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

Overview
Comment:Testing finds no cases where separator runs are faster than embedded btrees. So discard the separator runs related code.
Downloads: Tarball | ZIP archive
Timelines: family | ancestors | descendants | both | embedded-btree
Files: files | file ages | folders
SHA1: afcbe561ffbea8572cf6d1a908430bb77b156888
User & Date: dan 2012-06-27 16:02:23.686
Context
2012-06-27
16:10
Remove a broken assert from lsm_sorted.c. Leaf check-in: 50e87e23bb user: dan tags: embedded-btree
16:02
Testing finds no cases where separator runs are faster than embedded btrees. So discard the separator runs related code. check-in: afcbe561ff user: dan tags: embedded-btree
10:22
Ensure that "expensive-assert" code only runs when LSM_DEBUG_EXPENSIVE is defined. check-in: 2be9bac408 user: dan tags: embedded-btree
Changes
Unified Diff Ignore Whitespace Patch
Changes to src/lsmInt.h.
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
typedef struct Mempool Mempool;
typedef struct MetaPage MetaPage;
typedef struct MultiCursor MultiCursor;
typedef struct Page Page;
typedef struct Segment Segment;
typedef struct SegmentMerger SegmentMerger;
typedef struct Snapshot Snapshot;
typedef struct SortedRun SortedRun;
typedef struct TransMark TransMark;
typedef struct Tree Tree;
typedef struct TreeMark TreeMark;
typedef struct TreeVersion TreeVersion;
typedef struct TreeCursor TreeCursor;
typedef struct Merge Merge;
typedef struct MergeInput MergeInput;







<







68
69
70
71
72
73
74

75
76
77
78
79
80
81
typedef struct Mempool Mempool;
typedef struct MetaPage MetaPage;
typedef struct MultiCursor MultiCursor;
typedef struct Page Page;
typedef struct Segment Segment;
typedef struct SegmentMerger SegmentMerger;
typedef struct Snapshot Snapshot;

typedef struct TransMark TransMark;
typedef struct Tree Tree;
typedef struct TreeMark TreeMark;
typedef struct TreeVersion TreeVersion;
typedef struct TreeCursor TreeCursor;
typedef struct Merge Merge;
typedef struct MergeInput MergeInput;
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
  void *pLogCtx;

  /* Work done notification callback */
  void (*xWork)(lsm_db *, void *);
  void *pWorkCtx;
};

struct SortedRun {
  int iFirst;                     /* First page of this run */
  int iLast;                      /* Last page of this run */
  Pgno iRoot;                     /* Root page number (if any) */
  int nSize;                      /* Size of this run in pages */
};

struct Segment {
  SortedRun run;                  /* Main array */
  SortedRun sep;                  /* If sep.iFirst!=0, the separators array */
};

/*
** iSplitTopic/pSplitKey/nSplitKey:
**   If nRight>0, this buffer contains a copy of the largest key that has
**   already been written to the left-hand-side of the level.
*/
struct Level {
  Segment lhs;                    /* Left-hand (main) segment */







|






<
<
<
<
<







199
200
201
202
203
204
205
206
207
208
209
210
211
212





213
214
215
216
217
218
219
  void *pLogCtx;

  /* Work done notification callback */
  void (*xWork)(lsm_db *, void *);
  void *pWorkCtx;
};

struct Segment {
  int iFirst;                     /* First page of this run */
  int iLast;                      /* Last page of this run */
  Pgno iRoot;                     /* Root page number (if any) */
  int nSize;                      /* Size of this run in pages */
};






/*
** iSplitTopic/pSplitKey/nSplitKey:
**   If nRight>0, this buffer contains a copy of the largest key that has
**   already been written to the left-hand-side of the level.
*/
struct Level {
  Segment lhs;                    /* Left-hand (main) segment */
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
**
** It is assumed that code that uses an instance of this structure has
** access to the associated Level struct.
**
** bHierReadonly:
**   True if the b-tree hierarchy is currently read-only.
**
** aiOutputOff:
**   The byte offset to write to next within the last page of the output
**   segments main run (aiOutputOff[0]) or separators run (aiOutputOff[1]).
**   If either page is read-only, then the associated aiOutputOff[] entry
**   is set to a negative value.
*/
struct Merge {
  int nInput;                     /* Number of input runs being merged */
  MergeInput *aInput;             /* Array nInput entries in size */
  int nSkip;                      /* Number of separators entries to skip */
  int aiOutputOff[2];             /* Write offsets on run output pages */
  int abHierReadonly[2];          /* True if b-tree heirarchies are read-only */
};
struct MergeInput {
  Pgno iPg;                       /* Page on which next input is stored */
  int iCell;                      /* Cell containing next input to merge */
};

/* 







|
|
<
<
|





|
|







234
235
236
237
238
239
240
241
242


243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
**
** It is assumed that code that uses an instance of this structure has
** access to the associated Level struct.
**
** bHierReadonly:
**   True if the b-tree hierarchy is currently read-only.
**
** iOutputOff:
**   The byte offset to write to next within the last page of the 


**   output segment.
*/
struct Merge {
  int nInput;                     /* Number of input runs being merged */
  MergeInput *aInput;             /* Array nInput entries in size */
  int nSkip;                      /* Number of separators entries to skip */
  int iOutputOff;                 /* Write offset on output page */
  int bHierReadonly;              /* True if b-tree heirarchies are read-only */
};
struct MergeInput {
  Pgno iPg;                       /* Page on which next input is stored */
  int iCell;                      /* Cell containing next input to merge */
};

/* 
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
394
395
396
397
398
399
400
401
402
403

int lsmFsPageSize(FileSystem *);
void lsmFsSetPageSize(FileSystem *, int);

int lsmFsFileid(lsm_db *pDb, void **ppId, int *pnId);

/* Creating, populating, gobbling and deleting sorted runs. */
int lsmFsPhantom(FileSystem *, SortedRun *);
void lsmFsPhantomFree(FileSystem *pFS);
void lsmFsGobble(Snapshot *, SortedRun *, Page *);
int lsmFsSortedDelete(FileSystem *, Snapshot *, int, SortedRun *);
int lsmFsSortedFinish(FileSystem *, SortedRun *);
int lsmFsSortedAppend(FileSystem *, Snapshot *, SortedRun *, Page **);
int lsmFsPhantomMaterialize(FileSystem *, Snapshot *, SortedRun *);

/* Functions to retrieve the lsm_env pointer from a FileSystem or Page object */
lsm_env *lsmFsEnv(FileSystem *);
lsm_env *lsmPageEnv(Page *);
FileSystem *lsmPageFS(Page *);

int lsmFsSectorSize(FileSystem *);

void lsmSortedSplitkey(lsm_db *, Level *, int *);
int lsmFsSetupAppendList(lsm_db *db);

/* Reading sorted run content. */
int lsmFsDbPageGet(FileSystem *, Pgno, Page **);
int lsmFsDbPageNext(SortedRun *, Page *, int eDir, Page **);

int lsmFsPageWrite(Page *);
u8 *lsmFsPageData(Page *, int *);
int lsmFsPageRelease(Page *);
int lsmFsPagePersist(Page *);
void lsmFsPageRef(Page *);
Pgno lsmFsPageNumber(Page *);







<
<
|
|
|
|
|













|







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

int lsmFsPageSize(FileSystem *);
void lsmFsSetPageSize(FileSystem *, int);

int lsmFsFileid(lsm_db *pDb, void **ppId, int *pnId);

/* Creating, populating, gobbling and deleting sorted runs. */


void lsmFsGobble(Snapshot *, Segment *, Page *);
int lsmFsSortedDelete(FileSystem *, Snapshot *, int, Segment *);
int lsmFsSortedFinish(FileSystem *, Segment *);
int lsmFsSortedAppend(FileSystem *, Snapshot *, Segment *, Page **);
int lsmFsPhantomMaterialize(FileSystem *, Snapshot *, Segment *);

/* Functions to retrieve the lsm_env pointer from a FileSystem or Page object */
lsm_env *lsmFsEnv(FileSystem *);
lsm_env *lsmPageEnv(Page *);
FileSystem *lsmPageFS(Page *);

int lsmFsSectorSize(FileSystem *);

void lsmSortedSplitkey(lsm_db *, Level *, int *);
int lsmFsSetupAppendList(lsm_db *db);

/* Reading sorted run content. */
int lsmFsDbPageGet(FileSystem *, Pgno, Page **);
int lsmFsDbPageNext(Segment *, Page *, int eDir, Page **);

int lsmFsPageWrite(Page *);
u8 *lsmFsPageData(Page *, int *);
int lsmFsPageRelease(Page *);
int lsmFsPagePersist(Page *);
void lsmFsPageRef(Page *);
Pgno lsmFsPageNumber(Page *);
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
int lsmMCursorType(MultiCursor *, int *);
lsm_db *lsmMCursorDb(MultiCursor *);

int lsmSaveCursors(lsm_db *pDb);
int lsmRestoreCursors(lsm_db *pDb);

void lsmSortedDumpStructure(lsm_db *pDb, Snapshot *, int, int, const char *);
void lsmFsDumpBlockmap(lsm_db *, SortedRun *);
void lsmFsDumpBlocklists(lsm_db *);


void lsmPutU32(u8 *, u32);
u32 lsmGetU32(u8 *);

/*







<







461
462
463
464
465
466
467

468
469
470
471
472
473
474
int lsmMCursorType(MultiCursor *, int *);
lsm_db *lsmMCursorDb(MultiCursor *);

int lsmSaveCursors(lsm_db *pDb);
int lsmRestoreCursors(lsm_db *pDb);

void lsmSortedDumpStructure(lsm_db *pDb, Snapshot *, int, int, const char *);

void lsmFsDumpBlocklists(lsm_db *);


void lsmPutU32(u8 *, u32);
u32 lsmGetU32(u8 *);

/*
Changes to src/lsm_ckpt.c.
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
**   The checksum:
**
**     1. Checksum value 1.
**     2. Checksum value 2.
**
** In the above, a segment record is:
**
**     1. First page of main array,
**     2. Last page of main array,
**     3. Size of main array in pages,
**     4. First page of separators array (or 0),
**     5. Last page of separators array (or 0),
**     6. Root page of separators array (or 0).
*/

/*
** OVERSIZED CHECKPOINT BLOBS:
**
** There are two slots allocated for checkpoints at the start of each 
** database file. Each are 4096 bytes in size, so may accommodate 







|
|
|
|
<
<







62
63
64
65
66
67
68
69
70
71
72


73
74
75
76
77
78
79
**   The checksum:
**
**     1. Checksum value 1.
**     2. Checksum value 2.
**
** In the above, a segment record is:
**
**     1. First page of array,
**     2. Last page of array,
**     3. Root page of array (or 0),
**     4. Size of array in pages,


*/

/*
** OVERSIZED CHECKPOINT BLOBS:
**
** There are two slots allocated for checkpoints at the start of each 
** database file. Each are 4096 bytes in size, so may accommodate 
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
  Segment *pSeg, 
  CkptBuffer *p, 
  int *piOut, 
  int *pRc
){
  int iOut = *piOut;

  ckptSetValue(p, iOut++, pSeg->run.iFirst, pRc);
  ckptSetValue(p, iOut++, pSeg->run.iLast, pRc);
  ckptSetValue(p, iOut++, pSeg->run.iRoot, pRc);
  ckptSetValue(p, iOut++, pSeg->run.nSize, pRc);
  if( segmentHasSeparators(pSeg) ){
    ckptSetValue(p, iOut++, pSeg->sep.iFirst, pRc);
    ckptSetValue(p, iOut++, pSeg->sep.iLast, pRc);
    ckptSetValue(p, iOut++, pSeg->sep.iRoot, pRc);
    ckptSetValue(p, iOut++, pSeg->sep.nSize, pRc);
  }else{
    ckptSetValue(p, iOut++, 0, pRc);
    ckptSetValue(p, iOut++, 0, pRc);
    ckptSetValue(p, iOut++, 0, pRc);
    ckptSetValue(p, iOut++, 0, pRc);
  }

  *piOut = iOut;
}

static void ckptExportLevel(
  Level *pLevel,
  CkptBuffer *p,







|
|
|
|
<
<
<
<
<
<
<
<
<
<
<







217
218
219
220
221
222
223
224
225
226
227











228
229
230
231
232
233
234
  Segment *pSeg, 
  CkptBuffer *p, 
  int *piOut, 
  int *pRc
){
  int iOut = *piOut;

  ckptSetValue(p, iOut++, pSeg->iFirst, pRc);
  ckptSetValue(p, iOut++, pSeg->iLast, pRc);
  ckptSetValue(p, iOut++, pSeg->iRoot, pRc);
  ckptSetValue(p, iOut++, pSeg->nSize, pRc);












  *piOut = iOut;
}

static void ckptExportLevel(
  Level *pLevel,
  CkptBuffer *p,
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
static void ckptNewSegment(
  u32 *aIn,
  int *piIn,
  Segment *pSegment               /* Populate this structure */
){
  int iIn = *piIn;

  assert( pSegment->run.iFirst==0 && pSegment->run.iLast==0 );
  assert( pSegment->run.nSize==0 && pSegment->run.iRoot==0 );
  assert( pSegment->sep.iFirst==0 && pSegment->sep.iLast==0 );
  assert( pSegment->sep.nSize==0 && pSegment->sep.iRoot==0 );

  pSegment->run.iFirst = aIn[iIn++];
  pSegment->run.iLast = aIn[iIn++];
  pSegment->run.iRoot = aIn[iIn++];
  pSegment->run.nSize = aIn[iIn++];
  pSegment->sep.iFirst = aIn[iIn++];
  pSegment->sep.iLast = aIn[iIn++];
  pSegment->sep.iRoot = aIn[iIn++];
  pSegment->sep.nSize = aIn[iIn++];
  if( pSegment->sep.iFirst && pSegment->sep.nSize==0 ) pSegment->sep.nSize = 1;

  *piIn = iIn;
}

static int ckptSetupMerge(lsm_db *pDb, u32 *aInt, int *piIn, Level *pLevel){
  Merge *pMerge;                  /* Allocated Merge object */
  int nInput;                     /* Number of input segments in merge */







|
|
<
<
<
|
|
|
|
<
<
<
<
<







373
374
375
376
377
378
379
380
381



382
383
384
385





386
387
388
389
390
391
392
static void ckptNewSegment(
  u32 *aIn,
  int *piIn,
  Segment *pSegment               /* Populate this structure */
){
  int iIn = *piIn;

  assert( pSegment->iFirst==0 && pSegment->iLast==0 );
  assert( pSegment->nSize==0 && pSegment->iRoot==0 );



  pSegment->iFirst = aIn[iIn++];
  pSegment->iLast = aIn[iIn++];
  pSegment->iRoot = aIn[iIn++];
  pSegment->nSize = aIn[iIn++];






  *piIn = iIn;
}

static int ckptSetupMerge(lsm_db *pDb, u32 *aInt, int *piIn, Level *pLevel){
  Merge *pMerge;                  /* Allocated Merge object */
  int nInput;                     /* Number of input segments in merge */
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
  pMerge = (Merge *)lsmMallocZero(pDb->pEnv, nByte);
  if( !pMerge ) return LSM_NOMEM_BKPT;
  pLevel->pMerge = pMerge;

  /* Populate the Merge object. */
  pMerge->aInput = (MergeInput *)&pMerge[1];
  pMerge->nInput = nInput;
  pMerge->aiOutputOff[0] = -1;
  pMerge->aiOutputOff[1] = -1;
  pMerge->nSkip = (int)aInt[iIn++];
  pMerge->abHierReadonly[0] = 1;
  pMerge->abHierReadonly[1] = 1;
  for(i=0; i<nInput; i++){
    pMerge->aInput[i].iPg = (Pgno)aInt[iIn++];
    pMerge->aInput[i].iCell = (int)aInt[iIn++];
  }

  /* Set *piIn and return LSM_OK. */
  *piIn = iIn;







|
|

<
<







400
401
402
403
404
405
406
407
408
409


410
411
412
413
414
415
416
  pMerge = (Merge *)lsmMallocZero(pDb->pEnv, nByte);
  if( !pMerge ) return LSM_NOMEM_BKPT;
  pLevel->pMerge = pMerge;

  /* Populate the Merge object. */
  pMerge->aInput = (MergeInput *)&pMerge[1];
  pMerge->nInput = nInput;
  pMerge->iOutputOff = -1;
  pMerge->bHierReadonly = 1;
  pMerge->nSkip = (int)aInt[iIn++];


  for(i=0; i<nInput; i++){
    pMerge->aInput[i].iPg = (Pgno)aInt[iIn++];
    pMerge->aInput[i].iCell = (int)aInt[iIn++];
  }

  /* Set *piIn and return LSM_OK. */
  *piIn = iIn;
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
  return rc;
}

static int ckptImport(lsm_db *pDb, void *pCkpt, int nInt, int *pRc){
  int ret = 0;
  if( *pRc==LSM_OK ){
    Snapshot *pSnap = pDb->pWorker;
    FileSystem *pFS = pDb->pFS;
    u32 cksum[2] = {0, 0};
    u32 *aInt = (u32 *)pCkpt;

    lsmChecksumBytes((u8 *)aInt, sizeof(u32)*(nInt-2), 0, cksum);
    if( LSM_LITTLE_ENDIAN ){
      int i;
      for(i=0; i<nInt; i++) aInt[i] = BYTESWAP32(aInt[i]);







<







477
478
479
480
481
482
483

484
485
486
487
488
489
490
  return rc;
}

static int ckptImport(lsm_db *pDb, void *pCkpt, int nInt, int *pRc){
  int ret = 0;
  if( *pRc==LSM_OK ){
    Snapshot *pSnap = pDb->pWorker;

    u32 cksum[2] = {0, 0};
    u32 *aInt = (u32 *)pCkpt;

    lsmChecksumBytes((u8 *)aInt, sizeof(u32)*(nInt-2), 0, cksum);
    if( LSM_LITTLE_ENDIAN ){
      int i;
      for(i=0; i<nInt; i++) aInt[i] = BYTESWAP32(aInt[i]);
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
int lsmCheckpointLevels(
  lsm_db *pDb,                    /* Database handle */
  int *pnHdrLevel,                /* OUT: Levels to write to db header */
  void **paVal,                   /* OUT: Pointer to LEVELS blob */
  int *pnVal                      /* OUT: Size of LEVELS blob in bytes */
){
  int rc = LSM_OK;                /* Return code */
  const int SEGMENT_SIZE = 8;     /* Size of a checkpoint segment record */
  Level *p;                       /* Used to iterate through levels */
  int nFree;                      /* Free integers remaining in db header */
  int nHdr = 0;                   /* Number of levels stored in db header */
  int nLevels = 0;                /* Number of levels stored in LEVELS */
 
