/ Check-in [f4ac1bf2]
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:Use xFetch() to access temporary files in vdbesort.c. Use a single large allocation instead of many small allocations when accumulating records in vdbesort.c. This is an interim commit - it allocates a buffer the size of the page-cache every time data is sorted.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | orderby-planning
Files: files | file ages | folders
SHA1: f4ac1bf28c4ba395ccab8f1c9df72614a61095a7
User & Date: dan 2014-03-27 17:23:41
Context
2014-03-27
19:25
Instead of allocating a single large buffer at the beginning of each sort operation, start with a small buffer and extend it using realloc() as required. check-in: 81987c8c user: dan tags: orderby-planning
17:23
Use xFetch() to access temporary files in vdbesort.c. Use a single large allocation instead of many small allocations when accumulating records in vdbesort.c. This is an interim commit - it allocates a buffer the size of the page-cache every time data is sorted. check-in: f4ac1bf2 user: dan tags: orderby-planning
00:09
Merge the latest trunk changes and the fix for the crash on a corrupt database. check-in: 0b35346c user: drh tags: orderby-planning
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/vdbesort.c.

101
102
103
104
105
106
107


108
109
110
111
112
113
114
...
117
118
119
120
121
122
123

124
125
126
127
128
129
130
...
159
160
161
162
163
164
165

166
167
168
169
170
171
172
...
179
180
181
182
183
184
185







186
187
188
189
190
191
192
...
260
261
262
263
264
265
266



267
268
269
270
271
272
273
274
275
276
277
278

279
280
281
282
283
284
285
...
319
320
321
322
323
324
325

326
327
328
329
330
331
332
333
334
335







336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355


356
357
358
359
360
361
362
363
364
365
366
367
368
369
...
484
485
486
487
488
489
490




491
492
493
494
495
496
497
...
517
518
519
520
521
522
523

524

525
526
527
528
529
530
531

532
533
534
535
536
537
538
539
540
541
542

543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559





560
561
562
563
564
565
566
...
721
722
723
724
725
726
727























728
729
730
731
732
733
734
...
755
756
757
758
759
760
761







762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779


780
781
782
783
784
785
786
787
788
789
790
791

792
793
794
795
796
797










798
799
800
801




802
803

804
805
806
807
808
809
810
811
812
813
814
815
816
817
818

819
820
821
822
823
824
825
826
827
828

829


830
831
832
833
834
835
836
...
930
931
932
933
934
935
936



937
938
939
940
941
942
943
....
1035
1036
1037
1038
1039
1040
1041

1042

1043
1044
1045
1046
1047
1048
1049
  int mnPmaSize;                  /* Minimum PMA size, in bytes */
  int mxPmaSize;                  /* Maximum PMA size, in bytes.  0==no limit */
  VdbeSorterIter *aIter;          /* Array of iterators to merge */
  int *aTree;                     /* Current state of incremental merge */
  sqlite3_file *pTemp1;           /* PMA file 1 */
  SorterRecord *pRecord;          /* Head of in-memory record list */
  UnpackedRecord *pUnpacked;      /* Used to unpack keys */


};

/*
** The following type is an iterator for a PMA. It caches the current key in 
** variables nKey/aKey. If the iterator is at EOF, pFile==0.
*/
struct VdbeSorterIter {
................................................................................
  int nAlloc;                     /* Bytes of space at aAlloc */
  int nKey;                       /* Number of bytes in key */
  sqlite3_file *pFile;            /* File iterator is reading from */
  u8 *aAlloc;                     /* Allocated space */
  u8 *aKey;                       /* Pointer to current key */
  u8 *aBuffer;                    /* Current read buffer */
  int nBuffer;                    /* Size of read buffer in bytes */

};

