statement and all of its subqueries.
3246 ** For additional information on what it means to "expand" a SELECT
3247 ** statement, see the comment on the selectExpand worker callback above.
3248 **
3249 ** Expanding a SELECT statement is the first step in processing a
3250 ** SELECT statement. The SELECT statement must be expanded before
3251 ** name resolution is performed.
3252 **
3253 ** If anything goes wrong, an error message is written into pParse.
3254 ** The calling function can detect the problem by looking at pParse->nErr
3255 ** and/or pParse->db->mallocFailed.
3256 */
3257 static void sqlite3SelectExpand(Parse *pParse, Select *pSelect){
3258 Walker w;
3259 w.xSelectCallback = selectExpander;
3260 w.xExprCallback = exprWalkNoop;
3261 w.pParse = pParse;
3262 sqlite3WalkSelect(&w, pSelect);
3263 }
3264
3265
3266 #ifndef SQLITE_OMIT_SUBQUERY
3267 /*
3268 ** This is a Walker.xSelectCallback callback for the sqlite3SelectTypeInfo()
3269 ** interface.
3270 **
3271 ** For each FROM-clause subquery, add Column.zType and Column.zColl
3272 ** information to the Table structure that represents the result set
3273 ** of that subquery.
3274 **
3275 ** The Table structure that represents the result set was constructed
3276 ** by selectExpander() but the type and collation information was omitted
3277 ** at that point because identifiers had not yet been resolved. This
3278 ** routine is called after identifier resolution.
3279 */
3280 static int selectAddSubqueryTypeInfo(Walker *pWalker, Select *p){
3281 Parse *pParse;
3282 int i;
3283 SrcList *pTabList;
3284 struct SrcList_item *pFrom;
3285
3286 assert( p->selFlags & SF_Resolved );
3287 assert( (p->selFlags & SF_HasTypeInfo)==0 );
3288 p->selFlags |= SF_HasTypeInfo;
3289 pParse = pWalker->pParse;
3290 pTabList = p->pSrc;
3291 for(i=0, pFrom=pTabList->a; i<pTabList->nSrc; i++, pFrom++){
3292 Table *pTab = pFrom->pTab;
3293 if( ALWAYS(pTab!=0) && (pTab->tabFlags & TF_Ephemeral)!=0 ){
3294 /* A sub-query in the FROM clause of a SELECT */
3295 Select *pSel = pFrom->pSelect;
3296 assert( pSel );
3297 while( pSel->pPrior ) pSel = pSel->pPrior;
3298 selectAddColumnTypeAndCollation(pParse, pTab->nCol, pTab->aCol, pSel);
3299 }
3300 }
3301 return WRC_Continue;
3302 }
3303 #endif
3304
3305
3306 /*
3307 ** This routine adds datatype and collating sequence information to
3308 ** the Table structures of all FROM-clause subqueries in a
3309 ** SELECT statement.
3310 **
3311 ** Use this routine after name resolution.
3312 */
3313 static void sqlite3SelectAddTypeInfo(Parse *pParse, Select *pSelect){
3314 #ifndef SQLITE_OMIT_SUBQUERY
3315 Walker w;
3316 w.xSelectCallback = selectAddSubqueryTypeInfo;
3317 w.xExprCallback = exprWalkNoop;
3318 w.pParse = pParse;
3319 sqlite3WalkSelect(&w, pSelect);
3320 #endif
3321 }
3322
3323
3324 /*
3325 ** This routine sets of a SELECT statement for processing. The
3326 ** following is accomplished:
3327 **
3328 ** * VDBE Cursor numbers are assigned to all FROM-clause terms.
3329 ** * Ephemeral Table objects are created for all FROM-clause subqueries.
3330 ** * ON and USING clauses are shifted into WHERE statements
3331 ** * Wildcards "*" and "TABLE.*" in result sets are expanded.
3332 ** * Identifiers in expression are matched to tables.
3333 **
3334 ** This routine acts recursively on all subqueries within the SELECT.
3335 */
3336 void sqlite3SelectPrep(
3337 Parse *pParse, /* The parser context */
3338 Select *p, /* The SELECT statement being coded. */
3339 NameContext *pOuterNC /* Name context for container */
3340 ){
3341 sqlite3 *db;
3342 if( NEVER(p==0) ) return;
3343 db = pParse->db;
3344 if( p->selFlags & SF_HasTypeInfo ) return;
3345 sqlite3SelectExpand(pParse, p);
3346 if( pParse->nErr || db->mallocFailed ) return;
3347 sqlite3ResolveSelectNames(pParse, p, pOuterNC);
3348 if( pParse->nErr || db->mallocFailed ) return;
3349 sqlite3SelectAddTypeInfo(pParse, p);
3350 }
3351
3352 /*
3353 ** Reset the aggregate accumulator.
3354 **
3355 ** The aggregate accumulator is a set of memory cells that hold
3356 ** intermediate results while calculating an aggregate. This
3357 ** routine simply stores NULLs in all of those memory cells.
3358 */
3359 static void resetAccumulator(Parse *pParse, AggInfo *pAggInfo){
3360 Vdbe *v = pParse->pVdbe;
3361 int i;
3362 struct AggInfo_func *pFunc;
3363 if( pAggInfo->nFunc+pAggInfo->nColumn==0 ){
3364 return;
3365 }
3366 for(i=0; i<pAggInfo->nColumn; i++){
3367 sqlite3VdbeAddOp2(v, OP_Null, 0, pAggInfo->aCol[i].iMem);
3368 }
3369 for(pFunc=pAggInfo->aFunc, i=0; i<pAggInfo->nFunc; i++, pFunc++){
3370 sqlite3VdbeAddOp2(v, OP_Null, 0, pFunc->iMem);
3371 if( pFunc->iDistinct>=0 ){
3372 Expr *pE = pFunc->pExpr;
3373 assert( !ExprHasProperty(pE, EP_xIsSelect) );
3374 if( pE->x.pList==0 || pE->x.pList->nExpr!=1 ){
3375 sqlite3ErrorMsg(pParse, "DISTINCT aggregates must have exactly one "
3376 "argument");
3377 pFunc->iDistinct = -1;
3378 }else{
3379 KeyInfo *pKeyInfo = keyInfoFromExprList(pParse, pE->x.pList);
3380 sqlite3VdbeAddOp4(v, OP_OpenEphemeral, pFunc->iDistinct, 0, 0,
3381 (char*)pKeyInfo, P4_KEYINFO_HANDOFF);
3382 }
3383 }
3384 }
3385 }
3386
3387 /*
3388 ** Invoke the OP_AggFinalize opcode for every aggregate function
3389 ** in the AggInfo structure.
