forked from sqlcipher/sqlcipher
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathanalyze.c
More file actions
1889 lines (1761 loc) · 64 KB
/
analyze.c
File metadata and controls
1889 lines (1761 loc) · 64 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
** 2005-07-08
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code associated with the ANALYZE command.
**
** The ANALYZE command gather statistics about the content of tables
** and indices. These statistics are made available to the query planner
** to help it make better decisions about how to perform queries.
**
** The following system tables are or have been supported:
**
** CREATE TABLE sqlite_stat1(tbl, idx, stat);
** CREATE TABLE sqlite_stat2(tbl, idx, sampleno, sample);
** CREATE TABLE sqlite_stat3(tbl, idx, nEq, nLt, nDLt, sample);
** CREATE TABLE sqlite_stat4(tbl, idx, nEq, nLt, nDLt, sample);
**
** Additional tables might be added in future releases of SQLite.
** The sqlite_stat2 table is not created or used unless the SQLite version
** is between 3.6.18 and 3.7.8, inclusive, and unless SQLite is compiled
** with SQLITE_ENABLE_STAT2. The sqlite_stat2 table is deprecated.
** The sqlite_stat2 table is superseded by sqlite_stat3, which is only
** created and used by SQLite versions 3.7.9 and later and with
** SQLITE_ENABLE_STAT3 defined. The functionality of sqlite_stat3
** is a superset of sqlite_stat2. The sqlite_stat4 is an enhanced
** version of sqlite_stat3 and is only available when compiled with
** SQLITE_ENABLE_STAT4 and in SQLite versions 3.8.1 and later. It is
** not possible to enable both STAT3 and STAT4 at the same time. If they
** are both enabled, then STAT4 takes precedence.
**
** For most applications, sqlite_stat1 provides all the statistics required
** for the query planner to make good choices.
**
** Format of sqlite_stat1:
**
** There is normally one row per index, with the index identified by the
** name in the idx column. The tbl column is the name of the table to
** which the index belongs. In each such row, the stat column will be
** a string consisting of a list of integers. The first integer in this
** list is the number of rows in the index. (This is the same as the
** number of rows in the table, except for partial indices.) The second
** integer is the average number of rows in the index that have the same
** value in the first column of the index. The third integer is the average
** number of rows in the index that have the same value for the first two
** columns. The N-th integer (for N>1) is the average number of rows in
** the index which have the same value for the first N-1 columns. For
** a K-column index, there will be K+1 integers in the stat column. If
** the index is unique, then the last integer will be 1.
**
** The list of integers in the stat column can optionally be followed
** by the keyword "unordered". The "unordered" keyword, if it is present,
** must be separated from the last integer by a single space. If the
** "unordered" keyword is present, then the query planner assumes that
** the index is unordered and will not use the index for a range query.
**
** If the sqlite_stat1.idx column is NULL, then the sqlite_stat1.stat
** column contains a single integer which is the (estimated) number of
** rows in the table identified by sqlite_stat1.tbl.
**
** Format of sqlite_stat2:
**
** The sqlite_stat2 is only created and is only used if SQLite is compiled
** with SQLITE_ENABLE_STAT2 and if the SQLite version number is between
** 3.6.18 and 3.7.8. The "stat2" table contains additional information
** about the distribution of keys within an index. The index is identified by
** the "idx" column and the "tbl" column is the name of the table to which
** the index belongs. There are usually 10 rows in the sqlite_stat2
** table for each index.
**
** The sqlite_stat2 entries for an index that have sampleno between 0 and 9
** inclusive are samples of the left-most key value in the index taken at
** evenly spaced points along the index. Let the number of samples be S
** (10 in the standard build) and let C be the number of rows in the index.
** Then the sampled rows are given by:
**
** rownumber = (i*C*2 + C)/(S*2)
**
** For i between 0 and S-1. Conceptually, the index space is divided into
** S uniform buckets and the samples are the middle row from each bucket.
**
** The format for sqlite_stat2 is recorded here for legacy reference. This
** version of SQLite does not support sqlite_stat2. It neither reads nor
** writes the sqlite_stat2 table. This version of SQLite only supports
** sqlite_stat3.
