/ Check-in [2312eb6a]
Login
SQLite training in Houston TX on 2019-11-05 (details)
Part of the 2019 Tcl Conference

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

Overview
Comment:Performance improvements when reading large blobs, especially if SQLITE_DIRECT_OVERFLOW_READ is defined.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1: 2312eb6a9eb31051db4e0baf19e033ba39adc7b1
User & Date: drh 2014-04-04 18:49:19
References
2019-08-12
18:25 New ticket [3cf9bb22] Database corruption caused by REPLACE. artifact: 4c0dd574 user: drh
Context
2014-04-10
02:24
Modify the OR-clause handling so that it can safely deal with OR-clause with 17 or more terms. Fix for ticket [10fb063b1179be53ea0b53bb]. check-in: a67b5312 user: drh tags: trunk
2014-04-04
18:49
Performance improvements when reading large blobs, especially if SQLITE_DIRECT_OVERFLOW_READ is defined. check-in: 2312eb6a user: drh tags: trunk
18:20
Merge changes to the query planner that strive to ensure that any index usage that is a proper subset of some other index usage always has a slightly higher cost. check-in: 683dd379 user: drh tags: trunk
2014-04-01
01:24
Remove an unnecessary conditional, replacing it with an assert(). Improvements to comments. Closed-Leaf check-in: 834bf1c3 user: drh tags: overflow-pgno-cache
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/btree.c.

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
...
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
....
2559
2560
2561
2562
2563
2564
2565

2566
2567
2568
2569
2570
2571
2572
2573
....
3634
3635
3636
3637
3638
3639
3640
3641

3642
3643
3644
3645
3646
3647
3648
....
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
....
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
....
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935


3936
3937
3938
3939
3940
3941
3942
3943
3944

3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
....
3964
3965
3966
3967
3968
3969
3970



3971
3972
3973
3974
3975

3976
3977
3978
3979



3980
3981
3982
3983
3984
3985
3986
....
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010


4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021

4022








4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041

4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052



4053
4054


4055
4056
4057
4058
4059

4060
4061
4062
4063
4064
4065
4066
....
4074
4075
4076
4077
4078
4079
4080

4081
4082
4083
4084
4085
4086
4087

4088
4089
4090
4091
4092
4093
4094
....
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
....
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
....
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
....
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
....
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
....
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
....
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
....
4518
4519
4520
4521
4522
4523
4524
4525





4526
4527
4528
4529
4530
4531
4532
....
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
....
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
....
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
....
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
....
4792
4793
4794
4795
4796
4797
4798

4799
4800
4801
4802
4803
4804
4805
....
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
....
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
....
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
....
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
....
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
....
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
....
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
....
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
....
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
....
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469

8470
8471
8472
8473
8474
8475
8476
*/
#ifdef SQLITE_DEBUG
static int cursorHoldsMutex(BtCursor *p){
  return sqlite3_mutex_held(p->pBt->mutex);
}
#endif


#ifndef SQLITE_OMIT_INCRBLOB
/*
** Invalidate the overflow page-list cache for cursor pCur, if any.

*/
static void invalidateOverflowCache(BtCursor *pCur){
  assert( cursorHoldsMutex(pCur) );
  sqlite3_free(pCur->aOverflow);
  pCur->aOverflow = 0;
}

/*
** Invalidate the overflow page-list cache for all cursors opened
** on the shared btree structure pBt.
*/
static void invalidateAllOverflowCache(BtShared *pBt){
  BtCursor *p;
  assert( sqlite3_mutex_held(pBt->mutex) );
  for(p=pBt->pCursor; p; p=p->pNext){
    invalidateOverflowCache(p);
  }
}


/*
** This function is called before modifying the contents of a table
** to invalidate any incrblob cursors that are open on the
** row or one of the rows being modified.
**
** If argument isClearTable is true, then the entire contents of the
** table is about to be deleted. In this case invalidate all incrblob
................................................................................
  i64 iRow,               /* The rowid that might be changing */
  int isClearTable        /* True if all rows are being deleted */
){
  BtCursor *p;
  BtShared *pBt = pBtree->pBt;
  assert( sqlite3BtreeHoldsMutex(pBtree) );
  for(p=pBt->pCursor; p; p=p->pNext){
    if( p->isIncrblobHandle && (isClearTable || p->info.nKey==iRow) ){
      p->eState = CURSOR_INVALID;
    }
  }
}

#else
  /* Stub functions when INCRBLOB is omitted */
  #define invalidateOverflowCache(x)
  #define invalidateAllOverflowCache(x)
  #define invalidateIncrblobCursors(x,y,z)
#endif /* SQLITE_OMIT_INCRBLOB */

/*
** Set bit pgno of the BtShared.pHasContent bitvec. This is called 
** when a page that previously contained data becomes a free-list leaf 
** page.
................................................................................
** is capable of reading or writing to the databse.  Cursors that
** have been tripped into the CURSOR_FAULT state are not counted.
*/
static int countValidCursors(BtShared *pBt, int wrOnly){
  BtCursor *pCur;
  int r = 0;
  for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){

    if( (wrOnly==0 || pCur->wrFlag) && pCur->eState!=CURSOR_FAULT ) r++; 
  }
  return r;
}
#endif

/*
** If there are no outstanding cursors and we are not in the middle
................................................................................
  /* Now that no other errors can occur, finish filling in the BtCursor
  ** variables and link the cursor into the BtShared list.  */
  pCur->pgnoRoot = (Pgno)iTable;
  pCur->iPage = -1;
  pCur->pKeyInfo = pKeyInfo;
  pCur->pBtree = p;
  pCur->pBt = pBt;
  pCur->wrFlag = (u8)wrFlag;

  pCur->pNext = pBt->pCursor;
  if( pCur->pNext ){
    pCur->pNext->pPrev = pCur;
  }
  pBt->pCursor = pCur;
  pCur->eState = CURSOR_INVALID;
  return SQLITE_OK;
................................................................................
    if( pCur->pNext ){
      pCur->pNext->pPrev = pCur->pPrev;
    }
    for(i=0; i<=pCur->iPage; i++){
      releasePage(pCur->apPage[i]);
    }
    unlockBtreeIfUnused(pBt);
    invalidateOverflowCache(pCur);
    /* sqlite3_free(pCur); */
    sqlite3BtreeLeave(pBtree);
  }
  return SQLITE_OK;
}

/*
................................................................................
#endif
#ifdef _MSC_VER
  /* Use a real function in MSVC to work around bugs in that compiler. */
  static void getCellInfo(BtCursor *pCur){
    if( pCur->info.nSize==0 ){
      int iPage = pCur->iPage;
      btreeParseCell(pCur->apPage[iPage],pCur->aiIdx[iPage],&pCur->info);
      pCur->validNKey = 1;
    }else{
      assertCellInfo(pCur);
    }
  }
#else /* if not _MSC_VER */
  /* Use a macro in all other compilers so that the function is inlined */
#define getCellInfo(pCur)                                                      \
  if( pCur->info.nSize==0 ){                                                   \
    int iPage = pCur->iPage;                                                   \
    btreeParseCell(pCur->apPage[iPage],pCur->aiIdx[iPage],&pCur->info); \
    pCur->validNKey = 1;                                                       \
  }else{                                                                       \
    assertCellInfo(pCur);                                                      \
  }
#endif /* _MSC_VER */

#ifndef NDEBUG  /* The next routine used only within assert() statements */
/*
................................................................................
    memcpy(pBuf, pPayload, nByte);
  }
  return SQLITE_OK;
}

/*
** This function is used to read or overwrite payload information
** for the entry that the pCur cursor is pointing to. If the eOp
** parameter is 0, this is a read operation (data copied into
** buffer pBuf). If it is non-zero, a write (data copied from
** buffer pBuf).


**
** A total of "amt" bytes are read or written beginning at "offset".
** Data is read to or from the buffer pBuf.
**
** The content being read or written might appear on the main page
** or be scattered out on multiple overflow pages.
**
** If the BtCursor.isIncrblobHandle flag is set, and the current
** cursor entry uses one or more overflow pages, this function

** allocates space for and lazily popluates the overflow page-list 
** cache array (BtCursor.aOverflow). Subsequent calls use this
** cache to make seeking to the supplied offset more efficient.
**
** Once an overflow page-list cache has been allocated, it may be
** invalidated if some other cursor writes to the same table, or if
** the cursor is moved to a different row. Additionally, in auto-vacuum
** mode, the following events may invalidate an overflow page-list cache.
**
**   * An incremental vacuum,
................................................................................
){
  unsigned char *aPayload;
  int rc = SQLITE_OK;
  u32 nKey;
  int iIdx = 0;
  MemPage *pPage = pCur->apPage[pCur->iPage]; /* Btree page of current entry */
  BtShared *pBt = pCur->pBt;                  /* Btree this cursor belongs to */




  assert( pPage );
  assert( pCur->eState==CURSOR_VALID );
  assert( pCur->aiIdx[pCur->iPage]<pPage->nCell );
  assert( cursorHoldsMutex(pCur) );


  getCellInfo(pCur);
  aPayload = pCur->info.pCell + pCur->info.nHeader;
  nKey = (pPage->intKey ? 0 : (int)pCur->info.nKey);




  if( NEVER(offset+amt > nKey+pCur->info.nData) 
   || &aPayload[pCur->info.nLocal] > &pPage->aData[pBt->usableSize]
  ){
    /* Trying to read or write past the end of the data is an error */
    return SQLITE_CORRUPT_BKPT;
  }
................................................................................