  /* Number of free integers - 1024 less those used by the checkpoint header,
  ** less the 4 used for the log-pointer, less the 3 used for the 







|







663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
int lsmCheckpointLevels(
  lsm_db *pDb,                    /* Database handle */
  int *pnHdrLevel,                /* OUT: Levels to write to db header */
  void **paVal,                   /* OUT: Pointer to LEVELS blob */
  int *pnVal                      /* OUT: Size of LEVELS blob in bytes */
){
  int rc = LSM_OK;                /* Return code */
  const int SEGMENT_SIZE = 4;     /* Size of a checkpoint segment record */
  Level *p;                       /* Used to iterate through levels */
  int nFree;                      /* Free integers remaining in db header */
  int nHdr = 0;                   /* Number of levels stored in db header */
  int nLevels = 0;                /* Number of levels stored in LEVELS */
 
  /* Number of free integers - 1024 less those used by the checkpoint header,
  ** less the 4 used for the log-pointer, less the 3 used for the 
Changes to src/lsm_file.c.
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
*/
#include "lsmInt.h"

#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

/* 
** A "phantom" run under construction.
**
** Phantom runs are constructed entirely in memory, then written out to 
** disk. This is distinct from normal runs, for which each page is written
** to disk as soon as it is completely populated. 
**
** They are used when the in-memory tree is flushed to disk. In this case, 
** the main run is written directly to disk and the separators run 
** accumulated in memory as a phantom run and flushed to disk after the main
** run is completed. This allows the separators run to immediately follow
** the main run in the file - making the entire flush operation a single
** contiguous write.
**
** Before they are flushed to disk, the pages of phantom runs do not have
** page numbers. This means it is not possible to obtain pointers to them.
** In practice, this means that when creating a separators run, a phantom
** run is used to accumulate and write all leaf pages to disk, then a
** second pass is made to populate and append the b-tree hierarchy pages.
*/
typedef struct PhantomRun PhantomRun;
struct PhantomRun {
  SortedRun *pRun;                /* Accompanying SortedRun object */
  int nPhantom;                   /* Number of pages in run */
  int bRunFinished;               /* True if the associated run is finished */
  Page *pFirst;                   /* First page in phantom run */
  Page *pLast;                    /* Current last page in phantom run */
};

/*
** Maximum number of pages allowed to accumulate in memory when constructing
** a phantom run. If this limit is exceeded, the phantom run is flushed to
** disk even if it is not finished.
*/
#define FS_MAX_PHANTOM_PAGES 32


/*
** File-system object. Each database connection allocates a single instance
** of the following structure. It is used for all access to the database and
** log files.
**
** pLruFirst, pLruLast:







<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<







86
87
88
89
90
91
92




































93
94
95
96
97
98
99
*/
#include "lsmInt.h"

#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>






































/*
** File-system object. Each database connection allocates a single instance
** of the following structure. It is used for all access to the database and
** log files.
**
** pLruFirst, pLruLast:
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
struct FileSystem {
  lsm_db *pDb;                    /* Database handle that owns this object */
  lsm_env *pEnv;                  /* Environment pointer */
  char *zDb;                      /* Database file name */
  int nMetasize;                  /* Size of meta pages in bytes */
  int nPagesize;                  /* Database page-size in bytes */
  int nBlocksize;                 /* Database block-size in bytes */
  PhantomRun phantom;             /* Phantom run currently under construction */

  /* r/w file descriptors for both files. */
  lsm_file *fdDb;                 /* Database file */
  lsm_file *fdLog;                /* Log file */

  /* mmap() mode things */
  int bUseMmap;                   /* True to use mmap() to access db file */







<







112
113
114
115
116
117
118

119
120
121
122
123
124
125
struct FileSystem {
  lsm_db *pDb;                    /* Database handle that owns this object */
  lsm_env *pEnv;                  /* Environment pointer */
  char *zDb;                      /* Database file name */
  int nMetasize;                  /* Size of meta pages in bytes */
  int nPagesize;                  /* Database page-size in bytes */
  int nBlocksize;                 /* Database block-size in bytes */


  /* r/w file descriptors for both files. */
  lsm_file *fdDb;                 /* Database file */
  lsm_file *fdLog;                /* Log file */

  /* mmap() mode things */
  int bUseMmap;                   /* True to use mmap() to access db file */
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230

/*
** Number of pgsz byte pages omitted from the start of block 1. The start
** of block 1 contains two 4096 byte meta pages (8192 bytes in total).
*/
#define BLOCK1_HDR_SIZE(pgsz)  LSM_MAX(1, 8192/(pgsz))

/*
** Return true if the SortedRun passed as the second argument is a phantom
** run currently being constructed by FileSystem object pFS.
*/
#define isPhantom(pFS, pSorted) ((pSorted) && (pFS)->phantom.pRun==(pSorted))

/*
** Wrappers around the VFS methods of the lsm_env object:
**
**     lsmEnvOpen()
**     lsmEnvRead()
**     lsmEnvWrite()







<
<
<
<
<







175
176
177
178
179
180
181





182
183
184
185
186
187
188

/*
** Number of pgsz byte pages omitted from the start of block 1. The start
** of block 1 contains two 4096 byte meta pages (8192 bytes in total).
*/
#define BLOCK1_HDR_SIZE(pgsz)  LSM_MAX(1, 8192/(pgsz))







/*
** Wrappers around the VFS methods of the lsm_env object:
**
**     lsmEnvOpen()
**     lsmEnvRead()
**     lsmEnvWrite()
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
*/
void lsmFsClose(FileSystem *pFS){
  if( pFS ){
    Page *pPg;
    lsm_env *pEnv = pFS->pEnv;

    assert( pFS->nOut==0 );

    pPg = pFS->pLruFirst;
    while( pPg ){
      Page *pNext = pPg->pLruNext;
      if( pPg->flags & PAGE_FREE ) lsmFree(pEnv, pPg->aData);
      lsmFree(pEnv, pPg);
      pPg = pNext;
    }







<







372
373
374
375
376
377
378

379
380
381
382
383
384
385
*/
void lsmFsClose(FileSystem *pFS){
  if( pFS ){
    Page *pPg;
    lsm_env *pEnv = pFS->pEnv;

    assert( pFS->nOut==0 );

    pPg = pFS->pLruFirst;
    while( pPg ){
      Page *pNext = pPg->pLruNext;
      if( pPg->flags & PAGE_FREE ) lsmFree(pEnv, pPg->aData);
      lsmFree(pEnv, pPg);
      pPg = pNext;
    }
597
598
599
600
601
602
603



604
605
606
607
608
609
610
    pPg->pLruPrev->pLruNext = pPg;
  }else{
    pFS->pLruFirst = pPg;
  }
  pFS->pLruLast = pPg;
}




static void fsPageRemoveFromHash(FileSystem *pFS, Page *pPg){
  int iHash;
  Page **pp;

  iHash = fsHashKey(pFS->nHash, pPg->iPg);
  for(pp=&pFS->apHash[iHash]; *pp!=pPg; pp=&(*pp)->pHashNext);
  *pp = pPg->pHashNext;







>
>
>







554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
    pPg->pLruPrev->pLruNext = pPg;
  }else{
    pFS->pLruFirst = pPg;
  }
  pFS->pLruLast = pPg;
}

/*
** Remove page pPg from the hash table.
*/
static void fsPageRemoveFromHash(FileSystem *pFS, Page *pPg){
  int iHash;
  Page **pp;

  iHash = fsHashKey(pFS->nHash, pPg->iPg);
  for(pp=&pFS->apHash[iHash]; *pp!=pPg; pp=&(*pp)->pHashNext);
  *pp = pPg->pHashNext;
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
    *piNext = fsPageToBlock(pFS, lsmGetU32(&pLast->aData[pFS->nPagesize-4]));
    lsmFsPageRelease(pLast);
  }
  return rc;
}

static int fsRunEndsBetween(
  SortedRun *pRun, 
  SortedRun *pIgnore, 
  int iFirst, 
  int iLast
){
  return (pRun!=pIgnore && (
        (pRun->iFirst>=iFirst && pRun->iFirst<=iLast)
     || (pRun->iLast>=iFirst && pRun->iLast<=iLast)
  ));
}

static int fsLevelEndsBetween(
  Level *pLevel, 
  SortedRun *pIgnore, 
  int iFirst, 
  int iLast
){
  int i;

  if( fsRunEndsBetween(&pLevel->lhs.run, pIgnore, iFirst, iLast)
   || fsRunEndsBetween(&pLevel->lhs.sep, pIgnore, iFirst, iLast)
  ){
    return 1;
  }
  for(i=0; i<pLevel->nRight; i++){
    if( fsRunEndsBetween(&pLevel->aRhs[i].run, pIgnore, iFirst, iLast)
     || fsRunEndsBetween(&pLevel->aRhs[i].sep, pIgnore, iFirst, iLast)
    ){
      return 1;
    }
  }

  return 0;
}

static int fsFreeBlock(
  FileSystem *pFS, 
  Snapshot *pSnapshot, 
  SortedRun *pIgnore,             /* Ignore this run when searching */
  int iBlk
){
  int rc = LSM_OK;                /* Return code */
  int iFirst;                     /* First page on block iBlk */
  int iLast;                      /* Last page on block iBlk */
  int i;                          /* Used to iterate through append points */
  Level *pLevel;                  /* Used to iterate through levels */







|
|











|





|
<
<



|
<
<










|







739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
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
    *piNext = fsPageToBlock(pFS, lsmGetU32(&pLast->aData[pFS->nPagesize-4]));
    lsmFsPageRelease(pLast);
  }
  return rc;
}

static int fsRunEndsBetween(
  Segment *pRun, 
  Segment *pIgnore, 
  int iFirst, 
  int iLast
){
  return (pRun!=pIgnore && (
        (pRun->iFirst>=iFirst && pRun->iFirst<=iLast)
     || (pRun->iLast>=iFirst && pRun->iLast<=iLast)
  ));
}

static int fsLevelEndsBetween(
  Level *pLevel, 
  Segment *pIgnore, 
  int iFirst, 
  int iLast
){
  int i;

  if( fsRunEndsBetween(&pLevel->lhs, pIgnore, iFirst, iLast) ){


    return 1;
  }
  for(i=0; i<pLevel->nRight; i++){
    if( fsRunEndsBetween(&pLevel->aRhs[i], pIgnore, iFirst, iLast) ){


      return 1;
    }
  }

  return 0;
}

static int fsFreeBlock(
  FileSystem *pFS, 
  Snapshot *pSnapshot, 
  Segment *pIgnore,             /* Ignore this run when searching */
  int iBlk
){
  int rc = LSM_OK;                /* Return code */
  int iFirst;                     /* First page on block iBlk */
  int iLast;                      /* Last page on block iBlk */
  int i;                          /* Used to iterate through append points */
  Level *pLevel;                  /* Used to iterate through levels */
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875

/*
** Delete or otherwise recycle the blocks currently occupied by run pDel.
*/
int lsmFsSortedDelete(
  FileSystem *pFS, 
  Snapshot *pSnapshot,
  int bZero,                      /* True to zero the SortedRun structure */
  SortedRun *pDel
){
  if( pDel->iFirst ){
    int rc = LSM_OK;

    int iBlk;
    int iLastBlk;








|
|







816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831

/*
** Delete or otherwise recycle the blocks currently occupied by run pDel.
*/
int lsmFsSortedDelete(
  FileSystem *pFS, 
  Snapshot *pSnapshot,
  int bZero,                      /* True to zero the Segment structure */
  Segment *pDel
){
  if( pDel->iFirst ){
    int rc = LSM_OK;

    int iBlk;
    int iLastBlk;

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
      }else if( bZero==0 && pDel->iLast!=fsLastPageOnBlock(pFS, iLastBlk) ){
        break;
      }
      rc = fsFreeBlock(pFS, pSnapshot, pDel, iBlk);
      iBlk = iNext;
    }

    if( bZero ) memset(pDel, 0, sizeof(SortedRun));
  }
  return LSM_OK;
}

/*
** The pager reference passed as the only argument must refer to a sorted
** file page (not a log or meta page). This call indicates that the argument
** page is now the first page in its sorted file - all previous pages may
** be considered free.
*/
void lsmFsGobble(
  Snapshot *pSnapshot,
  SortedRun *pRun, 
  Page *pPg
){
  FileSystem *pFS = pPg->pFS;

  if( pPg->iPg!=pRun->iFirst ){
    int rc = LSM_OK;
    int iBlk = fsPageToBlock(pFS, pRun->iFirst);







|












|







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
      }else if( bZero==0 && pDel->iLast!=fsLastPageOnBlock(pFS, iLastBlk) ){
        break;
      }
      rc = fsFreeBlock(pFS, pSnapshot, pDel, iBlk);
      iBlk = iNext;
    }

    if( bZero ) memset(pDel, 0, sizeof(Segment));
  }
  return LSM_OK;
}

/*
** The pager reference passed as the only argument must refer to a sorted
** file page (not a log or meta page). This call indicates that the argument
** page is now the first page in its sorted file - all previous pages may
** be considered free.
*/
void lsmFsGobble(
  Snapshot *pSnapshot,
  Segment *pRun, 
  Page *pPg
){
  FileSystem *pFS = pPg->pFS;

  if( pPg->iPg!=pRun->iFirst ){
    int rc = LSM_OK;
    int iBlk = fsPageToBlock(pFS, pRun->iFirst);
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
** If the previous/next page does exist and is successfully loaded, *ppNext
** is set to point to it and LSM_OK is returned. Otherwise, if an error 
** occurs, *ppNext is set to NULL and and lsm error code returned.
**
** Page references returned by this function should be released by the 
** caller using lsmFsPageRelease().
*/
int lsmFsDbPageNext(SortedRun *pRun, Page *pPg, int eDir, Page **ppNext){
  FileSystem *pFS = pPg->pFS;
  int iPg = pPg->iPg;

  assert( eDir==1 || eDir==-1 );

  if( eDir<0 ){
    if( pRun && iPg==pRun->iFirst ){







|







902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
** If the previous/next page does exist and is successfully loaded, *ppNext
** is set to point to it and LSM_OK is returned. Otherwise, if an error 
** occurs, *ppNext is set to NULL and and lsm error code returned.
**
** Page references returned by this function should be released by the 
** caller using lsmFsPageRelease().
*/
int lsmFsDbPageNext(Segment *pRun, Page *pPg, int eDir, Page **ppNext){
  FileSystem *pFS = pPg->pFS;
  int iPg = pPg->iPg;

  assert( eDir==1 || eDir==-1 );

  if( eDir<0 ){
    if( pRun && iPg==pRun->iFirst ){
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278

  assert( db->pWorker );
  for(pLvl=lsmDbSnapshotLevel(db->pWorker); 
      rc==LSM_OK && pLvl; 
      pLvl=pLvl->pNext
  ){
    if( pLvl->nRight==0 ){
      addAppendPoint(db, pLvl->lhs.sep.iLast, &rc);
      addAppendPoint(db, pLvl->lhs.run.iLast, &rc);
    }else{
      int i;
      for(i=0; i<pLvl->nRight; i++){
        addAppendPoint(db, pLvl->aRhs[i].sep.iLast, &rc);
        addAppendPoint(db, pLvl->aRhs[i].run.iLast, &rc);
      }
    }
  }

  for(pLvl=lsmDbSnapshotLevel(db->pWorker); pLvl; pLvl=pLvl->pNext){
    int i;
    subAppendPoint(db, pLvl->lhs.sep.iFirst);
    subAppendPoint(db, pLvl->lhs.run.iFirst);
    for(i=0; i<pLvl->nRight; i++){
      subAppendPoint(db, pLvl->aRhs[i].sep.iFirst);
      subAppendPoint(db, pLvl->aRhs[i].run.iFirst);
    }
  }

  return rc;
}

int lsmFsPhantom(FileSystem *pFS, SortedRun *pRun){
  assert( pFS->phantom.pRun==0 );
  pFS->phantom.pRun = pRun;
  return LSM_OK;
}

void lsmFsPhantomFree(FileSystem *pFS){
  if( pFS->phantom.pRun ){
    Page *pPg;
    Page *pNext;
    for(pPg=pFS->phantom.pFirst; pPg; pPg=pNext){
      pNext = pPg->pHashNext;
      fsPageBufferFree(pPg);
    }
    memset(&pFS->phantom, 0, sizeof(PhantomRun));
  }
}

int lsmFsPhantomMaterialize(
  FileSystem *pFS, 
  Snapshot *pSnapshot, 
  SortedRun *p
){
  int rc = LSM_OK;
  if( isPhantom(pFS, p) ){
    PhantomRun *pPhantom = &pFS->phantom;
    Page *pPg;
    Page *pNext;
    int i;
    Pgno iFirst = 0;

    /* Search for an existing run in the database that this run can be
    ** appended to. See comments surrounding findAppendPoint() for details. */
    iFirst = findAppendPoint(pFS, pPhantom->nPhantom);

    /* If the array can not be written into any partially used block, 
    ** allocate a new block. The first page of the materialized run will
    ** be the second page of the new block (since the first is undersized
    ** and can not be used).  */
    if( iFirst==0 ){
      int iNew;                   /* New block */
      lsmBlockAllocate(pFS->pDb, &iNew);
      iFirst = fsFirstPageOnBlock(pFS, iNew) + 1;
    }

    p->iFirst = iFirst;
    p->iLast = iFirst + pPhantom->nPhantom - 1;
    assert( 0==fsIsFirst(pFS, p->iFirst) && 0==fsIsLast(pFS, p->iFirst) );
    assert( 0==fsIsFirst(pFS, p->iLast) && 0==fsIsLast(pFS, p->iLast) );
    assert( fsPageToBlock(pFS, p->iFirst)==fsPageToBlock(pFS, p->iLast) );

    i = iFirst;
    for(pPg=pPhantom->pFirst; pPg; pPg=pNext){
      int iHash;
      pNext = pPg->pHashNext;
      pPg->iPg = i++;
      pPg->nRef++;

      iHash = fsHashKey(pFS->nHash, pPg->iPg);
      pPg->pHashNext = pFS->apHash[iHash];
      pFS->apHash[iHash] = pPg;
      pFS->nOut++;
      lsmFsPageRelease(pPg);
    }
    assert( i==p->iLast+1 );

    p->nSize = pPhantom->nPhantom;
    memset(&pFS->phantom, 0, sizeof(PhantomRun));
  }
  return rc;
}

/*
** Append a page to file iFile. Return a reference to it. lsmFsPageWrite()
** has already been called on the returned reference.
*/
int lsmFsSortedAppend(
  FileSystem *pFS, 
  Snapshot *pSnapshot,
  SortedRun *p, 
  Page **ppOut
){
  int rc = LSM_OK;
  Page *pPg = 0;
  *ppOut = 0;

  if( isPhantom(pFS, p) ){
    const int nPagePerBlock = (pFS->nBlocksize / pFS->nPagesize);
    int nLimit = (nPagePerBlock - 2 - (fsFirstPageOnBlock(pFS, 1)-1) );

    if( pFS->phantom.nPhantom>=nLimit ){ 
      rc = lsmFsPhantomMaterialize(pFS, pSnapshot, p);
      if( rc!=LSM_OK ) return rc;
    }
  }

  if( isPhantom(pFS, p) ){
    rc = fsPageBuffer(pFS, 1, &pPg);
    if( rc==LSM_OK ){
      PhantomRun *pPhantom = &pFS->phantom;
      pPg->iPg = 0;
      pPg->nRef = 1;
      pPg->flags |= PAGE_DIRTY;
      pPg->pHashNext = 0;
      pPg->pLruNext = 0;
      pPg->pLruPrev = 0;
      pPg->pFS = pFS;
      if( pPhantom->pFirst ){
        assert( pPhantom->pLast );
        pPhantom->pLast->pHashNext = pPg;
      }else{
        pPhantom->pFirst = pPg;
      }
      pPhantom->pLast = pPg;
      pPhantom->nPhantom++;
    }
  }else{
    int iApp = 0;
    int iNext = 0;
    int iPrev = p->iLast;

    if( iPrev==0 ){
      iApp = findAppendPoint(pFS, 0);
    }else if( fsIsLast(pFS, iPrev) ){
      Page *pLast = 0;
      rc = fsPageGet(pFS, iPrev, 0, &pLast);
      if( rc!=LSM_OK ) return rc;
      iApp = lsmGetU32(&pLast->aData[pFS->nPagesize-4]);
      lsmFsPageRelease(pLast);
    }else{
      iApp = iPrev + 1;
    }