/*
** An instance of this structure is used to organize the stream of records
** being written to files by the merge-sort code into aligned, page-sized
** blocks.  Doing all I/O in aligned page-sized blocks helps I/O to go
** faster on many operating systems.
................................................................................
/*
** Free all memory belonging to the VdbeSorterIter object passed as the second
** argument. All structure fields are set to zero before returning.
*/
static void vdbeSorterIterZero(sqlite3 *db, VdbeSorterIter *pIter){
  sqlite3DbFree(db, pIter->aAlloc);
  sqlite3DbFree(db, pIter->aBuffer);

  memset(pIter, 0, sizeof(VdbeSorterIter));
}

/*
** Read nByte bytes of data from the stream of data iterated by object p.
** If successful, set *ppOut to point to a buffer containing the data
** and return SQLITE_OK. Otherwise, if an error occurs, return an SQLite
................................................................................
  sqlite3 *db,                    /* Database handle (for malloc) */
  VdbeSorterIter *p,              /* Iterator */
  int nByte,                      /* Bytes of data to read */
  u8 **ppOut                      /* OUT: Pointer to buffer containing data */
){
  int iBuf;                       /* Offset within buffer to read from */
  int nAvail;                     /* Bytes of data available in buffer */







  assert( p->aBuffer );

  /* If there is no more data to be read from the buffer, read the next 
  ** p->nBuffer bytes of data from the file into it. Or, if there are less
  ** than p->nBuffer bytes remaining in the PMA, read all remaining data.  */
  iBuf = p->iReadOff % p->nBuffer;
  if( iBuf==0 ){
................................................................................
/*
** Read a varint from the stream of data accessed by p. Set *pnOut to
** the value read.
*/
static int vdbeSorterIterVarint(sqlite3 *db, VdbeSorterIter *p, u64 *pnOut){
  int iBuf;




  iBuf = p->iReadOff % p->nBuffer;
  if( iBuf && (p->nBuffer-iBuf)>=9 ){
    p->iReadOff += sqlite3GetVarint(&p->aBuffer[iBuf], pnOut);
  }else{
    u8 aVarint[16], *a;
    int i = 0, rc;
    do{
      rc = vdbeSorterIterRead(db, p, 1, &a);
      if( rc ) return rc;
      aVarint[(i++)&0xf] = a[0];
    }while( (a[0]&0x80)!=0 );
    sqlite3GetVarint(aVarint, pnOut);

  }

  return SQLITE_OK;
}


/*
................................................................................
  const VdbeSorter *pSorter,      /* Sorter object */
  i64 iStart,                     /* Start offset in pFile */
  VdbeSorterIter *pIter,          /* Iterator to populate */
  i64 *pnByte                     /* IN/OUT: Increment this value by PMA size */
){
  int rc = SQLITE_OK;
  int nBuf;


  nBuf = sqlite3BtreeGetPageSize(db->aDb[0].pBt);

  assert( pSorter->iWriteOff>iStart );
  assert( pIter->aAlloc==0 );
  assert( pIter->aBuffer==0 );
  pIter->pFile = pSorter->pTemp1;
  pIter->iReadOff = iStart;
  pIter->nAlloc = 128;
  pIter->aAlloc = (u8 *)sqlite3DbMallocRaw(db, pIter->nAlloc);







  pIter->nBuffer = nBuf;
  pIter->aBuffer = (u8 *)sqlite3DbMallocRaw(db, nBuf);

  if( !pIter->aBuffer ){
    rc = SQLITE_NOMEM;
  }else{
    int iBuf;

    iBuf = iStart % nBuf;
    if( iBuf ){
      int nRead = nBuf - iBuf;
      if( (iStart + nRead) > pSorter->iWriteOff ){
        nRead = (int)(pSorter->iWriteOff - iStart);
      }
      rc = sqlite3OsRead(
          pSorter->pTemp1, &pIter->aBuffer[iBuf], nRead, iStart
      );
      assert( rc!=SQLITE_IOERR_SHORT_READ );
    }



    if( rc==SQLITE_OK ){
      u64 nByte;                       /* Size of PMA in bytes */
      pIter->iEof = pSorter->iWriteOff;
      rc = vdbeSorterIterVarint(db, pIter, &nByte);
      pIter->iEof = pIter->iReadOff + nByte;
      *pnByte += nByte;
    }
  }

  if( rc==SQLITE_OK ){
    rc = vdbeSorterIterNext(db, pIter);
  }
  return rc;
}
................................................................................