  /* Check if data must be read/written to/from the btree page itself. */
  if( offset<pCur->info.nLocal ){
    int a = amt;
    if( a+offset>pCur->info.nLocal ){
      a = pCur->info.nLocal - offset;
    }
    rc = copyPayload(&aPayload[offset], pBuf, a, eOp, pPage->pDbPage);
    offset = 0;
    pBuf += a;
    amt -= a;
  }else{
    offset -= pCur->info.nLocal;
  }

  if( rc==SQLITE_OK && amt>0 ){
    const u32 ovflSize = pBt->usableSize - 4;  /* Bytes content per ovfl page */
    Pgno nextPage;

    nextPage = get4byte(&aPayload[pCur->info.nLocal]);

#ifndef SQLITE_OMIT_INCRBLOB
    /* If the isIncrblobHandle flag is set and the BtCursor.aOverflow[]
    ** has not been allocated, allocate it now. The array is sized at


    ** one entry for each overflow page in the overflow chain. The
    ** page number of the first overflow page is stored in aOverflow[0],
    ** etc. A value of 0 in the aOverflow[] array means "not yet known"
    ** (the cache is lazily populated).
    */
    if( pCur->isIncrblobHandle && !pCur->aOverflow ){
      int nOvfl = (pCur->info.nPayload-pCur->info.nLocal+ovflSize-1)/ovflSize;
      pCur->aOverflow = (Pgno *)sqlite3MallocZero(sizeof(Pgno)*nOvfl);
      /* nOvfl is always positive.  If it were zero, fetchPayload would have
      ** been used instead of this routine. */
      if( ALWAYS(nOvfl) && !pCur->aOverflow ){

        rc = SQLITE_NOMEM;








      }
    }

    /* If the overflow page-list cache has been allocated and the
    ** entry for the first required overflow page is valid, skip
    ** directly to it.
    */
    if( pCur->aOverflow && pCur->aOverflow[offset/ovflSize] ){
      iIdx = (offset/ovflSize);
      nextPage = pCur->aOverflow[iIdx];
      offset = (offset%ovflSize);
    }
#endif

    for( ; rc==SQLITE_OK && amt>0 && nextPage; iIdx++){

#ifndef SQLITE_OMIT_INCRBLOB
      /* If required, populate the overflow page-list cache. */
      if( pCur->aOverflow ){

        assert(!pCur->aOverflow[iIdx] || pCur->aOverflow[iIdx]==nextPage);
        pCur->aOverflow[iIdx] = nextPage;
      }
#endif

      if( offset>=ovflSize ){
        /* The only reason to read this page is to obtain the page
        ** number for the next page in the overflow chain. The page
        ** data is not required. So first try to lookup the overflow
        ** page-list cache, if any, then fall back to the getOverflowPage()
        ** function.



        */
#ifndef SQLITE_OMIT_INCRBLOB


        if( pCur->aOverflow && pCur->aOverflow[iIdx+1] ){
          nextPage = pCur->aOverflow[iIdx+1];
        } else 
#endif
          rc = getOverflowPage(pBt, nextPage, 0, &nextPage);

        offset -= ovflSize;
      }else{
        /* Need to read this page properly. It contains some of the
        ** range of data that is being read (eOp==0) or written (eOp!=0).
        */
#ifdef SQLITE_DIRECT_OVERFLOW_READ
        sqlite3_file *fd;
................................................................................
        /* If all the following are true:
        **
        **   1) this is a read operation, and 
        **   2) data is required from the start of this overflow page, and
        **   3) the database is file-backed, and
        **   4) there is no open write-transaction, and
        **   5) the database is not a WAL database,

        **
        ** then data can be read directly from the database file into the
        ** output buffer, bypassing the page-cache altogether. This speeds
        ** up loading large records that span many overflow pages.
        */
        if( eOp==0                                             /* (1) */
         && offset==0                                          /* (2) */

         && pBt->inTransaction==TRANS_READ                     /* (4) */
         && (fd = sqlite3PagerFile(pBt->pPager))->pMethods     /* (3) */
         && pBt->pPage1->aData[19]==0x01                       /* (5) */
        ){
          u8 aSave[4];
          u8 *aWrite = &pBuf[-4];
          memcpy(aSave, aWrite, 4);
................................................................................
          memcpy(aWrite, aSave, 4);
        }else
#endif

        {
          DbPage *pDbPage;
          rc = sqlite3PagerAcquire(pBt->pPager, nextPage, &pDbPage,
              (eOp==0 ? PAGER_GET_READONLY : 0)
          );
          if( rc==SQLITE_OK ){
            aPayload = sqlite3PagerGetData(pDbPage);
            nextPage = get4byte(aPayload);
            rc = copyPayload(&aPayload[offset+4], pBuf, a, eOp, pDbPage);
            sqlite3PagerUnref(pDbPage);
            offset = 0;
          }
        }
        amt -= a;
        pBuf += a;
      }
................................................................................
  assert( pCur->eState==CURSOR_VALID );
  assert( pCur->iPage<BTCURSOR_MAX_DEPTH );
  assert( pCur->iPage>=0 );
  if( pCur->iPage>=(BTCURSOR_MAX_DEPTH-1) ){
    return SQLITE_CORRUPT_BKPT;
  }
  rc = getAndInitPage(pBt, newPgno, &pNewPage,
               pCur->wrFlag==0 ? PAGER_GET_READONLY : 0);
  if( rc ) return rc;
  pCur->apPage[i+1] = pNewPage;
  pCur->aiIdx[i+1] = 0;
  pCur->iPage++;

  pCur->info.nSize = 0;
  pCur->validNKey = 0;
  if( pNewPage->nCell<1 || pNewPage->intKey!=pCur->apPage[i]->intKey ){
    return SQLITE_CORRUPT_BKPT;
  }
  return SQLITE_OK;
}

#if 0
................................................................................
  );
#endif
  testcase( pCur->aiIdx[pCur->iPage-1] > pCur->apPage[pCur->iPage-1]->nCell );

  releasePage(pCur->apPage[pCur->iPage]);
  pCur->iPage--;
  pCur->info.nSize = 0;
  pCur->validNKey = 0;
}

/*
** Move the cursor to point to the root page of its b-tree structure.
**
** If the table has a virtual root page, then the cursor is moved to point
** to the virtual root page instead of the actual root page. A table has a
................................................................................
  if( pCur->iPage>=0 ){
    while( pCur->iPage ) releasePage(pCur->apPage[pCur->iPage--]);
  }else if( pCur->pgnoRoot==0 ){
    pCur->eState = CURSOR_INVALID;
    return SQLITE_OK;
  }else{
    rc = getAndInitPage(pCur->pBtree->pBt, pCur->pgnoRoot, &pCur->apPage[0],
                        pCur->wrFlag==0 ? PAGER_GET_READONLY : 0);
    if( rc!=SQLITE_OK ){
      pCur->eState = CURSOR_INVALID;
      return rc;
    }
    pCur->iPage = 0;
  }
  pRoot = pCur->apPage[0];
................................................................................
  assert( pRoot->intKey==1 || pRoot->intKey==0 );
  if( pRoot->isInit==0 || (pCur->pKeyInfo==0)!=pRoot->intKey ){
    return SQLITE_CORRUPT_BKPT;
  }

  pCur->aiIdx[0] = 0;
  pCur->info.nSize = 0;
  pCur->atLast = 0;
  pCur->validNKey = 0;

  if( pRoot->nCell>0 ){
    pCur->eState = CURSOR_VALID;
  }else if( !pRoot->leaf ){
    Pgno subpage;
    if( pRoot->pgno!=1 ) return SQLITE_CORRUPT_BKPT;
    subpage = get4byte(&pRoot->aData[pRoot->hdrOffset+8]);
................................................................................
    pgno = get4byte(&pPage->aData[pPage->hdrOffset+8]);
    pCur->aiIdx[pCur->iPage] = pPage->nCell;
    rc = moveToChild(pCur, pgno);
  }
  if( rc==SQLITE_OK ){
    pCur->aiIdx[pCur->iPage] = pPage->nCell-1;
    pCur->info.nSize = 0;
    pCur->validNKey = 0;
  }
  return rc;
}