**
** Format for sqlite_stat3:
**
** The sqlite_stat3 format is a subset of sqlite_stat4. Hence, the
** sqlite_stat4 format will be described first. Further information
** about sqlite_stat3 follows the sqlite_stat4 description.
**
** Format for sqlite_stat4:
**
** As with sqlite_stat2, the sqlite_stat4 table contains histogram data
** to aid the query planner in choosing good indices based on the values
** that indexed columns are compared against in the WHERE clauses of
** queries.
**
** The sqlite_stat4 table contains multiple entries for each index.
** The idx column names the index and the tbl column is the table of the
** index. If the idx and tbl columns are the same, then the sample is
** of the INTEGER PRIMARY KEY. The sample column is a blob which is the
** binary encoding of a key from the index. The nEq column is a
** list of integers. The first integer is the approximate number
** of entries in the index whose left-most column exactly matches
** the left-most column of the sample. The second integer in nEq
** is the approximate number of entries in the index where the
** first two columns match the first two columns of the sample.
** And so forth. nLt is another list of integers that show the approximate
** number of entries that are strictly less than the sample. The first
** integer in nLt contains the number of entries in the index where the
** left-most column is less than the left-most column of the sample.
** The K-th integer in the nLt entry is the number of index entries
** where the first K columns are less than the first K columns of the
** sample. The nDLt column is like nLt except that it contains the
** number of distinct entries in the index that are less than the
** sample.
**
** There can be an arbitrary number of sqlite_stat4 entries per index.
** The ANALYZE command will typically generate sqlite_stat4 tables
** that contain between 10 and 40 samples which are distributed across
** the key space, though not uniformly, and which include samples with
** large nEq values.
**
** Format for sqlite_stat3 redux:
**
** The sqlite_stat3 table is like sqlite_stat4 except that it only
** looks at the left-most column of the index. The sqlite_stat3.sample
** column contains the actual value of the left-most column instead
** of a blob encoding of the complete index key as is found in
** sqlite_stat4.sample. The nEq, nLt, and nDLt entries of sqlite_stat3
** all contain just a single integer which is the same as the first
** integer in the equivalent columns in sqlite_stat4.
*/
#ifndef SQLITE_OMIT_ANALYZE
#include "sqliteInt.h"
#if defined(SQLITE_ENABLE_STAT4)
# define IsStat4 1
# define IsStat3 0
#elif defined(SQLITE_ENABLE_STAT3)
# define IsStat4 0
# define IsStat3 1
#else
# define IsStat4 0
# define IsStat3 0
# undef SQLITE_STAT4_SAMPLES
# define SQLITE_STAT4_SAMPLES 1
#endif
#define IsStat34 (IsStat3+IsStat4) /* 1 for STAT3 or STAT4. 0 otherwise */
/*
** This routine generates code that opens the sqlite_statN tables.
** The sqlite_stat1 table is always relevant. sqlite_stat2 is now
** obsolete. sqlite_stat3 and sqlite_stat4 are only opened when
** appropriate compile-time options are provided.
**
** If the sqlite_statN tables do not previously exist, it is created.
**
** Argument zWhere may be a pointer to a buffer containing a table name,
** or it may be a NULL pointer. If it is not NULL, then all entries in
** the sqlite_statN tables associated with the named table are deleted.
** If zWhere==0, then code is generated to delete all stat table entries.
*/
static void openStatTable(
Parse *pParse, /* Parsing context */
int iDb, /* The database we are looking in */
int iStatCur, /* Open the sqlite_stat1 table on this cursor */
const char *zWhere, /* Delete entries for this table or index */
const char *zWhereType /* Either "tbl" or "idx" */
){
static const struct {
const char *zName;
const char *zCols;
} aTable[] = {
{ "sqlite_stat1", "tbl,idx,stat" },
#if defined(SQLITE_ENABLE_STAT4)
{ "sqlite_stat4", "tbl,idx,neq,nlt,ndlt,sample" },
{ "sqlite_stat3", 0 },
#elif defined(SQLITE_ENABLE_STAT3)
{ "sqlite_stat3", "tbl,idx,neq,nlt,ndlt,sample" },
{ "sqlite_stat4", 0 },
#else
{ "sqlite_stat3", 0 },
{ "sqlite_stat4", 0 },
#endif
};
int i;
sqlite3 *db = pParse->db;
Db *pDb;
Vdbe *v = sqlite3GetVdbe(pParse);
int aRoot[ArraySize(aTable)];
u8 aCreateTbl[ArraySize(aTable)];
if( v==0 ) return;
assert( sqlite3BtreeHoldsAllMutexes(db) );
assert( sqlite3VdbeDb(v)==db );
pDb = &db->aDb[iDb];
