/ Check-in [7eb022d7]
Login

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

Overview
Comment:Instead of the 4-byte fields, use regular varints for the poslist-size field in fts5_hash.c.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | fts5
Files: files | file ages | folders
SHA1: 7eb022d7e5fdb180af823c82c47c938e4a7a355f
User & Date: dan 2015-02-25 19:24:37
Context
2015-02-26
14:54
Fix an fts5 bug in large incremental merges. check-in: 208e3cb6 user: dan tags: fts5
2015-02-25
19:24
Instead of the 4-byte fields, use regular varints for the poslist-size field in fts5_hash.c. check-in: 7eb022d7 user: dan tags: fts5
2015-02-02
11:58
Ensure generated header file fts5parse.h is included in sqlite3.c. check-in: bc7be2fc user: dan tags: fts5
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to ext/fts5/fts5Int.h.

347
348
349
350
351
352
353


354
355
356
357
358
359
360
int sqlite3Fts5IndexReinit(Fts5Index *p);
int sqlite3Fts5IndexOptimize(Fts5Index *p);

int sqlite3Fts5IndexLoadConfig(Fts5Index *p);

int sqlite3Fts5GetVarint32(const unsigned char *p, u32 *v);
#define fts5GetVarint32(a,b) sqlite3Fts5GetVarint32(a,(u32*)&b)



/*
** End of interface to code in fts5_index.c.
**************************************************************************/

/**************************************************************************
** Interface to code in fts5_hash.c. 







>
>







347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
int sqlite3Fts5IndexReinit(Fts5Index *p);
int sqlite3Fts5IndexOptimize(Fts5Index *p);

int sqlite3Fts5IndexLoadConfig(Fts5Index *p);

int sqlite3Fts5GetVarint32(const unsigned char *p, u32 *v);
#define fts5GetVarint32(a,b) sqlite3Fts5GetVarint32(a,(u32*)&b)

int sqlite3Fts5GetVarintLen(u32 iVal);

/*
** End of interface to code in fts5_index.c.
**************************************************************************/

/**************************************************************************
** Interface to code in fts5_hash.c. 

Changes to ext/fts5/fts5_hash.c.

179
180
181
182
183
184
185

















186
187
188
189
190
191
192
...
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
...
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
...
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
...
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
...
478
479
480
481
482
483
484
485
486
487

488
489
490
491
492
493
494
495
496
497
  }

  sqlite3_free(apOld);
  pHash->nSlot = nNew;
  pHash->aSlot = apNew;
  return SQLITE_OK;
}


















int sqlite3Fts5HashWrite(
  Fts5Hash *pHash,
  i64 iRowid,                     /* Rowid for this entry */
  int iCol,                       /* Column token appears in (-ve -> delete) */
  int iPos,                       /* Position of token within column */
  const char *pToken, int nToken  /* Token to add or remove to or from index */
................................................................................
    memset(p, 0, sizeof(Fts5HashEntry));
    p->nAlloc = nByte;
    memcpy(p->zKey, pToken, nToken);
    p->zKey[nToken] = '\0';
    p->nData = nToken + 1 + sizeof(Fts5HashEntry);
    p->nData += sqlite3PutVarint(&((u8*)p)[p->nData], iRowid);
    p->iSzPoslist = p->nData;
    p->nData += 4;
    p->iRowid = iRowid;
    p->pHashNext = pHash->aSlot[iHash];
    pHash->aSlot[iHash] = p;
    pHash->nEntry++;
    nIncr += p->nData;
  }

  /* Check there is enough space to append a new entry. Worst case scenario
  ** is:
  **
  **     + 9 bytes for a new rowid,
  **     + 4 bytes reserved for the "poslist size" varint.
  **     + 1 byte for a "new column" byte,
  **     + 3 bytes for a new column number (16-bit max) as a varint,
  **     + 5 bytes for the new position offset (32-bit max).
  */
  if( (p->nAlloc - p->nData) < (9 + 4 + 1 + 3 + 5) ){
    int nNew = p->nAlloc * 2;
    Fts5HashEntry *pNew;
................................................................................
  }
  pPtr = (u8*)p;
  nIncr -= p->nData;