/* Move the cursor to the first entry in the table.  Return SQLITE_OK
** on success.  Set *pRes to 0 if the cursor actually points to something
** or set *pRes to 1 if the table is empty.
................................................................................
int sqlite3BtreeLast(BtCursor *pCur, int *pRes){
  int rc;
 
  assert( cursorHoldsMutex(pCur) );
  assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );

  /* If the cursor already points to the last entry, this is a no-op. */
  if( CURSOR_VALID==pCur->eState && pCur->atLast ){
#ifdef SQLITE_DEBUG
    /* This block serves to assert() that the cursor really does point 
    ** to the last entry in the b-tree. */
    int ii;
    for(ii=0; ii<pCur->iPage; ii++){
      assert( pCur->aiIdx[ii]==pCur->apPage[ii]->nCell );
    }
................................................................................
    if( CURSOR_INVALID==pCur->eState ){
      assert( pCur->pgnoRoot==0 || pCur->apPage[pCur->iPage]->nCell==0 );
      *pRes = 1;
    }else{
      assert( pCur->eState==CURSOR_VALID );
      *pRes = 0;
      rc = moveToRightmost(pCur);
      pCur->atLast = rc==SQLITE_OK ?1:0;





    }
  }
  return rc;
}

/* Move the cursor so that it points to an entry near the key 
** specified by pIdxKey or intKey.   Return a success code.
................................................................................
  assert( cursorHoldsMutex(pCur) );
  assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );
  assert( pRes );
  assert( (pIdxKey==0)==(pCur->pKeyInfo==0) );

  /* If the cursor is already positioned at the point we are trying
  ** to move to, then just return without doing any work */
  if( pCur->eState==CURSOR_VALID && pCur->validNKey 
   && pCur->apPage[0]->intKey 
  ){
    if( pCur->info.nKey==intKey ){
      *pRes = 0;
      return SQLITE_OK;
    }
    if( pCur->atLast && pCur->info.nKey<intKey ){
      *pRes = -1;
      return SQLITE_OK;
    }
  }

  if( pIdxKey ){
    xRecordCompare = sqlite3VdbeFindCompare(pIdxKey);
................................................................................
          lwr = idx+1;
          if( lwr>upr ){ c = -1; break; }
        }else if( nCellKey>intKey ){
          upr = idx-1;
          if( lwr>upr ){ c = +1; break; }
        }else{
          assert( nCellKey==intKey );
          pCur->validNKey = 1;
          pCur->info.nKey = nCellKey;
          pCur->aiIdx[pCur->iPage] = (u16)idx;
          if( !pPage->leaf ){
            lwr = idx;
            goto moveto_next_layer;
          }else{
            *pRes = 0;
................................................................................
          nCell = (int)pCur->info.nKey;
          pCellKey = sqlite3Malloc( nCell );
          if( pCellKey==0 ){
            rc = SQLITE_NOMEM;
            goto moveto_finish;
          }
          pCur->aiIdx[pCur->iPage] = (u16)idx;
          rc = accessPayload(pCur, 0, nCell, (unsigned char*)pCellKey, 0);
          if( rc ){
            sqlite3_free(pCellKey);
            goto moveto_finish;
          }
          c = xRecordCompare(nCell, pCellKey, pIdxKey, 0);
          sqlite3_free(pCellKey);
        }
................................................................................
    }
    pCur->aiIdx[pCur->iPage] = (u16)lwr;
    rc = moveToChild(pCur, chldPg);
    if( rc ) break;
  }
moveto_finish:
  pCur->info.nSize = 0;
  pCur->validNKey = 0;
  return rc;
}


/*
** Return TRUE if the cursor is not pointing at an entry of the table.
**
................................................................................
  MemPage *pPage;

  assert( cursorHoldsMutex(pCur) );
  assert( pRes!=0 );
  assert( *pRes==0 || *pRes==1 );
  assert( pCur->skipNext==0 || pCur->eState!=CURSOR_VALID );
  if( pCur->eState!=CURSOR_VALID ){

    rc = restoreCursorPosition(pCur);
    if( rc!=SQLITE_OK ){
      *pRes = 0;
      return rc;
    }
    if( CURSOR_INVALID==pCur->eState ){
      *pRes = 1;
................................................................................
  ** to be invalid here. This can only occur if a second cursor modifies
  ** the page while cursor pCur is holding a reference to it. Which can
  ** only happen if the database is corrupt in such a way as to link the
  ** page into more than one b-tree structure. */
  testcase( idx>pPage->nCell );

  pCur->info.nSize = 0;
  pCur->validNKey = 0;
  if( idx>=pPage->nCell ){
    if( !pPage->leaf ){
      rc = moveToChild(pCur, get4byte(&pPage->aData[pPage->hdrOffset+8]));
      if( rc ){
        *pRes = 0;
        return rc;
      }
................................................................................
  int rc;
  MemPage *pPage;

  assert( cursorHoldsMutex(pCur) );
  assert( pRes!=0 );
  assert( *pRes==0 || *pRes==1 );
  assert( pCur->skipNext==0 || pCur->eState!=CURSOR_VALID );
  pCur->atLast = 0;
  if( pCur->eState!=CURSOR_VALID ){
    if( ALWAYS(pCur->eState>=CURSOR_REQUIRESEEK) ){
      rc = btreeRestoreCursorPosition(pCur);
      if( rc!=SQLITE_OK ){
        *pRes = 0;
        return rc;
      }
................................................................................
        pCur->eState = CURSOR_INVALID;
        *pRes = 1;
        return SQLITE_OK;
      }
      moveToParent(pCur);
    }
    pCur->info.nSize = 0;
    pCur->validNKey = 0;

    pCur->aiIdx[pCur->iPage]--;
    pPage = pCur->apPage[pCur->iPage];
    if( pPage->intKey && !pPage->leaf ){
      rc = sqlite3BtreePrevious(pCur, pRes);
    }else{
      rc = SQLITE_OK;
................................................................................

  if( pCur->eState==CURSOR_FAULT ){
    assert( pCur->skipNext!=SQLITE_OK );
    return pCur->skipNext;
  }

  assert( cursorHoldsMutex(pCur) );
  assert( pCur->wrFlag && pBt->inTransaction==TRANS_WRITE
              && (pBt->btsFlags & BTS_READ_ONLY)==0 );
  assert( hasSharedCacheTableLock(p, pCur->pgnoRoot, pCur->pKeyInfo!=0, 2) );

  /* Assert that the caller has been consistent. If this cursor was opened
  ** expecting an index b-tree, then the caller should be inserting blob
  ** keys with no associated data. If the cursor was opened expecting an
  ** intkey table, the caller should be inserting integer keys with a
................................................................................
    /* If this is an insert into a table b-tree, invalidate any incrblob 
    ** cursors open on the row being replaced */
    invalidateIncrblobCursors(p, nKey, 0);

    /* If the cursor is currently on the last row and we are appending a
    ** new row onto the end, set the "loc" to avoid an unnecessary btreeMoveto()
    ** call */
    if( pCur->validNKey && nKey>0 && pCur->info.nKey==nKey-1 ){
      loc = -1;
    }
  }

  if( !loc ){
    rc = btreeMoveto(pCur, pKey, nKey, appendBias, &loc);
    if( rc ) return rc;
................................................................................
    assert( pPage->leaf );
  }
  insertCell(pPage, idx, newCell, szNew, 0, 0, &rc);
  assert( rc!=SQLITE_OK || pPage->nCell>0 || pPage->nOverflow>0 );

  /* If no error has occurred and pPage has an overflow cell, call balance() 
  ** to redistribute the cells within the tree. Since balance() may move
  ** the cursor, zero the BtCursor.info.nSize and BtCursor.validNKey
  ** variables.
  **
  ** Previous versions of SQLite called moveToRoot() to move the cursor
  ** back to the root page as balance() used to invalidate the contents
  ** of BtCursor.apPage[] and BtCursor.aiIdx[]. Instead of doing that,
  ** set the cursor state to "invalid". This makes common insert operations
  ** slightly faster.
................................................................................
  ** the b-tree if possible. If the cursor is left pointing to the last
  ** entry in the table, and the next row inserted has an integer key
  ** larger than the largest existing key, it is possible to insert the
  ** row without seeking the cursor. This can be a big performance boost.
  */
  pCur->info.nSize = 0;
  if( rc==SQLITE_OK && pPage->nOverflow ){
    pCur->validNKey = 0;
    rc = balance(pCur);