  if( !sqlite3TempInMemory(db) ){
    pgsz = sqlite3BtreeGetPageSize(db->aDb[0].pBt);
    pSorter->mnPmaSize = SORTER_MIN_WORKING * pgsz;
    mxCache = db->aDb[0].pSchema->cache_size;
    if( mxCache<SORTER_MIN_WORKING ) mxCache = SORTER_MIN_WORKING;
    pSorter->mxPmaSize = mxCache * pgsz;




  }

  return SQLITE_OK;
}

/*
** Free the list of sorted records starting at pRecord.
................................................................................
    sqlite3DbFree(db, pSorter->aIter);
    pSorter->aIter = 0;
  }
  if( pSorter->pTemp1 ){
    sqlite3OsCloseFree(pSorter->pTemp1);
    pSorter->pTemp1 = 0;
  }

  vdbeSorterRecordFree(db, pSorter->pRecord);

  pSorter->pRecord = 0;
  pSorter->iWriteOff = 0;
  pSorter->iReadOff = 0;
  pSorter->nInMemory = 0;
  pSorter->nTree = 0;
  pSorter->nPMA = 0;
  pSorter->aTree = 0;

}


/*
** Free any cursor components allocated by sqlite3VdbeSorterXXX routines.
*/
void sqlite3VdbeSorterClose(sqlite3 *db, VdbeCursor *pCsr){
  VdbeSorter *pSorter = pCsr->pSorter;
  if( pSorter ){
    sqlite3VdbeSorterReset(db, pSorter);
    sqlite3DbFree(db, pSorter->pUnpacked);

    sqlite3DbFree(db, pSorter);
    pCsr->pSorter = 0;
  }
}

/*
** Allocate space for a file-handle and open a temporary file. If successful,
** set *ppFile to point to the malloc'd file-handle and return SQLITE_OK.
** Otherwise, set *ppFile to 0 and return an SQLite error code.
*/
static int vdbeSorterOpenTempFile(sqlite3 *db, sqlite3_file **ppFile){
  int dummy;
  return sqlite3OsOpenMalloc(db->pVfs, 0, ppFile,
      SQLITE_OPEN_TEMP_JOURNAL |
      SQLITE_OPEN_READWRITE    | SQLITE_OPEN_CREATE |
      SQLITE_OPEN_EXCLUSIVE    | SQLITE_OPEN_DELETEONCLOSE, &dummy
  );





}

/*
** Merge the two sorted lists p1 and p2 into a single list.
** Set *ppOut to the head of the new list.
**
** In cases where key values are equal, keys from list p1 are considered
................................................................................
static void fileWriterWriteVarint(FileWriter *p, u64 iVal){
  int nByte; 
  u8 aByte[10];
  nByte = sqlite3PutVarint(aByte, iVal);
  fileWriterWrite(p, aByte, nByte);
}
























/*
** Write the current contents of the in-memory linked-list to a PMA. Return
** SQLITE_OK if successful, or an SQLite error code otherwise.
**
** The format of a PMA is:
**
**     * A varint. This varint contains the total number of bytes of content
................................................................................
  /* If the first temporary PMA file has not been opened, open it now. */
  if( rc==SQLITE_OK && pSorter->pTemp1==0 ){
    rc = vdbeSorterOpenTempFile(db, &pSorter->pTemp1);
    assert( rc!=SQLITE_OK || pSorter->pTemp1 );
    assert( pSorter->iWriteOff==0 );
    assert( pSorter->nPMA==0 );
  }








  if( rc==SQLITE_OK ){
    SorterRecord *p;
    SorterRecord *pNext = 0;

    fileWriterInit(db, pSorter->pTemp1, &writer, pSorter->iWriteOff);
    pSorter->nPMA++;
    fileWriterWriteVarint(&writer, pSorter->nInMemory);
    for(p=pSorter->pRecord; p; p=pNext){
      pNext = p->pNext;
      fileWriterWriteVarint(&writer, p->nVal);
      fileWriterWrite(&writer, p->pVal, p->nVal);
      sqlite3DbFree(db, p);
    }
    pSorter->pRecord = p;
    rc = fileWriterFinish(db, &writer, &pSorter->iWriteOff);
  }