    /* If this is the first page allocated, or if the page allocated is the
     ** last in the block, allocate a new block here.  */
    if( iApp==0 || fsIsLast(pFS, iApp) ){
      int iNew;                     /* New block number */

      lsmBlockAllocate(pFS->pDb, &iNew);
      if( iApp==0 ){
        iApp = fsFirstPageOnBlock(pFS, iNew);
      }else{
        iNext = fsFirstPageOnBlock(pFS, iNew);
      }
    }

    /* Grab the new page. */
    pPg = 0;
    rc = fsPageGet(pFS, iApp, 1, &pPg);
    assert( rc==LSM_OK || pPg==0 );

    /* If this is the first or last page of a block, fill in the pointer 
     ** value at the end of the new page. */
    if( rc==LSM_OK ){
      p->nSize++;
      p->iLast = iApp;
      if( p->iFirst==0 ) p->iFirst = iApp;
      pPg->flags |= PAGE_DIRTY;

      if( fsIsLast(pFS, iApp) ){
        lsmPutU32(&pPg->aData[pFS->nPagesize-4], iNext);
      }else 
        if( fsIsFirst(pFS, iApp) ){
          lsmPutU32(&pPg->aData[pFS->nPagesize-4], iPrev);
        }
    }
  }

  *ppOut = pPg;
  return rc;
}

/*
** Mark the sorted run passed as the second argument as finished. 
*/
int lsmFsSortedFinish(FileSystem *pFS, SortedRun *p){
  int rc = LSM_OK;
  if( p ){
    const int nPagePerBlock = (pFS->nBlocksize / pFS->nPagesize);

    if( pFS->phantom.pRun ) pFS->phantom.bRunFinished = 1;

    /* Check if the last page of this run happens to be the last of a block.
    ** If it is, then an extra block has already been allocated for this run.
    ** Shift this extra block back to the free-block list. 
    **
    ** Otherwise, add the first free page in the last block used by the run
    ** to the lAppend list.
    */







|
<



|
<






|
<

|
<



<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<










|





<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
|
|

|
|
|
|
|
|
|
|
|
|
|

|
|
|
|

|
|
|
|
|
|
|

|
|
|
|

|
|
|
|
|
|
|

|
|
|
|
|
|
<









|




<
<







1015
1016
1017
1018
1019
1020
1021
1022

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
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102

1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116


1117
1118
1119
1120
1121
1122
1123

  assert( db->pWorker );
  for(pLvl=lsmDbSnapshotLevel(db->pWorker); 
      rc==LSM_OK && pLvl; 
      pLvl=pLvl->pNext
  ){
    if( pLvl->nRight==0 ){
      addAppendPoint(db, pLvl->lhs.iLast, &rc);

    }else{
      int i;
      for(i=0; i<pLvl->nRight; i++){
        addAppendPoint(db, pLvl->aRhs[i].iLast, &rc);

      }
    }
  }

  for(pLvl=lsmDbSnapshotLevel(db->pWorker); pLvl; pLvl=pLvl->pNext){
    int i;
    subAppendPoint(db, pLvl->lhs.iFirst);

    for(i=0; i<pLvl->nRight; i++){
      subAppendPoint(db, pLvl->aRhs[i].iFirst);

    }
  }









































































  return rc;
}

/*
** Append a page to file iFile. Return a reference to it. lsmFsPageWrite()
** has already been called on the returned reference.
*/
int lsmFsSortedAppend(
  FileSystem *pFS, 
  Snapshot *pSnapshot,
  Segment *p, 
  Page **ppOut
){
  int rc = LSM_OK;
  Page *pPg = 0;
  *ppOut = 0;
































  int iApp = 0;
  int iNext = 0;
  int iPrev = p->iLast;

  if( iPrev==0 ){
    iApp = findAppendPoint(pFS, 0);
  }else if( fsIsLast(pFS, iPrev) ){
    Page *pLast = 0;
    rc = fsPageGet(pFS, iPrev, 0, &pLast);
    if( rc!=LSM_OK ) return rc;
    iApp = lsmGetU32(&pLast->aData[pFS->nPagesize-4]);
    lsmFsPageRelease(pLast);
  }else{
    iApp = iPrev + 1;
  }

  /* If this is the first page allocated, or if the page allocated is the
   ** last in the block, allocate a new block here.  */
  if( iApp==0 || fsIsLast(pFS, iApp) ){
    int iNew;                     /* New block number */

    lsmBlockAllocate(pFS->pDb, &iNew);
    if( iApp==0 ){
      iApp = fsFirstPageOnBlock(pFS, iNew);
    }else{
      iNext = fsFirstPageOnBlock(pFS, iNew);
    }
  }

  /* Grab the new page. */
  pPg = 0;
  rc = fsPageGet(pFS, iApp, 1, &pPg);
  assert( rc==LSM_OK || pPg==0 );

  /* If this is the first or last page of a block, fill in the pointer 
   ** value at the end of the new page. */
  if( rc==LSM_OK ){
    p->nSize++;
    p->iLast = iApp;
    if( p->iFirst==0 ) p->iFirst = iApp;
    pPg->flags |= PAGE_DIRTY;

    if( fsIsLast(pFS, iApp) ){
      lsmPutU32(&pPg->aData[pFS->nPagesize-4], iNext);
    }else 
      if( fsIsFirst(pFS, iApp) ){
        lsmPutU32(&pPg->aData[pFS->nPagesize-4], iPrev);
      }

  }

  *ppOut = pPg;
  return rc;
}

/*
** Mark the sorted run passed as the second argument as finished. 
*/
int lsmFsSortedFinish(FileSystem *pFS, Segment *p){
  int rc = LSM_OK;
  if( p ){
    const int nPagePerBlock = (pFS->nBlocksize / pFS->nPagesize);



    /* Check if the last page of this run happens to be the last of a block.
    ** If it is, then an extra block has already been allocated for this run.
    ** Shift this extra block back to the free-block list. 
    **
    ** Otherwise, add the first free page in the last block used by the run
    ** to the lAppend list.
    */
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
  *piParam = pFS->bUseMmap;
  return LSM_OK;
}

/*
** Helper function for lsmInfoArrayStructure().
*/
static SortedRun *startsWith(SortedRun *pRun, Pgno iFirst){
  return (iFirst==pRun->iFirst) ? pRun : 0;
}

/*
** This function implements the lsm_info(LSM_INFO_ARRAY_STRUCTURE) request.
** If successful, *pzOut is set to point to a nul-terminated string 
** containing the array structure and LSM_OK is returned. The caller should
** eventually free the string using lsmFree().
**
** If an error occurs, *pzOut is set to NULL and an LSM error code returned.
*/
int lsmInfoArrayStructure(lsm_db *pDb, Pgno iFirst, char **pzOut){
  int rc = LSM_OK;
  Snapshot *pWorker;              /* Worker snapshot */
  Snapshot *pRelease = 0;         /* Snapshot to release */
  SortedRun *pArray = 0;          /* Array to report on */
  Level *pLvl;                    /* Used to iterate through db levels */

  *pzOut = 0;
  if( iFirst==0 ) return LSM_ERROR;

  /* Obtain the worker snapshot */
  pWorker = pDb->pWorker;
  if( !pWorker ){
    pRelease = pWorker = lsmDbSnapshotWorker(pDb);
  }

  /* Search for the array that starts on page iFirst */
  for(pLvl=lsmDbSnapshotLevel(pWorker); pLvl && pArray==0; pLvl=pLvl->pNext){
    if( 0==(pArray = startsWith(&pLvl->lhs.sep, iFirst))
     && 0==(pArray = startsWith(&pLvl->lhs.run, iFirst))
    ){
      int i;
      for(i=0; i<pLvl->nRight; i++){
        if( (pArray = startsWith(&pLvl->aRhs[i].sep, iFirst)) ) break;
        if( (pArray = startsWith(&pLvl->aRhs[i].run, iFirst)) ) break;
      }
    }
  }

  if( pArray==0 ){
    /* Could not find the requested array. This is an error. */
    *pzOut = 0;







|















|













|
<
<


|
<







1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407


1408
1409
1410

1411
1412
1413
1414
1415
1416
1417
  *piParam = pFS->bUseMmap;
  return LSM_OK;
}

/*
** Helper function for lsmInfoArrayStructure().
*/
static Segment *startsWith(Segment *pRun, Pgno iFirst){
  return (iFirst==pRun->iFirst) ? pRun : 0;
}

/*
** This function implements the lsm_info(LSM_INFO_ARRAY_STRUCTURE) request.
** If successful, *pzOut is set to point to a nul-terminated string 
** containing the array structure and LSM_OK is returned. The caller should
** eventually free the string using lsmFree().
**
** If an error occurs, *pzOut is set to NULL and an LSM error code returned.
*/
int lsmInfoArrayStructure(lsm_db *pDb, Pgno iFirst, char **pzOut){
  int rc = LSM_OK;
  Snapshot *pWorker;              /* Worker snapshot */
  Snapshot *pRelease = 0;         /* Snapshot to release */
  Segment *pArray = 0;            /* Array to report on */
  Level *pLvl;                    /* Used to iterate through db levels */

  *pzOut = 0;
  if( iFirst==0 ) return LSM_ERROR;

  /* Obtain the worker snapshot */
  pWorker = pDb->pWorker;
  if( !pWorker ){
    pRelease = pWorker = lsmDbSnapshotWorker(pDb);
  }

  /* Search for the array that starts on page iFirst */
  for(pLvl=lsmDbSnapshotLevel(pWorker); pLvl && pArray==0; pLvl=pLvl->pNext){
    if( 0==(pArray = startsWith(&pLvl->lhs, iFirst)) ){


      int i;
      for(i=0; i<pLvl->nRight; i++){
        if( (pArray = startsWith(&pLvl->aRhs[i], iFirst)) ) break;

      }
    }
  }

  if( pArray==0 ){
    /* Could not find the requested array. This is an error. */
    *pzOut = 0;
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
    *pzOut = str.z;
  }

  lsmDbSnapshotRelease(pDb->pEnv, pRelease);
  return rc;
}

void lsmFsDumpBlockmap(lsm_db *pDb, SortedRun *p){
  if( p ){
    FileSystem *pFS = pDb->pFS;
    int iBlk;
    int iLastBlk;
    char *zMsg = 0;
    LsmString zBlk;

    lsmStringInit(&zBlk, pDb->pEnv);
    iBlk = fsPageToBlock(pFS, p->iFirst);
    iLastBlk = fsPageToBlock(pFS, p->iLast);

    while( iBlk ){
      lsmStringAppendf(&zBlk, " %d", iBlk);
      if( iBlk!=iLastBlk ){
        fsBlockNext(pFS, iBlk, &iBlk);
      }else{
        iBlk = 0;
      }
    }

    zMsg = lsmMallocPrintf(pDb->pEnv, "%d..%d: ", p->iFirst, p->iLast);
    lsmLogMessage(pDb, LSM_OK, "    % -15s %s", zMsg, zBlk.z);
    lsmFree(pDb->pEnv, zMsg);
    lsmStringClear(&zBlk);
  }
} 

#ifdef LSM_EXPENSIVE_DEBUG
/*
** Helper function for lsmFsIntegrityCheck()
*/
static void checkBlocks(
  FileSystem *pFS, 
  Segment *pSeg, 
  int bExtra,
  u8 *aUsed
){
  if( pSeg ){
    int i;
    for(i=0; i<2; i++){
      SortedRun *p = (i ? pSeg->pRun : pSeg->pSep);

      if( p && p->nSize>0 ){
        const int nPagePerBlock = (pFS->nBlocksize / pFS->nPagesize);

        int iBlk;
        int iLastBlk;
        iBlk = fsPageToBlock(pFS, p->iFirst);







<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<













|







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
    *pzOut = str.z;
  }

  lsmDbSnapshotRelease(pDb->pEnv, pRelease);
  return rc;
}





























#ifdef LSM_EXPENSIVE_DEBUG
/*
** Helper function for lsmFsIntegrityCheck()
*/
static void checkBlocks(
  FileSystem *pFS, 
  Segment *pSeg, 
  int bExtra,
  u8 *aUsed
){
  if( pSeg ){
    int i;
    for(i=0; i<2; i++){
      Segment *p = (i ? pSeg->pRun : pSeg->pSep);

      if( p && p->nSize>0 ){
        const int nPagePerBlock = (pFS->nBlocksize / pFS->nPagesize);

        int iBlk;
        int iLastBlk;
        iBlk = fsPageToBlock(pFS, p->iFirst);
Changes to src/lsm_main.c.
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428

  va_end(ap);
  return rc;
}

void lsmAppendSegmentList(LsmString *pStr, char *zPre, Segment *pSeg){
  lsmStringAppendf(pStr, "%s{%d %d %d %d %d %d}", zPre, 
        pSeg->sep.iFirst, pSeg->sep.iLast, pSeg->sep.iRoot,
        pSeg->run.iFirst, pSeg->run.iLast, pSeg->run.nSize
  );
}

int lsmStructList(
  lsm_db *pDb,                    /* Database handle */
  char **pzOut                    /* OUT: Nul-terminated string (tcl list) */
){







|
|







413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428

  va_end(ap);
  return rc;
}

void lsmAppendSegmentList(LsmString *pStr, char *zPre, Segment *pSeg){
  lsmStringAppendf(pStr, "%s{%d %d %d %d %d %d}", zPre, 
        0, 0, 0, 
        pSeg->iFirst, pSeg->iLast, pSeg->nSize
  );
}

int lsmStructList(
  lsm_db *pDb,                    /* Database handle */
  char **pzOut                    /* OUT: Nul-terminated string (tcl list) */
){
Changes to src/lsm_sorted.c.
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
**
**   * To iterate and/or seek within a single Segment (the combination of a 
**     main run and an optional sorted run).
**
**   * To iterate through the separators array of a segment.
*/
struct SegmentPtr {
  Segment *pSeg;                  /* Segment to access */
  SortedRun *pRun;                /* Points to either pSeg->run or pSeg->sep */

  /* Current page. See segmentPtrLoadPage(). */
  Page *pPg;                    /* Current page */
  u16 flags;                    /* Copy of page flags field */
  int nCell;                    /* Number of cells on pPg */
  int iPtr;                     /* Base cascade pointer */








|
<







142
143
144
145
146
147
148
149

150
151
152
153
154
155
156
**
**   * To iterate and/or seek within a single Segment (the combination of a 
**     main run and an optional sorted run).
**
**   * To iterate through the separators array of a segment.
*/
struct SegmentPtr {
  Segment *pSeg;                /* Segment to access */


  /* Current page. See segmentPtrLoadPage(). */
  Page *pPg;                    /* Current page */
  u16 flags;                    /* Copy of page flags field */
  int nCell;                    /* Number of cells on pPg */
  int iPtr;                     /* Base cascade pointer */

210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
typedef struct BtreePg BtreePg;
typedef struct BtreeCursor BtreeCursor;
struct BtreePg {
  Page *pPage;
  int iCell;
};
struct BtreeCursor {
  SortedRun *pRun;
  FileSystem *pFS;                /* File system to read pages from */
  int nDepth;                     /* Allocated size of aPg[] */
  int iPg;                        /* Current entry in aPg[]. -1 -> EOF. */
  BtreePg *aPg;                   /* Pages from root to current location */

  /* Cache of current entry. pKey==0 for EOF. */
  void *pKey;







|







209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
typedef struct BtreePg BtreePg;
typedef struct BtreeCursor BtreeCursor;
struct BtreePg {
  Page *pPage;
  int iCell;
};
struct BtreeCursor {
  Segment *pSeg;                  /* Iterate through this segments btree */
  FileSystem *pFS;                /* File system to read pages from */
  int nDepth;                     /* Allocated size of aPg[] */
  int iPg;                        /* Current entry in aPg[]. -1 -> EOF. */
  BtreePg *aPg;                   /* Pages from root to current location */

  /* Cache of current entry. pKey==0 for EOF. */
  void *pKey;
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
};

struct MergeWorker {
  lsm_db *pDb;                    /* Database handle */
  Level *pLevel;                  /* Worker snapshot Level being merged */
  MultiCursor *pCsr;              /* Cursor to read new segment contents from */
  int bFlush;                     /* True if this is an in-memory tree flush */
  Hierarchy aHier[2];
  Page *apPage[2];                /* Current output pages (0 is main run) */
  int nWork;                      /* Number of calls to mergeWorkerNextPage() */
};

#ifdef LSM_DEBUG_EXPENSIVE
static int assertPointersOk(lsm_db *, SortedRun *, SortedRun *, int);
static int assertBtreeOk(lsm_db *, SortedRun *);
#endif

struct FilePage { u8 *aData; int nData; };
static u8 *fsPageData(Page *pPg, int *pnData){
  *pnData = ((struct FilePage *)(pPg))->nData;
  return ((struct FilePage *)(pPg))->aData;
}







|
|




|
|







316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
};

struct MergeWorker {
  lsm_db *pDb;                    /* Database handle */
  Level *pLevel;                  /* Worker snapshot Level being merged */
  MultiCursor *pCsr;              /* Cursor to read new segment contents from */
  int bFlush;                     /* True if this is an in-memory tree flush */
  Hierarchy hier;                 /* B-tree hierarchy under construction */
  Page *pPage;                    /* Current output page */
  int nWork;                      /* Number of calls to mergeWorkerNextPage() */
};

#ifdef LSM_DEBUG_EXPENSIVE
static int assertPointersOk(lsm_db *, Segment *, Segment *, int);
static int assertBtreeOk(lsm_db *, Segment *);
#endif

struct FilePage { u8 *aData; int nData; };
static u8 *fsPageData(Page *pPg, int *pnData){
  *pnData = ((struct FilePage *)(pPg))->nData;
  return ((struct FilePage *)(pPg))->aData;
}
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
}

static int btreeCursorFirst(BtreeCursor *pCsr){
  int rc;

  Page *pPg = 0;
  FileSystem *pFS = pCsr->pFS;
  int iPg = pCsr->pRun->iRoot;

  do {
    rc = lsmFsDbPageGet(pFS, iPg, &pPg);
    assert( (rc==LSM_OK)==(pPg!=0) );
    if( rc==LSM_OK ){
      u8 *aData;
      int nData;







|







702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
}

static int btreeCursorFirst(BtreeCursor *pCsr){
  int rc;