    /* Must make sure nOverflow is reset to zero even if the balance()
    ** fails. Internal data structure corruption will result otherwise. 
    ** Also, set the cursor state to invalid. This stops saveCursorPosition()
    ** from trying to save the current position of the cursor.  */
    pCur->apPage[pCur->iPage]->nOverflow = 0;
................................................................................
  unsigned char *pCell;                /* Pointer to cell to delete */
  int iCellIdx;                        /* Index of cell to delete */
  int iCellDepth;                      /* Depth of node containing pCell */ 

  assert( cursorHoldsMutex(pCur) );
  assert( pBt->inTransaction==TRANS_WRITE );
  assert( (pBt->btsFlags & BTS_READ_ONLY)==0 );
  assert( pCur->wrFlag );
  assert( hasSharedCacheTableLock(p, pCur->pgnoRoot, pCur->pKeyInfo!=0, 2) );
  assert( !hasReadConflicts(p, pCur->pgnoRoot) );

  if( NEVER(pCur->aiIdx[pCur->iPage]>=pCur->apPage[pCur->iPage]->nCell) 
   || NEVER(pCur->eState!=CURSOR_VALID)
  ){
    return SQLITE_ERROR;  /* Something has gone awry. */
................................................................................
** parameters that attempt to write past the end of the existing data,
** no modifications are made and SQLITE_CORRUPT is returned.
*/
int sqlite3BtreePutData(BtCursor *pCsr, u32 offset, u32 amt, void *z){
  int rc;
  assert( cursorHoldsMutex(pCsr) );
  assert( sqlite3_mutex_held(pCsr->pBtree->db->mutex) );
  assert( pCsr->isIncrblobHandle );

  rc = restoreCursorPosition(pCsr);
  if( rc!=SQLITE_OK ){
    return rc;
  }
  assert( pCsr->eState!=CURSOR_REQUIRESEEK );
  if( pCsr->eState!=CURSOR_VALID ){
................................................................................
  /* Check some assumptions: 
  **   (a) the cursor is open for writing,
  **   (b) there is a read/write transaction open,
  **   (c) the connection holds a write-lock on the table (if required),
  **   (d) there are no conflicting read-locks, and
  **   (e) the cursor points at a valid row of an intKey table.
  */
  if( !pCsr->wrFlag ){
    return SQLITE_READONLY;
  }
  assert( (pCsr->pBt->btsFlags & BTS_READ_ONLY)==0
              && pCsr->pBt->inTransaction==TRANS_WRITE );
  assert( hasSharedCacheTableLock(pCsr->pBtree, pCsr->pgnoRoot, 0, 2) );
  assert( !hasReadConflicts(pCsr->pBtree, pCsr->pgnoRoot) );
  assert( pCsr->apPage[pCsr->iPage]->intKey );

  return accessPayload(pCsr, offset, amt, (unsigned char *)z, 1);
}

/* 
** Set a flag on this cursor to cache the locations of pages from the 
** overflow list for the current row. This is used by cursors opened
** for incremental blob IO only.
**
** This function sets a flag only. The actual page location cache
** (stored in BtCursor.aOverflow[]) is allocated and used by function
** accessPayload() (the worker function for sqlite3BtreeData() and
** sqlite3BtreePutData()).
*/
void sqlite3BtreeCacheOverflow(BtCursor *pCur){
  assert( cursorHoldsMutex(pCur) );
  assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );
  invalidateOverflowCache(pCur);
  pCur->isIncrblobHandle = 1;

}
#endif

/*
** Set both the "read version" (single byte at byte offset 18) and 
** "write version" (single byte at byte offset 19) fields in the database
** header to iVersion.







<
<

|
>

|
<
<
<
<













>







 







|






|
<
<







 







>
|







 







|
>







 







|







 







|









|
|







 







|
|
|
|
>
>







<
|
>
|
|
|







 







>
>
>





>




>
>
>







 







|













<
<
|
>
>
|
|
|
|

|

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







|




<



<

<
>



<







>
>
>

<
>
>
|

|
<

>







 







>





|

>







 







|




|







 







|






|







 







|







 







|







 







|
<







 







|







 







|







 







|
>
>
>
>
>







 







|






|







 







|







 







|







 







|







 







>







 







|







 







|







 







|







 







|







 







|







 







|







 







|







 







|







 







|







 







|












|
<
<
<
<
<
<
<

|
<
<
<
<
>







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
...
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497


498
499
500
501
502
503
504
....
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
....
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
....
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
....
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
....
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940

3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
....
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
....
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012


4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048

4049
4050
4051

4052

4053
4054
4055
4056

4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067

4068
4069
4070
4071
4072

4073
4074
4075
4076
4077
4078
4079
4080
4081
....
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
....
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
....
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
....
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
....
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
....
4409
4410
4411
4412
4413
4414
4415
4416

4417
4418
4419
4420
4421
4422
4423
....
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
....
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
....
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
....
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
....
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
....
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
....
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
....
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
....
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
....
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
....
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
....
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
....
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
....
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
....
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
....
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
....
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
....
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478







8479
8480




8481
8482
8483
8484
8485
8486
8487
8488
*/
#ifdef SQLITE_DEBUG
static int cursorHoldsMutex(BtCursor *p){
  return sqlite3_mutex_held(p->pBt->mutex);
}
#endif



/*
** Invalidate the overflow cache of the cursor passed as the first argument.
** on the shared btree structure pBt.
*/
#define invalidateOverflowCache(pCur) (pCur->curFlags &= ~BTCF_ValidOvfl)





/*
** Invalidate the overflow page-list cache for all cursors opened
** on the shared btree structure pBt.
*/
static void invalidateAllOverflowCache(BtShared *pBt){
  BtCursor *p;
  assert( sqlite3_mutex_held(pBt->mutex) );
  for(p=pBt->pCursor; p; p=p->pNext){
    invalidateOverflowCache(p);
  }
}

#ifndef SQLITE_OMIT_INCRBLOB
/*
** This function is called before modifying the contents of a table
** to invalidate any incrblob cursors that are open on the
** row or one of the rows being modified.
**
** If argument isClearTable is true, then the entire contents of the
** table is about to be deleted. In this case invalidate all incrblob
................................................................................
  i64 iRow,               /* The rowid that might be changing */
  int isClearTable        /* True if all rows are being deleted */
){
  BtCursor *p;
  BtShared *pBt = pBtree->pBt;
  assert( sqlite3BtreeHoldsMutex(pBtree) );
  for(p=pBt->pCursor; p; p=p->pNext){
    if( (p->curFlags & BTCF_Incrblob)!=0 && (isClearTable || p->info.nKey==iRow) ){
      p->eState = CURSOR_INVALID;
    }
  }
}

#else
  /* Stub function when INCRBLOB is omitted */


  #define invalidateIncrblobCursors(x,y,z)
#endif /* SQLITE_OMIT_INCRBLOB */

/*
** Set bit pgno of the BtShared.pHasContent bitvec. This is called 
** when a page that previously contained data becomes a free-list leaf 
** page.
................................................................................
** is capable of reading or writing to the databse.  Cursors that
** have been tripped into the CURSOR_FAULT state are not counted.
*/
static int countValidCursors(BtShared *pBt, int wrOnly){
  BtCursor *pCur;
  int r = 0;
  for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
    if( (wrOnly==0 || (pCur->curFlags & BTCF_WriteFlag)!=0)
     && pCur->eState!=CURSOR_FAULT ) r++; 
  }
  return r;
}
#endif

/*
** If there are no outstanding cursors and we are not in the middle
................................................................................
  /* Now that no other errors can occur, finish filling in the BtCursor
  ** variables and link the cursor into the BtShared list.  */
  pCur->pgnoRoot = (Pgno)iTable;
  pCur->iPage = -1;
  pCur->pKeyInfo = pKeyInfo;
  pCur->pBtree = p;
  pCur->pBt = pBt;
  assert( wrFlag==0 || wrFlag==BTCF_WriteFlag );
  pCur->curFlags = wrFlag;
  pCur->pNext = pBt->pCursor;
  if( pCur->pNext ){
    pCur->pNext->pPrev = pCur;
  }
  pBt->pCursor = pCur;
  pCur->eState = CURSOR_INVALID;
  return SQLITE_OK;
................................................................................
    if( pCur->pNext ){
      pCur->pNext->pPrev = pCur->pPrev;
    }
    for(i=0; i<=pCur->iPage; i++){
      releasePage(pCur->apPage[i]);
    }
    unlockBtreeIfUnused(pBt);
    sqlite3DbFree(pBtree->db, pCur->aOverflow);
    /* sqlite3_free(pCur); */
    sqlite3BtreeLeave(pBtree);
  }
  return SQLITE_OK;
}