3390 */
3391 static void finalizeAggFunctions(Parse *pParse, AggInfo *pAggInfo){
3392 Vdbe *v = pParse->pVdbe;
3393 int i;
3394 struct AggInfo_func *pF;
3395 for(i=0, pF=pAggInfo->aFunc; i<pAggInfo->nFunc; i++, pF++){
3396 ExprList *pList = pF->pExpr->x.pList;
3397 assert( !ExprHasProperty(pF->pExpr, EP_xIsSelect) );
3398 sqlite3VdbeAddOp4(v, OP_AggFinal, pF->iMem, pList ? pList->nExpr : 0, 0,
3399 (void*)pF->pFunc, P4_FUNCDEF);
3400 }
3401 }
3402
3403 /*
3404 ** Update the accumulator memory cells for an aggregate based on
3405 ** the current cursor position.
3406 */
3407 static void updateAccumulator(Parse *pParse, AggInfo *pAggInfo){
3408 Vdbe *v = pParse->pVdbe;
3409 int i;
3410 struct AggInfo_func *pF;
3411 struct AggInfo_col *pC;
3412
3413 pAggInfo->directMode = 1;
3414 sqlite3ExprCacheClear(pParse);
3415 for(i=0, pF=pAggInfo->aFunc; i<pAggInfo->nFunc; i++, pF++){
3416 int nArg;
3417 int addrNext = 0;
3418 int regAgg;
3419 ExprList *pList = pF->pExpr->x.pList;
3420 assert( !ExprHasProperty(pF->pExpr, EP_xIsSelect) );
3421 if( pList ){
3422 nArg = pList->nExpr;
3423 regAgg = sqlite3GetTempRange(pParse, nArg);
3424 sqlite3ExprCodeExprList(pParse, pList, regAgg, 0);
3425 }else{
3426 nArg = 0;
3427 regAgg = 0;
3428 }
3429 if( pF->iDistinct>=0 ){
3430 addrNext = sqlite3VdbeMakeLabel(v);
3431 assert( nArg==1 );
3432 codeDistinct(pParse, pF->iDistinct, addrNext, 1, regAgg);
3433 }
3434 if( pF->pFunc->flags & SQLITE_FUNC_NEEDCOLL ){
3435 CollSeq *pColl = 0;
3436 struct ExprList_item *pItem;
3437 int j;
3438 assert( pList!=0 ); /* pList!=0 if pF->pFunc has NEEDCOLL */
3439 for(j=0, pItem=pList->a; !pColl && j<nArg; j++, pItem++){
3440 pColl = sqlite3ExprCollSeq(pParse, pItem->pExpr);
3441 }
3442 if( !pColl ){
3443 pColl = pParse->db->pDfltColl;
3444 }
3445 sqlite3VdbeAddOp4(v, OP_CollSeq, 0, 0, 0, (char *)pColl, P4_COLLSEQ);
3446 }
3447 sqlite3VdbeAddOp4(v, OP_AggStep, 0, regAgg, pF->iMem,
3448 (void*)pF->pFunc, P4_FUNCDEF);
3449 sqlite3VdbeChangeP5(v, (u8)nArg);
3450 sqlite3ReleaseTempRange(pParse, regAgg, nArg);
3451 sqlite3ExprCacheAffinityChange(pParse, regAgg, nArg);
3452 if( addrNext ){
3453 sqlite3VdbeResolveLabel(v, addrNext);
3454 sqlite3ExprCacheClear(pParse);
3455 }
3456 }
3457 for(i=0, pC=pAggInfo->aCol; i<pAggInfo->nAccumulator; i++, pC++){
3458 sqlite3ExprCode(pParse, pC->pExpr, pC->iMem);
3459 }
3460 pAggInfo->directMode = 0;
3461 sqlite3ExprCacheClear(pParse);
3462 }
3463
3464 /*
3465 ** Generate code for the SELECT statement given in the p argument.
3466 **
3467 ** The results are distributed in various ways depending on the
3468 ** contents of the SelectDest structure pointed to by argument pDest
3469 ** as follows:
3470 **
3471 ** pDest->eDest Result
3472 ** ------------ -------------------------------------------
3473 ** SRT_Output Generate a row of output (using the OP_ResultRow
3474 ** opcode) for each row in the result set.
3475 **
3476 ** SRT_Mem Only valid if the result is a single column.
3477 ** Store the first column of the first result row
3478 ** in register pDest->iParm then abandon the rest
3479 ** of the query. This destination implies "LIMIT 1".
3480 **
3481 ** SRT_Set The result must be a single column. Store each
3482 ** row of result as the key in table pDest->iParm.
3483 ** Apply the affinity pDest->affinity before storing
3484 ** results. Used to implement "IN (SELECT ...)".
3485 **
3486 ** SRT_Union Store results as a key in a temporary table pDest->iParm.
3487 **
3488 ** SRT_Except Remove results from the temporary table pDest->iParm.
3489 **
3490 ** SRT_Table Store results in temporary table pDest->iParm.
3491 ** This is like SRT_EphemTab except that the table
3492 ** is assumed to already be open.
3493 **
3494 ** SRT_EphemTab Create an temporary table pDest->iParm and store
3495 ** the result there. The cursor is left open after
3496 ** returning. This is like SRT_Table except that
3497 ** this destination uses OP_OpenEphemeral to create
3498 ** the table first.
3499 **
3500 ** SRT_Coroutine Generate a co-routine that returns a new row of
3501 ** results each time it is invoked. The entry point
3502 ** of the co-routine is stored in register pDest->iParm.
