SQLite

Check-in [99efb235a0]
Login

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

Overview
Comment:In vdbesort.c, change the names of PmaReader variables "pIter" to "pReadr". Other related comment changes. The generated object code should be the same.
Downloads: Tarball | ZIP archive
Timelines: family | ancestors | descendants | both | threads
Files: files | file ages | folders
SHA1: 99efb235a08784020535a770d7e7d96d9ccee12d
User & Date: drh 2014-05-20 11:03:53.818
Context
2014-05-20
19:11
Improvements to the testability of the threads.c module. (check-in: 386e088868 user: drh tags: threads)
11:03
In vdbesort.c, change the names of PmaReader variables "pIter" to "pReadr". Other related comment changes. The generated object code should be the same. (check-in: 99efb235a0 user: drh tags: threads)
2014-05-19
23:17
Merge recent changes from trunk. (check-in: 6eefdad946 user: drh tags: threads)
Changes
Unified Diff Ignore Whitespace Patch
Changes to src/vdbesort.c.
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258

/*
** The MergeEngine object is used to combine two or more smaller PMAs into
** one big PMA using a merge operation.  Separate PMAs all need to be
** combined into one big PMA in order to be able to step through the sorted
** records in order.
**
** The aIter[] array contains a PmaReader object for each of the PMAs being
** merged.  An aIter[] object either points to a valid key or else is at EOF.
** For the purposes of the paragraphs below, we assume that the array is
** actually N elements in size, where N is the smallest power of 2 greater
** to or equal to the number of PMAs being merged. The extra aIter[] elements
** are treated as if they are empty (always at EOF).
**
** The aTree[] array is also N elements in size. The value of N is stored in
** the MergeEngine.nTree variable.
**
** The final (N/2) elements of aTree[] contain the results of comparing
** pairs of PMA keys together. Element i contains the result of 
** comparing aIter[2*i-N] and aIter[2*i-N+1]. Whichever key is smaller, the
** aTree element is set to the index of it. 
**
** For the purposes of this comparison, EOF is considered greater than any
** other key value. If the keys are equal (only possible with two EOF
** values), it doesn't matter which index is stored.
**
** The (N/4) elements of aTree[] that precede the final (N/2) described 
** above contains the index of the smallest of each block of 4 iterators.
** And so on. So that aTree[1] contains the index of the iterator that 
** currently points to the smallest key value. aTree[0] is unused.
**
** Example:
**
**     aIter[0] -> Banana
**     aIter[1] -> Feijoa
**     aIter[2] -> Elderberry
**     aIter[3] -> Currant
**     aIter[4] -> Grapefruit
**     aIter[5] -> Apple
**     aIter[6] -> Durian
**     aIter[7] -> EOF
**
**     aTree[] = { X, 5   0, 5    0, 3, 5, 6 }
**
** The current element is "Apple" (the value of the key indicated by 
** iterator 5). When the Next() operation is invoked, iterator 5 will
** be advanced to the next key in its segment. Say the next key is
** "Eggplant":
**
**     aIter[5] -> Eggplant
**
** The contents of aTree[] are updated first by comparing the new iterator
** 5 key to the current key of iterator 4 (still "Grapefruit"). The iterator
** 5 value is still smaller, so aTree[6] is set to 5. And so on up the tree.
** The value of iterator 6 - "Durian" - is now smaller than that of iterator
** 5, so aTree[3] is set to 6. Key 0 is smaller than key 6 (Banana<Durian),
** so the value written into element 1 of the array is 0. As follows:
**
**     aTree[] = { X, 0   0, 6    0, 3, 5, 6 }
**
** In other words, each time we advance to the next sorter element, log2(N)
** key comparison operations are required, where N is the number of segments
** being merged (rounded up to the next power of 2).
*/
struct MergeEngine {
  int nTree;                 /* Used size of aTree/aIter (power of 2) */
  int *aTree;                /* Current state of incremental merge */
  PmaReader *aIter;          /* Array of iterators to merge data from */
};

/*
** Exactly VdbeSorter.nTask instances of this object are allocated
** as part of each VdbeSorter object. Instances are never allocated any
** other way. VdbeSorter.nTask is set to the number of worker threads allowed
** (see SQLITE_CONFIG_WORKER_THREADS) plus one (the main thread).







|
|


|







|







|
|




|
|
|
|
|
|
|
|




|



|

|
|

|










|

|







184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258

/*
** The MergeEngine object is used to combine two or more smaller PMAs into
** one big PMA using a merge operation.  Separate PMAs all need to be
** combined into one big PMA in order to be able to step through the sorted
** records in order.
**
** The aReadr[] array contains a PmaReader object for each of the PMAs being
** merged.  An aReadr[] object either points to a valid key or else is at EOF.
** For the purposes of the paragraphs below, we assume that the array is
** actually N elements in size, where N is the smallest power of 2 greater
** to or equal to the number of PMAs being merged. The extra aReadr[] elements
** are treated as if they are empty (always at EOF).
**
** The aTree[] array is also N elements in size. The value of N is stored in
** the MergeEngine.nTree variable.
**
** The final (N/2) elements of aTree[] contain the results of comparing
** pairs of PMA keys together. Element i contains the result of 
** comparing aReadr[2*i-N] and aReadr[2*i-N+1]. Whichever key is smaller, the
** aTree element is set to the index of it. 
**
** For the purposes of this comparison, EOF is considered greater than any
** other key value. If the keys are equal (only possible with two EOF
** values), it doesn't matter which index is stored.
**
** The (N/4) elements of aTree[] that precede the final (N/2) described 
** above contains the index of the smallest of each block of 4 PmaReaders
** And so on. So that aTree[1] contains the index of the PmaReader that 
** currently points to the smallest key value. aTree[0] is unused.
**
** Example:
**
**     aReadr[0] -> Banana
**     aReadr[1] -> Feijoa
**     aReadr[2] -> Elderberry
**     aReadr[3] -> Currant
**     aReadr[4] -> Grapefruit
**     aReadr[5] -> Apple
**     aReadr[6] -> Durian
**     aReadr[7] -> EOF
**
**     aTree[] = { X, 5   0, 5    0, 3, 5, 6 }
**
** The current element is "Apple" (the value of the key indicated by 
** PmaReader 5). When the Next() operation is invoked, PmaReader 5 will
** be advanced to the next key in its segment. Say the next key is
** "Eggplant":
**
**     aReadr[5] -> Eggplant
**
** The contents of aTree[] are updated first by comparing the new PmaReader
** 5 key to the current key of PmaReader 4 (still "Grapefruit"). The PmaReader
** 5 value is still smaller, so aTree[6] is set to 5. And so on up the tree.
** The value of PmaReader 6 - "Durian" - is now smaller than that of PmaReader
** 5, so aTree[3] is set to 6. Key 0 is smaller than key 6 (Banana<Durian),
** so the value written into element 1 of the array is 0. As follows:
**
**     aTree[] = { X, 0   0, 6    0, 3, 5, 6 }
**
** In other words, each time we advance to the next sorter element, log2(N)
** key comparison operations are required, where N is the number of segments
** being merged (rounded up to the next power of 2).
*/
struct MergeEngine {
  int nTree;                 /* Used size of aTree/aReadr (power of 2) */
  int *aTree;                /* Current state of incremental merge */
  PmaReader *aReadr;         /* Array of PmaReaders to merge data from */
};