/* Create new statistic tables if they do not exist, or clear them
** if they do already exist.
*/
for(i=0; i<ArraySize(aTable); i++){
const char *zTab = aTable[i].zName;
Table *pStat;
if( (pStat = sqlite3FindTable(db, zTab, pDb->zName))==0 ){
if( aTable[i].zCols ){
/* The sqlite_statN table does not exist. Create it. Note that a
** side-effect of the CREATE TABLE statement is to leave the rootpage
** of the new table in register pParse->regRoot. This is important
** because the OpenWrite opcode below will be needing it. */
sqlite3NestedParse(pParse,
"CREATE TABLE %Q.%s(%s)", pDb->zName, zTab, aTable[i].zCols
);
aRoot[i] = pParse->regRoot;
aCreateTbl[i] = OPFLAG_P2ISREG;
}
}else{
/* The table already exists. If zWhere is not NULL, delete all entries
** associated with the table zWhere. If zWhere is NULL, delete the
** entire contents of the table. */
aRoot[i] = pStat->tnum;
aCreateTbl[i] = 0;
sqlite3TableLock(pParse, iDb, aRoot[i], 1, zTab);
if( zWhere ){
sqlite3NestedParse(pParse,
"DELETE FROM %Q.%s WHERE %s=%Q",
pDb->zName, zTab, zWhereType, zWhere
);
}else{
/* The sqlite_stat[134] table already exists. Delete all rows. */
sqlite3VdbeAddOp2(v, OP_Clear, aRoot[i], iDb);
}
}
}
/* Open the sqlite_stat[134] tables for writing. */
for(i=0; aTable[i].zCols; i++){
assert( i<ArraySize(aTable) );
sqlite3VdbeAddOp4Int(v, OP_OpenWrite, iStatCur+i, aRoot[i], iDb, 3);
sqlite3VdbeChangeP5(v, aCreateTbl[i]);
VdbeComment((v, aTable[i].zName));
}
}
/*
** Recommended number of samples for sqlite_stat4
*/
#ifndef SQLITE_STAT4_SAMPLES
# define SQLITE_STAT4_SAMPLES 24
#endif
/*
** Three SQL functions - stat_init(), stat_push(), and stat_get() -
** share an instance of the following structure to hold their state
** information.
*/
typedef struct Stat4Accum Stat4Accum;
typedef struct Stat4Sample Stat4Sample;
struct Stat4Sample {
tRowcnt *anEq; /* sqlite_stat4.nEq */
tRowcnt *anDLt; /* sqlite_stat4.nDLt */
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
tRowcnt *anLt; /* sqlite_stat4.nLt */
union {
i64 iRowid; /* Rowid in main table of the key */
u8 *aRowid; /* Key for WITHOUT ROWID tables */
} u;
u32 nRowid; /* Sizeof aRowid[] */
u8 isPSample; /* True if a periodic sample */
int iCol; /* If !isPSample, the reason for inclusion */
u32 iHash; /* Tiebreaker hash */
#endif
};
struct Stat4Accum {
tRowcnt nRow; /* Number of rows in the entire table */
tRowcnt nPSample; /* How often to do a periodic sample */
int nCol; /* Number of columns in index + pk/rowid */
int nKeyCol; /* Number of index columns w/o the pk/rowid */
int mxSample; /* Maximum number of samples to accumulate */
Stat4Sample current; /* Current row as a Stat4Sample */
u32 iPrn; /* Pseudo-random number used for sampling */
Stat4Sample *aBest; /* Array of nCol best samples */
int iMin; /* Index in a[] of entry with minimum score */
int nSample; /* Current number of samples */
int iGet; /* Index of current sample accessed by stat_get() */
Stat4Sample *a; /* Array of mxSample Stat4Sample objects */
sqlite3 *db; /* Database connection, for malloc() */
};
/* Reclaim memory used by a Stat4Sample
*/
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
static void sampleClear(sqlite3 *db, Stat4Sample *p){
assert( db!=0 );
if( p->nRowid ){
sqlite3DbFree(db, p->u.aRowid);
p->nRowid = 0;
}
}
#endif
/* Initialize the BLOB value of a ROWID
*/
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
static void sampleSetRowid(sqlite3 *db, Stat4Sample *p, int n, const u8 *pData){
assert( db!=0 );
if( p->nRowid ) sqlite3DbFree(db, p->u.aRowid);
p->u.aRowid = sqlite3DbMallocRaw(db, n);
if( p->u.aRowid ){
p->nRowid = n;
memcpy(p->u.aRowid, pData, n);
}else{
p->nRowid = 0;
}
}
#endif
/* Initialize the INTEGER value of a ROWID.
*/
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
static void sampleSetRowidInt64(sqlite3 *db, Stat4Sample *p, i64 iRowid){
assert( db!=0 );
if( p->nRowid ) sqlite3DbFree(db, p->u.aRowid);
p->nRowid = 0;
p->u.iRowid = iRowid;
}
#endif
/*
** Copy the contents of object (*pFrom) into (*pTo).
*/
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
static void sampleCopy(Stat4Accum *p, Stat4Sample *pTo, Stat4Sample *pFrom){
pTo->isPSample = pFrom->isPSample;
pTo->iCol = pFrom->iCol;
pTo->iHash = pFrom->iHash;
memcpy(pTo->anEq, pFrom->anEq, sizeof(tRowcnt)*p->nCol);
memcpy(pTo->anLt, pFrom->anLt, sizeof(tRowcnt)*p->nCol);
memcpy(pTo->anDLt, pFrom->anDLt, sizeof(tRowcnt)*p->nCol);
if( pFrom->nRowid ){
sampleSetRowid(p->db, pTo, pFrom->nRowid, pFrom->u.aRowid);
}else{
sampleSetRowidInt64(p->db, pTo, pFrom->u.iRowid);