  Page *pPg = 0;
  FileSystem *pFS = pCsr->pFS;
  int iPg = pCsr->pSeg->iRoot;

  do {
    rc = lsmFsDbPageGet(pFS, iPg, &pPg);
    assert( (rc==LSM_OK)==(pPg!=0) );
    if( rc==LSM_OK ){
      u8 *aData;
      int nData;
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
      Blob blob = {0,0,0};
      void *pSeek;
      int nSeek;
      int iTopicSeek;
      int dummy;

      int iPg = 0;
      int iLoad = pCsr->pRun->iRoot;

      rc = pageGetBtreeKey(pCsr->aPg[nDepth-1].pPage, 
          0, &dummy, &iTopicSeek, &pSeek, &nSeek, &pCsr->blob
      );

      do {
        Page *pPg;







|







806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
      Blob blob = {0,0,0};
      void *pSeek;
      int nSeek;
      int iTopicSeek;
      int dummy;

      int iPg = 0;
      int iLoad = pCsr->pSeg->iRoot;

      rc = pageGetBtreeKey(pCsr->aPg[nDepth-1].pPage, 
          0, &dummy, &iTopicSeek, &pSeek, &nSeek, &pCsr->blob
      );

      do {
        Page *pPg;
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
    }
  }
  return rc;
}

static int btreeCursorNew(
  lsm_db *pDb,
  SortedRun *pRun,
  BtreeCursor **ppCsr
){
  int rc = LSM_OK;
  BtreeCursor *pCsr;
  
  assert( pRun->iRoot );
  pCsr = lsmMallocZeroRc(pDb->pEnv, sizeof(BtreeCursor), &rc);
  if( pCsr ){
    pCsr->pFS = pDb->pFS;
    pCsr->pRun = pRun;
    pCsr->iPg = -1;
  }

  *ppCsr = pCsr;
  return rc;
}








|





|



|







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
    }
  }
  return rc;
}

static int btreeCursorNew(
  lsm_db *pDb,
  Segment *pSeg,
  BtreeCursor **ppCsr
){
  int rc = LSM_OK;
  BtreeCursor *pCsr;
  
  assert( pSeg->iRoot );
  pCsr = lsmMallocZeroRc(pDb->pEnv, sizeof(BtreeCursor), &rc);
  if( pCsr ){
    pCsr->pFS = pDb->pFS;
    pCsr->pSeg = pSeg;
    pCsr->iPg = -1;
  }

  *ppCsr = pCsr;
  return rc;
}

938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
  FileSystem *pFS,
  SegmentPtr *pPtr,              /* Load page into this SegmentPtr object */
  int iNew                       /* Page number of new page */
){
  Page *pPg = 0;                 /* The new page */
  int rc;                        /* Return Code */

  assert( pPtr->pSeg==0 
       || pPtr->pRun==&pPtr->pSeg->run 
       || pPtr->pRun==&pPtr->pSeg->sep 
  );
  rc = lsmFsDbPageGet(pFS, iNew, &pPg);
  assert( rc==LSM_OK || pPg==0 );
  segmentPtrSetPage(pPtr, pPg);

  return rc;
}








<
<
<
<







937
938
939
940
941
942
943




944
945
946
947
948
949
950
  FileSystem *pFS,
  SegmentPtr *pPtr,              /* Load page into this SegmentPtr object */
  int iNew                       /* Page number of new page */
){
  Page *pPg = 0;                 /* The new page */
  int rc;                        /* Return Code */





  rc = lsmFsDbPageGet(pFS, iNew, &pPg);
  assert( rc==LSM_OK || pPg==0 );
  segmentPtrSetPage(pPtr, pPg);

  return rc;
}

968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
  int eDir                       /* +1 for next(), -1 for prev() */
){
  Page *pNext;                   /* New page to load */
  int rc;                        /* Return code */

  assert( eDir==1 || eDir==-1 );
  assert( pPtr->pPg );
  assert( (pPtr->pSeg==0 && eDir>0)
       || pPtr->pRun==&pPtr->pSeg->run 
       || pPtr->pRun==&pPtr->pSeg->sep 
  );

  rc = lsmFsDbPageNext(pPtr->pRun, pPtr->pPg, eDir, &pNext);
  assert( rc==LSM_OK || pNext==0 );
  segmentPtrSetPage(pPtr, pNext);
  return rc;
}

static int segmentPtrLoadCell(
  SegmentPtr *pPtr,              /* Load page into this SegmentPtr object */







<
|
<
<

|







963
964
965
966
967
968
969

970


971
972
973
974
975
976
977
978
979
  int eDir                       /* +1 for next(), -1 for prev() */
){
  Page *pNext;                   /* New page to load */
  int rc;                        /* Return code */

  assert( eDir==1 || eDir==-1 );
  assert( pPtr->pPg );

  assert( pPtr->pSeg || eDir>0 );



  rc = lsmFsDbPageNext(pPtr->pSeg, pPtr->pPg, eDir, &pNext);
  assert( rc==LSM_OK || pNext==0 );
  segmentPtrSetPage(pPtr, pNext);
  return rc;
}

static int segmentPtrLoadCell(
  SegmentPtr *pPtr,              /* Load page into this SegmentPtr object */
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
  int i;
  Merge *pMerge = pLevel->pMerge;

  for(i=0; rc==LSM_OK && i<pLevel->nRight; i++){
    Page *pPg = 0;
    int iTopic;
    Blob blob = {0, 0, 0, 0};
    SortedRun *pRun = &pLevel->aRhs[i].run;

    assert( pRun->iFirst!=0 );

    rc = lsmFsDbPageGet(pDb->pFS, pMerge->aInput[i].iPg, &pPg);
    if( rc==LSM_OK ){
      rc = pageGetKeyCopy(pEnv, pPg, pMerge->aInput[i].iCell, &iTopic, &blob);
    }
    if( rc==LSM_OK ){
      int res = -1;
      if( pLevel->pSplitKey ){







<

|
<







1019
1020
1021
1022
1023
1024
1025

1026
1027

1028
1029
1030
1031
1032
1033
1034
  int i;
  Merge *pMerge = pLevel->pMerge;

  for(i=0; rc==LSM_OK && i<pLevel->nRight; i++){
    Page *pPg = 0;
    int iTopic;
    Blob blob = {0, 0, 0, 0};


    assert( pLevel->aRhs[i].iFirst!=0 );

    rc = lsmFsDbPageGet(pDb->pFS, pMerge->aInput[i].iPg, &pPg);
    if( rc==LSM_OK ){
      rc = pageGetKeyCopy(pEnv, pPg, pMerge->aInput[i].iCell, &iTopic, &blob);
    }
    if( rc==LSM_OK ){
      int res = -1;
      if( pLevel->pSplitKey ){
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
    int i;
    pCsr->aPtr[0].pSeg = &pLevel->lhs;
    pCsr->nPtr = nPtr;

    for(i=0; i<pLevel->nRight; i++){
      pCsr->aPtr[i+1].pSeg = &pLevel->aRhs[i];
    }
    for(i=0; i<pCsr->nPtr; i++){
      pCsr->aPtr[i].pRun = &pCsr->aPtr[i].pSeg->run;
    }
  }

  return rc;
}

static int levelCursorInitRun(
  lsm_db *pDb,
  SortedRun *pRun, 
  int (*xCmp)(void *, int, void *, int),
  LevelCursor *pCsr              /* Cursor structure to initialize */
){
  int rc = LSM_OK;

  memset(pCsr, 0, sizeof(LevelCursor));
  pCsr->pFS = pDb->pFS;
  pCsr->bIgnoreSeparators = 1;
  pCsr->xCmp = xCmp;
  pCsr->nPtr = 1;
  pCsr->aPtr = (SegmentPtr*)lsmMallocZeroRc(pDb->pEnv, 
      sizeof(SegmentPtr)*pCsr->nPtr, &rc
  );

  if( rc==LSM_OK ){
    pCsr->aPtr[0].pRun = pRun;
  }

  return rc;
}

static void segmentPtrReset(SegmentPtr *pPtr){
  lsmFsPageRelease(pPtr->pPg);







<
<
<







|















|







1078
1079
1080
1081
1082
1083
1084



1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
    int i;
    pCsr->aPtr[0].pSeg = &pLevel->lhs;
    pCsr->nPtr = nPtr;

    for(i=0; i<pLevel->nRight; i++){
      pCsr->aPtr[i+1].pSeg = &pLevel->aRhs[i];
    }



  }

  return rc;
}

static int levelCursorInitRun(
  lsm_db *pDb,
  Segment *pSeg, 
  int (*xCmp)(void *, int, void *, int),
  LevelCursor *pCsr              /* Cursor structure to initialize */
){
  int rc = LSM_OK;

  memset(pCsr, 0, sizeof(LevelCursor));
  pCsr->pFS = pDb->pFS;
  pCsr->bIgnoreSeparators = 1;
  pCsr->xCmp = xCmp;
  pCsr->nPtr = 1;
  pCsr->aPtr = (SegmentPtr*)lsmMallocZeroRc(pDb->pEnv, 
      sizeof(SegmentPtr)*pCsr->nPtr, &rc
  );

  if( rc==LSM_OK ){
    pCsr->aPtr[0].pSeg = pSeg;
  }

  return rc;
}

static void segmentPtrReset(SegmentPtr *pPtr){
  lsmFsPageRelease(pPtr->pPg);
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
  FileSystem *pFS, 
  SegmentPtr *pPtr, 
  int bLast, 
  int *pRc
){
  if( *pRc==LSM_OK ){
    Page *pNew = 0;
    Pgno iPg = (bLast ? pPtr->pRun->iLast : pPtr->pRun->iFirst);
    *pRc = lsmFsDbPageGet(pFS, iPg, &pNew);
    segmentPtrSetPage(pPtr, pNew);
  }
}

/*
** Try to move the segment pointer passed as the second argument so that it







|







1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
  FileSystem *pFS, 
  SegmentPtr *pPtr, 
  int bLast, 
  int *pRc
){
  if( *pRc==LSM_OK ){
    Page *pNew = 0;
    Pgno iPg = (bLast ? pPtr->pSeg->iLast : pPtr->pSeg->iFirst);
    *pRc = lsmFsDbPageGet(pFS, iPg, &pNew);
    segmentPtrSetPage(pPtr, pNew);
  }
}

/*
** Try to move the segment pointer passed as the second argument so that it
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
  for(eDir=-1; eDir<=1; eDir+=2){
    Page *pTest = pPtr->pPg;

    lsmFsPageRef(pTest);
    while( pTest ){
      Page *pNext;

      int rc = lsmFsDbPageNext(pPtr->pRun, pTest, eDir, &pNext);
      lsmFsPageRelease(pTest);
      pTest = pNext;
      assert( rc==LSM_OK );

      if( pTest ){
        int nData;
        u8 *aData = fsPageData(pTest, &nData);







|







1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
  for(eDir=-1; eDir<=1; eDir+=2){
    Page *pTest = pPtr->pPg;

    lsmFsPageRef(pTest);
    while( pTest ){
      Page *pNext;

      int rc = lsmFsDbPageNext(pPtr->pSeg, pTest, eDir, &pNext);
      lsmFsPageRelease(pTest);
      pTest = pNext;
      assert( rc==LSM_OK );

      if( pTest ){
        int nData;
        u8 *aData = fsPageData(pTest, &nData);
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
    ** page.  */
    res = iLastTopic - iTopic;
    if( res==0 ) res = pCsr->xCmp(pLastKey, nLastKey, pKey, nKey);
    if( res>=0 ) break;

    /* Advance to the next page that contains at least one key. */
    do {
      rc = lsmFsDbPageNext(pPtr->pRun, pPtr->pPg, 1, &pNext);
      if( pNext==0 ) break;
      assert( rc==LSM_OK );
      segmentPtrSetPage(pPtr, pNext);
    }while( (pPtr->nCell==0 || (pPtr->flags & SEGMENT_BTREE_FLAG)) );
    if( pNext==0 ) break;

    /* This should probably be an LSM_CORRUPT error. */







|







1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
    ** page.  */
    res = iLastTopic - iTopic;
    if( res==0 ) res = pCsr->xCmp(pLastKey, nLastKey, pKey, nKey);
    if( res>=0 ) break;

    /* Advance to the next page that contains at least one key. */
    do {
      rc = lsmFsDbPageNext(pPtr->pSeg, pPtr->pPg, 1, &pNext);
      if( pNext==0 ) break;
      assert( rc==LSM_OK );
      segmentPtrSetPage(pPtr, pNext);
    }while( (pPtr->nCell==0 || (pPtr->flags & SEGMENT_BTREE_FLAG)) );
    if( pNext==0 ) break;

    /* This should probably be an LSM_CORRUPT error. */
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
  ** loading (iPg+1) and testing that pKey/nKey is smaller than all
  ** the keys it houses.  */
#if 1
  assert( assertKeyLocation(pCsr, pPtr, pKey, nKey) );
#endif

  assert( pPtr->nCell>0 
       || pPtr->pRun->nSize==1 
       || lsmFsPageNumber(pPtr->pPg)==pPtr->pRun->iLast
  );
  if( pPtr->nCell==0 ){
    segmentPtrReset(pPtr);
  }else{
    iMin = 0;
    iMax = pPtr->nCell-1;








|
|







1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
  ** loading (iPg+1) and testing that pKey/nKey is smaller than all
  ** the keys it houses.  */
#if 1
  assert( assertKeyLocation(pCsr, pPtr, pKey, nKey) );
#endif

  assert( pPtr->nCell>0 
       || pPtr->pSeg->nSize==1 
       || lsmFsPageNumber(pPtr->pPg)==pPtr->pSeg->iLast
  );
  if( pPtr->nCell==0 ){
    segmentPtrReset(pPtr);
  }else{
    iMin = 0;
    iMax = pPtr->nCell-1;

1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
  }

  pCsr->iCurrentPtr = iBest;
}

static int seekInBtree(
  LevelCursor *pCsr,
  SortedRun *pRun,
  void *pKey, int nKey,           /* Key to seek to */
  Page **ppPg                     /* OUT: Leaf (sorted-run) page reference */
){
  int rc;
  int iPg;
  Page *pPg = 0;
  Blob blob = {0, 0, 0};
  int iTopic = 0;                 /* TODO: Fix me */

  iPg = pRun->iRoot;
  do {
    rc = lsmFsDbPageGet(pCsr->pFS, iPg, &pPg);
    assert( rc==LSM_OK || pPg==0 );
    if( rc==LSM_OK ){
      u8 *aData;                  /* Buffer containing page data */
      int nData;                  /* Size of aData[] in bytes */
      int iMin;







|









|







1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
  }

  pCsr->iCurrentPtr = iBest;
}

static int seekInBtree(
  LevelCursor *pCsr,
  Segment *pSeg,
  void *pKey, int nKey,           /* Key to seek to */
  Page **ppPg                     /* OUT: Leaf (sorted-run) page reference */
){
  int rc;
  int iPg;
  Page *pPg = 0;
  Blob blob = {0, 0, 0};
  int iTopic = 0;                 /* TODO: Fix me */

  iPg = pSeg->iRoot;
  do {
    rc = lsmFsDbPageGet(pCsr->pFS, iPg, &pPg);
    assert( rc==LSM_OK || pPg==0 );
    if( rc==LSM_OK ){
      u8 *aData;                  /* Buffer containing page data */
      int nData;                  /* Size of aData[] in bytes */
      int iMin;
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643

  sortedBlobFree(&blob);
  assert( (rc==LSM_OK)==(pPg!=0) );
  *ppPg = pPg;
  return rc;
}

static int seekInSeparators(
  LevelCursor *pCsr,
  SegmentPtr *pPtr,               /* Segment to seek within */
  void *pKey, int nKey,           /* Key to seek to */
  int *piPtr                      /* OUT: FC pointer */
){
  int rc;
  Page *pPg;
  SortedRun *pSep = &pPtr->pSeg->sep;

  rc = seekInBtree(pCsr, pSep, pKey, nKey, &pPg);

  if( rc==LSM_OK ){
    assert( pPtr->pRun==&pPtr->pSeg->run );
    assert( pPg );
    pPtr->pRun = pSep;
    segmentPtrSetPage(pPtr, pPg);
    rc = segmentPtrSeek(pCsr, pPtr, pKey, nKey, 0, piPtr);
    pPtr->pRun = &pPtr->pSeg->run;
  }

  return rc;
}

static int seekInSegment(
  LevelCursor *pCsr, 
  SegmentPtr *pPtr,
  void *pKey, int nKey,
  int iPg,                        /* Page to search */
  int eSeek,                      /* Search bias - see above */
  int *piPtr                      /* OUT: FC pointer */
){
  int iPtr = iPg;
  int rc = LSM_OK;

  assert( pPtr->pRun==&pPtr->pSeg->run );

  if( pPtr->pSeg->run.iRoot ){
    Page *pPg;
    assert( pPtr->pSeg->run.iRoot!=0 );
    rc = seekInBtree(pCsr, &pPtr->pSeg->run, pKey, nKey, &pPg);
    if( rc==LSM_OK ) segmentPtrSetPage(pPtr, pPg);
  }else{
    if( iPtr==0 ){
      iPtr = pPtr->pSeg->run.iFirst;
    }
    if( rc==LSM_OK ){
      rc = segmentPtrLoadPage(pCsr->pFS, pPtr, iPtr);
    }
  }

  if( rc==LSM_OK ){







<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<











<
<
|

|
|



|







1572
1573
1574
1575
1576
1577
1578
























1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589


1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604

  sortedBlobFree(&blob);
  assert( (rc==LSM_OK)==(pPg!=0) );
  *ppPg = pPg;
  return rc;
}

























static int seekInSegment(
  LevelCursor *pCsr, 
  SegmentPtr *pPtr,
  void *pKey, int nKey,
  int iPg,                        /* Page to search */
  int eSeek,                      /* Search bias - see above */
  int *piPtr                      /* OUT: FC pointer */
){
  int iPtr = iPg;
  int rc = LSM_OK;



  if( pPtr->pSeg->iRoot ){
    Page *pPg;
    assert( pPtr->pSeg->iRoot!=0 );
    rc = seekInBtree(pCsr, pPtr->pSeg, pKey, nKey, &pPg);
    if( rc==LSM_OK ) segmentPtrSetPage(pPtr, pPg);
  }else{
    if( iPtr==0 ){
      iPtr = pPtr->pSeg->iFirst;
    }
    if( rc==LSM_OK ){
      rc = segmentPtrLoadPage(pCsr->pFS, pPtr, iPtr);
    }
  }

  if( rc==LSM_OK ){
1835
1836
1837
1838
1839
1840
1841

1842
1843
1844
1845
1846
1847
1848
        *pp = pCsr->pNext;
        break;
      }
    }

    /* Free the allocation used to cache the current key, if any. */
    sortedBlobFree(&pCsr->key);


    /* Free the component cursors */
    mcursorFreeComponents(pCsr);