/*
** Exactly VdbeSorter.nTask instances of this object are allocated
** as part of each VdbeSorter object. Instances are never allocated any
** other way. VdbeSorter.nTask is set to the number of worker threads allowed
** (see SQLITE_CONFIG_WORKER_THREADS) plus one (the main thread).
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
**   largest record in the sorter.
*/
struct VdbeSorter {
  int mnPmaSize;                  /* Minimum PMA size, in bytes */
  int mxPmaSize;                  /* Maximum PMA size, in bytes.  0==no limit */
  int mxKeysize;                  /* Largest serialized key seen so far */
  int pgsz;                       /* Main database page size */
  PmaReader *pReader;             /* Read data from here after Rewind() */
  MergeEngine *pMerger;           /* Or here, if bUseThreads==0 */
  sqlite3 *db;                    /* Database connection */
  KeyInfo *pKeyInfo;              /* How to compare records */
  UnpackedRecord *pUnpacked;      /* Used by VdbeSorterCompare() */
  SorterList list;                /* List of in-memory records */
  int iMemory;                    /* Offset of free space in list.aMemory */
  int nMemory;                    /* Size of list.aMemory allocation in bytes */







|







300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
**   largest record in the sorter.
*/
struct VdbeSorter {
  int mnPmaSize;                  /* Minimum PMA size, in bytes */
  int mxPmaSize;                  /* Maximum PMA size, in bytes.  0==no limit */
  int mxKeysize;                  /* Largest serialized key seen so far */
  int pgsz;                       /* Main database page size */
  PmaReader *pReader;             /* Readr data from here after Rewind() */
  MergeEngine *pMerger;           /* Or here, if bUseThreads==0 */
  sqlite3 *db;                    /* Database connection */
  KeyInfo *pKeyInfo;              /* How to compare records */
  UnpackedRecord *pUnpacked;      /* Used by VdbeSorterCompare() */
  SorterList list;                /* List of in-memory records */
  int iMemory;                    /* Offset of free space in list.aMemory */
  int nMemory;                    /* Size of list.aMemory allocation in bytes */
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
/*
** An instance of the following object is used to read records out of a
** PMA, in sorted order.  The next key to be read is cached in nKey/aKey.
** pFile==0 at EOF.
*/
struct PmaReader {
  i64 iReadOff;                   /* Current read offset */
  i64 iEof;                       /* 1 byte past EOF for this iterator */
  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 entire file */
  IncrMerger *pIncr;              /* Incremental merger */
};







|


|







322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
/*
** An instance of the following object is used to read records out of a
** PMA, in sorted order.  The next key to be read is cached in nKey/aKey.
** pFile==0 at EOF.
*/
struct PmaReader {
  i64 iReadOff;                   /* Current read offset */
  i64 iEof;                       /* 1 byte past EOF for this PmaReader */
  int nAlloc;                     /* Bytes of space at aAlloc */
  int nKey;                       /* Number of bytes in key */
  sqlite3_file *pFile;            /* File we are 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 entire file */
  IncrMerger *pIncr;              /* Incremental merger */
};
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
static int vdbeIncrSwap(IncrMerger*);
static void vdbeIncrFree(IncrMerger *);

/*
** Free all memory belonging to the PmaReader object passed as the second
** argument. All structure fields are set to zero before returning.
*/
static void vdbePmaReaderClear(PmaReader *pIter){
  sqlite3_free(pIter->aAlloc);
  sqlite3_free(pIter->aBuffer);
  if( pIter->aMap ) sqlite3OsUnfetch(pIter->pFile, 0, pIter->aMap);
  vdbeIncrFree(pIter->pIncr);
  memset(pIter, 0, sizeof(PmaReader));
}

/*
** 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
** error code.
**
** The buffer indicated by *ppOut may only be considered valid until the
** next call to this function.
*/
static int vdbePmaReadBlob(
  PmaReader *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 ){







|
|
|
|
|
|












|







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
static int vdbeIncrSwap(IncrMerger*);
static void vdbeIncrFree(IncrMerger *);

/*
** Free all memory belonging to the PmaReader object passed as the second
** argument. All structure fields are set to zero before returning.
*/
static void vdbePmaReaderClear(PmaReader *pReadr){
  sqlite3_free(pReadr->aAlloc);
  sqlite3_free(pReadr->aBuffer);
  if( pReadr->aMap ) sqlite3OsUnfetch(pReadr->pFile, 0, pReadr->aMap);
  vdbeIncrFree(pReadr->pIncr);
  memset(pReadr, 0, sizeof(PmaReader));
}

/*
** 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
** error code.
**
** The buffer indicated by *ppOut may only be considered valid until the
** next call to this function.
*/
static int vdbePmaReadBlob(
  PmaReader *p,                   /* PmaReader from which to take the blob */
  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 ){
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
    if( (p->iEof - p->iReadOff) > (i64)p->nBuffer ){
      nRead = p->nBuffer;
    }else{
      nRead = (int)(p->iEof - p->iReadOff);
    }
    assert( nRead>0 );

    /* Read data from the file. Return early if an error occurs. */
    rc = sqlite3OsRead(p->pFile, p->aBuffer, nRead, p->iReadOff);
    assert( rc!=SQLITE_IOERR_SHORT_READ );
    if( rc!=SQLITE_OK ) return rc;
  }
  nAvail = p->nBuffer - iBuf; 

  if( nByte<=nAvail ){







|







491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
    if( (p->iEof - p->iReadOff) > (i64)p->nBuffer ){
      nRead = p->nBuffer;
    }else{
      nRead = (int)(p->iEof - p->iReadOff);
    }
    assert( nRead>0 );

    /* Readr data from the file. Return early if an error occurs. */
    rc = sqlite3OsRead(p->pFile, p->aBuffer, nRead, p->iReadOff);
    assert( rc!=SQLITE_IOERR_SHORT_READ );
    if( rc!=SQLITE_OK ) return rc;
  }
  nAvail = p->nBuffer - iBuf; 

  if( nByte<=nAvail ){
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
  if( pFile->iEof<=(i64)(pTask->pSorter->db->nMaxSorterMmap) ){
    rc = sqlite3OsFetch(pFile->pFd, 0, (int)pFile->iEof, (void**)pp);
  }
  return rc;
}

/*
** Seek iterator pIter to offset iOff within file pFile. Return SQLITE_OK 
** if successful, or an SQLite error code if an error occurs.
*/
static int vdbePmaReaderSeek(
  SortSubtask *pTask,             /* Task context */
  PmaReader *pIter,               /* Iterate to populate */
  SorterFile *pFile,              /* Sorter file to read from */
  i64 iOff                        /* Offset in pFile */
){
  int rc = SQLITE_OK;

  assert( pIter->pIncr==0 || pIter->pIncr->bEof==0 );

  if( pIter->aMap ){
    sqlite3OsUnfetch(pIter->pFile, 0, pIter->aMap);
    pIter->aMap = 0;
  }
  pIter->iReadOff = iOff;
  pIter->iEof = pFile->iEof;
  pIter->pFile = pFile->pFd;

  rc = vdbeSorterMapFile(pTask, pFile, &pIter->aMap);
  if( rc==SQLITE_OK && pIter->aMap==0 ){
    int pgsz = pTask->pSorter->pgsz;
    int iBuf = pIter->iReadOff % pgsz;
    if( pIter->aBuffer==0 ){
      pIter->aBuffer = (u8*)sqlite3Malloc(pgsz);
      if( pIter->aBuffer==0 ) rc = SQLITE_NOMEM;
      pIter->nBuffer = pgsz;
    }
    if( rc==SQLITE_OK && iBuf ){
      int nRead = pgsz - iBuf;
      if( (pIter->iReadOff + nRead) > pIter->iEof ){
        nRead = (int)(pIter->iEof - pIter->iReadOff);
      }
      rc = sqlite3OsRead(
          pIter->pFile, &pIter->aBuffer[iBuf], nRead, pIter->iReadOff
      );
      assert( rc!=SQLITE_IOERR_SHORT_READ );
    }
  }

  return rc;
}

/*
** Advance iterator pIter to the next key in its PMA. Return SQLITE_OK if
** no error occurs, or an SQLite error code if one does.
*/
static int vdbePmaReaderNext(PmaReader *pIter){
  int rc = SQLITE_OK;             /* Return Code */
  u64 nRec = 0;                   /* Size of record in bytes */