}
}
#endif
/*
** Reclaim all memory of a Stat4Accum structure.
*/
static void stat4Destructor(void *pOld){
Stat4Accum *p = (Stat4Accum*)pOld;
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
int i;
for(i=0; i<p->nCol; i++) sampleClear(p->db, p->aBest+i);
for(i=0; i<p->mxSample; i++) sampleClear(p->db, p->a+i);
sampleClear(p->db, &p->current);
#endif
sqlite3DbFree(p->db, p);
}
/*
** Implementation of the stat_init(N,K,C) SQL function. The three parameters
** are:
** N: The number of columns in the index including the rowid/pk (note 1)
** K: The number of columns in the index excluding the rowid/pk.
** C: The number of rows in the index (note 2)
**
** Note 1: In the special case of the covering index that implements a
** WITHOUT ROWID table, N is the number of PRIMARY KEY columns, not the
** total number of columns in the table.
**
** Note 2: C is only used for STAT3 and STAT4.
**
** For indexes on ordinary rowid tables, N==K+1. But for indexes on
** WITHOUT ROWID tables, N=K+P where P is the number of columns in the
** PRIMARY KEY of the table. The covering index that implements the
** original WITHOUT ROWID table as N==K as a special case.
**
** This routine allocates the Stat4Accum object in heap memory. The return
** value is a pointer to the Stat4Accum object. The datatype of the
** return value is BLOB, but it is really just a pointer to the Stat4Accum
** object.
*/
static void statInit(
sqlite3_context *context,
int argc,
sqlite3_value **argv
){
Stat4Accum *p;
int nCol; /* Number of columns in index being sampled */
int nKeyCol; /* Number of key columns */
int nColUp; /* nCol rounded up for alignment */
int n; /* Bytes of space to allocate */
sqlite3 *db; /* Database connection */
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
int mxSample = SQLITE_STAT4_SAMPLES;
#endif
/* Decode the three function arguments */
UNUSED_PARAMETER(argc);
nCol = sqlite3_value_int(argv[0]);
assert( nCol>0 );
nColUp = sizeof(tRowcnt)<8 ? (nCol+1)&~1 : nCol;
nKeyCol = sqlite3_value_int(argv[1]);
assert( nKeyCol<=nCol );
assert( nKeyCol>0 );
/* Allocate the space required for the Stat4Accum object */
n = sizeof(*p)
+ sizeof(tRowcnt)*nColUp /* Stat4Accum.anEq */
+ sizeof(tRowcnt)*nColUp /* Stat4Accum.anDLt */
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
+ sizeof(tRowcnt)*nColUp /* Stat4Accum.anLt */
+ sizeof(Stat4Sample)*(nCol+mxSample) /* Stat4Accum.aBest[], a[] */
+ sizeof(tRowcnt)*3*nColUp*(nCol+mxSample)
#endif
;
db = sqlite3_context_db_handle(context);
p = sqlite3DbMallocZero(db, n);
if( p==0 ){
sqlite3_result_error_nomem(context);
return;
}
p->db = db;
p->nRow = 0;
p->nCol = nCol;
p->nKeyCol = nKeyCol;
p->current.anDLt = (tRowcnt*)&p[1];
p->current.anEq = &p->current.anDLt[nColUp];
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
{
u8 *pSpace; /* Allocated space not yet assigned */
int i; /* Used to iterate through p->aSample[] */
p->iGet = -1;
p->mxSample = mxSample;
p->nPSample = (tRowcnt)(sqlite3_value_int64(argv[2])/(mxSample/3+1) + 1);
p->current.anLt = &p->current.anEq[nColUp];
p->iPrn = 0x689e962d*(u32)nCol ^ 0xd0944565*(u32)sqlite3_value_int(argv[2]);
/* Set up the Stat4Accum.a[] and aBest[] arrays */
p->a = (struct Stat4Sample*)&p->current.anLt[nColUp];
p->aBest = &p->a[mxSample];
pSpace = (u8*)(&p->a[mxSample+nCol]);
for(i=0; i<(mxSample+nCol); i++){
p->a[i].anEq = (tRowcnt *)pSpace; pSpace += (sizeof(tRowcnt) * nColUp);
p->a[i].anLt = (tRowcnt *)pSpace; pSpace += (sizeof(tRowcnt) * nColUp);
p->a[i].anDLt = (tRowcnt *)pSpace; pSpace += (sizeof(tRowcnt) * nColUp);
}
assert( (pSpace - (u8*)p)==n );
for(i=0; i<nCol; i++){
p->aBest[i].iCol = i;
}
}
#endif
/* Return a pointer to the allocated object to the caller. Note that
** only the pointer (the 2nd parameter) matters. The size of the object
** (given by the 3rd parameter) is never used and can be any positive
** value. */
sqlite3_result_blob(context, p, sizeof(*p), stat4Destructor);
}
static const FuncDef statInitFuncdef = {
2+IsStat34, /* nArg */
SQLITE_UTF8, /* funcFlags */
0, /* pUserData */
0, /* pNext */
statInit, /* xFunc */
0, /* xStep */
0, /* xFinalize */
"stat_init", /* zName */
0, /* pHash */
0 /* pDestructor */
};