3503 **
3504 ** SRT_Exists Store a 1 in memory cell pDest->iParm if the result
3505 ** set is not empty.
3506 **
3507 ** SRT_Discard Throw the results away. This is used by SELECT
3508 ** statements within triggers whose only purpose is
3509 ** the side-effects of functions.
3510 **
3511 ** This routine returns the number of errors. If any errors are
3512 ** encountered, then an appropriate error message is left in
3513 ** pParse->zErrMsg.
3514 **
3515 ** This routine does NOT free the Select structure passed in. The
3516 ** calling function needs to do that.
3517 */
3518 int sqlite3Select(
3519 Parse *pParse, /* The parser context */
3520 Select *p, /* The SELECT statement being coded. */
3521 SelectDest *pDest /* What to do with the query results */
3522 ){
3523 int i, j; /* Loop counters */
3524 WhereInfo *pWInfo; /* Return from sqlite3WhereBegin() */
3525 Vdbe *v; /* The virtual machine under construction */
3526 int isAgg; /* True for select lists like "count(*)" */
3527 ExprList *pEList; /* List of columns to extract. */
3528 SrcList *pTabList; /* List of tables to select from */
3529 Expr *pWhere; /* The WHERE clause. May be NULL */
3530 ExprList *pOrderBy; /* The ORDER BY clause. May be NULL */
3531 ExprList *pGroupBy; /* The GROUP BY clause. May be NULL */
3532 Expr *pHaving; /* The HAVING clause. May be NULL */
3533 int isDistinct; /* True if the DISTINCT keyword is present */
3534 int distinct; /* Table to use for the distinct set */
3535 int rc = 1; /* Value to return from this function */
3536 int addrSortIndex; /* Address of an OP_OpenEphemeral instruction */
3537 AggInfo sAggInfo; /* Information used by aggregate queries */
3538 int iEnd; /* Address of the end of the query */
3539 sqlite3 *db; /* The database connection */
3540
3541 db = pParse->db;
3542 if( p==0 || db->mallocFailed || pParse->nErr ){
3543 return 1;
3544 }
3545 if( sqlite3AuthCheck(pParse, SQLITE_SELECT, 0, 0, 0) ) return 1;
3546 memset(&sAggInfo, 0, sizeof(sAggInfo));
3547
3548 if( IgnorableOrderby(pDest) ){
3549 assert(pDest->eDest==SRT_Exists || pDest->eDest==SRT_Union ||
3550 pDest->eDest==SRT_Except || pDest->eDest==SRT_Discard);
3551 /* If ORDER BY makes no difference in the output then neither does
3552 ** DISTINCT so it can be removed too. */
3553 sqlite3ExprListDelete(db, p->pOrderBy);
3554 p->pOrderBy = 0;
3555 p->selFlags &= ~SF_Distinct;
3556 }
3557 sqlite3SelectPrep(pParse, p, 0);
3558 pOrderBy = p->pOrderBy;
3559 pTabList = p->pSrc;
3560 pEList = p->pEList;
3561 if( pParse->nErr || db->mallocFailed ){
3562 goto select_end;
3563 }
3564 isAgg = (p->selFlags & SF_Aggregate)!=0;
3565 assert( pEList!=0 );
3566
3567 /* Begin generating code.
3568 */
3569 v = sqlite3GetVdbe(pParse);
3570 if( v==0 ) goto select_end;
3571
3572 /* Generate code for all sub-queries in the FROM clause
3573 */
3574 #if !defined(SQLITE_OMIT_SUBQUERY) || !defined(SQLITE_OMIT_VIEW)
3575 for(i=0; !p->pPrior && i<pTabList->nSrc; i++){
3576 struct SrcList_item *pItem = &pTabList->a[i];
3577 SelectDest dest;
3578 Select *pSub = pItem->pSelect;
3579 int isAggSub;
3580
3581 if( pSub==0 || pItem->isPopulated ) continue;
3582
3583 /* Increment Parse.nHeight by the height of the largest expression
3584 ** tree refered to by this, the parent select. The child select
3585 ** may contain expression trees of at most
3586 ** (SQLITE_MAX_EXPR_DEPTH-Parse.nHeight) height. This is a bit
3587 ** more conservative than necessary, but much easier than enforcing
3588 ** an exact limit.
3589 */
3590 pParse->nHeight += sqlite3SelectExprHeight(p);
3591
3592 /* Check to see if the subquery can be absorbed into the parent. */
3593 isAggSub = (pSub->selFlags & SF_Aggregate)!=0;
3594 if( flattenSubquery(pParse, p, i, isAgg, isAggSub) ){
3595 if( isAggSub ){
3596 isAgg = 1;
3597 p->selFlags |= SF_Aggregate;
3598 }
3599 i = -1;
3600 }else{
3601 sqlite3SelectDestInit(&dest, SRT_EphemTab, pItem->iCursor);
3602 assert( pItem->isPopulated==0 );
3603 sqlite3Select(pParse, pSub, &dest);
3604 pItem->isPopulated = 1;
3605 }
3606 if( /*pParse->nErr ||*/ db->mallocFailed ){
3607 goto select_end;
3608 }
3609 pParse->nHeight -= sqlite3SelectExprHeight(p);
3610 pTabList = p->pSrc;
3611 if( !IgnorableOrderby(pDest) ){
3612 pOrderBy = p->pOrderBy;
3613 }
3614 }
3615 pEList = p->pEList;
3616 #endif
3617 pWhere = p->pWhere;
3618 pGroupBy = p->pGroupBy;
3619 pHaving = p->pHaving;
3620 isDistinct = (p->selFlags & SF_Distinct)!=0;
3621
3622 #ifndef SQLITE_OMIT_COMPOUND_SELECT
3623 /* If there is are a sequence of queries, do the earlier ones first.