    /* Free the cursor structure itself */
    lsmFree(pDb->pEnv, pCsr);
  }







>







1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
        *pp = pCsr->pNext;
        break;
      }
    }

    /* Free the allocation used to cache the current key, if any. */
    sortedBlobFree(&pCsr->key);
    sortedBlobFree(&pCsr->val);

    /* Free the component cursors */
    mcursorFreeComponents(pCsr);

    /* Free the cursor structure itself */
    lsmFree(pDb->pEnv, pCsr);
  }
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895

  assert( eMode==MULTICURSOR_ADDLEVEL_ALL
       || eMode==MULTICURSOR_ADDLEVEL_RHS
       || eMode==MULTICURSOR_ADDLEVEL_LHS_SEP
  );

  if( eMode==MULTICURSOR_ADDLEVEL_LHS_SEP ){
    assert( pLevel->lhs.run.iRoot );
    assert( pCsr->pBtCsr==0 );
    rc = btreeCursorNew(pCsr->pDb, &pLevel->lhs.run, &pCsr->pBtCsr);
    assert( (rc==LSM_OK)==(pCsr->pBtCsr!=0) );
  }else{
    int i;
    int nAdd = (eMode==MULTICURSOR_ADDLEVEL_RHS ? pLevel->nRight : 1);

    for(i=0; i<nAdd; i++){
      LevelCursor *pNew;







|

|







1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857

  assert( eMode==MULTICURSOR_ADDLEVEL_ALL
       || eMode==MULTICURSOR_ADDLEVEL_RHS
       || eMode==MULTICURSOR_ADDLEVEL_LHS_SEP
  );

  if( eMode==MULTICURSOR_ADDLEVEL_LHS_SEP ){
    assert( pLevel->lhs.iRoot );
    assert( pCsr->pBtCsr==0 );
    rc = btreeCursorNew(pCsr->pDb, &pLevel->lhs, &pCsr->pBtCsr);
    assert( (rc==LSM_OK)==(pCsr->pBtCsr!=0) );
  }else{
    int i;
    int nAdd = (eMode==MULTICURSOR_ADDLEVEL_RHS ? pLevel->nRight : 1);

    for(i=0; i<nAdd; i++){
      LevelCursor *pNew;
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927

      switch( eMode ){
        case MULTICURSOR_ADDLEVEL_ALL:
          rc = levelCursorInit(pDb, pLevel, pCsr->xCmp, pNew);
          break;

        case MULTICURSOR_ADDLEVEL_RHS:
          rc = levelCursorInitRun(pDb, &pLevel->aRhs[i].run, pCsr->xCmp, pNew);
          break;

        case MULTICURSOR_ADDLEVEL_LHS_SEP:
          rc = levelCursorInitRun(pDb, &pLevel->lhs.sep, pCsr->xCmp, pNew);
          break;
      }
      if( pCsr->flags & CURSOR_IGNORE_SYSTEM ){
        pNew->bIgnoreSystem = 1;
      }
      if( rc==LSM_OK ) pCsr->nSegCsr++;
    }







|
<
<
<
<







1871
1872
1873
1874
1875
1876
1877
1878




1879
1880
1881
1882
1883
1884
1885

      switch( eMode ){
        case MULTICURSOR_ADDLEVEL_ALL:
          rc = levelCursorInit(pDb, pLevel, pCsr->xCmp, pNew);
          break;

        case MULTICURSOR_ADDLEVEL_RHS:
          rc = levelCursorInitRun(pDb, &pLevel->aRhs[i], pCsr->xCmp, pNew);




          break;
      }
      if( pCsr->flags & CURSOR_IGNORE_SYSTEM ){
        pNew->bIgnoreSystem = 1;
      }
      if( rc==LSM_OK ) pCsr->nSegCsr++;
    }
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
** to the way the file.c module works some (the first and last in each block)
** are 4 bytes smaller than the others.
*/
static int mergeWorkerMoveHierarchy(
  MergeWorker *pMW,               /* Merge worker */
  int bSep                        /* True for separators run */
){
  SortedRun *pRun;                /* Run being modified */
  lsm_db *pDb = pMW->pDb;         /* Database handle */
  int rc = LSM_OK;                /* Return code */
  int i;
  int iRight = 0;
  Page **apHier = pMW->aHier[bSep].apHier;
  int nHier = pMW->aHier[bSep].nHier;

  assert( nHier>0 && pMW->pLevel->pMerge->abHierReadonly[bSep] );
  pRun = (bSep ? &pMW->pLevel->lhs.sep : &pMW->pLevel->lhs.run);

  for(i=0; rc==LSM_OK && i<nHier; i++){
    Page *pNew = 0;
    rc = lsmFsSortedAppend(pDb->pFS, pDb->pWorker, pRun, &pNew);
    assert( rc==LSM_OK );

    if( rc==LSM_OK ){
      u8 *a1; int n1;
      u8 *a2; int n2;

      a1 = fsPageData(pNew, &n1);







|




|
|

|
|



|







2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
** to the way the file.c module works some (the first and last in each block)
** are 4 bytes smaller than the others.
*/
static int mergeWorkerMoveHierarchy(
  MergeWorker *pMW,               /* Merge worker */
  int bSep                        /* True for separators run */
){
  Segment *pSeg;                  /* Segment being written */
  lsm_db *pDb = pMW->pDb;         /* Database handle */
  int rc = LSM_OK;                /* Return code */
  int i;
  int iRight = 0;
  Page **apHier = pMW->hier.apHier;
  int nHier = pMW->hier.nHier;

  assert( nHier>0 && pMW->pLevel->pMerge->bHierReadonly );
  pSeg = &pMW->pLevel->lhs;

  for(i=0; rc==LSM_OK && i<nHier; i++){
    Page *pNew = 0;
    rc = lsmFsSortedAppend(pDb->pFS, pDb->pWorker, pSeg, &pNew);
    assert( rc==LSM_OK );

    if( rc==LSM_OK ){
      u8 *a1; int n1;
      u8 *a2; int n2;

      a1 = fsPageData(pNew, &n1);
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
#ifdef LSM_DEBUG
  if( rc==LSM_OK ){
    for(i=0; i<nHier; i++) assert( lsmFsPageWritable(apHier[i]) );
  }
#endif

  if( rc==LSM_OK ){
    pMW->pLevel->pMerge->abHierReadonly[bSep] = 0;
  }
  return rc;
}

/*
** Allocate and populate the MergeWorker.apHier[] array.
*/
static int mergeWorkerLoadHierarchy(
  MergeWorker *pMW,
  int bSep
){
  int rc = LSM_OK;
  SortedRun *pRun;
  Hierarchy *p;
 
  pRun = (bSep ? &pMW->pLevel->lhs.sep : &pMW->pLevel->lhs.run);
  p = &pMW->aHier[bSep];

  if( p->apHier==0 && pRun->iRoot!=0 ){
    int bHierReadonly = pMW->pLevel->pMerge->abHierReadonly[bSep];
    FileSystem *pFS = pMW->pDb->pFS;
    lsm_env *pEnv = pMW->pDb->pEnv;
    Page **apHier = 0;
    int nHier = 0;
    int iPg = pRun->iRoot;

    do {
      Page *pPg = 0;
      u8 *aData;
      int nData;
      int flags;








|







|
<
<
<

|


|
|

|
|




|







2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696



2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
#ifdef LSM_DEBUG
  if( rc==LSM_OK ){
    for(i=0; i<nHier; i++) assert( lsmFsPageWritable(apHier[i]) );
  }
#endif

  if( rc==LSM_OK ){
    pMW->pLevel->pMerge->bHierReadonly = 0;
  }
  return rc;
}

/*
** Allocate and populate the MergeWorker.apHier[] array.
*/
static int mergeWorkerLoadHierarchy(MergeWorker *pMW){



  int rc = LSM_OK;
  Segment *pSeg;
  Hierarchy *p;
 
  pSeg = &pMW->pLevel->lhs;
  p = &pMW->hier;

  if( p->apHier==0 && pSeg->iRoot!=0 ){
    int bHierReadonly = pMW->pLevel->pMerge->bHierReadonly;
    FileSystem *pFS = pMW->pDb->pFS;
    lsm_env *pEnv = pMW->pDb->pEnv;
    Page **apHier = 0;
    int nHier = 0;
    int iPg = pSeg->iRoot;

    do {
      Page *pPg = 0;
      u8 *aData;
      int nData;
      int flags;

2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
** array page that contains the main record is added to the b-tree instead.
** In this case the record format is:
**
**         + 0x00 byte (1 byte) 
**         + Absolute pointer value (varint),
**         + Absolute page number of page containing key (varint).
**
** See function seekInSeparators() for the code that traverses b-tree pages.
*/
static int mergeWorkerPushHierarchy(
  MergeWorker *pMW,               /* Merge worker object */
  int bSep,                       /* True for separators, false otherwise */
  Pgno iKeyPg,                    /* Page that will contain pKey/nKey */
  int iTopic,                     /* Topic value for this key */
  void *pKey,                     /* Pointer to key buffer */
  int nKey                        /* Size of pKey buffer in bytes */
){
  lsm_db *pDb = pMW->pDb;         /* Database handle */
  int rc;                         /* Return Code */
  int iLevel;                     /* Level of b-tree hierachy to write to */
  int nData;                      /* Size of aData[] in bytes */
  u8 *aData;                      /* Page data for level iLevel */
  int iOff;                       /* Offset on b-tree page to write record to */
  int nRec;                       /* Initial number of records on b-tree page */
  Pgno iPtr;                      /* Pointer value to accompany pKey/nKey */
  int bIndirect;                  /* True to use an indirect record */

  Hierarchy *p;
  SortedRun *pRun;

  /* If there exists a b-tree hierarchy and it is not loaded into 
  ** memory, load it now.  */
  pRun = (bSep ? &pMW->pLevel->lhs.sep : &pMW->pLevel->lhs.run);
  p = &pMW->aHier[bSep];
  rc = mergeWorkerLoadHierarchy(pMW, bSep);

  /* Obtain the absolute pointer value to store along with the key in the
  ** page body. This pointer points to a page that contains keys that are
  ** smaller than pKey/nKey.  */
  if( p->nHier ){
    aData = fsPageData(p->apHier[0], &nData);
    iPtr = lsmGetU32(&aData[SEGMENT_POINTER_OFFSET(nData)]);
  }else{
    iPtr = pRun->iFirst;
  }

  if( p->nHier && pMW->pLevel->pMerge->abHierReadonly[bSep] ){
    rc = mergeWorkerMoveHierarchy(pMW, bSep);
    if( rc!=LSM_OK ) goto push_hierarchy_out;
  }

  /* Determine if the indirect format should be used. */
  bIndirect = (nKey*4 > lsmFsPageSize(pMW->pDb->pFS));








|




















|



|
|
|








|


|







2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
** array page that contains the main record is added to the b-tree instead.
** In this case the record format is:
**
**         + 0x00 byte (1 byte) 
**         + Absolute pointer value (varint),
**         + Absolute page number of page containing key (varint).
**
** See function seekInBtree() for the code that traverses b-tree pages.
*/
static int mergeWorkerPushHierarchy(
  MergeWorker *pMW,               /* Merge worker object */
  int bSep,                       /* True for separators, false otherwise */
  Pgno iKeyPg,                    /* Page that will contain pKey/nKey */
  int iTopic,                     /* Topic value for this key */
  void *pKey,                     /* Pointer to key buffer */
  int nKey                        /* Size of pKey buffer in bytes */
){
  lsm_db *pDb = pMW->pDb;         /* Database handle */
  int rc;                         /* Return Code */
  int iLevel;                     /* Level of b-tree hierachy to write to */
  int nData;                      /* Size of aData[] in bytes */
  u8 *aData;                      /* Page data for level iLevel */
  int iOff;                       /* Offset on b-tree page to write record to */
  int nRec;                       /* Initial number of records on b-tree page */
  Pgno iPtr;                      /* Pointer value to accompany pKey/nKey */
  int bIndirect;                  /* True to use an indirect record */

  Hierarchy *p;
  Segment *pSeg;

  /* If there exists a b-tree hierarchy and it is not loaded into 
  ** memory, load it now.  */
  pSeg = &pMW->pLevel->lhs;
  p = &pMW->hier;
  rc = mergeWorkerLoadHierarchy(pMW);

  /* Obtain the absolute pointer value to store along with the key in the
  ** page body. This pointer points to a page that contains keys that are
  ** smaller than pKey/nKey.  */
  if( p->nHier ){
    aData = fsPageData(p->apHier[0], &nData);
    iPtr = lsmGetU32(&aData[SEGMENT_POINTER_OFFSET(nData)]);
  }else{
    iPtr = pSeg->iFirst;
  }

  if( p->nHier && pMW->pLevel->pMerge->bHierReadonly ){
    rc = mergeWorkerMoveHierarchy(pMW, bSep);
    if( rc!=LSM_OK ) goto push_hierarchy_out;
  }

  /* Determine if the indirect format should be used. */
  bIndirect = (nKey*4 > lsmFsPageSize(pMW->pDb->pFS));

2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
      rc = lsmFsPageRelease(p->apHier[iLevel]);
    }

    /* Allocate a new page for apHier[iLevel]. */
    p->apHier[iLevel] = 0;
    if( rc==LSM_OK ){
      rc = lsmFsSortedAppend(
          pDb->pFS, pDb->pWorker, pRun, &p->apHier[iLevel]
      );
    }
    if( rc!=LSM_OK ) goto push_hierarchy_out;

    aData = fsPageData(p->apHier[iLevel], &nData);
    memset(aData, 0, nData);
    lsmPutU16(&aData[SEGMENT_FLAGS_OFFSET(nData)], SEGMENT_BTREE_FLAG);







|







2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
      rc = lsmFsPageRelease(p->apHier[iLevel]);
    }

    /* Allocate a new page for apHier[iLevel]. */
    p->apHier[iLevel] = 0;
    if( rc==LSM_OK ){
      rc = lsmFsSortedAppend(
          pDb->pFS, pDb->pWorker, pSeg, &p->apHier[iLevel]
      );
    }
    if( rc!=LSM_OK ) goto push_hierarchy_out;

    aData = fsPageData(p->apHier[iLevel], &nData);
    memset(aData, 0, nData);
    lsmPutU16(&aData[SEGMENT_FLAGS_OFFSET(nData)], SEGMENT_BTREE_FLAG);
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117

  /* Write the right-hand pointer of the right-most leaf page of the 
  ** b-tree heirarchy. */
  aData = fsPageData(p->apHier[0], &nData);
  lsmPutU32(&aData[SEGMENT_POINTER_OFFSET(nData)], iKeyPg);

  /* Ensure that the SortedRun.iRoot field is correct. */
  pRun->iRoot = lsmFsPageNumber(p->apHier[p->nHier-1]);

push_hierarchy_out:
  return rc;
}

/*
** The merge-worker object passed as the first argument to this function
** was used for an in-memory tree flush. If one was required, the separators 
** array has been assembled in-memory (as a "phantom"). In this case it
** consists of leaf nodes only, there are no b-tree nodes. This function 
** materializes the phantom run (writes it into the db file) and appends
** any required b-tree nodes.
*/
static int mergeWorkerBuildHierarchy(MergeWorker *pMW){
  int rc = LSM_OK;

  assert( pMW->bFlush );
  assert( pMW->pLevel->lhs.sep.iRoot==0 );

  if( pMW->apPage[1] ){
    SortedRun *pRun;              /* Separators run to materialize */
    lsm_db *db = pMW->pDb;
    Blob blob = {0, 0, 0};
    Page *pPg;
    int iLast;

    /* Write the leaf pages into the file. They now have page numbers,
    ** which can be used as pointers in the b-tree hierarchy.  */
    pRun = &pMW->pLevel->lhs.sep;
    rc = lsmFsPhantomMaterialize(db->pFS, db->pWorker, pRun);

    if( rc==LSM_OK ){
      rc = lsmFsDbPageGet(db->pFS, pRun->iFirst, &pPg);
    }

    iLast = pRun->iLast;
    while( rc==LSM_OK && lsmFsPageNumber(pPg)!=iLast ){
      Page *pNext = 0;

      rc = lsmFsDbPageNext(pRun, pPg, 1, &pNext);
      lsmFsPageRelease(pPg);
      pPg = pNext;

      if( rc==LSM_OK ){
        u8 *aData;
        int nData;
        aData = fsPageData(pPg, &nData);
        if( pageGetNRec(aData, nData)>0 ){
          u8 *pKey;
          int nKey;
          int iTopic;
          Pgno iPg = lsmFsPageNumber(pPg);

          pKey = pageGetKey(pPg, 0, &iTopic, &nKey, &blob);
          rc = mergeWorkerPushHierarchy(pMW, 1, iPg, iTopic, pKey, nKey);
        }
      }
    }

    if( pMW->aHier[1].nHier>0 ){
      Page *pRoot = pMW->aHier[1].apHier[pMW->aHier[1].nHier-1];
      pRun->iRoot = lsmFsPageNumber(pRoot);
    }else{
      pRun->iRoot = pRun->iFirst;
    }

    lsmFsPageRelease(pPg);
    sortedBlobFree(&blob);
  }
  return rc;
}

static int keyszToSkip(FileSystem *pFS, int nKey){
  int nPgsz;                /* Nominal database page size */
  nPgsz = lsmFsPageSize(pFS);
  return LSM_MIN(((nKey * 4) / nPgsz), 3);
}

/*
** Advance to the next page of an output run being populated by merge-worker
** pMW. If bSep is true, the separators run output is advanced by one page.
** Otherwise, the main run.
**
** The footer of the new page is initialized to indicate that it contains
** zero records. The flags field is cleared. The page footer pointer field
** is set to iFPtr.
**
** If successful, LSM_OK is returned. Otherwise, an error code.
*/
static int mergeWorkerNextPage(
  MergeWorker *pMW,               /* Merge worker object to append page to */
  int bSep,                       /* True to append to the separators array */
  int iFPtr                       /* Pointer value for footer of new page */
){
  int rc = LSM_OK;                /* Return code */
  Page *pNext = 0;                /* New page appended to run */
  lsm_db *pDb = pMW->pDb;         /* Database handle */
  SortedRun *pRun;                /* Run to append to */

  assert( bSep==0 || bSep==1 );

  pRun = (bSep ? &pMW->pLevel->lhs.sep : &pMW->pLevel->lhs.run);
  rc = lsmFsSortedAppend(pDb->pFS, pDb->pWorker, pRun, &pNext);
  assert( rc!=LSM_OK || bSep || pRun->iFirst>0 );

  if( rc==LSM_OK ){
    u8 *aData;                    /* Data buffer belonging to page pNext */
    int nData;                    /* Size of aData[] in bytes */

    lsmFsPageRelease(pMW->apPage[bSep]);
    pMW->apPage[bSep] = pNext;
    pMW->pLevel->pMerge->aiOutputOff[bSep] = 0;

    aData = fsPageData(pNext, &nData);
    lsmPutU16(&aData[SEGMENT_NRECORD_OFFSET(nData)], 0);
    lsmPutU16(&aData[SEGMENT_FLAGS_OFFSET(nData)], 0);
    lsmPutU32(&aData[SEGMENT_POINTER_OFFSET(nData)], iFPtr);

    if( bSep==0 ) pMW->nWork++;
  }

  return rc;
}

/*
** Write a blob of data into an output segment being populated by a 







|


<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<











<
<
<
|







<





|

<
<
|
|
|





|
|
|






|







2939
2940
2941
2942
2943
2944
2945
2946
2947
2948



































































2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959



2960
2961
2962
2963
2964
2965
2966
2967

2968
2969
2970
2971
2972
2973
2974


2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999

  /* Write the right-hand pointer of the right-most leaf page of the 
  ** b-tree heirarchy. */
  aData = fsPageData(p->apHier[0], &nData);
  lsmPutU32(&aData[SEGMENT_POINTER_OFFSET(nData)], iKeyPg);