  /* If this is a new rowid, append the 4-byte size field for the previous
  ** entry, and the new rowid for this entry.  */
  if( iRowid!=p->iRowid ){
    assert( p->iSzPoslist>0 );
    fts5Put4ByteVarint(&pPtr[p->iSzPoslist], p->nData - p->iSzPoslist - 4);
    p->nData += sqlite3PutVarint(&pPtr[p->nData], iRowid - p->iRowid);
    p->iSzPoslist = p->nData;
    p->nData += 4;
    p->iCol = 0;
    p->iPos = 0;
    p->iRowid = iRowid;
  }

  if( iCol>=0 ){
    /* Append a new column value, if necessary */
................................................................................

  rc = fts5HashEntrySort(pHash, 0, 0, &pList);
  if( rc==SQLITE_OK ){
    memset(pHash->aSlot, 0, sizeof(Fts5HashEntry*) * pHash->nSlot);
    while( pList ){
      Fts5HashEntry *pNext = pList->pScanNext;
      if( rc==SQLITE_OK ){
        const int nSz = pList->nData - pList->iSzPoslist - 4;
        const int nKey = strlen(pList->zKey);
        i64 iRowid = 0;
        u8 *pPtr = (u8*)pList;
        int iOff = sizeof(Fts5HashEntry) + nKey + 1;

        /* Fill in the final poslist size field */
        fts5Put4ByteVarint(&pPtr[pList->iSzPoslist], nSz);
        
        /* Issue the new-term callback */
        rc = xTerm(pCtx, pList->zKey, nKey);

        /* Issue the xEntry callbacks */
        while( rc==SQLITE_OK && iOff<pList->nData ){
          i64 iDelta;             /* Rowid delta value */
          int nPoslist;           /* Size of position list in bytes */
          int nVarint;
          iOff += getVarint(&pPtr[iOff], (u64*)&iDelta);
          iRowid += iDelta;
          nPoslist = fts5Get4ByteVarint(&pPtr[iOff], &nVarint);
          iOff += 4;
          rc = xEntry(pCtx, iRowid, &pPtr[iOff-nVarint], nPoslist+nVarint);
          iOff += nPoslist;
        }

        /* Issue the term-done callback */
        if( rc==SQLITE_OK ) rc = xTermDone(pCtx);
      }
      sqlite3_free(pList);
      pList = pNext;
................................................................................
  Fts5HashEntry *p;

  for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){
    if( memcmp(p->zKey, pTerm, nTerm)==0 && p->zKey[nTerm]==0 ) break;
  }

  if( p ){
    u8 *pPtr = (u8*)p;
    fts5Put4ByteVarint(&pPtr[p->iSzPoslist], p->nData - p->iSzPoslist - 4);
    *ppDoclist = &p->zKey[nTerm+1];
    *pnDoclist = p->nData - (sizeof(*p) + nTerm + 1);
  }else{
    *ppDoclist = 0;
    *pnDoclist = 0;
  }

................................................................................
  Fts5Hash *pHash,
  const char **pzTerm,            /* OUT: term (nul-terminated) */
  const char **ppDoclist,         /* OUT: pointer to doclist */
  int *pnDoclist                  /* OUT: size of doclist in bytes */
){
  Fts5HashEntry *p;
  if( (p = pHash->pScan) ){
    u8 *pPtr = (u8*)p;
    int nTerm = strlen(p->zKey);
    fts5Put4ByteVarint(&pPtr[p->iSzPoslist], p->nData - p->iSzPoslist - 4);

    *pzTerm = p->zKey;
    *ppDoclist = &p->zKey[nTerm+1];
    *pnDoclist = p->nData - (sizeof(*p) + nTerm + 1);
  }else{
    *pzTerm = 0;
    *ppDoclist = 0;
    *pnDoclist = 0;
  }
}








>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







 







|











|







 







|
<


|







 







<






|











|
<
|
|







 







|
<







 







<

<
>










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
...
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
...
268
269
270
271
272
273
274
275