#ifdef SQLITE_ENABLE_STAT4
/*
** pNew and pOld are both candidate non-periodic samples selected for
** the same column (pNew->iCol==pOld->iCol). Ignoring this column and
** considering only any trailing columns and the sample hash value, this
** function returns true if sample pNew is to be preferred over pOld.
** In other words, if we assume that the cardinalities of the selected
** column for pNew and pOld are equal, is pNew to be preferred over pOld.
**
** This function assumes that for each argument sample, the contents of
** the anEq[] array from pSample->anEq[pSample->iCol+1] onwards are valid.
*/
static int sampleIsBetterPost(
Stat4Accum *pAccum,
Stat4Sample *pNew,
Stat4Sample *pOld
){
int nCol = pAccum->nCol;
int i;
assert( pNew->iCol==pOld->iCol );
for(i=pNew->iCol+1; i<nCol; i++){
if( pNew->anEq[i]>pOld->anEq[i] ) return 1;
if( pNew->anEq[i]<pOld->anEq[i] ) return 0;
}
if( pNew->iHash>pOld->iHash ) return 1;
return 0;
}
#endif
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
/*
** Return true if pNew is to be preferred over pOld.
**
** This function assumes that for each argument sample, the contents of
** the anEq[] array from pSample->anEq[pSample->iCol] onwards are valid.
*/
static int sampleIsBetter(
Stat4Accum *pAccum,
Stat4Sample *pNew,
Stat4Sample *pOld
){
tRowcnt nEqNew = pNew->anEq[pNew->iCol];
tRowcnt nEqOld = pOld->anEq[pOld->iCol];
assert( pOld->isPSample==0 && pNew->isPSample==0 );
assert( IsStat4 || (pNew->iCol==0 && pOld->iCol==0) );
if( (nEqNew>nEqOld) ) return 1;
#ifdef SQLITE_ENABLE_STAT4
if( nEqNew==nEqOld ){
if( pNew->iCol<pOld->iCol ) return 1;
return (pNew->iCol==pOld->iCol && sampleIsBetterPost(pAccum, pNew, pOld));
}
return 0;
#else
return (nEqNew==nEqOld && pNew->iHash>pOld->iHash);
#endif
}
/*
** Copy the contents of sample *pNew into the p->a[] array. If necessary,
** remove the least desirable sample from p->a[] to make room.
*/
static void sampleInsert(Stat4Accum *p, Stat4Sample *pNew, int nEqZero){
Stat4Sample *pSample = 0;
int i;
assert( IsStat4 || nEqZero==0 );
#ifdef SQLITE_ENABLE_STAT4
if( pNew->isPSample==0 ){
Stat4Sample *pUpgrade = 0;
assert( pNew->anEq[pNew->iCol]>0 );
/* This sample is being added because the prefix that ends in column
** iCol occurs many times in the table. However, if we have already
** added a sample that shares this prefix, there is no need to add
** this one. Instead, upgrade the priority of the highest priority
** existing sample that shares this prefix. */
for(i=p->nSample-1; i>=0; i--){
Stat4Sample *pOld = &p->a[i];
if( pOld->anEq[pNew->iCol]==0 ){
if( pOld->isPSample ) return;
assert( pOld->iCol>pNew->iCol );
assert( sampleIsBetter(p, pNew, pOld) );
if( pUpgrade==0 || sampleIsBetter(p, pOld, pUpgrade) ){
pUpgrade = pOld;
}
}
}
if( pUpgrade ){
pUpgrade->iCol = pNew->iCol;
pUpgrade->anEq[pUpgrade->iCol] = pNew->anEq[pUpgrade->iCol];
goto find_new_min;
}
}
#endif
/* If necessary, remove sample iMin to make room for the new sample. */
if( p->nSample>=p->mxSample ){
Stat4Sample *pMin = &p->a[p->iMin];
tRowcnt *anEq = pMin->anEq;
tRowcnt *anLt = pMin->anLt;
tRowcnt *anDLt = pMin->anDLt;
sampleClear(p->db, pMin);
memmove(pMin, &pMin[1], sizeof(p->a[0])*(p->nSample-p->iMin-1));
pSample = &p->a[p->nSample-1];
pSample->nRowid = 0;
pSample->anEq = anEq;
pSample->anDLt = anDLt;
pSample->anLt = anLt;
p->nSample = p->mxSample-1;
}
/* The "rows less-than" for the rowid column must be greater than that
** for the last sample in the p->a[] array. Otherwise, the samples would
** be out of order. */
#ifdef SQLITE_ENABLE_STAT4
assert( p->nSample==0
|| pNew->anLt[p->nCol-1] > p->a[p->nSample-1].anLt[p->nCol-1] );
#endif
/* Insert the new sample */
pSample = &p->a[p->nSample];
sampleCopy(p, pSample, pNew);
p->nSample++;
/* Zero the first nEqZero entries in the anEq[] array. */
memset(pSample->anEq, 0, sizeof(tRowcnt)*nEqZero);
#ifdef SQLITE_ENABLE_STAT4
find_new_min:
#endif
if( p->nSample>=p->mxSample ){
int iMin = -1;
for(i=0; i<p->mxSample; i++){
if( p->a[i].isPSample ) continue;
if( iMin<0 || sampleIsBetter(p, &p->a[iMin], &p->a[i]) ){
iMin = i;
}
}
assert( iMin>=0 );
p->iMin = iMin;