  return rc;
}

/*
** Add a record to the sorter.
*/
int sqlite3VdbeSorterWrite(
  sqlite3 *db,                    /* Database handle */
  const VdbeCursor *pCsr,               /* Sorter cursor */
  Mem *pVal                       /* Memory cell containing record */
){
  VdbeSorter *pSorter = pCsr->pSorter;

  int rc = SQLITE_OK;             /* Return Code */
  SorterRecord *pNew;             /* New list element */

  assert( pSorter );
  pSorter->nInMemory += sqlite3VarintLen(pVal->n) + pVal->n;











  pNew = (SorterRecord *)sqlite3DbMallocRaw(db, pVal->n + sizeof(SorterRecord));
  if( pNew==0 ){
    rc = SQLITE_NOMEM;
  }else{




    pNew->pVal = (void *)&pNew[1];
    memcpy(pNew->pVal, pVal->z, pVal->n);

    pNew->nVal = pVal->n;
    pNew->pNext = pSorter->pRecord;
    pSorter->pRecord = pNew;
  }

  /* See if the contents of the sorter should now be written out. They
  ** are written out when either of the following are true:
  **
  **   * The total memory allocated for the in-memory list is greater 
  **     than (page-size * cache-size), or
  **
  **   * The total memory allocated for the in-memory list is greater 
  **     than (page-size * 10) and sqlite3HeapNearlyFull() returns true.
  */
  if( rc==SQLITE_OK && pSorter->mxPmaSize>0 && (

        (pSorter->nInMemory>pSorter->mxPmaSize)
     || (pSorter->nInMemory>pSorter->mnPmaSize && sqlite3HeapNearlyFull())
  )){
#ifdef SQLITE_DEBUG
    i64 nExpect = pSorter->iWriteOff
                + sqlite3VarintLen(pSorter->nInMemory)
                + pSorter->nInMemory;
#endif
    rc = vdbeSorterListToPMA(db, pCsr);
    pSorter->nInMemory = 0;

    assert( rc!=SQLITE_OK || (nExpect==pSorter->iWriteOff) );


  }

  return rc;
}

/*
** Helper function for sqlite3VdbeSorterRewind(). 
................................................................................
        break;
      }

      /* Open the second temp file, if it is not already open. */
      if( pTemp2==0 ){
        assert( iWrite2==0 );
        rc = vdbeSorterOpenTempFile(db, &pTemp2);



      }

      if( rc==SQLITE_OK ){
        int bEof = 0;
        fileWriterInit(db, pTemp2, &writer, iWrite2);
        fileWriterWriteVarint(&writer, nWrite);
        while( rc==SQLITE_OK && bEof==0 ){
................................................................................
      }
      *pbEof = (pSorter->aIter[pSorter->aTree[1]].pFile==0);
    }
  }else{
    SorterRecord *pFree = pSorter->pRecord;
    pSorter->pRecord = pFree->pNext;
    pFree->pNext = 0;

    vdbeSorterRecordFree(db, pFree);

    *pbEof = !pSorter->pRecord;
    rc = SQLITE_OK;
  }
  return rc;
}

/*







>
>







 







>







 







>







 







>
>
>
>
>
>
>







 







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







 







>










>
>
>
>
>
>
>
|
|

|
|
|
|

|
|
|
|
|
|
|
|
|
|
|
|
>
>
|
|
|
|
|
|
<







 







>
>
>
>







 







>
|
>







>











>











|
|


|

>
>
>
>
>







 







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







 







>
>
>
>
>
>
>












|





>
>












>






>
>
>
>
>
>
>
>
>
>
|
|
|
<
>
>
>
>
|

>
|
|
|
<










|
>
|
|
|

|
|
|

|
|
>
|
>
>







 







>
>
>







 







>
|
>







101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
...
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
...
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
...
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
...
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
...
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386

387
388
389
390
391
392
393
...
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
...
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
...
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
...
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880