  if( pIter->iReadOff>=pIter->iEof ){
    IncrMerger *pIncr = pIter->pIncr;
    int bEof = 1;
    if( pIncr ){
      rc = vdbeIncrSwap(pIncr);
      if( rc==SQLITE_OK && pIncr->bEof==0 ){
        rc = vdbePmaReaderSeek(
            pIncr->pTask, pIter, &pIncr->aFile[0], pIncr->iStartOff
        );
        bEof = 0;
      }
    }

    if( bEof ){
      /* This is an EOF condition */
      vdbePmaReaderClear(pIter);
      return rc;
    }
  }

  if( rc==SQLITE_OK ){
    rc = vdbePmaReadVarint(pIter, &nRec);
  }
  if( rc==SQLITE_OK ){
    pIter->nKey = (int)nRec;
    rc = vdbePmaReadBlob(pIter, (int)nRec, &pIter->aKey);
  }

  return rc;
}

/*
** Initialize iterator pIter to scan through the PMA stored in file pFile
** starting at offset iStart and ending at offset iEof-1. This function 
** leaves the iterator pointing to the first key in the PMA (or EOF if the 
** PMA is empty).
**
** If the pnByte parameter is NULL, then it is assumed that the file 
** contains a single PMA, and that that PMA omits the initial length varint.
*/
static int vdbePmaReaderInit(
  SortSubtask *pTask,             /* Task context */
  SorterFile *pFile,              /* Sorter file to read from */
  i64 iStart,                     /* Start offset in pFile */
  PmaReader *pIter,               /* Iterator to populate */
  i64 *pnByte                     /* IN/OUT: Increment this value by PMA size */
){
  int rc;

  assert( pFile->iEof>iStart );
  assert( pIter->aAlloc==0 && pIter->nAlloc==0 );
  assert( pIter->aBuffer==0 );
  assert( pIter->aMap==0 );

  rc = vdbePmaReaderSeek(pTask, pIter, pFile, iStart);
  if( rc==SQLITE_OK ){
    u64 nByte;                    /* Size of PMA in bytes */
    rc = vdbePmaReadVarint(pIter, &nByte);
    pIter->iEof = pIter->iReadOff + nByte;
    *pnByte += nByte;
  }

  if( rc==SQLITE_OK ){
    rc = vdbePmaReaderNext(pIter);
  }
  return rc;
}


/*
** Compare key1 (buffer pKey1, size nKey1 bytes) with key2 (buffer pKey2, 







|




|





|

|
|
|

|
|
|

|
|

|
|
|
|
|



|
|


|









|


|




|
|





|







|





|


|
|






|

|









|





|
|
|

|


|
|




|







595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
  if( pFile->iEof<=(i64)(pTask->pSorter->db->nMaxSorterMmap) ){
    rc = sqlite3OsFetch(pFile->pFd, 0, (int)pFile->iEof, (void**)pp);
  }
  return rc;
}

/*
** Seek PmaReader pReadr to offset iOff within file pFile. Return SQLITE_OK 
** if successful, or an SQLite error code if an error occurs.
*/
static int vdbePmaReaderSeek(
  SortSubtask *pTask,             /* Task context */
  PmaReader *pReadr,              /* Iterate to populate */
  SorterFile *pFile,              /* Sorter file to read from */
  i64 iOff                        /* Offset in pFile */
){
  int rc = SQLITE_OK;

  assert( pReadr->pIncr==0 || pReadr->pIncr->bEof==0 );

  if( pReadr->aMap ){
    sqlite3OsUnfetch(pReadr->pFile, 0, pReadr->aMap);
    pReadr->aMap = 0;
  }
  pReadr->iReadOff = iOff;
  pReadr->iEof = pFile->iEof;
  pReadr->pFile = pFile->pFd;

  rc = vdbeSorterMapFile(pTask, pFile, &pReadr->aMap);
  if( rc==SQLITE_OK && pReadr->aMap==0 ){
    int pgsz = pTask->pSorter->pgsz;
    int iBuf = pReadr->iReadOff % pgsz;
    if( pReadr->aBuffer==0 ){
      pReadr->aBuffer = (u8*)sqlite3Malloc(pgsz);
      if( pReadr->aBuffer==0 ) rc = SQLITE_NOMEM;
      pReadr->nBuffer = pgsz;
    }
    if( rc==SQLITE_OK && iBuf ){
      int nRead = pgsz - iBuf;
      if( (pReadr->iReadOff + nRead) > pReadr->iEof ){
        nRead = (int)(pReadr->iEof - pReadr->iReadOff);
      }
      rc = sqlite3OsRead(
          pReadr->pFile, &pReadr->aBuffer[iBuf], nRead, pReadr->iReadOff
      );
      assert( rc!=SQLITE_IOERR_SHORT_READ );
    }
  }

  return rc;
}

/*
** Advance PmaReader pReadr to the next key in its PMA. Return SQLITE_OK if
** no error occurs, or an SQLite error code if one does.
*/
static int vdbePmaReaderNext(PmaReader *pReadr){
  int rc = SQLITE_OK;             /* Return Code */
  u64 nRec = 0;                   /* Size of record in bytes */


  if( pReadr->iReadOff>=pReadr->iEof ){
    IncrMerger *pIncr = pReadr->pIncr;
    int bEof = 1;
    if( pIncr ){
      rc = vdbeIncrSwap(pIncr);
      if( rc==SQLITE_OK && pIncr->bEof==0 ){
        rc = vdbePmaReaderSeek(
            pIncr->pTask, pReadr, &pIncr->aFile[0], pIncr->iStartOff
        );
        bEof = 0;
      }
    }

    if( bEof ){
      /* This is an EOF condition */
      vdbePmaReaderClear(pReadr);
      return rc;
    }
  }

  if( rc==SQLITE_OK ){
    rc = vdbePmaReadVarint(pReadr, &nRec);
  }
  if( rc==SQLITE_OK ){
    pReadr->nKey = (int)nRec;
    rc = vdbePmaReadBlob(pReadr, (int)nRec, &pReadr->aKey);
  }

  return rc;
}

/*
** Initialize PmaReader pReadr to scan through the PMA stored in file pFile
** starting at offset iStart and ending at offset iEof-1. This function 
** leaves the PmaReader pointing to the first key in the PMA (or EOF if the 
** PMA is empty).
**
** If the pnByte parameter is NULL, then it is assumed that the file 
** contains a single PMA, and that that PMA omits the initial length varint.
*/
static int vdbePmaReaderInit(
  SortSubtask *pTask,             /* Task context */
  SorterFile *pFile,              /* Sorter file to read from */
  i64 iStart,                     /* Start offset in pFile */
  PmaReader *pReadr,              /* PmaReader to populate */
  i64 *pnByte                     /* IN/OUT: Increment this value by PMA size */
){
  int rc;

  assert( pFile->iEof>iStart );
  assert( pReadr->aAlloc==0 && pReadr->nAlloc==0 );
  assert( pReadr->aBuffer==0 );
  assert( pReadr->aMap==0 );

  rc = vdbePmaReaderSeek(pTask, pReadr, pFile, iStart);
  if( rc==SQLITE_OK ){
    u64 nByte;                    /* Size of PMA in bytes */
    rc = vdbePmaReadVarint(pReadr, &nByte);
    pReadr->iEof = pReadr->iReadOff + nByte;
    *pnByte += nByte;
  }

  if( rc==SQLITE_OK ){
    rc = vdbePmaReaderNext(pReadr);
  }
  return rc;
}


/*
** Compare key1 (buffer pKey1, size nKey1 bytes) with key2 (buffer pKey2, 
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
  if( pKey2 ){
    sqlite3VdbeRecordUnpack(pTask->pSorter->pKeyInfo, nKey2, pKey2, r2);
  }
  return sqlite3VdbeRecordCompare(nKey1, pKey1, r2, 0);
}

