SQLite

Check-in [45c73deb]
Login

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

Overview
Comment:Fix an fts5 problem with interleaving reads and writes in a single transaction.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA3-256: 45c73deb440496e848cb24d4c1326d4105dacfee8bbafb115e567051855e6518
User & Date: dan 2019-03-18 15:23:20
Context
2019-03-18
15:49
Fix a buffer overread that could occur when running fts5 prefix queries inside a transaction. (check-in: b3fa58dd user: dan tags: trunk)
15:23
Fix an fts5 problem with interleaving reads and writes in a single transaction. (check-in: 45c73deb user: dan tags: trunk)
10:30
Fix a typo in a comment. No changes to code. (check-in: c2f50aa4 user: drh tags: trunk)
Changes
Unified Diff Ignore Whitespace Patch
Changes to ext/fts5/fts5Int.h.
579
580
581
582
583
584
585

586
587
588
589
590
591
592
593
594
/*
** Empty (but do not delete) a hash table.
*/
void sqlite3Fts5HashClear(Fts5Hash*);

int sqlite3Fts5HashQuery(
  Fts5Hash*,                      /* Hash table to query */

  const char *pTerm, int nTerm,   /* Query term */
  const u8 **ppDoclist,           /* OUT: Pointer to doclist for pTerm */
  int *pnDoclist                  /* OUT: Size of doclist in bytes */
);

int sqlite3Fts5HashScanInit(
  Fts5Hash*,                      /* Hash table to query */
  const char *pTerm, int nTerm    /* Query prefix */
);







>

|







579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
/*
** Empty (but do not delete) a hash table.
*/
void sqlite3Fts5HashClear(Fts5Hash*);

int sqlite3Fts5HashQuery(
  Fts5Hash*,                      /* Hash table to query */
  int nPre,
  const char *pTerm, int nTerm,   /* Query term */
  void **ppObj,                   /* OUT: Pointer to doclist for pTerm */
  int *pnDoclist                  /* OUT: Size of doclist in bytes */
);

int sqlite3Fts5HashScanInit(
  Fts5Hash*,                      /* Hash table to query */
  const char *pTerm, int nTerm    /* Query prefix */
);
Changes to ext/fts5/fts5_hash.c.
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

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

static void fts5HashAddPoslistSize(Fts5Hash *pHash, Fts5HashEntry *p){





  if( p->iSzPoslist ){
    u8 *pPtr = (u8*)p;

    if( pHash->eDetail==FTS5_DETAIL_NONE ){
      assert( p->nData==p->iSzPoslist );
      if( p->bDel ){
        pPtr[p->nData++] = 0x00;
        if( p->bContent ){
          pPtr[p->nData++] = 0x00;
        }
      }
    }else{
      int nSz = (p->nData - p->iSzPoslist - 1);       /* Size in bytes */
      int nPos = nSz*2 + p->bDel;                     /* Value of nPos field */

      assert( p->bDel==0 || p->bDel==1 );
      if( nPos<=127 ){
        pPtr[p->iSzPoslist] = (u8)nPos;
      }else{
        int nByte = sqlite3Fts5GetVarintLen((u32)nPos);
        memmove(&pPtr[p->iSzPoslist + nByte], &pPtr[p->iSzPoslist + 1], nSz);
        sqlite3Fts5PutVarint(&pPtr[p->iSzPoslist], nPos);
        p->nData += (nByte-1);
      }
    }



    p->iSzPoslist = 0;
    p->bDel = 0;
    p->bContent = 0;

  }


}

/*
** Add an entry to the in-memory hash table. The key is the concatenation
** of bByte and (pToken/nToken). The value is (iRowid/iCol/iPos).
**
**     (bByte || pToken) -> (iRowid,iCol,iPos)







|
>
>
>
>
>

|
>

|

|

|



|









|



>
>
|
|
|
>
|
>
>







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

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

static int fts5HashAddPoslistSize(
  Fts5Hash *pHash, 
  Fts5HashEntry *p,
  Fts5HashEntry *p2
){
  int nRet = 0;
  if( p->iSzPoslist ){
    u8 *pPtr = p2 ? (u8*)p2 : (u8*)p;
    int nData = p->nData;
    if( pHash->eDetail==FTS5_DETAIL_NONE ){
      assert( nData==p->iSzPoslist );
      if( p->bDel ){
        pPtr[nData++] = 0x00;
        if( p->bContent ){
          pPtr[nData++] = 0x00;
        }
      }
    }else{
      int nSz = (nData - p->iSzPoslist - 1);       /* Size in bytes */
      int nPos = nSz*2 + p->bDel;                     /* Value of nPos field */