}
}
#endif /* SQLITE_ENABLE_STAT3_OR_STAT4 */
/*
** Field iChng of the index being scanned has changed. So at this point
** p->current contains a sample that reflects the previous row of the
** index. The value of anEq[iChng] and subsequent anEq[] elements are
** correct at this point.
*/
static void samplePushPrevious(Stat4Accum *p, int iChng){
#ifdef SQLITE_ENABLE_STAT4
int i;
/* Check if any samples from the aBest[] array should be pushed
** into IndexSample.a[] at this point. */
for(i=(p->nCol-2); i>=iChng; i--){
Stat4Sample *pBest = &p->aBest[i];
pBest->anEq[i] = p->current.anEq[i];
if( p->nSample<p->mxSample || sampleIsBetter(p, pBest, &p->a[p->iMin]) ){
sampleInsert(p, pBest, i);
}
}
/* Update the anEq[] fields of any samples already collected. */
for(i=p->nSample-1; i>=0; i--){
int j;
for(j=iChng; j<p->nCol; j++){
if( p->a[i].anEq[j]==0 ) p->a[i].anEq[j] = p->current.anEq[j];
}
}
#endif
#if defined(SQLITE_ENABLE_STAT3) && !defined(SQLITE_ENABLE_STAT4)
if( iChng==0 ){
tRowcnt nLt = p->current.anLt[0];
tRowcnt nEq = p->current.anEq[0];
/* Check if this is to be a periodic sample. If so, add it. */
if( (nLt/p->nPSample)!=(nLt+nEq)/p->nPSample ){
p->current.isPSample = 1;
sampleInsert(p, &p->current, 0);
p->current.isPSample = 0;
}else
/* Or if it is a non-periodic sample. Add it in this case too. */
if( p->nSample<p->mxSample
|| sampleIsBetter(p, &p->current, &p->a[p->iMin])
){
sampleInsert(p, &p->current, 0);
}
}
#endif
#ifndef SQLITE_ENABLE_STAT3_OR_STAT4
UNUSED_PARAMETER( p );
UNUSED_PARAMETER( iChng );