881
882
883
884
885
886
887
888
889
890

891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
....
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
....
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
  int mnPmaSize;                  /* Minimum PMA size, in bytes */
  int mxPmaSize;                  /* Maximum PMA size, in bytes.  0==no limit */
  VdbeSorterIter *aIter;          /* Array of iterators to merge */
  int *aTree;                     /* Current state of incremental merge */
  sqlite3_file *pTemp1;           /* PMA file 1 */
  SorterRecord *pRecord;          /* Head of in-memory record list */
  UnpackedRecord *pUnpacked;      /* Used to unpack keys */
  u8* aMemory;                    /* Block to allocate records from */
  int iMemory;                    /* Offset of free space in aMemory */
};

/*
** The following type is an iterator for a PMA. It caches the current key in 
** variables nKey/aKey. If the iterator is at EOF, pFile==0.
*/
struct VdbeSorterIter {
................................................................................
  int nAlloc;                     /* Bytes of space at aAlloc */
  int nKey;                       /* Number of bytes in key */
  sqlite3_file *pFile;            /* File iterator is reading from */
  u8 *aAlloc;                     /* Allocated space */
  u8 *aKey;                       /* Pointer to current key */
  u8 *aBuffer;                    /* Current read buffer */
  int nBuffer;                    /* Size of read buffer in bytes */
  u8 *aMap;                       /* Pointer to mapping of pFile */
};

/*
** An instance of this structure is used to organize the stream of records
** being written to files by the merge-sort code into aligned, page-sized
** blocks.  Doing all I/O in aligned page-sized blocks helps I/O to go
** faster on many operating systems.
................................................................................
/*
** Free all memory belonging to the VdbeSorterIter object passed as the second
** argument. All structure fields are set to zero before returning.
*/
static void vdbeSorterIterZero(sqlite3 *db, VdbeSorterIter *pIter){
  sqlite3DbFree(db, pIter->aAlloc);
  sqlite3DbFree(db, pIter->aBuffer);
  if( pIter->aMap ) sqlite3OsUnfetch(pIter->pFile, 0, pIter->aMap);
  memset(pIter, 0, sizeof(VdbeSorterIter));
}

/*
** Read nByte bytes of data from the stream of data iterated by object p.
** If successful, set *ppOut to point to a buffer containing the data
** and return SQLITE_OK. Otherwise, if an error occurs, return an SQLite
................................................................................
  sqlite3 *db,                    /* Database handle (for malloc) */
  VdbeSorterIter *p,              /* Iterator */
  int nByte,                      /* Bytes of data to read */
  u8 **ppOut                      /* OUT: Pointer to buffer containing data */
){
  int iBuf;                       /* Offset within buffer to read from */
  int nAvail;                     /* Bytes of data available in buffer */

  if( p->aMap ){
    *ppOut = &p->aMap[p->iReadOff];
    p->iReadOff += nByte;
    return SQLITE_OK;
  }

  assert( p->aBuffer );

  /* If there is no more data to be read from the buffer, read the next 
  ** p->nBuffer bytes of data from the file into it. Or, if there are less
  ** than p->nBuffer bytes remaining in the PMA, read all remaining data.  */
  iBuf = p->iReadOff % p->nBuffer;
  if( iBuf==0 ){
................................................................................
/*
** Read a varint from the stream of data accessed by p. Set *pnOut to
** the value read.
*/
static int vdbeSorterIterVarint(sqlite3 *db, VdbeSorterIter *p, u64 *pnOut){
  int iBuf;

  if( p->aMap ){
    p->iReadOff += sqlite3GetVarint(&p->aMap[p->iReadOff], pnOut);
  }else{
    iBuf = p->iReadOff % p->nBuffer;
    if( iBuf && (p->nBuffer-iBuf)>=9 ){
      p->iReadOff += sqlite3GetVarint(&p->aBuffer[iBuf], pnOut);
    }else{
      u8 aVarint[16], *a;
      int i = 0, rc;
      do{
        rc = vdbeSorterIterRead(db, p, 1, &a);
        if( rc ) return rc;
        aVarint[(i++)&0xf] = a[0];
      }while( (a[0]&0x80)!=0 );
      sqlite3GetVarint(aVarint, pnOut);
    }
  }