276
277
278
279
280
281
282
283
284
285
...
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
...
455
456
457
458
459
460
461
462

463
464
465
466
467
468
469
...
491
492
493
494
495
496
497

498

499
500
501
502
503
504
505
506
507
508
509
  }

  sqlite3_free(apOld);
  pHash->nSlot = nNew;
  pHash->aSlot = apNew;
  return SQLITE_OK;
}

static void fts5HashAddPoslistSize(Fts5HashEntry *p){
  if( p->iSzPoslist ){
    u8 *pPtr = (u8*)p;
    int nSz = p->nData - p->iSzPoslist - 1;

    if( nSz<=127 ){
      pPtr[p->iSzPoslist] = nSz;
    }else{
      int nByte = sqlite3Fts5GetVarintLen((u32)nSz);
      memmove(&pPtr[p->iSzPoslist + nByte], &pPtr[p->iSzPoslist + 1], nSz);
      sqlite3PutVarint(&pPtr[p->iSzPoslist], nSz);
      p->nData += (nByte-1);
    }
    p->iSzPoslist = 0;
  }
}

int sqlite3Fts5HashWrite(
  Fts5Hash *pHash,
  i64 iRowid,                     /* Rowid for this entry */
  int iCol,                       /* Column token appears in (-ve -> delete) */
  int iPos,                       /* Position of token within column */
  const char *pToken, int nToken  /* Token to add or remove to or from index */
................................................................................
    memset(p, 0, sizeof(Fts5HashEntry));
    p->nAlloc = nByte;
    memcpy(p->zKey, pToken, nToken);
    p->zKey[nToken] = '\0';
    p->nData = nToken + 1 + sizeof(Fts5HashEntry);
    p->nData += sqlite3PutVarint(&((u8*)p)[p->nData], iRowid);
    p->iSzPoslist = p->nData;
    p->nData += 1;
    p->iRowid = iRowid;
    p->pHashNext = pHash->aSlot[iHash];
    pHash->aSlot[iHash] = p;
    pHash->nEntry++;
    nIncr += p->nData;
  }

  /* Check there is enough space to append a new entry. Worst case scenario
  ** is:
  **
  **     + 9 bytes for a new rowid,
  **     + 4 byte reserved for the "poslist size" varint.
  **     + 1 byte for a "new column" byte,
  **     + 3 bytes for a new column number (16-bit max) as a varint,
  **     + 5 bytes for the new position offset (32-bit max).
  */
  if( (p->nAlloc - p->nData) < (9 + 4 + 1 + 3 + 5) ){
    int nNew = p->nAlloc * 2;
    Fts5HashEntry *pNew;
................................................................................
  }
  pPtr = (u8*)p;
  nIncr -= p->nData;

  /* If this is a new rowid, append the 4-byte size field for the previous
  ** entry, and the new rowid for this entry.  */
  if( iRowid!=p->iRowid ){
    fts5HashAddPoslistSize(p);

    p->nData += sqlite3PutVarint(&pPtr[p->nData], iRowid - p->iRowid);
    p->iSzPoslist = p->nData;
    p->nData += 1;
    p->iCol = 0;
    p->iPos = 0;
    p->iRowid = iRowid;
  }

  if( iCol>=0 ){
    /* Append a new column value, if necessary */
................................................................................

  rc = fts5HashEntrySort(pHash, 0, 0, &pList);
  if( rc==SQLITE_OK ){
    memset(pHash->aSlot, 0, sizeof(Fts5HashEntry*) * pHash->nSlot);
    while( pList ){
      Fts5HashEntry *pNext = pList->pScanNext;
      if( rc==SQLITE_OK ){

        const int nKey = strlen(pList->zKey);
        i64 iRowid = 0;
        u8 *pPtr = (u8*)pList;
        int iOff = sizeof(Fts5HashEntry) + nKey + 1;

        /* Fill in the final poslist size field */
        fts5HashAddPoslistSize(pList);
        
        /* Issue the new-term callback */
        rc = xTerm(pCtx, pList->zKey, nKey);