#endif
}
/*
** Implementation of the stat_push SQL function: stat_push(P,C,R)
** Arguments:
**
** P Pointer to the Stat4Accum object created by stat_init()
** C Index of left-most column to differ from previous row
** R Rowid for the current row. Might be a key record for
** WITHOUT ROWID tables.
**
** This SQL function always returns NULL. It's purpose it to accumulate
** statistical data and/or samples in the Stat4Accum object about the
** index being analyzed. The stat_get() SQL function will later be used to
** extract relevant information for constructing the sqlite_statN tables.
**
** The R parameter is only used for STAT3 and STAT4
*/
static void statPush(
sqlite3_context *context,
int argc,
sqlite3_value **argv
){
int i;
/* The three function arguments */
Stat4Accum *p = (Stat4Accum*)sqlite3_value_blob(argv[0]);
int iChng = sqlite3_value_int(argv[1]);
UNUSED_PARAMETER( argc );
UNUSED_PARAMETER( context );
assert( p->nCol>0 );
assert( iChng<p->nCol );
if( p->nRow==0 ){
/* This is the first call to this function. Do initialization. */
for(i=0; i<p->nCol; i++) p->current.anEq[i] = 1;
}else{
/* Second and subsequent calls get processed here */
samplePushPrevious(p, iChng);
/* Update anDLt[], anLt[] and anEq[] to reflect the values that apply
** to the current row of the index. */
for(i=0; i<iChng; i++){
p->current.anEq[i]++;
}
for(i=iChng; i<p->nCol; i++){
p->current.anDLt[i]++;
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
p->current.anLt[i] += p->current.anEq[i];
#endif
p->current.anEq[i] = 1;
}
}
p->nRow++;
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
if( sqlite3_value_type(argv[2])==SQLITE_INTEGER ){
sampleSetRowidInt64(p->db, &p->current, sqlite3_value_int64(argv[2]));
}else{
sampleSetRowid(p->db, &p->current, sqlite3_value_bytes(argv[2]),
sqlite3_value_blob(argv[2]));
}
p->current.iHash = p->iPrn = p->iPrn*1103515245 + 12345;
#endif
#ifdef SQLITE_ENABLE_STAT4
{
tRowcnt nLt = p->current.anLt[p->nCol-1];
/* Check if this is to be a periodic sample. If so, add it. */
if( (nLt/p->nPSample)!=(nLt+1)/p->nPSample ){
p->current.isPSample = 1;
p->current.iCol = 0;
sampleInsert(p, &p->current, p->nCol-1);
p->current.isPSample = 0;
}
/* Update the aBest[] array. */
for(i=0; i<(p->nCol-1); i++){
p->current.iCol = i;
if( i>=iChng || sampleIsBetterPost(p, &p->current, &p->aBest[i]) ){
sampleCopy(p, &p->aBest[i], &p->current);
}
}
}
#endif
}
static const FuncDef statPushFuncdef = {
2+IsStat34, /* nArg */
SQLITE_UTF8, /* funcFlags */
0, /* pUserData */
0, /* pNext */
statPush, /* xFunc */
0, /* xStep */
0, /* xFinalize */
"stat_push", /* zName */
0, /* pHash */
0 /* pDestructor */
};
#define STAT_GET_STAT1 0 /* "stat" column of stat1 table */
#define STAT_GET_ROWID 1 /* "rowid" column of stat[34] entry */
#define STAT_GET_NEQ 2 /* "neq" column of stat[34] entry */
#define STAT_GET_NLT 3 /* "nlt" column of stat[34] entry */
#define STAT_GET_NDLT 4 /* "ndlt" column of stat[34] entry */
/*
** Implementation of the stat_get(P,J) SQL function. This routine is
** used to query statistical information that has been gathered into
** the Stat4Accum object by prior calls to stat_push(). The P parameter
** has type BLOB but it is really just a pointer to the Stat4Accum object.
** The content to returned is determined by the parameter J
** which is one of the STAT_GET_xxxx values defined above.
**
** If neither STAT3 nor STAT4 are enabled, then J is always
** STAT_GET_STAT1 and is hence omitted and this routine becomes
** a one-parameter function, stat_get(P), that always returns the
** stat1 table entry information.
*/
static void statGet(
sqlite3_context *context,
int argc,
sqlite3_value **argv
){
Stat4Accum *p = (Stat4Accum*)sqlite3_value_blob(argv[0]);
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
/* STAT3 and STAT4 have a parameter on this routine. */
int eCall = sqlite3_value_int(argv[1]);
assert( argc==2 );
assert( eCall==STAT_GET_STAT1 || eCall==STAT_GET_NEQ
|| eCall==STAT_GET_ROWID || eCall==STAT_GET_NLT
|| eCall==STAT_GET_NDLT
);
if( eCall==STAT_GET_STAT1 )
#else
assert( argc==1 );