      assert( p->bDel==0 || p->bDel==1 );
      if( nPos<=127 ){
        pPtr[p->iSzPoslist] = (u8)nPos;
      }else{
        int nByte = sqlite3Fts5GetVarintLen((u32)nPos);
        memmove(&pPtr[p->iSzPoslist + nByte], &pPtr[p->iSzPoslist + 1], nSz);
        sqlite3Fts5PutVarint(&pPtr[p->iSzPoslist], nPos);
        nData += (nByte-1);
      }
    }

    nRet = nData - p->nData;
    if( p2==0 ){
      p->iSzPoslist = 0;
      p->bDel = 0;
      p->bContent = 0;
      p->nData = nData;
    }
  }
  return nRet;
}

/*
** Add an entry to the in-memory hash table. The key is the concatenation
** of bByte and (pToken/nToken). The value is (iRowid/iCol/iPos).
**
**     (bByte || pToken) -> (iRowid,iCol,iPos)
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
  assert( (p->nAlloc - p->nData) >= (9 + 4 + 1 + 3 + 5) );

  pPtr = (u8*)p;

  /* 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(pHash, p);
    p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iRowid - p->iRowid);
    p->iRowid = iRowid;
    bNew = 1;
    p->iSzPoslist = p->nData;
    if( pHash->eDetail!=FTS5_DETAIL_NONE ){
      p->nData += 1;
      p->iCol = (pHash->eDetail==FTS5_DETAIL_FULL ? 0 : -1);







|







335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
  assert( (p->nAlloc - p->nData) >= (9 + 4 + 1 + 3 + 5) );

  pPtr = (u8*)p;

  /* 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(pHash, p, 0);
    p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iRowid - p->iRowid);
    p->iRowid = iRowid;
    bNew = 1;
    p->iSzPoslist = p->nData;
    if( pHash->eDetail!=FTS5_DETAIL_NONE ){
      p->nData += 1;
      p->iCol = (pHash->eDetail==FTS5_DETAIL_FULL ? 0 : -1);
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
}

/*
** Query the hash table for a doclist associated with term pTerm/nTerm.
*/
int sqlite3Fts5HashQuery(
  Fts5Hash *pHash,                /* Hash table to query */

  const char *pTerm, int nTerm,   /* Query term */
  const u8 **ppDoclist,           /* OUT: Pointer to doclist for pTerm */
  int *pnDoclist                  /* OUT: Size of doclist in bytes */
){
  unsigned int iHash = fts5HashKey(pHash->nSlot, (const u8*)pTerm, nTerm);
  char *zKey = 0;
  Fts5HashEntry *p;

  for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){
    zKey = fts5EntryKey(p);
    assert( p->nKey+1==(int)strlen(zKey) );
    if( nTerm==p->nKey+1 && memcmp(zKey, pTerm, nTerm)==0 ) break;
  }

  if( p ){






    fts5HashAddPoslistSize(pHash, p);
    *ppDoclist = (const u8*)&zKey[nTerm+1];

    *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm + 1);


  }else{
    *ppDoclist = 0;
    *pnDoclist = 0;
  }

  return SQLITE_OK;
}

int sqlite3Fts5HashScanInit(







>

|













>
>
>
>
>
>
|
|
>
|
>
>

|







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
}

/*
** Query the hash table for a doclist associated with term pTerm/nTerm.
*/
int sqlite3Fts5HashQuery(
  Fts5Hash *pHash,                /* Hash table to query */
  int nPre,
  const char *pTerm, int nTerm,   /* Query term */
  void **ppOut,                   /* OUT: Pointer to new object */
  int *pnDoclist                  /* OUT: Size of doclist in bytes */
){
  unsigned int iHash = fts5HashKey(pHash->nSlot, (const u8*)pTerm, nTerm);
  char *zKey = 0;
  Fts5HashEntry *p;

  for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){
    zKey = fts5EntryKey(p);
    assert( p->nKey+1==(int)strlen(zKey) );
    if( nTerm==p->nKey+1 && memcmp(zKey, pTerm, nTerm)==0 ) break;
  }