/*
** This function is called to compare two iterator keys when merging 
** multiple b-tree segments. Parameter iOut is the index of the aTree[] 
** value to recalculate.
*/
static int vdbeSorterDoCompare(
  SortSubtask *pTask, 
  MergeEngine *pMerger, 
  int iOut







|







744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
  if( pKey2 ){
    sqlite3VdbeRecordUnpack(pTask->pSorter->pKeyInfo, nKey2, pKey2, r2);
  }
  return sqlite3VdbeRecordCompare(nKey1, pKey1, r2, 0);
}

/*
** This function is called to compare two PmaReader keys when merging 
** multiple b-tree segments. Parameter iOut is the index of the aTree[] 
** value to recalculate.
*/
static int vdbeSorterDoCompare(
  SortSubtask *pTask, 
  MergeEngine *pMerger, 
  int iOut
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
    i1 = (iOut - pMerger->nTree/2) * 2;
    i2 = i1 + 1;
  }else{
    i1 = pMerger->aTree[iOut*2];
    i2 = pMerger->aTree[iOut*2+1];
  }

  p1 = &pMerger->aIter[i1];
  p2 = &pMerger->aIter[i2];

  if( p1->pFile==0 ){
    iRes = i2;
  }else if( p2->pFile==0 ){
    iRes = i1;
  }else{
    int res;







|
|







769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
    i1 = (iOut - pMerger->nTree/2) * 2;
    i2 = i1 + 1;
  }else{
    i1 = pMerger->aTree[iOut*2];
    i2 = pMerger->aTree[iOut*2+1];
  }

  p1 = &pMerger->aReadr[i1];
  p2 = &pMerger->aReadr[i2];

  if( p1->pFile==0 ){
    iRes = i2;
  }else if( p2->pFile==0 ){
    iRes = i1;
  }else{
    int res;
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
}
#else
# define vdbeSorterJoinAll(x,rcin) (rcin)
# define vdbeSorterJoinThread(pTask) SQLITE_OK
#endif

/*
** Allocate a new MergeEngine object with space for nIter iterators.
*/
static MergeEngine *vdbeMergeEngineNew(int nIter){
  int N = 2;                      /* Smallest power of two >= nIter */
  int nByte;                      /* Total bytes of space to allocate */
  MergeEngine *pNew;              /* Pointer to allocated object to return */

  assert( nIter<=SORTER_MAX_MERGE_COUNT );

  while( N<nIter ) N += N;
  nByte = sizeof(MergeEngine) + N * (sizeof(int) + sizeof(PmaReader));

  pNew = sqlite3FaultSim(100) ? 0 : (MergeEngine*)sqlite3MallocZero(nByte);
  if( pNew ){
    pNew->nTree = N;
    pNew->aIter = (PmaReader*)&pNew[1];
    pNew->aTree = (int*)&pNew->aIter[N];
  }
  return pNew;
}

/*
** Free the MergeEngine object passed as the only argument.
*/
static void vdbeMergeEngineFree(MergeEngine *pMerger){
  int i;
  if( pMerger ){
    for(i=0; i<pMerger->nTree; i++){
      vdbePmaReaderClear(&pMerger->aIter[i]);
    }
  }
  sqlite3_free(pMerger);
}

/*
** Free all resources associated with the IncrMerger object indicated by







|

|
|



|

|





|
|











|







1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
}
#else
# define vdbeSorterJoinAll(x,rcin) (rcin)
# define vdbeSorterJoinThread(pTask) SQLITE_OK
#endif

/*
** Allocate a new MergeEngine object with space for nReader PmaReaders.
*/
static MergeEngine *vdbeMergeEngineNew(int nReader){
  int N = 2;                      /* Smallest power of two >= nReader */
  int nByte;                      /* Total bytes of space to allocate */
  MergeEngine *pNew;              /* Pointer to allocated object to return */

  assert( nReader<=SORTER_MAX_MERGE_COUNT );

  while( N<nReader ) N += N;
  nByte = sizeof(MergeEngine) + N * (sizeof(int) + sizeof(PmaReader));

  pNew = sqlite3FaultSim(100) ? 0 : (MergeEngine*)sqlite3MallocZero(nByte);
  if( pNew ){
    pNew->nTree = N;
    pNew->aReadr = (PmaReader*)&pNew[1];
    pNew->aTree = (int*)&pNew->aReadr[N];
  }
  return pNew;
}

/*
** Free the MergeEngine object passed as the only argument.
*/
static void vdbeMergeEngineFree(MergeEngine *pMerger){
  int i;
  if( pMerger ){
    for(i=0; i<pMerger->nTree; i++){
      vdbePmaReaderClear(&pMerger->aReadr[i]);
    }
  }
  sqlite3_free(pMerger);
}

/*
** Free all resources associated with the IncrMerger object indicated by
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
  vdbeSorterWorkDebug(pTask, "exit");
  assert( rc!=SQLITE_OK || pList->pList==0 );
  assert( rc!=SQLITE_OK || pTask->file.iEof==iSz );
  return rc;
}

/*
** Advance the MergeEngine iterator passed as the second argument to
** the next entry. Set *pbEof to true if this means the iterator has 
** reached EOF.
**
** Return SQLITE_OK if successful or an error code if an error occurs.
*/
static int vdbeSorterNext(
  SortSubtask *pTask, 
  MergeEngine *pMerger, 
  int *pbEof
){
  int rc;
  int iPrev = pMerger->aTree[1];/* Index of iterator to advance */

  /* Advance the current iterator */
  rc = vdbePmaReaderNext(&pMerger->aIter[iPrev]);

  /* Update contents of aTree[] */
  if( rc==SQLITE_OK ){
    int i;                      /* Index of aTree[] to recalculate */
    PmaReader *pIter1;     /* First iterator to compare */
    PmaReader *pIter2;     /* Second iterator to compare */
    u8 *pKey2;                  /* To pIter2->aKey, or 0 if record cached */

    /* Find the first two iterators to compare. The one that was just
    ** advanced (iPrev) and the one next to it in the array.  */
    pIter1 = &pMerger->aIter[(iPrev & 0xFFFE)];
    pIter2 = &pMerger->aIter[(iPrev | 0x0001)];
    pKey2 = pIter2->aKey;

    for(i=(pMerger->nTree+iPrev)/2; i>0; i=i/2){
      /* Compare pIter1 and pIter2. Store the result in variable iRes. */
      int iRes;
      if( pIter1->pFile==0 ){
        iRes = +1;
      }else if( pIter2->pFile==0 ){
        iRes = -1;
      }else{
        iRes = vdbeSorterCompare(pTask, 
            pIter1->aKey, pIter1->nKey, pKey2, pIter2->nKey
        );
      }