3624 */
3625 if( p->pPrior ){
3626 if( p->pRightmost==0 ){
3627 Select *pLoop, *pRight = 0;
3628 int cnt = 0;
3629 int mxSelect;
3630 for(pLoop=p; pLoop; pLoop=pLoop->pPrior, cnt++){
3631 pLoop->pRightmost = p;
3632 pLoop->pNext = pRight;
3633 pRight = pLoop;
3634 }
3635 mxSelect = db->aLimit[SQLITE_LIMIT_COMPOUND_SELECT];
3636 if( mxSelect && cnt>mxSelect ){
3637 sqlite3ErrorMsg(pParse, "too many terms in compound SELECT");
3638 return 1;
3639 }
3640 }
3641 return multiSelect(pParse, p, pDest);
3642 }
3643 #endif
3644
3645 /* If writing to memory or generating a set
3646 ** only a single column may be output.
3647 */
3648 #ifndef SQLITE_OMIT_SUBQUERY
3649 if( checkForMultiColumnSelectError(pParse, pDest, pEList->nExpr) ){
3650 goto select_end;
3651 }
3652 #endif
3653
3654 /* If possible, rewrite the query to use GROUP BY instead of DISTINCT.
3655 ** GROUP BY might use an index, DISTINCT never does.
3656 */
3657 assert( p->pGroupBy==0 || (p->selFlags & SF_Aggregate)!=0 );
3658 if( (p->selFlags & (SF_Distinct|SF_Aggregate))==SF_Distinct ){
3659 p->pGroupBy = sqlite3ExprListDup(db, p->pEList, 0);
3660 pGroupBy = p->pGroupBy;
3661 p->selFlags &= ~SF_Distinct;
3662 isDistinct = 0;
3663 }
3664
3665 /* If there is an ORDER BY clause, then this sorting
3666 ** index might end up being unused if the data can be
3667 ** extracted in pre-sorted order. If that is the case, then the
3668 ** OP_OpenEphemeral instruction will be changed to an OP_Noop once
3669 ** we figure out that the sorting index is not needed. The addrSortIndex
3670 ** variable is used to facilitate that change.
3671 */
3672 if( pOrderBy ){
3673 KeyInfo *pKeyInfo;
3674 pKeyInfo = keyInfoFromExprList(pParse, pOrderBy);
3675 pOrderBy->iECursor = pParse->nTab++;
3676 p->addrOpenEphm[2] = addrSortIndex =
3677 sqlite3VdbeAddOp4(v, OP_OpenEphemeral,
3678 pOrderBy->iECursor, pOrderBy->nExpr+2, 0,
3679 (char*)pKeyInfo, P4_KEYINFO_HANDOFF);
3680 }else{
3681 addrSortIndex = -1;
3682 }
3683
3684 /* If the output is destined for a temporary table, open that table.
3685 */
3686 if( pDest->eDest==SRT_EphemTab ){
3687 sqlite3VdbeAddOp2(v, OP_OpenEphemeral, pDest->iParm, pEList->nExpr);
3688 }
3689
3690 /* Set the limiter.
3691 */
3692 iEnd = sqlite3VdbeMakeLabel(v);
3693 computeLimitRegisters(pParse, p, iEnd);
3694
3695 /* Open a virtual index to use for the distinct set.
3696 */
3697 if( isDistinct ){
3698 KeyInfo *pKeyInfo;
3699 assert( isAgg || pGroupBy );
3700 distinct = pParse->nTab++;
3701 pKeyInfo = keyInfoFromExprList(pParse, p->pEList);
3702 sqlite3VdbeAddOp4(v, OP_OpenEphemeral, distinct, 0, 0,
3703 (char*)pKeyInfo, P4_KEYINFO_HANDOFF);
3704 }else{
3705 distinct = -1;
3706 }
3707
3708 /* Aggregate and non-aggregate queries are handled differently */
3709 if( !isAgg && pGroupBy==0 ){
3710 /* This case is for non-aggregate queries
3711 ** Begin the database scan
3712 */
3713 pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pOrderBy, 0);
3714 if( pWInfo==0 ) goto select_end;
3715
3716 /* If sorting index that was created by a prior OP_OpenEphemeral
3717 ** instruction ended up not being needed, then change the OP_OpenEphemeral
3718 ** into an OP_Noop.
3719 */
3720 if( addrSortIndex>=0 && pOrderBy==0 ){
3721 sqlite3VdbeChangeToNoop(v, addrSortIndex, 1);
3722 p->addrOpenEphm[2] = -1;
3723 }
3724
3725 /* Use the standard inner loop
3726 */
3727 assert(!isDistinct);
3728 selectInnerLoop(pParse, p, pEList, 0, 0, pOrderBy, -1, pDest,
3729 pWInfo->iContinue, pWInfo->iBreak);
3730
3731 /* End the database scan loop.
3732 */
3733 sqlite3WhereEnd(pWInfo);
3734 }else{
3735 /* This is the processing for aggregate queries */
3736 NameContext sNC; /* Name context for processing aggregate information */
3737 int iAMem; /* First Mem address for storing current GROUP BY */
3738 int iBMem; /* First Mem address for previous GROUP BY */
3739 int iUseFlag; /* Mem address holding flag indicating that at least
3740 ** one row of the input to the aggregator has been
3741 ** processed */
3742 int iAbortFlag; /* Mem address which causes query abort if positive */
3743 int groupBySort; /* Rows come from source in GROUP BY order */
3744 int addrEnd; /* End of processing for this SELECT */
3745
3746 /* Remove any and all aliases between the result set and the
3747 ** GROUP BY clause.
3748 */
3749 if( pGroupBy ){
3750 int k; /* Loop counter */
3751 struct ExprList_item *pItem; /* For looping over expression in a list */
3752
3753 for(k=p->pEList->nExpr, pItem=p->pEList->a; k>0; k--, pItem++){
3754 pItem->iAlias = 0;
3755 }
3756 for(k=pGroupBy->nExpr, pItem=pGroupBy->a; k>0; k--, pItem++){
3757 pItem->iAlias = 0;
3758 }
3759 }
3760
3761
3762 /* Create a label to jump to when we want to abort the query */
3763 addrEnd = sqlite3VdbeMakeLabel(v);
3764
3765 /* Convert TK_COLUMN nodes into TK_AGG_COLUMN and make entries in
3766 ** sAggInfo for all TK_AGG_FUNCTION nodes in expressions of the
3767 ** SELECT statement.