  if( p ){
    int nHashPre = sizeof(Fts5HashEntry) + nTerm + 1;
    int nList = p->nData - nHashPre;
    u8 *pRet = (u8*)(*ppOut = sqlite3_malloc64(nPre + nList + 10));
    if( pRet ){
      Fts5HashEntry *pFaux = (Fts5HashEntry*)&pRet[nPre-nHashPre];
      memcpy(&pRet[nPre], &((u8*)p)[nHashPre], nList);
      nList += fts5HashAddPoslistSize(pHash, p, pFaux);
      *pnDoclist = nList;
    }else{
      *pnDoclist = 0;
      return SQLITE_NOMEM;
    }
  }else{
    *ppOut = 0;
    *pnDoclist = 0;
  }

  return SQLITE_OK;
}

int sqlite3Fts5HashScanInit(
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
  const u8 **ppDoclist,           /* OUT: pointer to doclist */
  int *pnDoclist                  /* OUT: size of doclist in bytes */
){
  Fts5HashEntry *p;
  if( (p = pHash->pScan) ){
    char *zKey = fts5EntryKey(p);
    int nTerm = (int)strlen(zKey);
    fts5HashAddPoslistSize(pHash, p);
    *pzTerm = zKey;
    *ppDoclist = (const u8*)&zKey[nTerm+1];
    *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm + 1);
  }else{
    *pzTerm = 0;
    *ppDoclist = 0;
    *pnDoclist = 0;
  }
}







|









542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
  const u8 **ppDoclist,           /* OUT: pointer to doclist */
  int *pnDoclist                  /* OUT: size of doclist in bytes */
){
  Fts5HashEntry *p;
  if( (p = pHash->pScan) ){
    char *zKey = fts5EntryKey(p);
    int nTerm = (int)strlen(zKey);
    fts5HashAddPoslistSize(pHash, p, 0);
    *pzTerm = zKey;
    *ppDoclist = (const u8*)&zKey[nTerm+1];
    *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm + 1);
  }else{
    *pzTerm = 0;
    *ppDoclist = 0;
    *pnDoclist = 0;
  }
}
Changes to ext/fts5/fts5_index.c.
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463

2464
2465
2466
2467
2468


2469
2470
2471






2472
2473
2474





2475
2476

2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
*/
static void fts5SegIterHashInit(
  Fts5Index *p,                   /* FTS5 backend */
  const u8 *pTerm, int nTerm,     /* Term to seek to */
  int flags,                      /* Mask of FTS5INDEX_XXX flags */
  Fts5SegIter *pIter              /* Object to populate */
){
  const u8 *pList = 0;
  int nList = 0;
  const u8 *z = 0;
  int n = 0;


  assert( p->pHash );
  assert( p->rc==SQLITE_OK );

  if( pTerm==0 || (flags & FTS5INDEX_QUERY_SCAN) ){


    p->rc = sqlite3Fts5HashScanInit(p->pHash, (const char*)pTerm, nTerm);
    sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &pList, &nList);
    n = (z ? (int)strlen((const char*)z) : 0);






  }else{
    pIter->flags |= FTS5_SEGITER_ONETERM;
    sqlite3Fts5HashQuery(p->pHash, (const char*)pTerm, nTerm, &pList, &nList);





    z = pTerm;
    n = nTerm;

  }

  if( pList ){
    Fts5Data *pLeaf;
    sqlite3Fts5BufferSet(&p->rc, &pIter->term, n, z);
    pLeaf = fts5IdxMalloc(p, sizeof(Fts5Data));
    if( pLeaf==0 ) return;
    pLeaf->p = (u8*)pList;
    pLeaf->nn = pLeaf->szLeaf = nList;
    pIter->pLeaf = pLeaf;
    pIter->iLeafOffset = fts5GetVarint(pLeaf->p, (u64*)&pIter->iRowid);
    pIter->iEndofDoclist = pLeaf->nn;

    if( flags & FTS5INDEX_QUERY_DESC ){
      pIter->flags |= FTS5_SEGITER_REVERSE;