      /* If pIter1 contained the smaller value, set aTree[i] to its index.
      ** Then set pIter2 to the next iterator to compare to pIter1. In this
      ** case there is no cache of pIter2 in pTask->pUnpacked, so set
      ** pKey2 to point to the record belonging to pIter2.
      **
      ** Alternatively, if pIter2 contains the smaller of the two values,
      ** set aTree[i] to its index and update pIter1. If vdbeSorterCompare()
      ** was actually called above, then pTask->pUnpacked now contains
      ** a value equivalent to pIter2. So set pKey2 to NULL to prevent
      ** vdbeSorterCompare() from decoding pIter2 again.
      **
      ** If the two values were equal, then the value from the oldest
      ** PMA should be considered smaller. The VdbeSorter.aIter[] array
      ** is sorted from oldest to newest, so pIter1 contains older values
      ** than pIter2 iff (pIter1<pIter2).  */
      if( iRes<0 || (iRes==0 && pIter1<pIter2) ){
        pMerger->aTree[i] = (int)(pIter1 - pMerger->aIter);
        pIter2 = &pMerger->aIter[ pMerger->aTree[i ^ 0x0001] ];
        pKey2 = pIter2->aKey;
      }else{
        if( pIter1->pFile ) pKey2 = 0;
        pMerger->aTree[i] = (int)(pIter2 - pMerger->aIter);
        pIter1 = &pMerger->aIter[ pMerger->aTree[i ^ 0x0001] ];
      }
    }
    *pbEof = (pMerger->aIter[pMerger->aTree[1]].pFile==0);
  }

  return (rc==SQLITE_OK ? pTask->pUnpacked->errCode : rc);
}

#if SQLITE_MAX_WORKER_THREADS>0
/*







|
|










|

|
|




|
|
|

|

|
|
|


|

|

|



|



|
|
|
|

|
|

|
|


|
|
|
|
|
|
|

|
|
|


|







1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
  vdbeSorterWorkDebug(pTask, "exit");
  assert( rc!=SQLITE_OK || pList->pList==0 );
  assert( rc!=SQLITE_OK || pTask->file.iEof==iSz );
  return rc;
}

/*
** Advance the MergeEngine PmaReader passed as the second argument to
** the next entry. Set *pbEof to true if this means the PmaReader has 
** reached EOF.
**
** Return SQLITE_OK if successful or an error code if an error occurs.
*/
static int vdbeSorterNext(
  SortSubtask *pTask, 
  MergeEngine *pMerger, 
  int *pbEof
){
  int rc;
  int iPrev = pMerger->aTree[1];/* Index of PmaReader to advance */

  /* Advance the current PmaReader */
  rc = vdbePmaReaderNext(&pMerger->aReadr[iPrev]);

  /* Update contents of aTree[] */
  if( rc==SQLITE_OK ){
    int i;                      /* Index of aTree[] to recalculate */
    PmaReader *pReadr1;         /* First PmaReader to compare */
    PmaReader *pReadr2;         /* Second PmaReader to compare */
    u8 *pKey2;                  /* To pReadr2->aKey, or 0 if record cached */

    /* Find the first two PmaReaders to compare. The one that was just
    ** advanced (iPrev) and the one next to it in the array.  */
    pReadr1 = &pMerger->aReadr[(iPrev & 0xFFFE)];
    pReadr2 = &pMerger->aReadr[(iPrev | 0x0001)];
    pKey2 = pReadr2->aKey;

    for(i=(pMerger->nTree+iPrev)/2; i>0; i=i/2){
      /* Compare pReadr1 and pReadr2. Store the result in variable iRes. */
      int iRes;
      if( pReadr1->pFile==0 ){
        iRes = +1;
      }else if( pReadr2->pFile==0 ){
        iRes = -1;
      }else{
        iRes = vdbeSorterCompare(pTask, 
            pReadr1->aKey, pReadr1->nKey, pKey2, pReadr2->nKey
        );
      }

      /* If pReadr1 contained the smaller value, set aTree[i] to its index.
      ** Then set pReadr2 to the next PmaReader to compare to pReadr1. In this
      ** case there is no cache of pReadr2 in pTask->pUnpacked, so set
      ** pKey2 to point to the record belonging to pReadr2.
      **
      ** Alternatively, if pReadr2 contains the smaller of the two values,
      ** set aTree[i] to its index and update pReadr1. If vdbeSorterCompare()
      ** was actually called above, then pTask->pUnpacked now contains
      ** a value equivalent to pReadr2. So set pKey2 to NULL to prevent
      ** vdbeSorterCompare() from decoding pReadr2 again.
      **
      ** If the two values were equal, then the value from the oldest
      ** PMA should be considered smaller. The VdbeSorter.aReadr[] array
      ** is sorted from oldest to newest, so pReadr1 contains older values
      ** than pReadr2 iff (pReadr1<pReadr2).  */
      if( iRes<0 || (iRes==0 && pReadr1<pReadr2) ){
        pMerger->aTree[i] = (int)(pReadr1 - pMerger->aReadr);
        pReadr2 = &pMerger->aReadr[ pMerger->aTree[i ^ 0x0001] ];
        pKey2 = pReadr2->aKey;
      }else{
        if( pReadr1->pFile ) pKey2 = 0;
        pMerger->aTree[i] = (int)(pReadr2 - pMerger->aReadr);
        pReadr1 = &pMerger->aReadr[ pMerger->aTree[i ^ 0x0001] ];
      }
    }
    *pbEof = (pMerger->aReadr[pMerger->aTree[1]].pFile==0);
  }

  return (rc==SQLITE_OK ? pTask->pUnpacked->errCode : rc);
}

#if SQLITE_MAX_WORKER_THREADS>0
/*
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
  assert( pIncr->bEof==0 );

  vdbeSorterPopulateDebug(pTask, "enter");

  vdbePmaWriterInit(pOut->pFd, &writer, pTask->pSorter->pgsz, iStart);
  while( rc==SQLITE_OK ){
    int dummy;
    PmaReader *pReader = &pMerger->aIter[ pMerger->aTree[1] ];
    int nKey = pReader->nKey;
    i64 iEof = writer.iWriteOff + writer.iBufEnd;

    /* Check if the output file is full or if the input has been exhausted.
    ** In either case exit the loop. */
    if( pReader->pFile==0 ) break;
    if( (iEof + nKey + sqlite3VarintLen(nKey))>(iStart + pIncr->mxSz) ) break;







|







1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
  assert( pIncr->bEof==0 );

  vdbeSorterPopulateDebug(pTask, "enter");

  vdbePmaWriterInit(pOut->pFd, &writer, pTask->pSorter->pgsz, iStart);
  while( rc==SQLITE_OK ){
    int dummy;
    PmaReader *pReader = &pMerger->aReadr[ pMerger->aTree[1] ];
    int nKey = pReader->nKey;
    i64 iEof = writer.iWriteOff + writer.iBufEnd;

    /* Check if the output file is full or if the input has been exhausted.
    ** In either case exit the loop. */
    if( pReader->pFile==0 ) break;
    if( (iEof + nKey + sqlite3VarintLen(nKey))>(iStart + pIncr->mxSz) ) break;
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
  return vdbeSorterCreateThread(pIncr->pTask, vdbeIncrPopulateThread, p);
}
#endif

/*
** This function is called when the PmaReader corresponding to pIncr has
** finished reading the contents of aFile[0]. Its purpose is to "refill"
** aFile[0] such that the iterator should start rereading it from the
** beginning.
**
** For single-threaded objects, this is accomplished by literally reading 
** keys from pIncr->pMerger and repopulating aFile[0]. 
**
** For multi-threaded objects, all that is required is to wait until the 
** background thread is finished (if it is not already) and then swap 







|







1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
  return vdbeSorterCreateThread(pIncr->pTask, vdbeIncrPopulateThread, p);
}
#endif

/*
** This function is called when the PmaReader corresponding to pIncr has
** finished reading the contents of aFile[0]. Its purpose is to "refill"
** aFile[0] such that the PmaReader should start rereading it from the
** beginning.
**
** For single-threaded objects, this is accomplished by literally reading 
** keys from pIncr->pMerger and repopulating aFile[0]. 
**
** For multi-threaded objects, all that is required is to wait until the 
** background thread is finished (if it is not already) and then swap 
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
  pIncr->pTask->file2.iEof -= pIncr->mxSz;
}
#endif /* SQLITE_MAX_WORKER_THREADS>0 */

#define INCRINIT_NORMAL 0
#define INCRINIT_TASK   1
#define INCRINIT_ROOT   2
static int vdbePmaReaderIncrInit(PmaReader *pIter, int eMode);