3768 */
3769 memset(&sNC, 0, sizeof(sNC));
3770 sNC.pParse = pParse;
3771 sNC.pSrcList = pTabList;
3772 sNC.pAggInfo = &sAggInfo;
3773 sAggInfo.nSortingColumn = pGroupBy ? pGroupBy->nExpr+1 : 0;
3774 sAggInfo.pGroupBy = pGroupBy;
3775 sqlite3ExprAnalyzeAggList(&sNC, pEList);
3776 sqlite3ExprAnalyzeAggList(&sNC, pOrderBy);
3777 if( pHaving ){
3778 sqlite3ExprAnalyzeAggregates(&sNC, pHaving);
3779 }
3780 sAggInfo.nAccumulator = sAggInfo.nColumn;
3781 for(i=0; i<sAggInfo.nFunc; i++){
3782 assert( !ExprHasProperty(sAggInfo.aFunc[i].pExpr, EP_xIsSelect) );
3783 sqlite3ExprAnalyzeAggList(&sNC, sAggInfo.aFunc[i].pExpr->x.pList);
3784 }
3785 if( db->mallocFailed ) goto select_end;
3786
3787 /* Processing for aggregates with GROUP BY is very different and
3788 ** much more complex than aggregates without a GROUP BY.
3789 */
3790 if( pGroupBy ){
3791 KeyInfo *pKeyInfo; /* Keying information for the group by clause */
3792 int j1; /* A-vs-B comparision jump */
3793 int addrOutputRow; /* Start of subroutine that outputs a result row */
3794 int regOutputRow; /* Return address register for output subroutine */
3795 int addrSetAbort; /* Set the abort flag and return */
3796 int addrTopOfLoop; /* Top of the input loop */
3797 int addrSortingIdx; /* The OP_OpenEphemeral for the sorting index */
3798 int addrReset; /* Subroutine for resetting the accumulator */
3799 int regReset; /* Return address register for reset subroutine */
3800
3801 /* If there is a GROUP BY clause we might need a sorting index to
3802 ** implement it. Allocate that sorting index now. If it turns out
3803 ** that we do not need it after all, the OpenEphemeral instruction
3804 ** will be converted into a Noop.
3805 */
3806 sAggInfo.sortingIdx = pParse->nTab++;
3807 pKeyInfo = keyInfoFromExprList(pParse, pGroupBy);
3808 addrSortingIdx = sqlite3VdbeAddOp4(v, OP_OpenEphemeral,
3809 sAggInfo.sortingIdx, sAggInfo.nSortingColumn,
3810 0, (char*)pKeyInfo, P4_KEYINFO_HANDOFF);
3811
3812 /* Initialize memory locations used by GROUP BY aggregate processing
3813 */
3814 iUseFlag = ++pParse->nMem;
3815 iAbortFlag = ++pParse->nMem;
3816 regOutputRow = ++pParse->nMem;
3817 addrOutputRow = sqlite3VdbeMakeLabel(v);
3818 regReset = ++pParse->nMem;
3819 addrReset = sqlite3VdbeMakeLabel(v);
3820 iAMem = pParse->nMem + 1;
3821 pParse->nMem += pGroupBy->nExpr;
3822 iBMem = pParse->nMem + 1;
3823 pParse->nMem += pGroupBy->nExpr;
3824 sqlite3VdbeAddOp2(v, OP_Integer, 0, iAbortFlag);
3825 VdbeComment((v, "clear abort flag"));
3826 sqlite3VdbeAddOp2(v, OP_Integer, 0, iUseFlag);
3827 VdbeComment((v, "indicate accumulator empty"));
3828
3829 /* Begin a loop that will extract all source rows in GROUP BY order.
3830 ** This might involve two separate loops with an OP_Sort in between, or
3831 ** it might be a single loop that uses an index to extract information
3832 ** in the right order to begin with.
3833 */
3834 sqlite3VdbeAddOp2(v, OP_Gosub, regReset, addrReset);
3835 pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pGroupBy, 0);
3836 if( pWInfo==0 ) goto select_end;
3837 if( pGroupBy==0 ){
3838 /* The optimizer is able to deliver rows in group by order so
3839 ** we do not have to sort. The OP_OpenEphemeral table will be
3840 ** cancelled later because we still need to use the pKeyInfo
3841 */
3842 pGroupBy = p->pGroupBy;
3843 groupBySort = 0;
3844 }else{
3845 /* Rows are coming out in undetermined order. We have to push
3846 ** each row into a sorting index, terminate the first loop,
3847 ** then loop over the sorting index in order to get the output
3848 ** in sorted order
3849 */
3850 int regBase;
3851 int regRecord;
3852 int nCol;
3853 int nGroupBy;
3854
3855 groupBySort = 1;
3856 nGroupBy = pGroupBy->nExpr;
3857 nCol = nGroupBy + 1;
3858 j = nGroupBy+1;
3859 for(i=0; i<sAggInfo.nColumn; i++){
3860 if( sAggInfo.aCol[i].iSorterColumn>=j ){
3861 nCol++;
3862 j++;
3863 }
3864 }
3865 regBase = sqlite3GetTempRange(pParse, nCol);
3866 sqlite3ExprCacheClear(pParse);
3867 sqlite3ExprCodeExprList(pParse, pGroupBy, regBase, 0);
3868 sqlite3VdbeAddOp2(v, OP_Sequence, sAggInfo.sortingIdx,regBase+nGroupBy);
3869 j = nGroupBy+1;
3870 for(i=0; i<sAggInfo.nColumn; i++){
3871 struct AggInfo_col *pCol = &sAggInfo.aCol[i];
3872 if( pCol->iSorterColumn>=j ){
3873 int r1 = j + regBase;
3874 int r2;
3875
3876 r2 = sqlite3ExprCodeGetColumn(pParse,
3877 pCol->pTab, pCol->iColumn, pCol->iTable, r1, 0);
3878 if( r1!=r2 ){
3879 sqlite3VdbeAddOp2(v, OP_SCopy, r2, r1);
3880 }
3881 j++;
3882 }
3883 }
3884 regRecord = sqlite3GetTempReg(pParse);
3885 sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nCol, regRecord);
3886 sqlite3VdbeAddOp2(v, OP_IdxInsert, sAggInfo.sortingIdx, regRecord);
3887 sqlite3ReleaseTempReg(pParse, regRecord);
3888 sqlite3ReleaseTempRange(pParse, regBase, nCol);
3889 sqlite3WhereEnd(pWInfo);
3890 sqlite3VdbeAddOp2(v, OP_Sort, sAggInfo.sortingIdx, addrEnd);
3891 VdbeComment((v, "GROUP BY sort"));
3892 sAggInfo.useSortingIdx = 1;
3893 sqlite3ExprCacheClear(pParse);
3894 }
3895
3896 /* Evaluate the current GROUP BY terms and store in b0, b1, b2...