  return SQLITE_OK;
}


/*
................................................................................
  const VdbeSorter *pSorter,      /* Sorter object */
  i64 iStart,                     /* Start offset in pFile */
  VdbeSorterIter *pIter,          /* Iterator to populate */
  i64 *pnByte                     /* IN/OUT: Increment this value by PMA size */
){
  int rc = SQLITE_OK;
  int nBuf;
  void *pMap;

  nBuf = sqlite3BtreeGetPageSize(db->aDb[0].pBt);

  assert( pSorter->iWriteOff>iStart );
  assert( pIter->aAlloc==0 );
  assert( pIter->aBuffer==0 );
  pIter->pFile = pSorter->pTemp1;
  pIter->iReadOff = iStart;
  pIter->nAlloc = 128;
  pIter->aAlloc = (u8 *)sqlite3DbMallocRaw(db, pIter->nAlloc);

  /* See if this PMA can be read using xFetch. */
  rc = sqlite3OsFetch(pIter->pFile, 0, pSorter->iWriteOff, &pMap);
  if( rc!=SQLITE_OK ) return rc;
  if( pMap ){
    pIter->aMap = (u8*)pMap;
  }else{
    pIter->nBuffer = nBuf;
    pIter->aBuffer = (u8 *)sqlite3DbMallocRaw(db, nBuf);

    if( !pIter->aBuffer ){
      rc = SQLITE_NOMEM;
    }else{
      int iBuf;

      iBuf = iStart % nBuf;
      if( iBuf ){
        int nRead = nBuf - iBuf;
        if( (iStart + nRead) > pSorter->iWriteOff ){
          nRead = (int)(pSorter->iWriteOff - iStart);
        }
        rc = sqlite3OsRead(
            pSorter->pTemp1, &pIter->aBuffer[iBuf], nRead, iStart
        );
        assert( rc!=SQLITE_IOERR_SHORT_READ );
      }
    }
  }

  if( rc==SQLITE_OK ){
    u64 nByte;                       /* Size of PMA in bytes */
    pIter->iEof = pSorter->iWriteOff;
    rc = vdbeSorterIterVarint(db, pIter, &nByte);
    pIter->iEof = pIter->iReadOff + nByte;
    *pnByte += nByte;

  }

  if( rc==SQLITE_OK ){
    rc = vdbeSorterIterNext(db, pIter);
  }
  return rc;
}
................................................................................

  if( !sqlite3TempInMemory(db) ){
    pgsz = sqlite3BtreeGetPageSize(db->aDb[0].pBt);
    pSorter->mnPmaSize = SORTER_MIN_WORKING * pgsz;
    mxCache = db->aDb[0].pSchema->cache_size;
    if( mxCache<SORTER_MIN_WORKING ) mxCache = SORTER_MIN_WORKING;
    pSorter->mxPmaSize = mxCache * pgsz;

    pSorter->aMemory = (u8*)sqlite3DbMallocRaw(db, pSorter->mxPmaSize);
    assert( pSorter->iMemory==0 );
    if( !pSorter->aMemory ) return SQLITE_NOMEM;
  }

  return SQLITE_OK;
}

/*
** Free the list of sorted records starting at pRecord.
................................................................................
    sqlite3DbFree(db, pSorter->aIter);
    pSorter->aIter = 0;
  }
  if( pSorter->pTemp1 ){
    sqlite3OsCloseFree(pSorter->pTemp1);
    pSorter->pTemp1 = 0;
  }
  if( pSorter->aMemory==0 ){
    vdbeSorterRecordFree(db, pSorter->pRecord);
  }
  pSorter->pRecord = 0;
  pSorter->iWriteOff = 0;
  pSorter->iReadOff = 0;
  pSorter->nInMemory = 0;
  pSorter->nTree = 0;
  pSorter->nPMA = 0;
  pSorter->aTree = 0;
  pSorter->iMemory = 0;
}


/*
** Free any cursor components allocated by sqlite3VdbeSorterXXX routines.
*/
void sqlite3VdbeSorterClose(sqlite3 *db, VdbeCursor *pCsr){
  VdbeSorter *pSorter = pCsr->pSorter;
  if( pSorter ){
    sqlite3VdbeSorterReset(db, pSorter);
    sqlite3DbFree(db, pSorter->pUnpacked);
    sqlite3DbFree(db, pSorter->aMemory);
    sqlite3DbFree(db, pSorter);
    pCsr->pSorter = 0;
  }
}