/*
** Initialize the merger argument passed as the second argument. Once this
** function returns, the first key of merged data may be read from the merger
** object in the usual fashion.
**
** If argument eMode is INCRINIT_ROOT, then it is assumed that any IncrMerge
** objects attached to the PmaReader objects that the merger reads from have
** already been populated, but that they have not yet populated aFile[0] and
** set the PmaReader objects up to read from it. In this case all that is
** required is to call vdbePmaReaderNext() on each iterator to point it at
** its first key.
**
** Otherwise, if eMode is any value other than INCRINIT_ROOT, then use 
** vdbePmaReaderIncrInit() to initialize each PmaReader that feeds data 
** to pMerger.
**
** SQLITE_OK is returned if successful, or an SQLite error code otherwise.
*/
static int vdbeIncrInitMerger(
  SortSubtask *pTask, 
  MergeEngine *pMerger, 
  int eMode                       /* One of the INCRINIT_XXX constants */
){
  int rc = SQLITE_OK;             /* Return code */
  int i;                          /* For iterating through PmaReader objects */
  int nTree = pMerger->nTree;

  for(i=0; rc==SQLITE_OK && i<nTree; i++){
    if( eMode==INCRINIT_ROOT ){
      /* Iterators should be normally initialized in order, as if they are
      ** reading from the same temp file this makes for more linear file IO.
      ** However, in the INCRINIT_ROOT case, if iterator aIter[nTask-1] is
      ** in use it will block the vdbePmaReaderNext() call while it uses
      ** the main thread to fill its buffer. So calling PmaReaderNext()
      ** on this iterator before any of the multi-threaded iterators takes
      ** better advantage of multi-processor hardware. */
      rc = vdbePmaReaderNext(&pMerger->aIter[nTree-i-1]);
    }else{
      rc = vdbePmaReaderIncrInit(&pMerger->aIter[i], INCRINIT_NORMAL);
    }
  }

  for(i=pMerger->nTree-1; rc==SQLITE_OK && i>0; i--){
    rc = vdbeSorterDoCompare(pTask, pMerger, i);
  }

  return (rc==SQLITE_OK ? pTask->pUnpacked->errCode : rc);
}

/*
** If the PmaReader passed as the first argument is not an incremental-reader
** (if pIter->pIncr==0), then this function is a no-op. Otherwise, it serves
** to open and/or initialize the temp file related fields of the IncrMerge
** object at (pIter->pIncr).
**
** If argument eMode is set to INCRINIT_NORMAL, then PmaReader iterators
** in the sub-tree headed by pIter are also initialized. Data is then loaded
** into the buffers belonging to this iterator, pIter, and it is set to
** point to the first key in its range.
**
** If argument eMode is set to INCRINIT_TASK, then PmaReader is guaranteed
** to be a multi-threaded iterator and this function is being called in a
** background thread. In this case all iterators in the sub-tree are 
** initialized as for INCRINIT_NORMAL and the aFile[1] buffer belonging to
** pIter is populated. However, the iterator itself is not set up to point
** to its first key. A call to vdbePmaReaderNext() is still required to do
** that. 
**
** The reason this function does not call vdbePmaReaderNext() immediately 
** in the INCRINIT_TASK case is that vdbePmaReaderNext() assumes that has
** to block on thread (pTask->thread) before accessing aFile[1]. But, since
** this entire function is being run by thread (pTask->thread), that will
** lead to the current background thread attempting to join itself.
**
** Finally, if argument eMode is set to INCRINIT_ROOT, it may be assumed
** that pIter->pIncr is a multi-threaded IncrMerge objects, and that all
** child-trees have already been initialized using IncrInit(INCRINIT_TASK).
** In this case vdbePmaReaderNext() is called on all child iterators and
** the current iterator set to point to the first key in its range.
**
** SQLITE_OK is returned if successful, or an SQLite error code otherwise.
*/
static int vdbePmaReaderIncrInit(PmaReader *pIter, int eMode){
  int rc = SQLITE_OK;
  IncrMerger *pIncr = pIter->pIncr;
  if( pIncr ){
    SortSubtask *pTask = pIncr->pTask;
    sqlite3 *db = pTask->pSorter->db;

    rc = vdbeIncrInitMerger(pTask, pIncr->pMerger, eMode);

    /* Set up the required files for pIncr. A multi-theaded IncrMerge object







|










|














|




|

|


|

|

|












|

|

|
|
|



|
|

|










|

|
|



|

|







1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
  pIncr->pTask->file2.iEof -= pIncr->mxSz;
}
#endif /* SQLITE_MAX_WORKER_THREADS>0 */

#define INCRINIT_NORMAL 0
#define INCRINIT_TASK   1
#define INCRINIT_ROOT   2
static int vdbePmaReaderIncrInit(PmaReader *pReadr, int eMode);

/*
** Initialize the merger argument passed as the second argument. Once this
** function returns, the first key of merged data may be read from the merger
** object in the usual fashion.
**
** If argument eMode is INCRINIT_ROOT, then it is assumed that any IncrMerge
** objects attached to the PmaReader objects that the merger reads from have
** already been populated, but that they have not yet populated aFile[0] and
** set the PmaReader objects up to read from it. In this case all that is
** required is to call vdbePmaReaderNext() on each PmaReader to point it at
** its first key.
**
** Otherwise, if eMode is any value other than INCRINIT_ROOT, then use 
** vdbePmaReaderIncrInit() to initialize each PmaReader that feeds data 
** to pMerger.
**
** SQLITE_OK is returned if successful, or an SQLite error code otherwise.
*/
static int vdbeIncrInitMerger(
  SortSubtask *pTask, 
  MergeEngine *pMerger, 
  int eMode                       /* One of the INCRINIT_XXX constants */
){
  int rc = SQLITE_OK;             /* Return code */
  int i;                          /* For looping over PmaReader objects */
  int nTree = pMerger->nTree;

  for(i=0; rc==SQLITE_OK && i<nTree; i++){
    if( eMode==INCRINIT_ROOT ){
      /* PmaReaders should be normally initialized in order, as if they are
      ** reading from the same temp file this makes for more linear file IO.
      ** However, in the INCRINIT_ROOT case, if PmaReader aReadr[nTask-1] is
      ** in use it will block the vdbePmaReaderNext() call while it uses
      ** the main thread to fill its buffer. So calling PmaReaderNext()
      ** on this PmaReader before any of the multi-threaded PmaReaders takes
      ** better advantage of multi-processor hardware. */
      rc = vdbePmaReaderNext(&pMerger->aReadr[nTree-i-1]);
    }else{
      rc = vdbePmaReaderIncrInit(&pMerger->aReadr[i], INCRINIT_NORMAL);
    }
  }

  for(i=pMerger->nTree-1; rc==SQLITE_OK && i>0; i--){
    rc = vdbeSorterDoCompare(pTask, pMerger, i);
  }

  return (rc==SQLITE_OK ? pTask->pUnpacked->errCode : rc);
}

/*
** If the PmaReader passed as the first argument is not an incremental-reader
** (if pReadr->pIncr==0), then this function is a no-op. Otherwise, it serves
** to open and/or initialize the temp file related fields of the IncrMerge
** object at (pReadr->pIncr).
**
** If argument eMode is set to INCRINIT_NORMAL, then PmaReader PmaReaders
** in the sub-tree headed by pReadr are also initialized. Data is then loaded
** into the buffers belonging to this PmaReader, pReadr, and it is set to
** point to the first key in its range.
**
** If argument eMode is set to INCRINIT_TASK, then PmaReader is guaranteed
** to be a multi-threaded PmaReader and this function is being called in a
** background thread. In this case all PmaReaders in the sub-tree are 
** initialized as for INCRINIT_NORMAL and the aFile[1] buffer belonging to
** pReadr is populated. However, the PmaReader itself is not set up to point
** to its first key. A call to vdbePmaReaderNext() is still required to do
** that. 
**
** The reason this function does not call vdbePmaReaderNext() immediately 
** in the INCRINIT_TASK case is that vdbePmaReaderNext() assumes that has
** to block on thread (pTask->thread) before accessing aFile[1]. But, since
** this entire function is being run by thread (pTask->thread), that will
** lead to the current background thread attempting to join itself.
**
** Finally, if argument eMode is set to INCRINIT_ROOT, it may be assumed
** that pReadr->pIncr is a multi-threaded IncrMerge objects, and that all
** child-trees have already been initialized using IncrInit(INCRINIT_TASK).
** In this case vdbePmaReaderNext() is called on all child PmaReaders and
** the current PmaReader set to point to the first key in its range.
**
** SQLITE_OK is returned if successful, or an SQLite error code otherwise.
*/
static int vdbePmaReaderIncrInit(PmaReader *pReadr, int eMode){
  int rc = SQLITE_OK;
  IncrMerger *pIncr = pReadr->pIncr;
  if( pIncr ){
    SortSubtask *pTask = pIncr->pTask;
    sqlite3 *db = pTask->pSorter->db;

    rc = vdbeIncrInitMerger(pTask, pIncr->pMerger, eMode);