  /* Ensure that the SortedRun.iRoot field is correct. */
  pSeg->iRoot = lsmFsPageNumber(p->apHier[p->nHier-1]);

push_hierarchy_out:



































































  return rc;
}

static int keyszToSkip(FileSystem *pFS, int nKey){
  int nPgsz;                /* Nominal database page size */
  nPgsz = lsmFsPageSize(pFS);
  return LSM_MIN(((nKey * 4) / nPgsz), 3);
}

/*
** Advance to the next page of an output run being populated by merge-worker



** pMW. The footer of the new page is initialized to indicate that it contains
** zero records. The flags field is cleared. The page footer pointer field
** is set to iFPtr.
**
** If successful, LSM_OK is returned. Otherwise, an error code.
*/
static int mergeWorkerNextPage(
  MergeWorker *pMW,               /* Merge worker object to append page to */

  int iFPtr                       /* Pointer value for footer of new page */
){
  int rc = LSM_OK;                /* Return code */
  Page *pNext = 0;                /* New page appended to run */
  lsm_db *pDb = pMW->pDb;         /* Database handle */
  Segment *pSeg;                  /* Run to append to */



  pSeg = &pMW->pLevel->lhs;
  rc = lsmFsSortedAppend(pDb->pFS, pDb->pWorker, pSeg, &pNext);
  assert( rc!=LSM_OK || pSeg->iFirst>0 );

  if( rc==LSM_OK ){
    u8 *aData;                    /* Data buffer belonging to page pNext */
    int nData;                    /* Size of aData[] in bytes */

    lsmFsPageRelease(pMW->pPage);
    pMW->pPage = pNext;
    pMW->pLevel->pMerge->iOutputOff = 0;

    aData = fsPageData(pNext, &nData);
    lsmPutU16(&aData[SEGMENT_NRECORD_OFFSET(nData)], 0);
    lsmPutU16(&aData[SEGMENT_FLAGS_OFFSET(nData)], 0);
    lsmPutU32(&aData[SEGMENT_POINTER_OFFSET(nData)], iFPtr);

    pMW->nWork++;
  }

  return rc;
}

/*
** Write a blob of data into an output segment being populated by a 
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
    Merge *pMerge = pMW->pLevel->pMerge;
    int nCopy;                    /* Number of bytes to copy */
    u8 *aData;                    /* Pointer to buffer of current output page */
    int nData;                    /* Size of aData[] in bytes */
    int nRec;                     /* Number of records on current output page */
    int iOff;                     /* Offset in aData[] to write to */

    assert( lsmFsPageWritable(pMW->apPage[bSep]) );
   
    aData = fsPageData(pMW->apPage[bSep], &nData);
    nRec = pageGetNRec(aData, nData);
    iOff = pMerge->aiOutputOff[bSep];
    nCopy = LSM_MIN(nRem, SEGMENT_EOF(nData, nRec) - iOff);

    memcpy(&aData[iOff], &aWrite[nWrite-nRem], nCopy);
    nRem -= nCopy;

    if( nRem>0 ){
      rc = mergeWorkerNextPage(pMW, bSep, iFPtr);
    }else{
      pMerge->aiOutputOff[bSep] = iOff + nCopy;
    }
  }

  return rc;
}


static int mergeWorkerWrite(
  MergeWorker *pMW,               /* Merge worker object to write into */
  int bSep,                       /* True to write to separators array */
  int eType,                      /* One of SORTED_SEPARATOR, WRITE or DELETE */
  void *pKey, int nKey,           /* Key value */
  MultiCursor *pCsr,              /* Read value (if any) from here */
  int iPtr,                       /* Absolute value of page pointer, or 0 */
  int *piPtrOut                   /* OUT: Pointer to write to separators */
){
  int rc = LSM_OK;                /* Return code */
  Merge *pMerge;                  /* Persistent part of level merge state */
  int nHdr;                       /* Space required for this record header */
  Page *pPg;                      /* Page to write to */
  u8 *aData;                      /* Data buffer for page pWriter->pPage */
  int nData;                      /* Size of buffer aData[] in bytes */
  int nRec;                       /* Number of records on page pPg */
  int iFPtr;                      /* Value of pointer in footer of pPg */
  int iRPtr;                      /* Value of pointer written into record */
  int iOff;                       /* Current write offset within page pPg */
  SortedRun *pRun;                /* Run being written to */
  int flags = 0;                  /* If != 0, flags value for page footer */
  void *pVal;
  int nVal;

  assert( bSep==0 || bSep==1 );
  assert( bSep==0 || rtIsSeparator(eType) );

  pMerge = pMW->pLevel->pMerge;    
  pRun = (bSep ? &pMW->pLevel->lhs.sep : &pMW->pLevel->lhs.run);

  pPg = pMW->apPage[bSep];
  aData = fsPageData(pPg, &nData);
  nRec = pageGetNRec(aData, nData);
  iFPtr = pageGetPtr(aData, nData);

  /* If iPtr is 0, set it to the same value as the absolute pointer 
  ** stored as part of the previous record.  */
  if( iPtr==0 ){







|

|

|






|

|









<
















|




<
<
<

|

|







3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045

3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066



3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
    Merge *pMerge = pMW->pLevel->pMerge;
    int nCopy;                    /* Number of bytes to copy */
    u8 *aData;                    /* Pointer to buffer of current output page */
    int nData;                    /* Size of aData[] in bytes */
    int nRec;                     /* Number of records on current output page */
    int iOff;                     /* Offset in aData[] to write to */

    assert( lsmFsPageWritable(pMW->pPage) );
   
    aData = fsPageData(pMW->pPage, &nData);
    nRec = pageGetNRec(aData, nData);
    iOff = pMerge->iOutputOff;
    nCopy = LSM_MIN(nRem, SEGMENT_EOF(nData, nRec) - iOff);

    memcpy(&aData[iOff], &aWrite[nWrite-nRem], nCopy);
    nRem -= nCopy;

    if( nRem>0 ){
      rc = mergeWorkerNextPage(pMW, iFPtr);
    }else{
      pMerge->iOutputOff = iOff + nCopy;
    }
  }

  return rc;
}


static int mergeWorkerWrite(
  MergeWorker *pMW,               /* Merge worker object to write into */

  int eType,                      /* One of SORTED_SEPARATOR, WRITE or DELETE */
  void *pKey, int nKey,           /* Key value */
  MultiCursor *pCsr,              /* Read value (if any) from here */
  int iPtr,                       /* Absolute value of page pointer, or 0 */
  int *piPtrOut                   /* OUT: Pointer to write to separators */
){
  int rc = LSM_OK;                /* Return code */
  Merge *pMerge;                  /* Persistent part of level merge state */
  int nHdr;                       /* Space required for this record header */
  Page *pPg;                      /* Page to write to */
  u8 *aData;                      /* Data buffer for page pWriter->pPage */
  int nData;                      /* Size of buffer aData[] in bytes */
  int nRec;                       /* Number of records on page pPg */
  int iFPtr;                      /* Value of pointer in footer of pPg */
  int iRPtr;                      /* Value of pointer written into record */
  int iOff;                       /* Current write offset within page pPg */
  Segment *pSeg;                  /* Segment being written */
  int flags = 0;                  /* If != 0, flags value for page footer */
  void *pVal;
  int nVal;




  pMerge = pMW->pLevel->pMerge;    
  pSeg = &pMW->pLevel->lhs;

  pPg = pMW->pPage;
  aData = fsPageData(pPg, &nData);
  nRec = pageGetNRec(aData, nData);
  iFPtr = pageGetPtr(aData, nData);

  /* If iPtr is 0, set it to the same value as the absolute pointer 
  ** stored as part of the previous record.  */
  if( iPtr==0 ){
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
  rc = lsmMCursorValue(pCsr, &pVal, &nVal);
  if( rc==LSM_OK ){
    nHdr = 1 + lsmVarintLen32(iRPtr) + lsmVarintLen32(nKey);
    if( rtIsWrite(eType) ) nHdr += lsmVarintLen32(nVal);

    /* If the entire header will not fit on page pPg, or if page pPg is 
     ** marked read-only, advance to the next page of the output run. */
    iOff = pMerge->aiOutputOff[bSep];
    if( iOff<0 || iOff+nHdr > SEGMENT_EOF(nData, nRec+1) ){
      iFPtr = iFPtr + (nRec ? pageGetRecordPtr(aData, nData, nRec-1) : 0);
      iRPtr = iPtr - iFPtr;
      iOff = 0;
      nRec = 0;
      rc = mergeWorkerNextPage(pMW, bSep, iFPtr);
      pPg = pMW->apPage[bSep];
    }
  }

  /* If this record header will be the first on the page, and the page is 
  ** not the very first in the entire run, special actions may need to be 
  ** taken:
  **
  **   * If currently writing the main run, *piPtrOut should be set to
  **     the current page number. The caller will add a key to the separators
  **     array that points to the current page.
  **
  **   * If currently writing the separators array, push a copy of the key
  **     into the b-tree hierarchy.
  */
  if( rc==LSM_OK && nRec==0 && pRun->iFirst!=pRun->iLast ){
    assert( pMerge->nSkip>=0 );

    if( (bSep && pMW->bFlush==0) || (bSep==0 && pMerge->nSkip==0) ){
      Pgno iPg = lsmFsPageNumber(pPg);
      rc = mergeWorkerPushHierarchy(pMW, bSep, iPg, rtTopic(eType), pKey, nKey);
    }
    if( bSep==0 ){
      if( pMerge->nSkip ){
        pMerge->nSkip--;
        flags = PGFTR_SKIP_THIS_FLAG;
      }else{
        *piPtrOut = lsmFsPageNumber(pPg);
        pMerge->nSkip = keyszToSkip(pMW->pDb->pFS, nKey);
      }
      if( pMerge->nSkip ) flags |= PGFTR_SKIP_NEXT_FLAG;
    }
  }

  /* Update the output segment */
  if( rc==LSM_OK ){
    aData = fsPageData(pPg, &nData);

    /* Update the page footer. */
    lsmPutU16(&aData[SEGMENT_NRECORD_OFFSET(nData)], nRec+1);
    lsmPutU16(&aData[SEGMENT_CELLPTR_OFFSET(nData, nRec)], iOff);
    if( flags ) lsmPutU16(&aData[SEGMENT_FLAGS_OFFSET(nData)], flags);

    /* Write the entry header into the current page. */
    aData[iOff++] = eType;                                               /* 1 */
    iOff += lsmVarintPut32(&aData[iOff], iRPtr);                         /* 2 */
    iOff += lsmVarintPut32(&aData[iOff], nKey);                          /* 3 */
    if( rtIsWrite(eType) ) iOff += lsmVarintPut32(&aData[iOff], nVal);   /* 4 */
    pMerge->aiOutputOff[bSep] = iOff;

    /* Write the key and data into the segment. */
    assert( iFPtr==pageGetPtr(aData, nData) );
    rc = mergeWorkerData(pMW, bSep, iFPtr+iRPtr, pKey, nKey);
    if( rc==LSM_OK && rtIsWrite(eType) ){
      if( rtTopic(eType)==0 ) rc = lsmMCursorValue(pCsr, &pVal, &nVal);
      if( rc==LSM_OK ){
        rc = mergeWorkerData(pMW, bSep, iFPtr+iRPtr, pVal, nVal);
      }
    }
  }

  return rc;
}








|





|
|














|


|

|

<
|
|
|
|
|
|
|
|
<
















|



|



|







3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135

3136
3137
3138
3139
3140
3141
3142
3143

3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
  rc = lsmMCursorValue(pCsr, &pVal, &nVal);
  if( rc==LSM_OK ){
    nHdr = 1 + lsmVarintLen32(iRPtr) + lsmVarintLen32(nKey);
    if( rtIsWrite(eType) ) nHdr += lsmVarintLen32(nVal);

    /* If the entire header will not fit on page pPg, or if page pPg is 
     ** marked read-only, advance to the next page of the output run. */
    iOff = pMerge->iOutputOff;
    if( iOff<0 || iOff+nHdr > SEGMENT_EOF(nData, nRec+1) ){
      iFPtr = iFPtr + (nRec ? pageGetRecordPtr(aData, nData, nRec-1) : 0);
      iRPtr = iPtr - iFPtr;
      iOff = 0;
      nRec = 0;
      rc = mergeWorkerNextPage(pMW, iFPtr);
      pPg = pMW->pPage;
    }
  }

  /* If this record header will be the first on the page, and the page is 
  ** not the very first in the entire run, special actions may need to be 
  ** taken:
  **
  **   * If currently writing the main run, *piPtrOut should be set to
  **     the current page number. The caller will add a key to the separators
  **     array that points to the current page.
  **
  **   * If currently writing the separators array, push a copy of the key
  **     into the b-tree hierarchy.
  */
  if( rc==LSM_OK && nRec==0 && pSeg->iFirst!=pSeg->iLast ){
    assert( pMerge->nSkip>=0 );

    if( pMerge->nSkip==0 ){
      Pgno iPg = lsmFsPageNumber(pPg);
      rc = mergeWorkerPushHierarchy(pMW, 0, iPg, rtTopic(eType), pKey, nKey);
    }

    if( pMerge->nSkip ){
      pMerge->nSkip--;
      flags = PGFTR_SKIP_THIS_FLAG;
    }else{
      *piPtrOut = lsmFsPageNumber(pPg);
      pMerge->nSkip = keyszToSkip(pMW->pDb->pFS, nKey);
    }
    if( pMerge->nSkip ) flags |= PGFTR_SKIP_NEXT_FLAG;

  }

  /* Update the output segment */
  if( rc==LSM_OK ){
    aData = fsPageData(pPg, &nData);

    /* Update the page footer. */
    lsmPutU16(&aData[SEGMENT_NRECORD_OFFSET(nData)], nRec+1);
    lsmPutU16(&aData[SEGMENT_CELLPTR_OFFSET(nData, nRec)], iOff);
    if( flags ) lsmPutU16(&aData[SEGMENT_FLAGS_OFFSET(nData)], flags);

    /* Write the entry header into the current page. */
    aData[iOff++] = eType;                                               /* 1 */
    iOff += lsmVarintPut32(&aData[iOff], iRPtr);                         /* 2 */
    iOff += lsmVarintPut32(&aData[iOff], nKey);                          /* 3 */
    if( rtIsWrite(eType) ) iOff += lsmVarintPut32(&aData[iOff], nVal);   /* 4 */
    pMerge->iOutputOff = iOff;

    /* Write the key and data into the segment. */
    assert( iFPtr==pageGetPtr(aData, nData) );
    rc = mergeWorkerData(pMW, 0, iFPtr+iRPtr, pKey, nKey);
    if( rc==LSM_OK && rtIsWrite(eType) ){
      if( rtTopic(eType)==0 ) rc = lsmMCursorValue(pCsr, &pVal, &nVal);
      if( rc==LSM_OK ){
        rc = mergeWorkerData(pMW, 0, iFPtr+iRPtr, pVal, nVal);
      }
    }
  }

  return rc;
}

3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384

3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
    if( bBtree && pMerge->nInput ){
      assert( i==pCsr->nSegCsr );
      btreeCursorPosition(pCsr->pBtCsr, &pMerge->aInput[i]);
    }
  }

  lsmMCursorClose(pCsr);
  lsmFsPageRelease(pMW->apPage[0]);
  lsmFsPageRelease(pMW->apPage[1]);

  for(i=0; i<2; i++){
    Hierarchy *p = &pMW->aHier[i];
    int iPg;
    for(iPg=0; iPg<p->nHier; iPg++){
      lsmFsPageRelease(p->apHier[iPg]);
    }
    lsmFree(pMW->pDb->pEnv, p->apHier);
    p->apHier = 0;
    p->nHier = 0;
  }

  pMW->pCsr = 0;
  pMW->apPage[0] = 0;
  pMW->apPage[1] = 0;
}

static int mergeWorkerFirstPage(MergeWorker *pMW){
  int rc;                         /* Return code */
  SortedRun *pRun;                /* Run containing sep. keys to merge in */
  Page *pPg = 0;                  /* First page of run pRun */
  int iFPtr;                      /* Pointer value read from footer of pPg */
  MultiCursor *pCsr = pMW->pCsr;

  assert( pMW->apPage[0]==0 );

  if( pCsr->pBtCsr ){
    rc = LSM_OK;
    iFPtr = pMW->pLevel->pNext->lhs.run.iFirst;
  }else{

    pRun = pMW->pCsr->aSegCsr[pMW->pCsr->nSegCsr-1].aPtr[0].pRun;
    rc = lsmFsDbPageGet(pMW->pDb->pFS, pRun->iFirst, &pPg);
    if( rc==LSM_OK ){
      u8 *aData;                    /* Buffer for page pPg */
      int nData;                    /* Size of aData[] in bytes */
      aData = fsPageData(pPg, &nData);
      iFPtr = pageGetPtr(aData, nData);
      lsmFsPageRelease(pPg);
    }
  }

  if( rc==LSM_OK ){
    rc = mergeWorkerNextPage(pMW, 0, iFPtr);
  }

  return rc;
}

static int mergeWorkerStep(MergeWorker *pMW){
  lsm_db *pDb = pMW->pDb;       /* Database handle */







|
|


|










|
|




<
|



|



|

>
|
|










|







3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249

3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
    if( bBtree && pMerge->nInput ){
      assert( i==pCsr->nSegCsr );
      btreeCursorPosition(pCsr->pBtCsr, &pMerge->aInput[i]);
    }
  }

  lsmMCursorClose(pCsr);
  lsmFsPageRelease(pMW->pPage);
  lsmFsPageRelease(pMW->pPage);

  for(i=0; i<2; i++){
    Hierarchy *p = &pMW->hier;
    int iPg;
    for(iPg=0; iPg<p->nHier; iPg++){
      lsmFsPageRelease(p->apHier[iPg]);
    }
    lsmFree(pMW->pDb->pEnv, p->apHier);
    p->apHier = 0;
    p->nHier = 0;
  }

  pMW->pCsr = 0;
  pMW->pPage = 0;
  pMW->pPage = 0;
}

static int mergeWorkerFirstPage(MergeWorker *pMW){
  int rc;                         /* Return code */