/*
** Allocate space for a file-handle and open a temporary file. If successful,
** set *ppFile to point to the malloc'd file-handle and return SQLITE_OK.
** Otherwise, set *ppFile to 0 and return an SQLite error code.
*/
static int vdbeSorterOpenTempFile(sqlite3 *db, sqlite3_file **ppFile){
  int rc;
  rc = sqlite3OsOpenMalloc(db->pVfs, 0, ppFile,
      SQLITE_OPEN_TEMP_JOURNAL |
      SQLITE_OPEN_READWRITE    | SQLITE_OPEN_CREATE |
      SQLITE_OPEN_EXCLUSIVE    | SQLITE_OPEN_DELETEONCLOSE, &rc
  );
  if( rc==SQLITE_OK ){
    i64 max = SQLITE_MAX_MMAP_SIZE;
    sqlite3OsFileControlHint( *ppFile, SQLITE_FCNTL_MMAP_SIZE, (void*)&max);
  }
  return rc;
}

/*
** Merge the two sorted lists p1 and p2 into a single list.
** Set *ppOut to the head of the new list.
**
** In cases where key values are equal, keys from list p1 are considered
................................................................................
static void fileWriterWriteVarint(FileWriter *p, u64 iVal){
  int nByte; 
  u8 aByte[10];
  nByte = sqlite3PutVarint(aByte, iVal);
  fileWriterWrite(p, aByte, nByte);
}

#if SQLITE_MAX_MMAP_SIZE>0
/*
** The first argument is a file-handle open on a temporary file. The file
** is guaranteed to be nByte bytes or smaller in size. This function 
** attempts to extend the file to nByte bytes in size and to ensure that
** the VFS has memory mapped it.
**
** Whether or not the file does end up memory mapped of course depends on 
** the specific VFS implementation.
*/
static int vdbeSorterExtendFile(sqlite3_file *pFile, i64 nByte){
  int rc = sqlite3OsTruncate(pFile, nByte);
  if( rc==SQLITE_OK ){
    void *p = 0;
    sqlite3OsFetch(pFile, 0, nByte, &p);
    sqlite3OsUnfetch(pFile, 0, p);
  }
  return rc;
}
#else
# define vdbeSorterExtendFile(x,y) SQLITE_OK
#endif

/*
** Write the current contents of the in-memory linked-list to a PMA. Return
** SQLITE_OK if successful, or an SQLite error code otherwise.
**
** The format of a PMA is:
**
**     * A varint. This varint contains the total number of bytes of content
................................................................................
  /* If the first temporary PMA file has not been opened, open it now. */
  if( rc==SQLITE_OK && pSorter->pTemp1==0 ){
    rc = vdbeSorterOpenTempFile(db, &pSorter->pTemp1);
    assert( rc!=SQLITE_OK || pSorter->pTemp1 );
    assert( pSorter->iWriteOff==0 );
    assert( pSorter->nPMA==0 );
  }

  /* Try to get the file to memory map */
  if( rc==SQLITE_OK ){
    rc = vdbeSorterExtendFile(
        pSorter->pTemp1, pSorter->iWriteOff + pSorter->nInMemory + 9
    );
  }

  if( rc==SQLITE_OK ){
    SorterRecord *p;
    SorterRecord *pNext = 0;

    fileWriterInit(db, pSorter->pTemp1, &writer, pSorter->iWriteOff);
    pSorter->nPMA++;
    fileWriterWriteVarint(&writer, pSorter->nInMemory);
    for(p=pSorter->pRecord; p; p=pNext){
      pNext = p->pNext;
      fileWriterWriteVarint(&writer, p->nVal);
      fileWriterWrite(&writer, p->pVal, p->nVal);
      if( pSorter->aMemory==0 ) sqlite3DbFree(db, p);
    }
    pSorter->pRecord = p;
    rc = fileWriterFinish(db, &writer, &pSorter->iWriteOff);
  }

  if( pSorter->aMemory ) pSorter->pRecord = 0;
  assert( pSorter->pRecord==0 || rc!=SQLITE_OK );
  return rc;
}