#endif
{
/* Return the value to store in the "stat" column of the sqlite_stat1
** table for this index.
**
** The value is a string composed of a list of integers describing
** the index. The first integer in the list is the total number of
** entries in the index. There is one additional integer in the list
** for each indexed column. This additional integer is an estimate of
** the number of rows matched by a stabbing query on the index using
** a key with the corresponding number of fields. In other words,
** if the index is on columns (a,b) and the sqlite_stat1 value is
** "100 10 2", then SQLite estimates that:
**
** * the index contains 100 rows,
** * "WHERE a=?" matches 10 rows, and
** * "WHERE a=? AND b=?" matches 2 rows.
**
** If D is the count of distinct values and K is the total number of
** rows, then each estimate is computed as:
**
** I = (K+D-1)/D
*/
char *z;
int i;
char *zRet = sqlite3MallocZero( (p->nKeyCol+1)*25 );
if( zRet==0 ){
sqlite3_result_error_nomem(context);
return;
}
sqlite3_snprintf(24, zRet, "%llu", (u64)p->nRow);
z = zRet + sqlite3Strlen30(zRet);
for(i=0; i<p->nKeyCol; i++){
u64 nDistinct = p->current.anDLt[i] + 1;
u64 iVal = (p->nRow + nDistinct - 1) / nDistinct;
sqlite3_snprintf(24, z, " %llu", iVal);
z += sqlite3Strlen30(z);
assert( p->current.anEq[i] );
}
assert( z[0]=='\0' && z>zRet );
sqlite3_result_text(context, zRet, -1, sqlite3_free);
}
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
else if( eCall==STAT_GET_ROWID ){
if( p->iGet<0 ){
samplePushPrevious(p, 0);
p->iGet = 0;
}
if( p->iGet<p->nSample ){
Stat4Sample *pS = p->a + p->iGet;
if( pS->nRowid==0 ){
sqlite3_result_int64(context, pS->u.iRowid);
}else{
sqlite3_result_blob(context, pS->u.aRowid, pS->nRowid,
SQLITE_TRANSIENT);
}
}
}else{
tRowcnt *aCnt = 0;
assert( p->iGet<p->nSample );
switch( eCall ){
case STAT_GET_NEQ: aCnt = p->a[p->iGet].anEq; break;
case STAT_GET_NLT: aCnt = p->a[p->iGet].anLt; break;
default: {
aCnt = p->a[p->iGet].anDLt;
p->iGet++;
break;
}
}
if( IsStat3 ){
sqlite3_result_int64(context, (i64)aCnt[0]);
}else{
char *zRet = sqlite3MallocZero(p->nCol * 25);
if( zRet==0 ){
sqlite3_result_error_nomem(context);
}else{
int i;
char *z = zRet;
for(i=0; i<p->nCol; i++){
sqlite3_snprintf(24, z, "%llu ", (u64)aCnt[i]);
z += sqlite3Strlen30(z);
}
assert( z[0]=='\0' && z>zRet );
z[-1] = '\0';
sqlite3_result_text(context, zRet, -1, sqlite3_free);
}
}
}
#endif /* SQLITE_ENABLE_STAT3_OR_STAT4 */
#ifndef SQLITE_DEBUG
UNUSED_PARAMETER( argc );
#endif
}
static const FuncDef statGetFuncdef = {
1+IsStat34, /* nArg */
SQLITE_UTF8, /* funcFlags */
0, /* pUserData */
0, /* pNext */
statGet, /* xFunc */
0, /* xStep */
0, /* xFinalize */
"stat_get", /* zName */
0, /* pHash */
0 /* pDestructor */
};
static void callStatGet(Vdbe *v, int regStat4, int iParam, int regOut){
assert( regOut!=regStat4 && regOut!=regStat4+1 );
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
sqlite3VdbeAddOp2(v, OP_Integer, iParam, regStat4+1);
#elif SQLITE_DEBUG
assert( iParam==STAT_GET_STAT1 );
#else
UNUSED_PARAMETER( iParam );
#endif
sqlite3VdbeAddOp3(v, OP_Function, 0, regStat4, regOut);
sqlite3VdbeChangeP4(v, -1, (char*)&statGetFuncdef, P4_FUNCDEF);
sqlite3VdbeChangeP5(v, 1 + IsStat34);
}
/*
** Generate code to do an analysis of all indices associated with
** a single table.
*/
static void analyzeOneTable(
Parse *pParse, /* Parser context */
Table *pTab, /* Table whose indices are to be analyzed */
Index *pOnlyIdx, /* If not NULL, only analyze this one index */
int iStatCur, /* Index of VdbeCursor that writes the sqlite_stat1 table */
int iMem, /* Available memory locations begin here */
int iTab /* Next available cursor */
){
sqlite3 *db = pParse->db; /* Database handle */
Index *pIdx; /* An index to being analyzed */
int iIdxCur; /* Cursor open on index being analyzed */
int iTabCur; /* Table cursor */
Vdbe *v; /* The virtual machine being built up */
int i; /* Loop counter */
int jZeroRows = -1; /* Jump from here if number of rows is zero */
int iDb; /* Index of database containing pTab */
u8 needTableCnt = 1; /* True to count the table */
int regNewRowid = iMem++; /* Rowid for the inserted record */
int regStat4 = iMem++; /* Register to hold Stat4Accum object */
int regChng = iMem++; /* Index of changed index field */
#ifdef SQLITE_ENABLE_STAT3_OR_STAT4
int regRowid = iMem++; /* Rowid argument passed to stat_push() */
#endif
int regTemp = iMem++; /* Temporary use register */
int regTabname = iMem++; /* Register containing table name */
int regIdxname = iMem++; /* Register containing index name */
int regStat1 = iMem++; /* Value for the stat column of sqlite_stat1 */
int regPrev = iMem; /* MUST BE LAST (see below) */
pParse->nMem = MAX(pParse->nMem, iMem);
v = sqlite3GetVdbe(pParse);
if( v==0 || NEVER(pTab==0) ){
return;
}
if( pTab->tnum==0 ){
/* Do not gather statistics on views or virtual tables */
return;
}
if( sqlite3_strnicmp(pTab->zName, "sqlite_", 7)==0 ){
/* Do not gather statistics on system tables */
return;
}
assert( sqlite3BtreeHoldsAllMutexes(db) );
iDb = sqlite3SchemaToIndex(db, pTab->pSchema);
assert( iDb>=0 );
assert( sqlite3SchemaMutexHeld(db, iDb, 0) );