3897 ** (b0 is memory location iBMem+0, b1 is iBMem+1, and so forth)
3898 ** Then compare the current GROUP BY terms against the GROUP BY terms
3899 ** from the previous row currently stored in a0, a1, a2...
3900 */
3901 addrTopOfLoop = sqlite3VdbeCurrentAddr(v);
3902 sqlite3ExprCacheClear(pParse);
3903 for(j=0; j<pGroupBy->nExpr; j++){
3904 if( groupBySort ){
3905 sqlite3VdbeAddOp3(v, OP_Column, sAggInfo.sortingIdx, j, iBMem+j);
3906 }else{
3907 sAggInfo.directMode = 1;
3908 sqlite3ExprCode(pParse, pGroupBy->a[j].pExpr, iBMem+j);
3909 }
3910 }
3911 sqlite3VdbeAddOp4(v, OP_Compare, iAMem, iBMem, pGroupBy->nExpr,
3912 (char*)pKeyInfo, P4_KEYINFO);
3913 j1 = sqlite3VdbeCurrentAddr(v);
3914 sqlite3VdbeAddOp3(v, OP_Jump, j1+1, 0, j1+1);
3915
3916 /* Generate code that runs whenever the GROUP BY changes.
3917 ** Changes in the GROUP BY are detected by the previous code
3918 ** block. If there were no changes, this block is skipped.
3919 **
3920 ** This code copies current group by terms in b0,b1,b2,...
3921 ** over to a0,a1,a2. It then calls the output subroutine
3922 ** and resets the aggregate accumulator registers in preparation
3923 ** for the next GROUP BY batch.
3924 */
3925 sqlite3ExprCodeMove(pParse, iBMem, iAMem, pGroupBy->nExpr);
3926 sqlite3VdbeAddOp2(v, OP_Gosub, regOutputRow, addrOutputRow);
3927 VdbeComment((v, "output one row"));
3928 sqlite3VdbeAddOp2(v, OP_IfPos, iAbortFlag, addrEnd);
3929 VdbeComment((v, "check abort flag"));
3930 sqlite3VdbeAddOp2(v, OP_Gosub, regReset, addrReset);
3931 VdbeComment((v, "reset accumulator"));
3932
3933 /* Update the aggregate accumulators based on the content of
3934 ** the current row
3935 */
3936 sqlite3VdbeJumpHere(v, j1);
3937 updateAccumulator(pParse, &sAggInfo);
3938 sqlite3VdbeAddOp2(v, OP_Integer, 1, iUseFlag);
3939 VdbeComment((v, "indicate data in accumulator"));
3940
3941 /* End of the loop
3942 */
3943 if( groupBySort ){
3944 sqlite3VdbeAddOp2(v, OP_Next, sAggInfo.sortingIdx, addrTopOfLoop);
3945 }else{
3946 sqlite3WhereEnd(pWInfo);
3947 sqlite3VdbeChangeToNoop(v, addrSortingIdx, 1);
3948 }
3949
3950 /* Output the final row of result
3951 */
3952 sqlite3VdbeAddOp2(v, OP_Gosub, regOutputRow, addrOutputRow);
3953 VdbeComment((v, "output final row"));
3954
3955 /* Jump over the subroutines
3956 */
3957 sqlite3VdbeAddOp2(v, OP_Goto, 0, addrEnd);
3958
3959 /* Generate a subroutine that outputs a single row of the result
3960 ** set. This subroutine first looks at the iUseFlag. If iUseFlag
3961 ** is less than or equal to zero, the subroutine is a no-op. If
3962 ** the processing calls for the query to abort, this subroutine
3963 ** increments the iAbortFlag memory location before returning in
3964 ** order to signal the caller to abort.
3965 */
3966 addrSetAbort = sqlite3VdbeCurrentAddr(v);
3967 sqlite3VdbeAddOp2(v, OP_Integer, 1, iAbortFlag);
3968 VdbeComment((v, "set abort flag"));
3969 sqlite3VdbeAddOp1(v, OP_Return, regOutputRow);
3970 sqlite3VdbeResolveLabel(v, addrOutputRow);
3971 addrOutputRow = sqlite3VdbeCurrentAddr(v);
3972 sqlite3VdbeAddOp2(v, OP_IfPos, iUseFlag, addrOutputRow+2);
3973 VdbeComment((v, "Groupby result generator entry point"));
3974 sqlite3VdbeAddOp1(v, OP_Return, regOutputRow);
3975 finalizeAggFunctions(pParse, &sAggInfo);
3976 sqlite3ExprIfFalse(pParse, pHaving, addrOutputRow+1, SQLITE_JUMPIFNULL);
3977 selectInnerLoop(pParse, p, p->pEList, 0, 0, pOrderBy,
3978 distinct, pDest,
3979 addrOutputRow+1, addrSetAbort);
3980 sqlite3VdbeAddOp1(v, OP_Return, regOutputRow);
3981 VdbeComment((v, "end groupby result generator"));
3982
3983 /* Generate a subroutine that will reset the group-by accumulator
3984 */
3985 sqlite3VdbeResolveLabel(v, addrReset);
3986 resetAccumulator(pParse, &sAggInfo);
3987 sqlite3VdbeAddOp1(v, OP_Return, regReset);
3988
3989 } /* endif pGroupBy. Begin aggregate queries without GROUP BY: */
3990 else {
3991 ExprList *pDel = 0;
3992 #ifndef SQLITE_OMIT_BTREECOUNT
3993 Table *pTab;
3994 if( (pTab = isSimpleCount(p, &sAggInfo))!=0 ){
3995 /* If isSimpleCount() returns a pointer to a Table structure, then
3996 ** the SQL statement is of the form:
3997 **
3998 ** SELECT count(*) FROM <tbl>
3999 **
4000 ** where the Table structure returned represents table <tbl>.