/*
** Add a record to the sorter.
*/
int sqlite3VdbeSorterWrite(
  sqlite3 *db,                    /* Database handle */
  const VdbeCursor *pCsr,               /* Sorter cursor */
  Mem *pVal                       /* Memory cell containing record */
){
  VdbeSorter *pSorter = pCsr->pSorter;
  SorterRecord sRecord;           /* Used for aMemory overflow record */
  int rc = SQLITE_OK;             /* Return Code */
  SorterRecord *pNew;             /* New list element */

  assert( pSorter );
  pSorter->nInMemory += sqlite3VarintLen(pVal->n) + pVal->n;

  if( pSorter->aMemory ){
    int nReq = sizeof(SorterRecord) + pVal->n;
    if( (pSorter->iMemory+nReq) > pSorter->mxPmaSize ){
      pNew = &sRecord;
      pNew->pVal = pVal->z;
    }else{
      pNew = &pSorter->aMemory[pSorter->iMemory];
      pSorter->iMemory += ROUND8(nReq);
    }
  }else{
    pNew = (SorterRecord *)sqlite3DbMallocRaw(db, pVal->n+sizeof(SorterRecord));
    if( pNew==0 ){
      return SQLITE_NOMEM;

    }
  }

  if( pNew!=&sRecord ){
    pNew->pVal = (void*)&pNew[1];
    memcpy(pNew->pVal, pVal->z, pVal->n);
  }
  pNew->nVal = pVal->n;
  pNew->pNext = pSorter->pRecord;
  pSorter->pRecord = pNew;


  /* See if the contents of the sorter should now be written out. They
  ** are written out when either of the following are true:
  **
  **   * The total memory allocated for the in-memory list is greater 
  **     than (page-size * cache-size), or
  **
  **   * The total memory allocated for the in-memory list is greater 
  **     than (page-size * 10) and sqlite3HeapNearlyFull() returns true.
  */
  if( pSorter->mxPmaSize>0 ){
    if( (pNew==&sRecord) || (pSorter->aMemory==0 && (
        (pSorter->nInMemory > pSorter->mxPmaSize)
     || (pSorter->nInMemory > pSorter->mnPmaSize && sqlite3HeapNearlyFull())
    ))){
#ifdef SQLITE_DEBUG
      i64 nExpect = pSorter->iWriteOff
        + sqlite3VarintLen(pSorter->nInMemory)
        + pSorter->nInMemory;
#endif
      rc = vdbeSorterListToPMA(db, pCsr);
      pSorter->nInMemory = 0;
      pSorter->iMemory = 0;
      assert( rc!=SQLITE_OK || (nExpect==pSorter->iWriteOff) );
      assert( rc!=SQLITE_OK || pSorter->pRecord==0 );
    }
  }

  return rc;
}

/*
** Helper function for sqlite3VdbeSorterRewind(). 
................................................................................
        break;
      }

      /* Open the second temp file, if it is not already open. */
      if( pTemp2==0 ){
        assert( iWrite2==0 );
        rc = vdbeSorterOpenTempFile(db, &pTemp2);
        if( rc==SQLITE_OK ){
          rc = vdbeSorterExtendFile(pTemp2, pSorter->iWriteOff);
        }
      }

      if( rc==SQLITE_OK ){
        int bEof = 0;
        fileWriterInit(db, pTemp2, &writer, iWrite2);
        fileWriterWriteVarint(&writer, nWrite);
        while( rc==SQLITE_OK && bEof==0 ){
................................................................................
      }
      *pbEof = (pSorter->aIter[pSorter->aTree[1]].pFile==0);
    }
  }else{
    SorterRecord *pFree = pSorter->pRecord;
    pSorter->pRecord = pFree->pNext;
    pFree->pNext = 0;
    if( pSorter->aMemory==0 ){
      vdbeSorterRecordFree(db, pFree);
    }
    *pbEof = !pSorter->pRecord;
    rc = SQLITE_OK;
  }
  return rc;
}

/*