        /* Issue the xEntry callbacks */
        while( rc==SQLITE_OK && iOff<pList->nData ){
          i64 iDelta;             /* Rowid delta value */
          int nPoslist;           /* Size of position list in bytes */
          int nVarint;
          iOff += getVarint(&pPtr[iOff], (u64*)&iDelta);
          iRowid += iDelta;
          nVarint = fts5GetVarint32(&pPtr[iOff], nPoslist);

          rc = xEntry(pCtx, iRowid, &pPtr[iOff], nPoslist+nVarint);
          iOff += nVarint+nPoslist;
        }

        /* Issue the term-done callback */
        if( rc==SQLITE_OK ) rc = xTermDone(pCtx);
      }
      sqlite3_free(pList);
      pList = pNext;
................................................................................
  Fts5HashEntry *p;

  for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){
    if( memcmp(p->zKey, pTerm, nTerm)==0 && p->zKey[nTerm]==0 ) break;
  }

  if( p ){
    fts5HashAddPoslistSize(p);

    *ppDoclist = &p->zKey[nTerm+1];
    *pnDoclist = p->nData - (sizeof(*p) + nTerm + 1);
  }else{
    *ppDoclist = 0;
    *pnDoclist = 0;
  }

................................................................................
  Fts5Hash *pHash,
  const char **pzTerm,            /* OUT: term (nul-terminated) */
  const char **ppDoclist,         /* OUT: pointer to doclist */
  int *pnDoclist                  /* OUT: size of doclist in bytes */
){
  Fts5HashEntry *p;
  if( (p = pHash->pScan) ){

    int nTerm = strlen(p->zKey);

    fts5HashAddPoslistSize(p);
    *pzTerm = p->zKey;
    *ppDoclist = &p->zKey[nTerm+1];
    *pnDoclist = p->nData - (sizeof(*p) + nTerm + 1);
  }else{
    *pzTerm = 0;
    *ppDoclist = 0;
    *pnDoclist = 0;
  }
}

Changes to ext/fts5/fts5_index.c.

367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
...
662
663
664
665
666
667
668








669
670
671
672
673
674
675

/*
** An object of type Fts5SegWriter is used to write to segments.
*/
struct Fts5PageWriter {
  int pgno;                       /* Page number for this page */
  Fts5Buffer buf;                 /* Buffer containing page data */
  Fts5Buffer term;              /* Buffer containing previous term on page */
};
struct Fts5SegWriter {
  int iIdx;                       /* Index to write to */
  int iSegid;                     /* Segid to write to */
  int nWriter;                    /* Number of entries in aWriter */
  Fts5PageWriter *aWriter;        /* Array of PageWriter objects */
  i64 iPrevRowid;                 /* Previous docid written to current leaf */
................................................................................
    p -= 2;
    n = sqlite3GetVarint(p, &v64);
    *v = (u32)v64;
    assert( n>3 && n<=9 );
    return n;
  }
}









/*
** Allocate and return a buffer at least nByte bytes in size.
**
** If an OOM error is encountered, return NULL and set the error code in
** the Fts5Index handle passed as the first argument.
*/







|







 







>
>
>
>
>
>
>
>







367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
...
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683

/*
** An object of type Fts5SegWriter is used to write to segments.
*/
struct Fts5PageWriter {
  int pgno;                       /* Page number for this page */
  Fts5Buffer buf;                 /* Buffer containing page data */
  Fts5Buffer term;                /* Buffer containing previous term on page */
};
struct Fts5SegWriter {
  int iIdx;                       /* Index to write to */
  int iSegid;                     /* Segid to write to */
  int nWriter;                    /* Number of entries in aWriter */
  Fts5PageWriter *aWriter;        /* Array of PageWriter objects */
  i64 iPrevRowid;                 /* Previous docid written to current leaf */
................................................................................
    p -= 2;
    n = sqlite3GetVarint(p, &v64);
    *v = (u32)v64;
    assert( n>3 && n<=9 );
    return n;
  }
}

int sqlite3Fts5GetVarintLen(u32 iVal){
  if( iVal<(1 << 7 ) ) return 1;
  if( iVal<(1 << 14) ) return 2;
  if( iVal<(1 << 21) ) return 3;
  if( iVal<(1 << 28) ) return 4;
  return 5;
}

/*
** Allocate and return a buffer at least nByte bytes in size.
**
** If an OOM error is encountered, return NULL and set the error code in
** the Fts5Index handle passed as the first argument.
*/

Changes to ext/fts5/tool/loadfts5.tcl.