/*
................................................................................
#endif
#ifdef _MSC_VER
  /* Use a real function in MSVC to work around bugs in that compiler. */
  static void getCellInfo(BtCursor *pCur){
    if( pCur->info.nSize==0 ){
      int iPage = pCur->iPage;
      btreeParseCell(pCur->apPage[iPage],pCur->aiIdx[iPage],&pCur->info);
      pCur->curFlags |= BTCF_ValidNKey;
    }else{
      assertCellInfo(pCur);
    }
  }
#else /* if not _MSC_VER */
  /* Use a macro in all other compilers so that the function is inlined */
#define getCellInfo(pCur)                                                      \
  if( pCur->info.nSize==0 ){                                                   \
    int iPage = pCur->iPage;                                                   \
    btreeParseCell(pCur->apPage[iPage],pCur->aiIdx[iPage],&pCur->info);        \
    pCur->curFlags |= BTCF_ValidNKey;                                          \
  }else{                                                                       \
    assertCellInfo(pCur);                                                      \
  }
#endif /* _MSC_VER */

#ifndef NDEBUG  /* The next routine used only within assert() statements */
/*
................................................................................
    memcpy(pBuf, pPayload, nByte);
  }
  return SQLITE_OK;
}

/*
** This function is used to read or overwrite payload information
** for the entry that the pCur cursor is pointing to. The eOp
** argument is interpreted as follows:
**
**   0: The operation is a read. Populate the overflow cache.
**   1: The operation is a write. Populate the overflow cache.
**   2: The operation is a read. Do not populate the overflow cache.
**
** A total of "amt" bytes are read or written beginning at "offset".
** Data is read to or from the buffer pBuf.
**
** The content being read or written might appear on the main page
** or be scattered out on multiple overflow pages.
**

** If the current cursor entry uses one or more overflow pages and the
** eOp argument is not 2, this function may allocate space for and lazily 
** popluates the overflow page-list cache array (BtCursor.aOverflow). 
** Subsequent calls use this cache to make seeking to the supplied offset 
** more efficient.
**
** Once an overflow page-list cache has been allocated, it may be
** invalidated if some other cursor writes to the same table, or if
** the cursor is moved to a different row. Additionally, in auto-vacuum
** mode, the following events may invalidate an overflow page-list cache.
**
**   * An incremental vacuum,
................................................................................
){
  unsigned char *aPayload;
  int rc = SQLITE_OK;
  u32 nKey;
  int iIdx = 0;
  MemPage *pPage = pCur->apPage[pCur->iPage]; /* Btree page of current entry */
  BtShared *pBt = pCur->pBt;                  /* Btree this cursor belongs to */
#ifdef SQLITE_DIRECT_OVERFLOW_READ
  int bEnd;                                   /* True if reading to end of data */
#endif

  assert( pPage );
  assert( pCur->eState==CURSOR_VALID );
  assert( pCur->aiIdx[pCur->iPage]<pPage->nCell );
  assert( cursorHoldsMutex(pCur) );
  assert( eOp!=2 || offset==0 );      /* Always start from beginning for eOp==2 */

  getCellInfo(pCur);
  aPayload = pCur->info.pCell + pCur->info.nHeader;
  nKey = (pPage->intKey ? 0 : (int)pCur->info.nKey);
#ifdef SQLITE_DIRECT_OVERFLOW_READ
  bEnd = (offset+amt==nKey+pCur->info.nData);
#endif

  if( NEVER(offset+amt > nKey+pCur->info.nData) 
   || &aPayload[pCur->info.nLocal] > &pPage->aData[pBt->usableSize]
  ){
    /* Trying to read or write past the end of the data is an error */
    return SQLITE_CORRUPT_BKPT;
  }
................................................................................

  /* Check if data must be read/written to/from the btree page itself. */
  if( offset<pCur->info.nLocal ){
    int a = amt;
    if( a+offset>pCur->info.nLocal ){
      a = pCur->info.nLocal - offset;
    }
    rc = copyPayload(&aPayload[offset], pBuf, a, (eOp & 0x01), pPage->pDbPage);
    offset = 0;
    pBuf += a;
    amt -= a;
  }else{
    offset -= pCur->info.nLocal;
  }

  if( rc==SQLITE_OK && amt>0 ){
    const u32 ovflSize = pBt->usableSize - 4;  /* Bytes content per ovfl page */
    Pgno nextPage;

    nextPage = get4byte(&aPayload[pCur->info.nLocal]);



    /* If the BtCursor.aOverflow[] has not been allocated, allocate it now.
    ** Except, do not allocate aOverflow[] for eOp==2.
    **
    ** The aOverflow[] array is sized at one entry for each overflow page
    ** in the overflow chain. The page number of the first overflow page is
    ** stored in aOverflow[0], etc. A value of 0 in the aOverflow[] array
    ** means "not yet known" (the cache is lazily populated).
    */
    if( eOp!=2 && (pCur->curFlags & BTCF_ValidOvfl)==0 ){
      int nOvfl = (pCur->info.nPayload-pCur->info.nLocal+ovflSize-1)/ovflSize;
      if( nOvfl>pCur->nOvflAlloc ){
        Pgno *aNew = (Pgno*)sqlite3DbRealloc(
            pCur->pBtree->db, pCur->aOverflow, nOvfl*2*sizeof(Pgno)
        );
        if( aNew==0 ){
          rc = SQLITE_NOMEM;
        }else{
          pCur->nOvflAlloc = nOvfl*2;
          pCur->aOverflow = aNew;
        }
      }
      if( rc==SQLITE_OK ){
        memset(pCur->aOverflow, 0, nOvfl*sizeof(Pgno));
        pCur->curFlags |= BTCF_ValidOvfl;
      }
    }

    /* If the overflow page-list cache has been allocated and the
    ** entry for the first required overflow page is valid, skip
    ** directly to it.
    */
    if( (pCur->curFlags & BTCF_ValidOvfl)!=0 && pCur->aOverflow[offset/ovflSize] ){
      iIdx = (offset/ovflSize);
      nextPage = pCur->aOverflow[iIdx];
      offset = (offset%ovflSize);
    }


    for( ; rc==SQLITE_OK && amt>0 && nextPage; iIdx++){


      /* If required, populate the overflow page-list cache. */

      if( (pCur->curFlags & BTCF_ValidOvfl)!=0 ){
        assert(!pCur->aOverflow[iIdx] || pCur->aOverflow[iIdx]==nextPage);
        pCur->aOverflow[iIdx] = nextPage;
      }


      if( offset>=ovflSize ){
        /* The only reason to read this page is to obtain the page
        ** number for the next page in the overflow chain. The page
        ** data is not required. So first try to lookup the overflow
        ** page-list cache, if any, then fall back to the getOverflowPage()
        ** function.
        **
        ** Note that the aOverflow[] array must be allocated because eOp!=2
        ** here.  If eOp==2, then offset==0 and this branch is never taken.
        */

        assert( eOp!=2 );
        assert( pCur->curFlags & BTCF_ValidOvfl );
        if( pCur->aOverflow[iIdx+1] ){
          nextPage = pCur->aOverflow[iIdx+1];
        }else{

          rc = getOverflowPage(pBt, nextPage, 0, &nextPage);
        }
        offset -= ovflSize;
      }else{
        /* Need to read this page properly. It contains some of the
        ** range of data that is being read (eOp==0) or written (eOp!=0).
        */
#ifdef SQLITE_DIRECT_OVERFLOW_READ
        sqlite3_file *fd;
................................................................................
        /* If all the following are true:
        **
        **   1) this is a read operation, and 
        **   2) data is required from the start of this overflow page, and
        **   3) the database is file-backed, and
        **   4) there is no open write-transaction, and
        **   5) the database is not a WAL database,
        **   6) all data from the page is being read.
        **
        ** then data can be read directly from the database file into the
        ** output buffer, bypassing the page-cache altogether. This speeds
        ** up loading large records that span many overflow pages.
        */
        if( (eOp&0x01)==0                                      /* (1) */
         && offset==0                                          /* (2) */
         && (bEnd || a==ovflSize)                              /* (6) */
         && pBt->inTransaction==TRANS_READ                     /* (4) */
         && (fd = sqlite3PagerFile(pBt->pPager))->pMethods     /* (3) */
         && pBt->pPage1->aData[19]==0x01                       /* (5) */
        ){
          u8 aSave[4];
          u8 *aWrite = &pBuf[-4];
          memcpy(aSave, aWrite, 4);
................................................................................
          memcpy(aWrite, aSave, 4);
        }else
#endif