4001 **
4002 ** This statement is so common that it is optimized specially. The
4003 ** OP_Count instruction is executed either on the intkey table that
4004 ** contains the data for table <tbl> or on one of its indexes. It
4005 ** is better to execute the op on an index, as indexes are almost
4006 ** always spread across less pages than their corresponding tables.
4007 */
4008 const int iDb = sqlite3SchemaToIndex(pParse->db, pTab->pSchema);
4009 const int iCsr = pParse->nTab++; /* Cursor to scan b-tree */
4010 Index *pIdx; /* Iterator variable */
4011 KeyInfo *pKeyInfo = 0; /* Keyinfo for scanned index */
4012 Index *pBest = 0; /* Best index found so far */
4013 int iRoot = pTab->tnum; /* Root page of scanned b-tree */
4014
4015 sqlite3CodeVerifySchema(pParse, iDb);
4016 sqlite3TableLock(pParse, iDb, pTab->tnum, 0, pTab->zName);
4017
4018 /* Search for the index that has the least amount of columns. If
4019 ** there is such an index, and it has less columns than the table
4020 ** does, then we can assume that it consumes less space on disk and
4021 ** will therefore be cheaper to scan to determine the query result.
4022 ** In this case set iRoot to the root page number of the index b-tree
4023 ** and pKeyInfo to the KeyInfo structure required to navigate the
4024 ** index.
4025 **
4026 ** In practice the KeyInfo structure will not be used. It is only
4027 ** passed to keep OP_OpenRead happy.
4028 */
4029 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){
4030 if( !pBest || pIdx->nColumn<pBest->nColumn ){
4031 pBest = pIdx;
4032 }
4033 }
4034 if( pBest && pBest->nColumn<pTab->nCol ){
4035 iRoot = pBest->tnum;
4036 pKeyInfo = sqlite3IndexKeyinfo(pParse, pBest);
4037 }
4038
4039 /* Open a read-only cursor, execute the OP_Count, close the cursor. */
4040 sqlite3VdbeAddOp3(v, OP_OpenRead, iCsr, iRoot, iDb);
4041 if( pKeyInfo ){
4042 sqlite3VdbeChangeP4(v, -1, (char *)pKeyInfo, P4_KEYINFO_HANDOFF);
4043 }
4044 sqlite3VdbeAddOp2(v, OP_Count, iCsr, sAggInfo.aFunc[0].iMem);
4045 sqlite3VdbeAddOp1(v, OP_Close, iCsr);
4046 }else
4047 #endif /* SQLITE_OMIT_BTREECOUNT */
4048 {
4049 /* Check if the query is of one of the following forms:
4050 **
4051 ** SELECT min(x) FROM ...
4052 ** SELECT max(x) FROM ...
4053 **
4054 ** If it is, then ask the code in where.c to attempt to sort results
4055 ** as if there was an "ORDER ON x" or "ORDER ON x DESC" clause.
4056 ** If where.c is able to produce results sorted in this order, then
4057 ** add vdbe code to break out of the processing loop after the
4058 ** first iteration (since the first iteration of the loop is
4059 ** guaranteed to operate on the row with the minimum or maximum
4060 ** value of x, the only row required).
4061 **
4062 ** A special flag must be passed to sqlite3WhereBegin() to slightly
4063 ** modify behaviour as follows:
4064 **
4065 ** + If the query is a "SELECT min(x)", then the loop coded by
4066 ** where.c should not iterate over any values with a NULL value
4067 ** for x.
4068 **
4069 ** + The optimizer code in where.c (the thing that decides which
4070 ** index or indices to use) should place a different priority on
4071 ** satisfying the 'ORDER BY' clause than it does in other cases.
4072 ** Refer to code and comments in where.c for details.
4073 */
4074 ExprList *pMinMax = 0;
4075 u8 flag = minMaxQuery(p);
4076 if( flag ){
4077 assert( !ExprHasProperty(p->pEList->a[0].pExpr, EP_xIsSelect) );
4078 pMinMax = sqlite3ExprListDup(db, p->pEList->a[0].pExpr->x.pList,0);
4079 pDel = pMinMax;
4080 if( pMinMax && !db->mallocFailed ){
4081 pMinMax->a[0].sortOrder = flag!=WHERE_ORDERBY_MIN ?1:0;
4082 pMinMax->a[0].pExpr->op = TK_COLUMN;
4083 }
4084 }
4085
4086 /* This case runs if the aggregate has no GROUP BY clause. The
4087 ** processing is much simpler since there is only a single row
4088 ** of output.
4089 */
4090 resetAccumulator(pParse, &sAggInfo);
4091 pWInfo = sqlite3WhereBegin(pParse, pTabList, pWhere, &pMinMax, flag);
4092 if( pWInfo==0 ){
4093 sqlite3ExprListDelete(db, pDel);
4094 goto select_end;
4095 }
4096 updateAccumulator(pParse, &sAggInfo);
4097 if( !pMinMax && flag ){
4098 sqlite3VdbeAddOp2(v, OP_Goto, 0, pWInfo->iBreak);
4099 VdbeComment((v, "%s() by index",
4100 (flag==WHERE_ORDERBY_MIN?"min":"max")));
4101 }
4102 sqlite3WhereEnd(pWInfo);
4103 finalizeAggFunctions(pParse, &sAggInfo);
4104 }
4105
4106 pOrderBy = 0;
4107 sqlite3ExprIfFalse(pParse, pHaving, addrEnd, SQLITE_JUMPIFNULL);
4108 selectInnerLoop(pParse, p, p->pEList, 0, 0, 0, -1,
4109 pDest, addrEnd, addrEnd);
4110 sqlite3ExprListDelete(db, pDel);
4111 }
4112 sqlite3VdbeResolveLabel(v, addrEnd);
4113
4114 } /* endif aggregate query */
4115
4116 /* If there is an ORDER BY clause, then we need to sort the results
4117 ** and send them to the callback one by one.