32
33
34
35
36
37
38

39
40
41
42
43
44
45
46
47

48
49
50
51
52
53
54
..
60
61
62
63
64
65
66




67
68
69
70
71
72
73
..
82
83
84
85
86
87
88
89


90
91
92
93
94
95
96
proc usage {} {
  puts stderr "Usage: $::argv0 ?SWITCHES? DATABASE PATH"
  puts stderr ""
  puts stderr "Switches are:"
  puts stderr "  -fts4        (use fts4 instead of fts5)"
  puts stderr "  -fts5        (use fts5)"
  puts stderr "  -porter      (use porter tokenizer)"

  puts stderr "  -limit N     (load no more than N documents)"
  puts stderr "  -automerge N (set the automerge parameter to N)"
  puts stderr "  -crisismerge N (set the crisismerge parameter to N)"
  exit 1
}

set O(vtab)       fts5
set O(tok)        ""
set O(limit)      0

set O(automerge)  -1
set O(crisismerge)  -1

if {[llength $argv]<2} usage
set nOpt [expr {[llength $argv]-2}]
for {set i 0} {$i < $nOpt} {incr i} {
  set arg [lindex $argv $i]
................................................................................
    -fts5 {
      set O(vtab) fts5
    }

    -porter {
      set O(tok) ", tokenize=porter"
    }





    -limit {
      if { [incr i]>=$nOpt } usage
      set O(limit) [lindex $argv $i]
    }
    
    -automerge {
................................................................................

    default {
      usage
    }
  }
}

sqlite3 db [lindex $argv end-1]


db func loadfile loadfile

db transaction {
  catch {
    db eval "CREATE VIRTUAL TABLE t1 USING $O(vtab) (path, content$O(tok))"
  }
  if {$O(automerge)>=0} {







>









>







 







>
>
>
>







 







|
>
>







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
..
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
..
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
proc usage {} {
  puts stderr "Usage: $::argv0 ?SWITCHES? DATABASE PATH"
  puts stderr ""
  puts stderr "Switches are:"
  puts stderr "  -fts4        (use fts4 instead of fts5)"
  puts stderr "  -fts5        (use fts5)"
  puts stderr "  -porter      (use porter tokenizer)"
  puts stderr "  -delete      (delete the database file before starting)"
  puts stderr "  -limit N     (load no more than N documents)"
  puts stderr "  -automerge N (set the automerge parameter to N)"
  puts stderr "  -crisismerge N (set the crisismerge parameter to N)"
  exit 1
}

set O(vtab)       fts5
set O(tok)        ""
set O(limit)      0
set O(delete)     0
set O(automerge)  -1
set O(crisismerge)  -1

if {[llength $argv]<2} usage
set nOpt [expr {[llength $argv]-2}]
for {set i 0} {$i < $nOpt} {incr i} {
  set arg [lindex $argv $i]
................................................................................
    -fts5 {
      set O(vtab) fts5
    }

    -porter {
      set O(tok) ", tokenize=porter"
    }

    -delete {
      set O(delete) 1
    }

    -limit {
      if { [incr i]>=$nOpt } usage
      set O(limit) [lindex $argv $i]
    }
    
    -automerge {
................................................................................

    default {
      usage
    }
  }
}

set dbfile [lindex $argv end-1]
if {$O(delete)} { file delete -force $dbfile }
sqlite3 db $dbfile
db func loadfile loadfile

db transaction {
  catch {
    db eval "CREATE VIRTUAL TABLE t1 USING $O(vtab) (path, content$O(tok))"
  }
  if {$O(automerge)>=0} {