        {
          DbPage *pDbPage;
          rc = sqlite3PagerAcquire(pBt->pPager, nextPage, &pDbPage,
              ((eOp&0x01)==0 ? PAGER_GET_READONLY : 0)
          );
          if( rc==SQLITE_OK ){
            aPayload = sqlite3PagerGetData(pDbPage);
            nextPage = get4byte(aPayload);
            rc = copyPayload(&aPayload[offset+4], pBuf, a, (eOp&0x01), pDbPage);
            sqlite3PagerUnref(pDbPage);
            offset = 0;
          }
        }
        amt -= a;
        pBuf += a;
      }
................................................................................
  assert( pCur->eState==CURSOR_VALID );
  assert( pCur->iPage<BTCURSOR_MAX_DEPTH );
  assert( pCur->iPage>=0 );
  if( pCur->iPage>=(BTCURSOR_MAX_DEPTH-1) ){
    return SQLITE_CORRUPT_BKPT;
  }
  rc = getAndInitPage(pBt, newPgno, &pNewPage,
               (pCur->curFlags & BTCF_WriteFlag)==0 ? PAGER_GET_READONLY : 0);
  if( rc ) return rc;
  pCur->apPage[i+1] = pNewPage;
  pCur->aiIdx[i+1] = 0;
  pCur->iPage++;

  pCur->info.nSize = 0;
  pCur->curFlags &= ~(BTCF_ValidNKey|BTCF_ValidOvfl);
  if( pNewPage->nCell<1 || pNewPage->intKey!=pCur->apPage[i]->intKey ){
    return SQLITE_CORRUPT_BKPT;
  }
  return SQLITE_OK;
}

#if 0
................................................................................
  );
#endif
  testcase( pCur->aiIdx[pCur->iPage-1] > pCur->apPage[pCur->iPage-1]->nCell );

  releasePage(pCur->apPage[pCur->iPage]);
  pCur->iPage--;
  pCur->info.nSize = 0;
  pCur->curFlags &= ~(BTCF_ValidNKey|BTCF_ValidOvfl);
}

/*
** Move the cursor to point to the root page of its b-tree structure.
**
** If the table has a virtual root page, then the cursor is moved to point
** to the virtual root page instead of the actual root page. A table has a
................................................................................
  if( pCur->iPage>=0 ){
    while( pCur->iPage ) releasePage(pCur->apPage[pCur->iPage--]);
  }else if( pCur->pgnoRoot==0 ){
    pCur->eState = CURSOR_INVALID;
    return SQLITE_OK;
  }else{
    rc = getAndInitPage(pCur->pBtree->pBt, pCur->pgnoRoot, &pCur->apPage[0],
                 (pCur->curFlags & BTCF_WriteFlag)==0 ? PAGER_GET_READONLY : 0);
    if( rc!=SQLITE_OK ){
      pCur->eState = CURSOR_INVALID;
      return rc;
    }
    pCur->iPage = 0;
  }
  pRoot = pCur->apPage[0];
................................................................................
  assert( pRoot->intKey==1 || pRoot->intKey==0 );
  if( pRoot->isInit==0 || (pCur->pKeyInfo==0)!=pRoot->intKey ){
    return SQLITE_CORRUPT_BKPT;
  }

  pCur->aiIdx[0] = 0;
  pCur->info.nSize = 0;
  pCur->curFlags &= ~(BTCF_AtLast|BTCF_ValidNKey|BTCF_ValidOvfl);


  if( pRoot->nCell>0 ){
    pCur->eState = CURSOR_VALID;
  }else if( !pRoot->leaf ){
    Pgno subpage;
    if( pRoot->pgno!=1 ) return SQLITE_CORRUPT_BKPT;
    subpage = get4byte(&pRoot->aData[pRoot->hdrOffset+8]);
................................................................................
    pgno = get4byte(&pPage->aData[pPage->hdrOffset+8]);
    pCur->aiIdx[pCur->iPage] = pPage->nCell;
    rc = moveToChild(pCur, pgno);
  }
  if( rc==SQLITE_OK ){
    pCur->aiIdx[pCur->iPage] = pPage->nCell-1;
    pCur->info.nSize = 0;
    pCur->curFlags &= ~BTCF_ValidNKey;
  }
  return rc;
}

/* Move the cursor to the first entry in the table.  Return SQLITE_OK
** on success.  Set *pRes to 0 if the cursor actually points to something
** or set *pRes to 1 if the table is empty.
................................................................................
int sqlite3BtreeLast(BtCursor *pCur, int *pRes){
  int rc;
 
  assert( cursorHoldsMutex(pCur) );
  assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );

  /* If the cursor already points to the last entry, this is a no-op. */
  if( CURSOR_VALID==pCur->eState && (pCur->curFlags & BTCF_AtLast)!=0 ){
#ifdef SQLITE_DEBUG
    /* This block serves to assert() that the cursor really does point 
    ** to the last entry in the b-tree. */
    int ii;
    for(ii=0; ii<pCur->iPage; ii++){
      assert( pCur->aiIdx[ii]==pCur->apPage[ii]->nCell );
    }
................................................................................
    if( CURSOR_INVALID==pCur->eState ){
      assert( pCur->pgnoRoot==0 || pCur->apPage[pCur->iPage]->nCell==0 );
      *pRes = 1;
    }else{
      assert( pCur->eState==CURSOR_VALID );
      *pRes = 0;
      rc = moveToRightmost(pCur);
      if( rc==SQLITE_OK ){
        pCur->curFlags |= BTCF_AtLast;
      }else{
        pCur->curFlags &= ~BTCF_AtLast;
      }
   
    }
  }
  return rc;
}

/* Move the cursor so that it points to an entry near the key 
** specified by pIdxKey or intKey.   Return a success code.
................................................................................
  assert( cursorHoldsMutex(pCur) );
  assert( sqlite3_mutex_held(pCur->pBtree->db->mutex) );
  assert( pRes );
  assert( (pIdxKey==0)==(pCur->pKeyInfo==0) );

  /* If the cursor is already positioned at the point we are trying
  ** to move to, then just return without doing any work */
  if( pCur->eState==CURSOR_VALID && (pCur->curFlags & BTCF_ValidNKey)!=0
   && pCur->apPage[0]->intKey 
  ){
    if( pCur->info.nKey==intKey ){
      *pRes = 0;
      return SQLITE_OK;
    }
    if( (pCur->curFlags & BTCF_AtLast)!=0 && pCur->info.nKey<intKey ){
      *pRes = -1;
      return SQLITE_OK;
    }
  }

  if( pIdxKey ){
    xRecordCompare = sqlite3VdbeFindCompare(pIdxKey);
................................................................................
          lwr = idx+1;
          if( lwr>upr ){ c = -1; break; }
        }else if( nCellKey>intKey ){
          upr = idx-1;
          if( lwr>upr ){ c = +1; break; }
        }else{
          assert( nCellKey==intKey );
          pCur->curFlags |= BTCF_ValidNKey;
          pCur->info.nKey = nCellKey;
          pCur->aiIdx[pCur->iPage] = (u16)idx;
          if( !pPage->leaf ){
            lwr = idx;
            goto moveto_next_layer;
          }else{
            *pRes = 0;
................................................................................
          nCell = (int)pCur->info.nKey;
          pCellKey = sqlite3Malloc( nCell );
          if( pCellKey==0 ){
            rc = SQLITE_NOMEM;
            goto moveto_finish;
          }
          pCur->aiIdx[pCur->iPage] = (u16)idx;
          rc = accessPayload(pCur, 0, nCell, (unsigned char*)pCellKey, 2);
          if( rc ){
            sqlite3_free(pCellKey);
            goto moveto_finish;
          }
          c = xRecordCompare(nCell, pCellKey, pIdxKey, 0);
          sqlite3_free(pCellKey);
        }
................................................................................
    }
    pCur->aiIdx[pCur->iPage] = (u16)lwr;
    rc = moveToChild(pCur, chldPg);
    if( rc ) break;
  }
moveto_finish:
  pCur->info.nSize = 0;
  pCur->curFlags &= ~(BTCF_ValidNKey|BTCF_ValidOvfl);
  return rc;
}