    /* Set up the required files for pIncr. A multi-theaded IncrMerge object
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
        }
      }
    }

#if SQLITE_MAX_WORKER_THREADS>0
    if( rc==SQLITE_OK && pIncr->bUseThread ){
      /* Use the current thread to populate aFile[1], even though this
      ** iterator is multi-threaded. The reason being that this function
      ** is already running in background thread pIncr->pTask->thread. */
      assert( eMode==INCRINIT_ROOT || eMode==INCRINIT_TASK );
      rc = vdbeIncrPopulate(pIncr);
    }
#endif

    if( rc==SQLITE_OK && eMode!=INCRINIT_TASK ){
      rc = vdbePmaReaderNext(pIter);
    }
  }
  return rc;
}

#if SQLITE_MAX_WORKER_THREADS>0
/*







|







|







1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
        }
      }
    }

#if SQLITE_MAX_WORKER_THREADS>0
    if( rc==SQLITE_OK && pIncr->bUseThread ){
      /* Use the current thread to populate aFile[1], even though this
      ** PmaReader is multi-threaded. The reason being that this function
      ** is already running in background thread pIncr->pTask->thread. */
      assert( eMode==INCRINIT_ROOT || eMode==INCRINIT_TASK );
      rc = vdbeIncrPopulate(pIncr);
    }
#endif

    if( rc==SQLITE_OK && eMode!=INCRINIT_TASK ){
      rc = vdbePmaReaderNext(pReadr);
    }
  }
  return rc;
}

#if SQLITE_MAX_WORKER_THREADS>0
/*
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
  return pRet;
}

/*
** Use a background thread to invoke vdbePmaReaderIncrInit(INCRINIT_TASK) 
** on the the PmaReader object passed as the first argument.
**
** This call will initialize the various fields of the pIter->pIncr 
** structure and, if it is a multi-threaded IncrMerger, launch a 
** background thread to populate aFile[1].
*/
static int vdbePmaReaderBgIncrInit(PmaReader *pIter){
  void *pCtx = (void*)pIter;
  return vdbeSorterCreateThread(pIter->pIncr->pTask, vdbePmaReaderBgInit, pCtx);
}
#endif

/*
** Allocate a new MergeEngine object to merge the contents of nPMA level-0
** PMAs from pTask->file. If no error occurs, set *ppOut to point to
** the new object and return SQLITE_OK. Or, if an error does occur, set *ppOut
** to NULL and return an SQLite error code.
**
** When this function is called, *piOffset is set to the offset of the
** first PMA to read from pTask->file. Assuming no error occurs, it is 
** set to the offset immediately following the last byte of the last
** PMA before returning. If an error does occur, then the final value of
** *piOffset is undefined.
*/
static int vdbeMergeEngineLevel0(
  SortSubtask *pTask,             /* Sorter task to read from */
  int nPMA,                       /* Number of PMAs to read */
  i64 *piOffset,                  /* IN/OUT: Read offset in pTask->file */
  MergeEngine **ppOut             /* OUT: New merge-engine */
){
  MergeEngine *pNew;              /* Merge engine to return */
  i64 iOff = *piOffset;
  int i;
  int rc = SQLITE_OK;

  *ppOut = pNew = vdbeMergeEngineNew(nPMA);
  if( pNew==0 ) rc = SQLITE_NOMEM;

  for(i=0; i<nPMA && rc==SQLITE_OK; i++){
    i64 nDummy;
    PmaReader *pIter = &pNew->aIter[i];
    rc = vdbePmaReaderInit(pTask, &pTask->file, iOff, pIter, &nDummy);
    iOff = pIter->iEof;
  }

  if( rc!=SQLITE_OK ){
    vdbeMergeEngineFree(pNew);
    *ppOut = 0;
  }
  *piOffset = iOff;







|



|
|
|


















|












|
|
|







1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
  return pRet;
}

/*
** Use a background thread to invoke vdbePmaReaderIncrInit(INCRINIT_TASK) 
** on the the PmaReader object passed as the first argument.
**
** This call will initialize the various fields of the pReadr->pIncr 
** structure and, if it is a multi-threaded IncrMerger, launch a 
** background thread to populate aFile[1].
*/
static int vdbePmaReaderBgIncrInit(PmaReader *pReadr){
  void *pCtx = (void*)pReadr;
  return vdbeSorterCreateThread(pReadr->pIncr->pTask, vdbePmaReaderBgInit, pCtx);
}
#endif

/*
** Allocate a new MergeEngine object to merge the contents of nPMA level-0
** PMAs from pTask->file. If no error occurs, set *ppOut to point to
** the new object and return SQLITE_OK. Or, if an error does occur, set *ppOut
** to NULL and return an SQLite error code.
**
** When this function is called, *piOffset is set to the offset of the
** first PMA to read from pTask->file. Assuming no error occurs, it is 
** set to the offset immediately following the last byte of the last
** PMA before returning. If an error does occur, then the final value of
** *piOffset is undefined.
*/
static int vdbeMergeEngineLevel0(
  SortSubtask *pTask,             /* Sorter task to read from */
  int nPMA,                       /* Number of PMAs to read */
  i64 *piOffset,                  /* IN/OUT: Readr offset in pTask->file */
  MergeEngine **ppOut             /* OUT: New merge-engine */
){
  MergeEngine *pNew;              /* Merge engine to return */
  i64 iOff = *piOffset;
  int i;
  int rc = SQLITE_OK;

  *ppOut = pNew = vdbeMergeEngineNew(nPMA);
  if( pNew==0 ) rc = SQLITE_NOMEM;

  for(i=0; i<nPMA && rc==SQLITE_OK; i++){
    i64 nDummy;
    PmaReader *pReadr = &pNew->aReadr[i];
    rc = vdbePmaReaderInit(pTask, &pTask->file, iOff, pReadr, &nDummy);
    iOff = pReadr->iEof;
  }

  if( rc!=SQLITE_OK ){
    vdbeMergeEngineFree(pNew);
    *ppOut = 0;
  }
  *piOffset = iOff;
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121

  for(i=1; i<nDepth; i++){
    nDiv = nDiv * SORTER_MAX_MERGE_COUNT;
  }

  for(i=1; i<nDepth && rc==SQLITE_OK; i++){
    int iIter = (iSeq / nDiv) % SORTER_MAX_MERGE_COUNT;
    PmaReader *pIter = &p->aIter[iIter];

    if( pIter->pIncr==0 ){
      MergeEngine *pNew = vdbeMergeEngineNew(SORTER_MAX_MERGE_COUNT);
      if( pNew==0 ){
        rc = SQLITE_NOMEM;
      }else{
        rc = vdbeIncrNew(pTask, pNew, &pIter->pIncr);
      }
    }
    if( rc==SQLITE_OK ){
      p = pIter->pIncr->pMerger;
      nDiv = nDiv / SORTER_MAX_MERGE_COUNT;
    }
  }

  if( rc==SQLITE_OK ){
    p->aIter[iSeq % SORTER_MAX_MERGE_COUNT].pIncr = pIncr;
  }else{
    vdbeIncrFree(pIncr);
  }
  return rc;
}