4118 */
4119 if( pOrderBy ){
4120 generateSortTail(pParse, p, v, pEList->nExpr, pDest);
4121 }
4122
4123 /* Jump here to skip this query
4124 */
4125 sqlite3VdbeResolveLabel(v, iEnd);
4126
4127 /* The SELECT was successfully coded. Set the return code to 0
4128 ** to indicate no errors.
4129 */
4130 rc = 0;
4131
4132 /* Control jumps to here if an error is encountered above, or upon
4133 ** successful coding of the SELECT.
4134 */
4135 select_end:
4136
4137 /* Identify column names if results of the SELECT are to be output.
4138 */
4139 if( rc==SQLITE_OK && pDest->eDest==SRT_Output ){
4140 generateColumnNames(pParse, pTabList, pEList);
4141 }
4142
4143 sqlite3DbFree(db, sAggInfo.aCol);
4144 sqlite3DbFree(db, sAggInfo.aFunc);
4145 return rc;
4146 }
4147
4148 #if defined(SQLITE_DEBUG)
4149 /*
4150 *******************************************************************************
4151 ** The following code is used for testing and debugging only. The code
4152 ** that follows does not appear in normal builds.
4153 **
4154 ** These routines are used to print out the content of all or part of a
4155 ** parse structures such as Select or Expr. Such printouts are useful
4156 ** for helping to understand what is happening inside the code generator
4157 ** during the execution of complex SELECT statements.
4158 **
4159 ** These routine are not called anywhere from within the normal
4160 ** code base. Then are intended to be called from within the debugger
4161 ** or from temporary "printf" statements inserted for debugging.
4162 */
4163 void sqlite3PrintExpr(Expr *p){
4164 if( !ExprHasProperty(p, EP_IntValue) && p->u.zToken ){
4165 sqlite3DebugPrintf("(%s", p->u.zToken);
4166 }else{
4167 sqlite3DebugPrintf("(%d", p->op);
4168 }
4169 if( p->pLeft ){
4170 sqlite3DebugPrintf(" ");
4171 sqlite3PrintExpr(p->pLeft);
4172 }
4173 if( p->pRight ){
4174 sqlite3DebugPrintf(" ");
4175 sqlite3PrintExpr(p->pRight);
4176 }
4177 sqlite3DebugPrintf(")");
4178 }
4179 void sqlite3PrintExprList(ExprList *pList){
4180 int i;
4181 for(i=0; i<pList->nExpr; i++){
4182 sqlite3PrintExpr(pList->a[i].pExpr);
4183 if( i<pList->nExpr-1 ){
4184 sqlite3DebugPrintf(", ");
4185 }
4186 }
4187 }
4188 void sqlite3PrintSelect(Select *p, int indent){
4189 sqlite3DebugPrintf("%*sSELECT(%p) ", indent, "", p);
4190 sqlite3PrintExprList(p->pEList);
4191 sqlite3DebugPrintf("\n");
4192 if( p->pSrc ){
4193 char *zPrefix;
4194 int i;
4195 zPrefix = "FROM";
4196 for(i=0; i<p->pSrc->nSrc; i++){
4197 struct SrcList_item *pItem = &p->pSrc->a[i];
4198 sqlite3DebugPrintf("%*s ", indent+6, zPrefix);
4199 zPrefix = "";
4200 if( pItem->pSelect ){
4201 sqlite3DebugPrintf("(\n");
4202 sqlite3PrintSelect(pItem->pSelect, indent+10);
4203 sqlite3DebugPrintf("%*s)", indent+8, "");
4204 }else if( pItem->zName ){
4205 sqlite3DebugPrintf("%s", pItem->zName);
4206 }
4207 if( pItem->pTab ){
4208 sqlite3DebugPrintf("(table: %s)", pItem->pTab->zName);
4209 }
4210 if( pItem->zAlias ){
4211 sqlite3DebugPrintf(" AS %s", pItem->zAlias);
4212 }
4213 if( i<p->pSrc->nSrc-1 ){
4214 sqlite3DebugPrintf(",");
4215 }
4216 sqlite3DebugPrintf("\n");
4217 }
4218 }
4219 if( p->pWhere ){
4220 sqlite3DebugPrintf("%*s WHERE ", indent, "");
4221 sqlite3PrintExpr(p->pWhere);
4222 sqlite3DebugPrintf("\n");
4223 }
4224 if( p->pGroupBy ){
4225 sqlite3DebugPrintf("%*s GROUP BY ", indent, "");
4226 sqlite3PrintExprList(p->pGroupBy);
4227 sqlite3DebugPrintf("\n");
4228 }
4229 if( p->pHaving ){
4230 sqlite3DebugPrintf("%*s HAVING ", indent, "");
4231 sqlite3PrintExpr(p->pHaving);
4232 sqlite3DebugPrintf("\n");
4233 }
4234 if( p->pOrderBy ){
4235 sqlite3DebugPrintf("%*s ORDER BY ", indent, "");
4236 sqlite3PrintExprList(p->pOrderBy);
4237 sqlite3DebugPrintf("\n");
4238 }
4239 }
4240 /* End of the structure debug printing code
4241 *****************************************************************************/
4242 #endif /* defined(SQLITE_TEST) || defined(SQLITE_DEBUG) */
OLDNEW
« no previous file with comments | « third_party/sqlite/src/rowset.c ('k') | third_party/sqlite/src/shell.c » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698