/*
** Return TRUE if the cursor is not pointing at an entry of the table.
**
................................................................................
  MemPage *pPage;

  assert( cursorHoldsMutex(pCur) );
  assert( pRes!=0 );
  assert( *pRes==0 || *pRes==1 );
  assert( pCur->skipNext==0 || pCur->eState!=CURSOR_VALID );
  if( pCur->eState!=CURSOR_VALID ){
    invalidateOverflowCache(pCur);
    rc = restoreCursorPosition(pCur);
    if( rc!=SQLITE_OK ){
      *pRes = 0;
      return rc;
    }
    if( CURSOR_INVALID==pCur->eState ){
      *pRes = 1;
................................................................................
  ** to be invalid here. This can only occur if a second cursor modifies
  ** the page while cursor pCur is holding a reference to it. Which can
  ** only happen if the database is corrupt in such a way as to link the
  ** page into more than one b-tree structure. */
  testcase( idx>pPage->nCell );

  pCur->info.nSize = 0;
  pCur->curFlags &= ~(BTCF_ValidNKey|BTCF_ValidOvfl);
  if( idx>=pPage->nCell ){
    if( !pPage->leaf ){
      rc = moveToChild(pCur, get4byte(&pPage->aData[pPage->hdrOffset+8]));
      if( rc ){
        *pRes = 0;
        return rc;
      }
................................................................................
  int rc;
  MemPage *pPage;

  assert( cursorHoldsMutex(pCur) );
  assert( pRes!=0 );
  assert( *pRes==0 || *pRes==1 );
  assert( pCur->skipNext==0 || pCur->eState!=CURSOR_VALID );
  pCur->curFlags &= ~(BTCF_AtLast|BTCF_ValidOvfl);
  if( pCur->eState!=CURSOR_VALID ){
    if( ALWAYS(pCur->eState>=CURSOR_REQUIRESEEK) ){
      rc = btreeRestoreCursorPosition(pCur);
      if( rc!=SQLITE_OK ){
        *pRes = 0;
        return rc;
      }
................................................................................
        pCur->eState = CURSOR_INVALID;
        *pRes = 1;
        return SQLITE_OK;
      }
      moveToParent(pCur);
    }
    pCur->info.nSize = 0;
    pCur->curFlags &= ~(BTCF_ValidNKey|BTCF_ValidOvfl);

    pCur->aiIdx[pCur->iPage]--;
    pPage = pCur->apPage[pCur->iPage];
    if( pPage->intKey && !pPage->leaf ){
      rc = sqlite3BtreePrevious(pCur, pRes);
    }else{
      rc = SQLITE_OK;
................................................................................

  if( pCur->eState==CURSOR_FAULT ){
    assert( pCur->skipNext!=SQLITE_OK );
    return pCur->skipNext;
  }

  assert( cursorHoldsMutex(pCur) );
  assert( (pCur->curFlags & BTCF_WriteFlag)!=0 && pBt->inTransaction==TRANS_WRITE
              && (pBt->btsFlags & BTS_READ_ONLY)==0 );
  assert( hasSharedCacheTableLock(p, pCur->pgnoRoot, pCur->pKeyInfo!=0, 2) );

  /* Assert that the caller has been consistent. If this cursor was opened
  ** expecting an index b-tree, then the caller should be inserting blob
  ** keys with no associated data. If the cursor was opened expecting an
  ** intkey table, the caller should be inserting integer keys with a
................................................................................
    /* If this is an insert into a table b-tree, invalidate any incrblob 
    ** cursors open on the row being replaced */
    invalidateIncrblobCursors(p, nKey, 0);

    /* If the cursor is currently on the last row and we are appending a
    ** new row onto the end, set the "loc" to avoid an unnecessary btreeMoveto()
    ** call */
    if( (pCur->curFlags&BTCF_ValidNKey)!=0 && nKey>0 && pCur->info.nKey==nKey-1 ){
      loc = -1;
    }
  }

  if( !loc ){
    rc = btreeMoveto(pCur, pKey, nKey, appendBias, &loc);
    if( rc ) return rc;
................................................................................
    assert( pPage->leaf );
  }
  insertCell(pPage, idx, newCell, szNew, 0, 0, &rc);
  assert( rc!=SQLITE_OK || pPage->nCell>0 || pPage->nOverflow>0 );

  /* If no error has occurred and pPage has an overflow cell, call balance() 
  ** to redistribute the cells within the tree. Since balance() may move
  ** the cursor, zero the BtCursor.info.nSize and BTCF_ValidNKey
  ** variables.
  **
  ** Previous versions of SQLite called moveToRoot() to move the cursor
  ** back to the root page as balance() used to invalidate the contents
  ** of BtCursor.apPage[] and BtCursor.aiIdx[]. Instead of doing that,
  ** set the cursor state to "invalid". This makes common insert operations
  ** slightly faster.
................................................................................
  ** the b-tree if possible. If the cursor is left pointing to the last
  ** entry in the table, and the next row inserted has an integer key
  ** larger than the largest existing key, it is possible to insert the
  ** row without seeking the cursor. This can be a big performance boost.
  */
  pCur->info.nSize = 0;
  if( rc==SQLITE_OK && pPage->nOverflow ){
    pCur->curFlags &= ~(BTCF_ValidNKey);
    rc = balance(pCur);

    /* Must make sure nOverflow is reset to zero even if the balance()
    ** fails. Internal data structure corruption will result otherwise. 
    ** Also, set the cursor state to invalid. This stops saveCursorPosition()
    ** from trying to save the current position of the cursor.  */
    pCur->apPage[pCur->iPage]->nOverflow = 0;
................................................................................
  unsigned char *pCell;                /* Pointer to cell to delete */
  int iCellIdx;                        /* Index of cell to delete */
  int iCellDepth;                      /* Depth of node containing pCell */ 

  assert( cursorHoldsMutex(pCur) );
  assert( pBt->inTransaction==TRANS_WRITE );
  assert( (pBt->btsFlags & BTS_READ_ONLY)==0 );
  assert( pCur->curFlags & BTCF_WriteFlag );
  assert( hasSharedCacheTableLock(p, pCur->pgnoRoot, pCur->pKeyInfo!=0, 2) );
  assert( !hasReadConflicts(p, pCur->pgnoRoot) );

  if( NEVER(pCur->aiIdx[pCur->iPage]>=pCur->apPage[pCur->iPage]->nCell) 
   || NEVER(pCur->eState!=CURSOR_VALID)
  ){
    return SQLITE_ERROR;  /* Something has gone awry. */
................................................................................
** parameters that attempt to write past the end of the existing data,
** no modifications are made and SQLITE_CORRUPT is returned.
*/
int sqlite3BtreePutData(BtCursor *pCsr, u32 offset, u32 amt, void *z){
  int rc;
  assert( cursorHoldsMutex(pCsr) );
  assert( sqlite3_mutex_held(pCsr->pBtree->db->mutex) );
  assert( pCsr->curFlags & BTCF_Incrblob );

  rc = restoreCursorPosition(pCsr);
  if( rc!=SQLITE_OK ){
    return rc;
  }
  assert( pCsr->eState!=CURSOR_REQUIRESEEK );
  if( pCsr->eState!=CURSOR_VALID ){
................................................................................
  /* Check some assumptions: 
  **   (a) the cursor is open for writing,
  **   (b) there is a read/write transaction open,
  **   (c) the connection holds a write-lock on the table (if required),
  **   (d) there are no conflicting read-locks, and
  **   (e) the cursor points at a valid row of an intKey table.
  */
  if( (pCsr->curFlags & BTCF_WriteFlag)==0 ){
    return SQLITE_READONLY;
  }
  assert( (pCsr->pBt->btsFlags & BTS_READ_ONLY)==0
              && pCsr->pBt->inTransaction==TRANS_WRITE );
  assert( hasSharedCacheTableLock(pCsr->pBtree, pCsr->pgnoRoot, 0, 2) );
  assert( !hasReadConflicts(pCsr->pBtree, pCsr->pgnoRoot) );
  assert( pCsr->apPage[pCsr->iPage]->intKey );

  return accessPayload(pCsr, offset, amt, (unsigned char *)z, 1);
}

/* 
** Mark this cursor as an incremental blob cursor.







*/
void sqlite3BtreeIncrblobCursor(BtCursor *pCur){




  pCur->curFlags |= BTCF_Incrblob;
}
#endif

/*
** Set both the "read version" (single byte at byte offset 18) and 
** "write version" (single byte at byte offset 19) fields in the database
** header to iVersion.

Changes to src/btree.h.

186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
int sqlite3BtreeDataSize(BtCursor*, u32 *pSize);
int sqlite3BtreeData(BtCursor*, u32 offset, u32 amt, void*);

char *sqlite3BtreeIntegrityCheck(Btree*, int *aRoot, int nRoot, int, int*);
struct Pager *sqlite3BtreePager(Btree*);

int sqlite3BtreePutData(BtCursor*, u32 offset, u32 amt, void*);
void sqlite3BtreeCacheOverflow(BtCursor *);
void sqlite3BtreeClearCursor(BtCursor *);
int sqlite3BtreeSetVersion(Btree *pBt, int iVersion);
void sqlite3BtreeCursorHints(BtCursor *, unsigned int mask);

#ifndef NDEBUG
int sqlite3BtreeCursorIsValid(BtCursor*);
#endif







|







186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
int sqlite3BtreeDataSize(BtCursor*, u32 *pSize);
int sqlite3BtreeData(BtCursor*, u32 offset, u32 amt, void*);

char *sqlite3BtreeIntegrityCheck(Btree*, int *aRoot, int nRoot, int, int*);
struct Pager *sqlite3BtreePager(Btree*);

int sqlite3BtreePutData(BtCursor*, u32 offset, u32 amt, void*);
void sqlite3BtreeIncrblobCursor(BtCursor *);
void sqlite3BtreeClearCursor(BtCursor *);
int sqlite3BtreeSetVersion(Btree *pBt, int iVersion);
void sqlite3BtreeCursorHints(BtCursor *, unsigned int mask);

#ifndef NDEBUG
int sqlite3BtreeCursorIsValid(BtCursor*);
#endif

Changes to src/btreeInt.h.