  Page *pPg = 0;                  /* First page of run pSeg */
  int iFPtr;                      /* Pointer value read from footer of pPg */
  MultiCursor *pCsr = pMW->pCsr;

  assert( pMW->pPage==0 );

  if( pCsr->pBtCsr ){
    rc = LSM_OK;
    iFPtr = pMW->pLevel->pNext->lhs.iFirst;
  }else{
    Segment *pSeg;
    pSeg = pMW->pCsr->aSegCsr[pMW->pCsr->nSegCsr-1].aPtr[0].pSeg;
    rc = lsmFsDbPageGet(pMW->pDb->pFS, pSeg->iFirst, &pPg);
    if( rc==LSM_OK ){
      u8 *aData;                    /* Buffer for page pPg */
      int nData;                    /* Size of aData[] in bytes */
      aData = fsPageData(pPg, &nData);
      iFPtr = pageGetPtr(aData, nData);
      lsmFsPageRelease(pPg);
    }
  }

  if( rc==LSM_OK ){
    rc = mergeWorkerNextPage(pMW, iFPtr);
  }

  return rc;
}

static int mergeWorkerStep(MergeWorker *pMW){
  lsm_db *pDb = pMW->pDb;       /* Database handle */
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518

  /* If this is a separator key and we know that the output pointer has not
  ** changed, there is no point in writing an output record. Otherwise,
  ** proceed. */
  if( rtIsSeparator(eType)==0 || iPtr!=0 ){
    int iSPtr = 0;                /* Separators require a pointer here */

    if( pMW->apPage[0]==0 ){
      rc = mergeWorkerFirstPage(pMW);
    }

    /* Write the record into the main run. */
    if( rc==LSM_OK ){
      rc = mergeWorkerWrite(pMW, 0, eType, pKey, nKey, pCsr, iPtr, &iSPtr);
    }

    /* If the call to mergeWorkerWrite() above started a new page, then
    ** add a SORTED_SEPARATOR key to the separators run.  */
#if 0
    if( rc==LSM_OK && iSPtr ){

      /* If the separators array has not been started, start it now. */
      if( pMW->apPage[1]==0 ){
        assert( pSeg->run.iFirst!=0 );
        rc = mergeWorkerNextPage(pMW, 1, pSeg->run.iFirst);
        if( !pMW->bFlush ) pSeg->sep.iRoot = pSeg->sep.iFirst;
      }

      if( rc==LSM_OK ){
        int eSType;                 /* Type of record for separators array */

        /* Figure out how many (if any) keys to skip from this point. */
        assert( pMW->apPage[1] && (pSeg->sep.iFirst || pMW->bFlush) );
        pMW->pLevel->pMerge->nSkip = keyszToSkip(pDb->pFS, nKey);

        /* Write the key into the separators array. */
        eSType = rtTopic(eType) | SORTED_SEPARATOR;
        rc = mergeWorkerWrite(pMW, 1, eSType, pKey, nKey, 0, 0, iSPtr, 0);
      }
    }
#endif
  }

  /* Advance the cursor to the next input record (assuming one exists). */
  assert( lsmMCursorValid(pMW->pCsr) );
  if( rc==LSM_OK ) rc = lsmMCursorNext(pMW->pCsr);
  assert( pMW->apPage[1]==0 );

  /* If the cursor is at EOF, the merge is finished. Release all page
  ** references currently held by the merge worker and inform the 
  ** FileSystem object that no further pages will be appended to either 
  ** the main or separators array. 
  */
  if( rc==LSM_OK && !lsmMCursorValid(pMW->pCsr) ){
    if( pSeg->run.iFirst ){
      rc = lsmFsSortedFinish(pDb->pFS, &pSeg->run);
    }
    if( rc==LSM_OK && pMW->bFlush ){
      rc = mergeWorkerBuildHierarchy(pMW);
    }
    if( rc==LSM_OK && pSeg->sep.iFirst ){
      rc = lsmFsSortedFinish(pDb->pFS, &pSeg->sep);
    }

#ifdef LSM_DEBUG_EXPENSIVE
    if( rc==LSM_OK ){
      rc = assertBtreeOk(pDb, &pSeg->run);
      if( pMW->pCsr->pBtCsr ){
        SortedRun *pNext = &pMW->pLevel->pNext->lhs.run;
        rc = assertPointersOk(pDb, &pSeg->run, pNext, 0);
      }
    }
#endif

    mergeWorkerShutdown(pMW);
  }
  return rc;







|





|

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<





|







|
|
<
<
<
<
<
<




|

|
|







3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332


























3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347






3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362

  /* If this is a separator key and we know that the output pointer has not
  ** changed, there is no point in writing an output record. Otherwise,
  ** proceed. */
  if( rtIsSeparator(eType)==0 || iPtr!=0 ){
    int iSPtr = 0;                /* Separators require a pointer here */

    if( pMW->pPage==0 ){
      rc = mergeWorkerFirstPage(pMW);
    }

    /* Write the record into the main run. */
    if( rc==LSM_OK ){
      rc = mergeWorkerWrite(pMW, eType, pKey, nKey, pCsr, iPtr, &iSPtr);
    }


























  }

  /* Advance the cursor to the next input record (assuming one exists). */
  assert( lsmMCursorValid(pMW->pCsr) );
  if( rc==LSM_OK ) rc = lsmMCursorNext(pMW->pCsr);
  assert( pMW->pPage==0 );

  /* If the cursor is at EOF, the merge is finished. Release all page
  ** references currently held by the merge worker and inform the 
  ** FileSystem object that no further pages will be appended to either 
  ** the main or separators array. 
  */
  if( rc==LSM_OK && !lsmMCursorValid(pMW->pCsr) ){
    if( pSeg->iFirst ){
      rc = lsmFsSortedFinish(pDb->pFS, pSeg);






    }

#ifdef LSM_DEBUG_EXPENSIVE
    if( rc==LSM_OK ){
      rc = assertBtreeOk(pDb, pSeg);
      if( pMW->pCsr->pBtCsr ){
        Segment *pNext = &pMW->pLevel->pNext->lhs;
        rc = assertPointersOk(pDb, pSeg, pNext, 0);
      }
    }
#endif

    mergeWorkerShutdown(pMW);
  }
  return rc;
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
    lsmDbSnapshotSetLevel(pDb->pWorker, pNew);

    rc = multiCursorNew(pDb, pDb->pWorker, (pDb->pTV!=0), 0, &pCsr);
    if( rc==LSM_OK ){
      if( pNext ){
        assert( pNext->pMerge==0 || pNext->nRight>0 );
        if( pNext->pMerge==0 ){
          if( pNext->lhs.run.iRoot ){
            rc = multiCursorAddLevel(pCsr, pNext, MULTICURSOR_ADDLEVEL_LHS_SEP);

            /* This call moves any blocks occupied by separators array pDel 
            ** to the pending list. We do this here, even though pDel will be 
            ** read while building the new level, so that the blocks will be 
            ** included in the "FREELIST" entry visited by the cursor (and 
            ** written into the new top level).  */
            if( rc==LSM_OK ){
              pDel = &pNext->lhs;
              rc = lsmFsSortedDelete(pDb->pFS, pDb->pWorker, 0, &pDel->sep);
            }
          }
          iLeftPtr = pNext->lhs.run.iFirst;
        }
      }else{
        /* The new level will be the only level in the LSM. There is no reason
         ** to write out delete keys in this case.  */
        multiCursorIgnoreDelete(pCsr);
      }
    }







|

<
<
<
<
<
<


<


|







3404
3405
3406
3407
3408
3409
3410
3411
3412






3413
3414

3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
    lsmDbSnapshotSetLevel(pDb->pWorker, pNew);

    rc = multiCursorNew(pDb, pDb->pWorker, (pDb->pTV!=0), 0, &pCsr);
    if( rc==LSM_OK ){
      if( pNext ){
        assert( pNext->pMerge==0 || pNext->nRight>0 );
        if( pNext->pMerge==0 ){
          if( pNext->lhs.iRoot ){
            rc = multiCursorAddLevel(pCsr, pNext, MULTICURSOR_ADDLEVEL_LHS_SEP);






            if( rc==LSM_OK ){
              pDel = &pNext->lhs;

            }
          }
          iLeftPtr = pNext->lhs.iFirst;
        }
      }else{
        /* The new level will be the only level in the LSM. There is no reason
         ** to write out delete keys in this case.  */
        multiCursorIgnoreDelete(pCsr);
      }
    }
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
    pNew->pMerge = &merge;
    mergeworker.pDb = pDb;
    mergeworker.pLevel = pNew;
    mergeworker.pCsr = pCsr;

    /* Mark the separators array for the new level as a "phantom". */
    mergeworker.bFlush = 1;
    lsmFsPhantom(pDb->pFS, &pNew->lhs.sep);

    /* Allocate the first page of the output segment. */
    rc = mergeWorkerNextPage(&mergeworker, 0, iLeftPtr);

    /* Do the work to create the new merged segment on disk */
    if( rc==LSM_OK ) rc = lsmMCursorFirst(pCsr);
    while( rc==LSM_OK && mergeWorkerDone(&mergeworker)==0 ){
      rc = mergeWorkerStep(&mergeworker);
    }

    lsmFsPhantomFree(pDb->pFS);
    mergeWorkerShutdown(&mergeworker);
    pNew->pMerge = 0;
  }
  lsmFreelistDeltaEnd(pDb);

  /* Link the new level into the top of the tree. Delete the separators
  ** array (if any) that was merged into the new level. */
  if( rc==LSM_OK ){
    if( pDel ){
      /* lsmFsSortedDelete() has already been called on pDel->sep. So all
      ** that is required here is to zero it (so that it is not used by
      ** future LSM searches). */
      memset(&pDel->sep, 0, sizeof(SortedRun));
      pDel->run.iRoot = 0;
    }
  }else{
    lsmDbSnapshotSetLevel(pDb->pWorker, pNext);
    sortedFreeLevel(pDb->pEnv, pNew);
  }

  if( rc==LSM_OK ){







<


|







<





|
<


<
<
<
<
|







3442
3443
3444
3445
3446
3447
3448

3449
3450
3451
3452
3453
3454
3455
3456
3457
3458

3459
3460
3461
3462
3463
3464

3465
3466




3467
3468
3469
3470
3471
3472
3473
3474
    pNew->pMerge = &merge;
    mergeworker.pDb = pDb;
    mergeworker.pLevel = pNew;
    mergeworker.pCsr = pCsr;

    /* Mark the separators array for the new level as a "phantom". */
    mergeworker.bFlush = 1;


    /* Allocate the first page of the output segment. */
    rc = mergeWorkerNextPage(&mergeworker, iLeftPtr);

    /* Do the work to create the new merged segment on disk */
    if( rc==LSM_OK ) rc = lsmMCursorFirst(pCsr);
    while( rc==LSM_OK && mergeWorkerDone(&mergeworker)==0 ){
      rc = mergeWorkerStep(&mergeworker);
    }


    mergeWorkerShutdown(&mergeworker);
    pNew->pMerge = 0;
  }
  lsmFreelistDeltaEnd(pDb);

  /* Link the new level into the top of the tree. */

  if( rc==LSM_OK ){
    if( pDel ){




      pDel->iRoot = 0;
    }
  }else{
    lsmDbSnapshotSetLevel(pDb->pWorker, pNext);
    sortedFreeLevel(pDb->pEnv, pNew);
  }

  if( rc==LSM_OK ){
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
    pTopLevel = lsmDbSnapshotLevel(pDb->pWorker);
    pNew->pNext = p;
    for(pp=&pTopLevel; *pp!=pLevel; pp=&((*pp)->pNext));
    *pp = pNew;
    lsmDbSnapshotSetLevel(pDb->pWorker, pTopLevel);

    /* Determine whether or not the next separators will be linked in */
    if( pNext && pNext->pMerge==0 && pNext->lhs.run.iRoot ){
      bUseNext = 1;
    }
  }

  /* Allocate the merge object */
  nByte = sizeof(Merge) + sizeof(MergeInput) * (nMerge + bUseNext);
  pMerge = (Merge *)lsmMallocZeroRc(pDb->pEnv, nByte, &rc);
  if( pMerge ){
    pMerge->aInput = (MergeInput *)&pMerge[1];
    pMerge->nInput = nMerge + bUseNext;
    pNew->pMerge = pMerge;
  }

  *ppNew = pNew;
  return rc;
}

static int mergeWorkerLoadOutputPage(MergeWorker *pMW, int bSep){
  int rc = LSM_OK;                /* Return code */
  SortedRun *pRun;                /* Run to load page from */
  Level *pLevel;

  pLevel = pMW->pLevel;
  pRun = (bSep ? &pLevel->lhs.sep : &pLevel->lhs.run);
  if( pRun->iLast ){
    Page *pPg;
    rc = lsmFsDbPageGet(pMW->pDb->pFS, pRun->iLast, &pPg);

    while( rc==LSM_OK ){
      Page *pNext;
      u8 *aData;
      int nData;
      aData = fsPageData(pPg, &nData);
      if( (pageGetFlags(aData, nData) & SEGMENT_BTREE_FLAG)==0 ) break;
      rc = lsmFsDbPageNext(pRun, pPg, -1, &pNext);
      lsmFsPageRelease(pPg);
      pPg = pNext;
    }

    if( rc==LSM_OK ){
      pMW->apPage[bSep] = pPg;
      if( pLevel->pMerge->aiOutputOff[bSep]>=0 ) rc = lsmFsPageWrite(pPg);
    }
  }
  return rc;
}

static int mergeWorkerInit(
  lsm_db *pDb,                    /* Db connection to do merge work */







|

















|

|



|
|

|







|





|
|







3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
    pTopLevel = lsmDbSnapshotLevel(pDb->pWorker);
    pNew->pNext = p;
    for(pp=&pTopLevel; *pp!=pLevel; pp=&((*pp)->pNext));
    *pp = pNew;
    lsmDbSnapshotSetLevel(pDb->pWorker, pTopLevel);

    /* Determine whether or not the next separators will be linked in */
    if( pNext && pNext->pMerge==0 && pNext->lhs.iRoot ){
      bUseNext = 1;
    }
  }

  /* Allocate the merge object */
  nByte = sizeof(Merge) + sizeof(MergeInput) * (nMerge + bUseNext);
  pMerge = (Merge *)lsmMallocZeroRc(pDb->pEnv, nByte, &rc);
  if( pMerge ){
    pMerge->aInput = (MergeInput *)&pMerge[1];
    pMerge->nInput = nMerge + bUseNext;
    pNew->pMerge = pMerge;
  }

  *ppNew = pNew;
  return rc;
}

static int mergeWorkerLoadOutputPage(MergeWorker *pMW){
  int rc = LSM_OK;                /* Return code */
  Segment *pSeg;                  /* Run to load page from */
  Level *pLevel;

  pLevel = pMW->pLevel;
  pSeg = &pLevel->lhs;
  if( pSeg->iLast ){
    Page *pPg;
    rc = lsmFsDbPageGet(pMW->pDb->pFS, pSeg->iLast, &pPg);

    while( rc==LSM_OK ){
      Page *pNext;
      u8 *aData;
      int nData;
      aData = fsPageData(pPg, &nData);
      if( (pageGetFlags(aData, nData) & SEGMENT_BTREE_FLAG)==0 ) break;
      rc = lsmFsDbPageNext(pSeg, pPg, -1, &pNext);
      lsmFsPageRelease(pPg);
      pPg = pNext;
    }

    if( rc==LSM_OK ){
      pMW->pPage = pPg;
      if( pLevel->pMerge->iOutputOff>=0 ) rc = lsmFsPageWrite(pPg);
    }
  }
  return rc;
}

static int mergeWorkerInit(
  lsm_db *pDb,                    /* Db connection to do merge work */
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
    multiCursorReadSeparators(pCsr);
  }else{
    multiCursorIgnoreDelete(pCsr);
  }
  assert( rc!=LSM_OK || pMerge->nInput==(pCsr->nSegCsr+(pCsr->pBtCsr!=0)) );
  pMW->pCsr = pCsr;

  /* Load each of the output pages into memory. */
  if( rc==LSM_OK ) rc = mergeWorkerLoadOutputPage(pMW, 0);
  if( rc==LSM_OK ) rc = mergeWorkerLoadOutputPage(pMW, 1);

  /* Position the cursor. */
  if( rc==LSM_OK ){
    if( pMW->apPage[0]==0 ){
      /* The output array is still empty. So position the cursor at the very 
      ** start of the input.  */
      rc = multiCursorEnd(pCsr, 0);
    }else{
      /* The output array is non-empty. Position the cursor based on the
      ** page/cell data saved in the Merge.aInput[] array.  */
      int i;







|
|
<



|







3664
3665
3666
3667
3668
3669
3670
3671
3672

3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
    multiCursorReadSeparators(pCsr);
  }else{
    multiCursorIgnoreDelete(pCsr);
  }
  assert( rc!=LSM_OK || pMerge->nInput==(pCsr->nSegCsr+(pCsr->pBtCsr!=0)) );
  pMW->pCsr = pCsr;

  /* Load the current output page into memory. */
  if( rc==LSM_OK ) rc = mergeWorkerLoadOutputPage(pMW);


  /* Position the cursor. */
  if( rc==LSM_OK ){
    if( pMW->pPage==0 ){
      /* The output array is still empty. So position the cursor at the very 
      ** start of the input.  */
      rc = multiCursorEnd(pCsr, 0);
    }else{
      /* The output array is non-empty. Position the cursor based on the
      ** page/cell data saved in the Merge.aInput[] array.  */
      int i;
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
      ** from rhs segments for which the content has been completely merged
      ** into the lhs of the level.
      */
      if( rc==LSM_OK ){
        if( mergeWorkerDone(&mergeworker)==0 ){
          int iGobble = mergeworker.pCsr->aTree[1] - CURSOR_DATA_SEGMENT;
          if( iGobble<pLevel->nRight ){
            Level *pLevel = mergeworker.pCsr->aSegCsr[iGobble].pLevel;
            SegmentPtr *pGobble = &mergeworker.pCsr->aSegCsr[iGobble].aPtr[0];

            if( (pGobble->flags & PGFTR_SKIP_THIS_FLAG)==0 
             && pGobble->pRun->iRoot==0
            ){
              lsmFsGobble(pWorker, pGobble->pRun, pGobble->pPg);
            }
          }

        }else if( pLevel->lhs.run.iFirst==0 ){
          /* If the new level is completely empty, remove it from the 
          ** database snapshot. This can only happen if all input keys were
          ** annihilated. Since keys are only annihilated if the new level
          ** is the last in the linked list (contains the most ancient of
          ** database content), this guarantees that pLevel->pNext==0.  */ 

          Level *pTop;          /* Top level of worker snapshot */
          Level **pp;           /* Read/write iterator for Level.pNext list */
          assert( pLevel->pNext==0 );
          assert( segmentHasSeparators(&pLevel->lhs)==0 );

          /* Remove the level from the worker snapshot. */
          pTop = lsmDbSnapshotLevel(pWorker);
          for(pp=&pTop; *pp!=pLevel; pp=&((*pp)->pNext));
          *pp = pLevel->pNext;
          lsmDbSnapshotSetLevel(pWorker, pTop);

          /* Free the Level structure. */
          lsmFsSortedDelete(pDb->pFS, pWorker, 1, &pLevel->lhs.run);
          sortedFreeLevel(pDb->pEnv, pLevel);
        }else{
          int i;