/*







|

|




|



|





|







2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121

  for(i=1; i<nDepth; i++){
    nDiv = nDiv * SORTER_MAX_MERGE_COUNT;
  }

  for(i=1; i<nDepth && rc==SQLITE_OK; i++){
    int iIter = (iSeq / nDiv) % SORTER_MAX_MERGE_COUNT;
    PmaReader *pReadr = &p->aReadr[iIter];

    if( pReadr->pIncr==0 ){
      MergeEngine *pNew = vdbeMergeEngineNew(SORTER_MAX_MERGE_COUNT);
      if( pNew==0 ){
        rc = SQLITE_NOMEM;
      }else{
        rc = vdbeIncrNew(pTask, pNew, &pReadr->pIncr);
      }
    }
    if( rc==SQLITE_OK ){
      p = pReadr->pIncr->pMerger;
      nDiv = nDiv / SORTER_MAX_MERGE_COUNT;
    }
  }

  if( rc==SQLITE_OK ){
    p->aReadr[iSeq % SORTER_MAX_MERGE_COUNT].pIncr = pIncr;
  }else{
    vdbeIncrFree(pIncr);
  }
  return rc;
}

/*
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
          }
        }
      }

      if( rc==SQLITE_OK ){
#if SQLITE_MAX_WORKER_THREADS>0
        if( pMain!=0 ){
          rc = vdbeIncrNew(pTask, pRoot, &pMain->aIter[iTask].pIncr);
        }else
#endif
        {
          assert( pMain==0 );
          pMain = pRoot;
        }
      }else{







|







2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
          }
        }
      }

      if( rc==SQLITE_OK ){
#if SQLITE_MAX_WORKER_THREADS>0
        if( pMain!=0 ){
          rc = vdbeIncrNew(pTask, pRoot, &pMain->aReadr[iTask].pIncr);
        }else
#endif
        {
          assert( pMain==0 );
          pMain = pRoot;
        }
      }else{
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262

  rc = vdbeSorterMergeTreeBuild(pSorter, &pMain);
  if( rc==SQLITE_OK ){
#if SQLITE_MAX_WORKER_THREADS
    assert( pSorter->bUseThreads==0 || pSorter->nTask>1 );
    if( pSorter->bUseThreads ){
      int iTask;
      PmaReader *pIter;
      SortSubtask *pLast = &pSorter->aTask[pSorter->nTask-1];
      rc = vdbeSortAllocUnpacked(pLast);
      if( rc==SQLITE_OK ){
        pIter = (PmaReader*)sqlite3DbMallocZero(db, sizeof(PmaReader));
        pSorter->pReader = pIter;
        if( pIter==0 ) rc = SQLITE_NOMEM;
      }
      if( rc==SQLITE_OK ){
        rc = vdbeIncrNew(pLast, pMain, &pIter->pIncr);
        if( rc==SQLITE_OK ){
          vdbeIncrSetThreads(pIter->pIncr);
          for(iTask=0; iTask<(pSorter->nTask-1); iTask++){
            IncrMerger *pIncr;
            if( (pIncr = pMain->aIter[iTask].pIncr) ){
              vdbeIncrSetThreads(pIncr);
              assert( pIncr->pTask!=pLast );
            }
          }
          for(iTask=0; rc==SQLITE_OK && iTask<pSorter->nTask; iTask++){
            PmaReader *p = &pMain->aIter[iTask];
            assert( p->pIncr==0 || p->pIncr->pTask==&pSorter->aTask[iTask] );
            if( p->pIncr ){ 
              if( iTask==pSorter->nTask-1 ){
                rc = vdbePmaReaderIncrInit(p, INCRINIT_TASK);
              }else{
                rc = vdbePmaReaderBgIncrInit(p);
              }
            }
          }
        }
        pMain = 0;
      }
      if( rc==SQLITE_OK ){
        rc = vdbePmaReaderIncrInit(pIter, INCRINIT_ROOT);
      }
    }else
#endif
    {
      rc = vdbeIncrInitMerger(pTask0, pMain, INCRINIT_NORMAL);
      pSorter->pMerger = pMain;
      pMain = 0;







|



|
|
|


|

|


|





|













|







2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262

  rc = vdbeSorterMergeTreeBuild(pSorter, &pMain);
  if( rc==SQLITE_OK ){
#if SQLITE_MAX_WORKER_THREADS
    assert( pSorter->bUseThreads==0 || pSorter->nTask>1 );
    if( pSorter->bUseThreads ){
      int iTask;
      PmaReader *pReadr;
      SortSubtask *pLast = &pSorter->aTask[pSorter->nTask-1];
      rc = vdbeSortAllocUnpacked(pLast);
      if( rc==SQLITE_OK ){
        pReadr = (PmaReader*)sqlite3DbMallocZero(db, sizeof(PmaReader));
        pSorter->pReader = pReadr;
        if( pReadr==0 ) rc = SQLITE_NOMEM;
      }
      if( rc==SQLITE_OK ){
        rc = vdbeIncrNew(pLast, pMain, &pReadr->pIncr);
        if( rc==SQLITE_OK ){
          vdbeIncrSetThreads(pReadr->pIncr);
          for(iTask=0; iTask<(pSorter->nTask-1); iTask++){
            IncrMerger *pIncr;
            if( (pIncr = pMain->aReadr[iTask].pIncr) ){
              vdbeIncrSetThreads(pIncr);
              assert( pIncr->pTask!=pLast );
            }
          }
          for(iTask=0; rc==SQLITE_OK && iTask<pSorter->nTask; iTask++){
            PmaReader *p = &pMain->aReadr[iTask];
            assert( p->pIncr==0 || p->pIncr->pTask==&pSorter->aTask[iTask] );
            if( p->pIncr ){ 
              if( iTask==pSorter->nTask-1 ){
                rc = vdbePmaReaderIncrInit(p, INCRINIT_TASK);
              }else{
                rc = vdbePmaReaderBgIncrInit(p);
              }
            }
          }
        }
        pMain = 0;
      }
      if( rc==SQLITE_OK ){
        rc = vdbePmaReaderIncrInit(pReadr, INCRINIT_ROOT);
      }
    }else
#endif
    {
      rc = vdbeIncrInitMerger(pTask0, pMain, INCRINIT_NORMAL);
      pSorter->pMerger = pMain;
      pMain = 0;
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
    PmaReader *pReader;
#if SQLITE_MAX_WORKER_THREADS>0
    if( pSorter->bUseThreads ){
      pReader = pSorter->pReader;
    }else
#endif
    /*if( !pSorter->bUseThreads )*/{
      pReader = &pSorter->pMerger->aIter[pSorter->pMerger->aTree[1]];
    }
    *pnKey = pReader->nKey;
    pKey = pReader->aKey;
  }else{
    *pnKey = pSorter->list.pList->nVal;
    pKey = SRVAL(pSorter->list.pList);
  }







|







2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
    PmaReader *pReader;
#if SQLITE_MAX_WORKER_THREADS>0
    if( pSorter->bUseThreads ){
      pReader = pSorter->pReader;
    }else
#endif
    /*if( !pSorter->bUseThreads )*/{
      pReader = &pSorter->pMerger->aReadr[pSorter->pMerger->aTree[1]];
    }
    *pnKey = pReader->nKey;
    pKey = pReader->aKey;
  }else{
    *pnKey = pSorter->list.pList->nVal;
    pKey = SRVAL(pSorter->list.pList);
  }