<



>





>
>



>
>
>
>
>
>

<
|
>
>
>
>
>


>


|
<

<
<
<







2453
2454
2455
2456
2457
2458
2459

2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480

2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492

2493



2494
2495
2496
2497
2498
2499
2500
*/
static void fts5SegIterHashInit(
  Fts5Index *p,                   /* FTS5 backend */
  const u8 *pTerm, int nTerm,     /* Term to seek to */
  int flags,                      /* Mask of FTS5INDEX_XXX flags */
  Fts5SegIter *pIter              /* Object to populate */
){

  int nList = 0;
  const u8 *z = 0;
  int n = 0;
  Fts5Data *pLeaf = 0;

  assert( p->pHash );
  assert( p->rc==SQLITE_OK );

  if( pTerm==0 || (flags & FTS5INDEX_QUERY_SCAN) ){
    const u8 *pList = 0;

    p->rc = sqlite3Fts5HashScanInit(p->pHash, (const char*)pTerm, nTerm);
    sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &pList, &nList);
    n = (z ? (int)strlen((const char*)z) : 0);
    if( pList ){
      pLeaf = fts5IdxMalloc(p, sizeof(Fts5Data));
      if( pLeaf ){
        pLeaf->p = pList;
      }
    }
  }else{

    p->rc = sqlite3Fts5HashQuery(p->pHash, sizeof(Fts5Data), 
        (const char*)pTerm, nTerm, (void**)&pLeaf, &nList
    );
    if( pLeaf ){
      pLeaf->p = (u8*)&pLeaf[1];
    }
    z = pTerm;
    n = nTerm;
    pIter->flags |= FTS5_SEGITER_ONETERM;
  }

  if( pLeaf ){

    sqlite3Fts5BufferSet(&p->rc, &pIter->term, n, z);



    pLeaf->nn = pLeaf->szLeaf = nList;
    pIter->pLeaf = pLeaf;
    pIter->iLeafOffset = fts5GetVarint(pLeaf->p, (u64*)&pIter->iRowid);
    pIter->iEndofDoclist = pLeaf->nn;

    if( flags & FTS5INDEX_QUERY_DESC ){
      pIter->flags |= FTS5_SEGITER_REVERSE;
Changes to ext/fts5/test/fts5aa.test.
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
  INSERT INTO n1 VALUES('a b c d');
}

proc funk {} {
  db eval { UPDATE n1_config SET v=50 WHERE k='version' }
  set fd [db incrblob main n1_data block 10]
  fconfigure $fd -encoding binary -translation binary
  puts -nonewline $fd "\x44\x45"
  close $fd
}
db func funk funk

# This test case corrupts the structure record within the first invocation
# of function funk(). Which used to cause the bm25() function to throw an
# exception. But since bm25() can now used the cached structure record,







|







423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
  INSERT INTO n1 VALUES('a b c d');
}

proc funk {} {
  db eval { UPDATE n1_config SET v=50 WHERE k='version' }
  set fd [db incrblob main n1_data block 10]
  fconfigure $fd -encoding binary -translation binary
#  puts -nonewline $fd "\x44\x45"
  close $fd
}
db func funk funk

# This test case corrupts the structure record within the first invocation
# of function funk(). Which used to cause the bm25() function to throw an
# exception. But since bm25() can now used the cached structure record,
598
599
600
601
602
603
604



















605
606
607
608
609
}
do_execsql_test 23.1 {
  SELECT * FROM t11, t10 WHERE t11.x = t10.x AND t10.rowid IS NULL;
}
do_execsql_test 23.2 {
  SELECT * FROM t11, t10 WHERE t10.rowid IS NULL;
}




















}

expand_all_sql db
finish_test







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





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
}
do_execsql_test 23.1 {
  SELECT * FROM t11, t10 WHERE t11.x = t10.x AND t10.rowid IS NULL;
}
do_execsql_test 23.2 {
  SELECT * FROM t11, t10 WHERE t10.rowid IS NULL;
}

#-------------------------------------------------------------------------
do_execsql_test 24.0 {
  CREATE VIRTUAL TABLE t12 USING fts5(x, detail=%DETAIL%);
  INSERT INTO t12 VALUES('aaaa');
}
do_execsql_test 24.1 {
  BEGIN;
    DELETE FROM t12 WHERE rowid=1;
    SELECT * FROM t12('aaaa');
    INSERT INTO t12 VALUES('aaaa');
  END;
}
do_execsql_test 24.2 {
  INSERT INTO t12(t12) VALUES('integrity-check');
}
do_execsql_test 24.3 {
    SELECT * FROM t12('aaaa');
} {aaaa}

}

expand_all_sql db
finish_test