          /* Free the separators of the next level, if required. */
          if( pLevel->pMerge->nInput > pLevel->nRight ){
            assert( pLevel->pNext );
            assert( pLevel->pNext->lhs.run.iRoot );
            lsmFsSortedDelete(pDb->pFS, pWorker, 1, &pLevel->pNext->lhs.sep);
            pLevel->pNext->lhs.run.iRoot = 0;
          }

          /* Free the right-hand-side of pLevel */
          for(i=0; i<pLevel->nRight; i++){
            lsmFsSortedDelete(pDb->pFS, pWorker, 1, &pLevel->aRhs[i].run);
            lsmFsSortedDelete(pDb->pFS, pWorker, 1, &pLevel->aRhs[i].sep);
          }
          lsmFree(pDb->pEnv, pLevel->aRhs);
          pLevel->nRight = 0;
          pLevel->aRhs = 0;

          /* Free the Merge object */
          lsmFree(pDb->pEnv, pLevel->pMerge);







<

<

|

|



|









<








|






<
|
<
|




|
<







3799
3800
3801
3802
3803
3804
3805

3806

3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823

3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838

3839

3840
3841
3842
3843
3844
3845

3846
3847
3848
3849
3850
3851
3852
      ** from rhs segments for which the content has been completely merged
      ** into the lhs of the level.
      */
      if( rc==LSM_OK ){
        if( mergeWorkerDone(&mergeworker)==0 ){
          int iGobble = mergeworker.pCsr->aTree[1] - CURSOR_DATA_SEGMENT;
          if( iGobble<pLevel->nRight ){

            SegmentPtr *pGobble = &mergeworker.pCsr->aSegCsr[iGobble].aPtr[0];

            if( (pGobble->flags & PGFTR_SKIP_THIS_FLAG)==0 
             && pGobble->pSeg->iRoot==0
            ){
              lsmFsGobble(pWorker, pGobble->pSeg, pGobble->pPg);
            }
          }

        }else if( pLevel->lhs.iFirst==0 ){
          /* If the new level is completely empty, remove it from the 
          ** database snapshot. This can only happen if all input keys were
          ** annihilated. Since keys are only annihilated if the new level
          ** is the last in the linked list (contains the most ancient of
          ** database content), this guarantees that pLevel->pNext==0.  */ 

          Level *pTop;          /* Top level of worker snapshot */
          Level **pp;           /* Read/write iterator for Level.pNext list */
          assert( pLevel->pNext==0 );


          /* Remove the level from the worker snapshot. */
          pTop = lsmDbSnapshotLevel(pWorker);
          for(pp=&pTop; *pp!=pLevel; pp=&((*pp)->pNext));
          *pp = pLevel->pNext;
          lsmDbSnapshotSetLevel(pWorker, pTop);

          /* Free the Level structure. */
          lsmFsSortedDelete(pDb->pFS, pWorker, 1, &pLevel->lhs);
          sortedFreeLevel(pDb->pEnv, pLevel);
        }else{
          int i;

          /* Free the separators of the next level, if required. */
          if( pLevel->pMerge->nInput > pLevel->nRight ){

            assert( pLevel->pNext->lhs.iRoot );

            pLevel->pNext->lhs.iRoot = 0;
          }

          /* Free the right-hand-side of pLevel */
          for(i=0; i<pLevel->nRight; i++){
            lsmFsSortedDelete(pDb->pFS, pWorker, 1, &pLevel->aRhs[i]);

          }
          lsmFree(pDb->pEnv, pLevel->aRhs);
          pLevel->nRight = 0;
          pLevel->aRhs = 0;

          /* Free the Merge object */
          lsmFree(pDb->pEnv, pLevel->pMerge);
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
}

/*
** Return a string representation of the segment passed as the only argument.
** Space for the returned string is allocated using lsmMalloc(), and should
** be freed by the caller using lsmFree().
*/
static char *segToString(lsm_env *pEnv, SortedRun *pRun, int nMin){
  int nSize = pRun->nSize;
  Pgno iRoot = pRun->iRoot;
  Pgno iFirst = pRun->iFirst;
  Pgno iLast = pRun->iLast;
  char *z;

  char *z1;
  char *z2;
  int nPad;

  z1 = lsmMallocPrintf(pEnv, "%d.%d", iFirst, iLast);







|
|
|
|
|







3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
}

/*
** Return a string representation of the segment passed as the only argument.
** Space for the returned string is allocated using lsmMalloc(), and should
** be freed by the caller using lsmFree().
*/
static char *segToString(lsm_env *pEnv, Segment *pSeg, int nMin){
  int nSize = pSeg->nSize;
  Pgno iRoot = pSeg->iRoot;
  Pgno iFirst = pSeg->iFirst;
  Pgno iLast = pSeg->iLast;
  char *z;

  char *z1;
  char *z2;
  int nPad;

  z1 = lsmMallocPrintf(pEnv, "%d.%d", iFirst, iLast);
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
}

static int fileToString(
  lsm_env *pEnv,                  /* For xMalloc() */
  char *aBuf, 
  int nBuf, 
  int nMin,
  SortedRun *pRun
){
  int i = 0;
  char *zSeg;

  zSeg = segToString(pEnv, pRun, nMin);
  i += sqlite4_snprintf(&aBuf[i], nBuf-i, "%s", zSeg);
  lsmFree(pEnv, zSeg);

  return i;
}

void sortedDumpPage(lsm_db *pDb, SortedRun *pRun, Page *pPg, int bVals){
  Blob blob = {0, 0, 0};         /* Blob used for keys */
  LsmString s;
  int i;

  int nRec;
  int iPtr;
  int flags;







|




|






|







4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
}

static int fileToString(
  lsm_env *pEnv,                  /* For xMalloc() */
  char *aBuf, 
  int nBuf, 
  int nMin,
  Segment *pSeg
){
  int i = 0;
  char *zSeg;

  zSeg = segToString(pEnv, pSeg, nMin);
  i += sqlite4_snprintf(&aBuf[i], nBuf-i, "%s", zSeg);
  lsmFree(pEnv, zSeg);

  return i;
}

void sortedDumpPage(lsm_db *pDb, Segment *pRun, Page *pPg, int bVals){
  Blob blob = {0, 0, 0};         /* Blob used for keys */
  LsmString s;
  int i;

  int nRec;
  int iPtr;
  int flags;
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
    lsmFsPageRelease(pPg);
  }

  lsmDbSnapshotRelease(pDb->pEnv, pRelease);
  return rc;
}

void sortedDumpSegment(lsm_db *pDb, SortedRun *pRun, int bVals){
  assert( pDb->xLog );
  if( pRun && pRun->iFirst ){
    char *zSeg;
    Page *pPg;

    zSeg = segToString(pDb->pEnv, pRun, 0);
    lsmLogMessage(pDb, LSM_OK, "Segment: %s", zSeg);







|







4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
    lsmFsPageRelease(pPg);
  }

  lsmDbSnapshotRelease(pDb->pEnv, pRelease);
  return rc;
}

void sortedDumpSegment(lsm_db *pDb, Segment *pRun, int bVals){
  assert( pDb->xLog );
  if( pRun && pRun->iFirst ){
    char *zSeg;
    Page *pPg;

    zSeg = segToString(pDb->pEnv, pRun, 0);
    lsmLogMessage(pDb, LSM_OK, "Segment: %s", zSeg);
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
    lsmLogMessage(pDb, LSM_OK, "Database structure (%s)", zWhy);

    for(pLevel=pTopLevel; pLevel; pLevel=pLevel->pNext){
      char zLeft[1024];
      char zRight[1024];
      int i = 0;

      SortedRun *aLeft[24];  
      SortedRun *aRight[24];

      int nLeft = 0;
      int nRight = 0;

      Segment *pSeg = &pLevel->lhs;
      if( segmentHasSeparators(pSeg) ){
        aLeft[nLeft++] = &pSeg->sep;
      }
      aLeft[nLeft++] = &pSeg->run;

      for(i=0; i<pLevel->nRight; i++){
        if( segmentHasSeparators(&pLevel->aRhs[i]) ){
          aRight[nRight++] = &pLevel->aRhs[i].sep;
        }
        aRight[nRight++] = &pLevel->aRhs[i].run;
      }

      for(i=0; i<nLeft || i<nRight; i++){
        int iPad = 0;
        char zLevel[32];
        zLeft[0] = '\0';
        zRight[0] = '\0';







|
|





<
|
|
<
<

<
|
<
<







4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327

4328
4329


4330

4331


4332
4333
4334
4335
4336
4337
4338
    lsmLogMessage(pDb, LSM_OK, "Database structure (%s)", zWhy);

    for(pLevel=pTopLevel; pLevel; pLevel=pLevel->pNext){
      char zLeft[1024];
      char zRight[1024];
      int i = 0;

      Segment *aLeft[24];  
      Segment *aRight[24];

      int nLeft = 0;
      int nRight = 0;

      Segment *pSeg = &pLevel->lhs;

      aLeft[nLeft++] = pSeg;



      for(i=0; i<pLevel->nRight; i++){

        aRight[nRight++] = &pLevel->aRhs[i];


      }

      for(i=0; i<nLeft || i<nRight; i++){
        int iPad = 0;
        char zLevel[32];
        zLeft[0] = '\0';
        zRight[0] = '\0';
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
            zLevel, iPad, "", zLeft, zRight
        );
      }

      iLevel++;
    }

#if 0
    lsmLogMessage(pDb, LSM_OK, "Block map", zWhy);
    for(pLevel=pDb->pLevel; pLevel; pLevel=pLevel->pNext){
      int iRhs;
      lsmFsDumpBlockmap(pDb, pLevel->lhs.pSep);
      lsmFsDumpBlockmap(pDb, pLevel->lhs.pRun);
      for(iRhs=0; iRhs<pLevel->nRight; iRhs++){
        lsmFsDumpBlockmap(pDb, pLevel->aRhs[iRhs].pSep);
        lsmFsDumpBlockmap(pDb, pLevel->aRhs[iRhs].pRun);
      }
    }
    lsmFsDumpBlocklists(pDb);
#endif

    if( bKeys ){
      for(pLevel=pTopLevel; pLevel; pLevel=pLevel->pNext){
        int i;
        sortedDumpSegment(pDb, &pLevel->lhs.sep, 0);
        sortedDumpSegment(pDb, &pLevel->lhs.run, bVals);
        for(i=0; i<pLevel->nRight; i++){
          if( pLevel->aRhs[i].sep.iFirst>0 ){
            sortedDumpSegment(pDb, &pLevel->aRhs[i].sep, 0);
          }
          sortedDumpSegment(pDb, &pLevel->aRhs[i].run, bVals);
        }
      }
    }
  }

  if( pSnap==0 ){
    lsmDbSnapshotRelease(pDb->pEnv, pDump);







<
<
<
<
<
<
<
<
<
<
<
<
<
<



|
<

<
|
<
<







4358
4359
4360
4361
4362
4363
4364














4365
4366
4367
4368

4369

4370


4371
4372
4373
4374
4375
4376
4377
            zLevel, iPad, "", zLeft, zRight
        );
      }

      iLevel++;
    }















    if( bKeys ){
      for(pLevel=pTopLevel; pLevel; pLevel=pLevel->pNext){
        int i;
        sortedDumpSegment(pDb, &pLevel->lhs, bVals);

        for(i=0; i<pLevel->nRight; i++){

          sortedDumpSegment(pDb, &pLevel->aRhs[i], bVals);


        }
      }
    }
  }

  if( pSnap==0 ){
    lsmDbSnapshotRelease(pDb->pEnv, pDump);
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
  int rc = LSM_OK;
  Level *p;

  assert( pDb->pWorker );
  for(p=lsmDbSnapshotLevel(pDb->pWorker); p && rc==LSM_OK; p=p->pNext){
    Merge *pMerge = p->pMerge;
    if( pMerge ){
      pMerge->aiOutputOff[0] = -1;
      pMerge->aiOutputOff[1] = -1;
      pMerge->abHierReadonly[0] = 1;
      pMerge->abHierReadonly[1] = 1;
    }
  }

  return LSM_OK;
}

void lsmSortedSaveTreeCursors(lsm_db *pDb){
  MultiCursor *pCsr;
  for(pCsr=pDb->pCsr; pCsr; pCsr=pCsr->pNext){
    lsmTreeCursorSave(pCsr->pTreeCsr);
  }
}

#ifdef LSM_DEBUG_EXPENSIVE */
/*
** This function is only included in the build if LSM_DEBUG_EXPENSIVE is 
** defined. Its only purpose is to evaluate various assert() statements to 
** verify that the database is well formed in certain respects.
**
** More specifically, it checks that the array pOne contains the required 
** pointers to pTwo. Array pTwo must be a main array. pOne may be either a 
** separators array or another main array. If pOne does not contain the 
** correct set of pointers, an assert() statement fails.
*/
static int assertPointersOk(
  lsm_db *pDb,                    /* Database handle */
  SortedRun *pOne,                /* Run containing pointers */
  SortedRun *pTwo,                /* Run containing pointer targets */
  int bRhs                        /* True if pTwo may have been Gobble()d */
){
  int rc = LSM_OK;                /* Error code */
  SegmentPtr ptr1;                /* Iterates through pOne */
  SegmentPtr ptr2;                /* Iterates through pTwo */
  Pgno iPrev;

  assert( pOne && pTwo );

  memset(&ptr1, 0, sizeof(ptr1));
  memset(&ptr2, 0, sizeof(ptr1));
  ptr1.pRun = pOne;
  ptr2.pRun = pTwo;
  segmentPtrEndPage(pDb->pFS, &ptr1, 0, &rc);
  segmentPtrEndPage(pDb->pFS, &ptr2, 0, &rc);

  /* Check that the footer pointer of the first page of pOne points to
  ** the first page of pTwo. */
  iPrev = pTwo->iFirst;
  if( ptr1.iPtr!=iPrev && !bRhs ){







|
<
|
<













|












|
|











|
|







4392
4393
4394
4395
4396
4397
4398
4399

4400

4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
  int rc = LSM_OK;
  Level *p;

  assert( pDb->pWorker );
  for(p=lsmDbSnapshotLevel(pDb->pWorker); p && rc==LSM_OK; p=p->pNext){
    Merge *pMerge = p->pMerge;
    if( pMerge ){
      pMerge->iOutputOff = -1;

      pMerge->bHierReadonly = 1;

    }
  }

  return LSM_OK;
}

void lsmSortedSaveTreeCursors(lsm_db *pDb){
  MultiCursor *pCsr;
  for(pCsr=pDb->pCsr; pCsr; pCsr=pCsr->pNext){
    lsmTreeCursorSave(pCsr->pTreeCsr);
  }
}

#ifdef LSM_DEBUG_EXPENSIVE
/*
** This function is only included in the build if LSM_DEBUG_EXPENSIVE is 
** defined. Its only purpose is to evaluate various assert() statements to 
** verify that the database is well formed in certain respects.
**
** More specifically, it checks that the array pOne contains the required 
** pointers to pTwo. Array pTwo must be a main array. pOne may be either a 
** separators array or another main array. If pOne does not contain the 
** correct set of pointers, an assert() statement fails.
*/
static int assertPointersOk(
  lsm_db *pDb,                    /* Database handle */
  Segment *pOne,                  /* Segment containing pointers */
  Segment *pTwo,                  /* Segment containing pointer targets */
  int bRhs                        /* True if pTwo may have been Gobble()d */
){
  int rc = LSM_OK;                /* Error code */
  SegmentPtr ptr1;                /* Iterates through pOne */
  SegmentPtr ptr2;                /* Iterates through pTwo */
  Pgno iPrev;

  assert( pOne && pTwo );

  memset(&ptr1, 0, sizeof(ptr1));
  memset(&ptr2, 0, sizeof(ptr1));
  ptr1.pSeg = pOne;
  ptr2.pSeg = pTwo;
  segmentPtrEndPage(pDb->pFS, &ptr1, 0, &rc);
  segmentPtrEndPage(pDb->pFS, &ptr2, 0, &rc);

  /* Check that the footer pointer of the first page of pOne points to
  ** the first page of pTwo. */
  iPrev = pTwo->iFirst;
  if( ptr1.iPtr!=iPrev && !bRhs ){
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
** verify that the database is well formed in certain respects.
**
** More specifically, it checks that the b-tree embedded in array pRun
** contains the correct keys. If not, an assert() fails.
*/
static int assertBtreeOk(
  lsm_db *pDb,
  SortedRun *pRun
){
  int rc = LSM_OK;                /* Return code */
  if( pRun->iRoot ){
    Blob blob = {0, 0, 0};        /* Buffer used to cache overflow keys */
    FileSystem *pFS = pDb->pFS;   /* File system to read from */
    Page *pPg = 0;                /* Main run page */
    BtreeCursor *pCsr = 0;        /* Btree cursor */

    rc = btreeCursorNew(pDb, pRun, &pCsr);
    if( rc==LSM_OK ){
      rc = btreeCursorFirst(pCsr);
    }
    if( rc==LSM_OK ){
      rc = lsmFsDbPageGet(pFS, pRun->iFirst, &pPg);
    }

    while( rc==LSM_OK ){
      Page *pNext;
      u8 *aData;
      int nData;
      int flags;

      rc = lsmFsDbPageNext(pRun, pPg, 1, &pNext);
      lsmFsPageRelease(pPg);
      pPg = pNext;
      if( pPg==0 ) break;
      aData = fsPageData(pPg, &nData);
      flags = pageGetFlags(aData, nData);
      if( rc==LSM_OK 
       && 0==((SEGMENT_BTREE_FLAG|PGFTR_SKIP_THIS_FLAG) & flags)







|


|





|




|








|







4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
** verify that the database is well formed in certain respects.
**
** More specifically, it checks that the b-tree embedded in array pRun
** contains the correct keys. If not, an assert() fails.
*/
static int assertBtreeOk(
  lsm_db *pDb,
  Segment *pSeg
){
  int rc = LSM_OK;                /* Return code */
  if( pSeg->iRoot ){
    Blob blob = {0, 0, 0};        /* Buffer used to cache overflow keys */
    FileSystem *pFS = pDb->pFS;   /* File system to read from */
    Page *pPg = 0;                /* Main run page */
    BtreeCursor *pCsr = 0;        /* Btree cursor */

    rc = btreeCursorNew(pDb, pSeg, &pCsr);
    if( rc==LSM_OK ){
      rc = btreeCursorFirst(pCsr);
    }
    if( rc==LSM_OK ){
      rc = lsmFsDbPageGet(pFS, pSeg->iFirst, &pPg);
    }

    while( rc==LSM_OK ){
      Page *pNext;
      u8 *aData;
      int nData;
      int flags;

      rc = lsmFsDbPageNext(pSeg, pPg, 1, &pNext);
      lsmFsPageRelease(pPg);
      pPg = pNext;
      if( pPg==0 ) break;
      aData = fsPageData(pPg, &nData);
      flags = pageGetFlags(aData, nData);
      if( rc==LSM_OK 
       && 0==((SEGMENT_BTREE_FLAG|PGFTR_SKIP_THIS_FLAG) & flags)