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
** found at self->pBt->mutex. 
*/
struct BtCursor {
  Btree *pBtree;            /* The Btree to which this cursor belongs */
  BtShared *pBt;            /* The BtShared this cursor points to */
  BtCursor *pNext, *pPrev;  /* Forms a linked list of all cursors */
  struct KeyInfo *pKeyInfo; /* Argument passed to comparison function */
#ifndef SQLITE_OMIT_INCRBLOB
  Pgno *aOverflow;          /* Cache of overflow page locations */
#endif
  Pgno pgnoRoot;            /* The root page of this tree */
  CellInfo info;            /* A parse of the cell we are pointing at */
  i64 nKey;        /* Size of pKey, or last integer key */
  void *pKey;      /* Saved key that was cursor's last known position */


  int skipNext;    /* Prev() is noop if negative. Next() is noop if positive */
  u8 wrFlag;                /* True if writable */
  u8 atLast;                /* Cursor pointing to the last entry */
  u8 validNKey;             /* True if info.nKey is valid */

  u8 eState;                /* One of the CURSOR_XXX constants (see below) */
#ifndef SQLITE_OMIT_INCRBLOB
  u8 isIncrblobHandle;      /* True if this cursor is an incr. io handle */
#endif
  u8 hints;                             /* As configured by CursorSetHints() */
  i16 iPage;                            /* Index of current page in apPage */
  u16 aiIdx[BTCURSOR_MAX_DEPTH];        /* Current index in apPage[i] */
  MemPage *apPage[BTCURSOR_MAX_DEPTH];  /* Pages from root to current page */
};










/*
** Potential values for BtCursor.eState.
**
** CURSOR_INVALID:
**   Cursor does not point to a valid entry. This can happen (for example) 
**   because the table is empty or because BtreeCursorFirst() has not been
**   called.







<

<
<

|
|
>
>

<
<
<
>

<
<
<






>
>
>
>
>
>
>
>
>







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
** found at self->pBt->mutex. 
*/
struct BtCursor {
  Btree *pBtree;            /* The Btree to which this cursor belongs */
  BtShared *pBt;            /* The BtShared this cursor points to */
  BtCursor *pNext, *pPrev;  /* Forms a linked list of all cursors */
  struct KeyInfo *pKeyInfo; /* Argument passed to comparison function */

  Pgno *aOverflow;          /* Cache of overflow page locations */


  CellInfo info;            /* A parse of the cell we are pointing at */
  i64 nKey;                 /* Size of pKey, or last integer key */
  void *pKey;               /* Saved key that was cursor last known position */
  Pgno pgnoRoot;            /* The root page of this tree */
  int nOvflAlloc;           /* Allocated size of aOverflow[] array */
  int skipNext;    /* Prev() is noop if negative. Next() is noop if positive */



  u8 curFlags;              /* zero or more BTCF_* flags defined below */
  u8 eState;                /* One of the CURSOR_XXX constants (see below) */



  u8 hints;                             /* As configured by CursorSetHints() */
  i16 iPage;                            /* Index of current page in apPage */
  u16 aiIdx[BTCURSOR_MAX_DEPTH];        /* Current index in apPage[i] */
  MemPage *apPage[BTCURSOR_MAX_DEPTH];  /* Pages from root to current page */
};

/*
** Legal values for BtCursor.curFlags
*/
#define BTCF_WriteFlag    0x01   /* True if a write cursor */
#define BTCF_ValidNKey    0x02   /* True if info.nKey is valid */
#define BTCF_ValidOvfl    0x04   /* True if aOverflow is valid */
#define BTCF_AtLast       0x08   /* Cursor is pointing ot the last entry */
#define BTCF_Incrblob     0x10   /* True if an incremental I/O handle */

/*
** Potential values for BtCursor.eState.
**
** CURSOR_INVALID:
**   Cursor does not point to a valid entry. This can happen (for example) 
**   because the table is empty or because BtreeCursorFirst() has not been
**   called.

Changes to src/test_btree.c.

47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
*/
void sqlite3BtreeCursorList(Btree *p){
#ifdef SQLITE_DEBUG
  BtCursor *pCur;
  BtShared *pBt = p->pBt;
  for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
    MemPage *pPage = pCur->apPage[pCur->iPage];
    char *zMode = pCur->wrFlag ? "rw" : "ro";
    sqlite3DebugPrintf("CURSOR %p rooted at %4d(%s) currently at %d.%d%s\n",
       pCur, pCur->pgnoRoot, zMode,
       pPage ? pPage->pgno : 0, pCur->aiIdx[pCur->iPage],
       (pCur->eState==CURSOR_VALID) ? "" : " eof"
    );
  }
#endif
}







|








47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
*/
void sqlite3BtreeCursorList(Btree *p){
#ifdef SQLITE_DEBUG
  BtCursor *pCur;
  BtShared *pBt = p->pBt;
  for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
    MemPage *pPage = pCur->apPage[pCur->iPage];
    char *zMode = (pCur->curFlags & BTCF_WriteFlag) ? "rw" : "ro";
    sqlite3DebugPrintf("CURSOR %p rooted at %4d(%s) currently at %d.%d%s\n",
       pCur, pCur->pgnoRoot, zMode,
       pPage ? pPage->pgno : 0, pCur->aiIdx[pCur->iPage],
       (pCur->eState==CURSOR_VALID) ? "" : " eof"
    );
  }
#endif
}

Changes to src/vdbeblob.c.

73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
      rc = SQLITE_ERROR;
      sqlite3_finalize(p->pStmt);
      p->pStmt = 0;
    }else{
      p->iOffset = pC->aType[p->iCol + pC->nField];
      p->nByte = sqlite3VdbeSerialTypeLen(type);
      p->pCsr =  pC->pCursor;
      sqlite3BtreeEnterCursor(p->pCsr);
      sqlite3BtreeCacheOverflow(p->pCsr);
      sqlite3BtreeLeaveCursor(p->pCsr);
    }
  }

  if( rc==SQLITE_ROW ){
    rc = SQLITE_OK;
  }else if( p->pStmt ){
    rc = sqlite3_finalize(p->pStmt);







|
<
<







73
74
75
76
77
78
79
80


81
82
83
84
85
86
87
      rc = SQLITE_ERROR;
      sqlite3_finalize(p->pStmt);
      p->pStmt = 0;
    }else{
      p->iOffset = pC->aType[p->iCol + pC->nField];
      p->nByte = sqlite3VdbeSerialTypeLen(type);
      p->pCsr =  pC->pCursor;
      sqlite3BtreeIncrblobCursor(p->pCsr);


    }
  }

  if( rc==SQLITE_ROW ){
    rc = SQLITE_OK;
  }else if( p->pStmt ){
    rc = sqlite3_finalize(p->pStmt);

Changes to test/func.test.

1297
1298
1299
1300
1301
1302
1303

1304
1305
1306
1307
1308

1309
1310
1311
1312
1313
1314
1315
do_test func-29.3 {
  db close
  sqlite3 db test.db
  sqlite3_db_status db CACHE_MISS 1
  db eval {SELECT typeof(+x) FROM t29 ORDER BY id}
} {integer null real blob text}
if {[permutation] != "mmap"} {

  do_test func-29.4 {
    set x [lindex [sqlite3_db_status db CACHE_MISS 1] 1]
    if {$x>100} {set x many}
    set x
  } {many}

}
do_test func-29.5 {
  db close
  sqlite3 db test.db
  sqlite3_db_status db CACHE_MISS 1
  db eval {SELECT sum(length(x)) FROM t29}
} {1000009}







>
|
|
|
|
|
>







1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
do_test func-29.3 {
  db close
  sqlite3 db test.db
  sqlite3_db_status db CACHE_MISS 1
  db eval {SELECT typeof(+x) FROM t29 ORDER BY id}
} {integer null real blob text}
if {[permutation] != "mmap"} {
  ifcapable !direct_read {
    do_test func-29.4 {
      set x [lindex [sqlite3_db_status db CACHE_MISS 1] 1]
      if {$x>100} {set x many}
      set x
    } {many}
  }
}
do_test func-29.5 {
  db close
  sqlite3 db test.db
  sqlite3_db_status db CACHE_MISS 1
  db eval {SELECT sum(length(x)) FROM t29}
} {1000009}