SQLite4
Check-in [daa6942834]
Not logged in

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

Overview
Comment:Add (partially working) code for deleting keys to lsm_tree.c. Required for range-deletes.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | range-delete
Files: files | file ages | folders
SHA1: daa69428346be101a84b93ad7eebfa5ab6674e9b
User & Date: dan 2012-10-03 20:05:09
Context
2012-10-06
20:38
Add tests for range-deletes. Fix some things. Still doesn't work properly. check-in: 178f7d5eca user: dan tags: range-delete
2012-10-03
20:05
Add (partially working) code for deleting keys to lsm_tree.c. Required for range-deletes. check-in: daa6942834 user: dan tags: range-delete
09:24
Minor changes to the lsmperf.tcl script. check-in: 45e59053e7 user: dan tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/lsm.h.

376
377
378
379
380
381
382
383
384



385
386
387
388
389
390
391
/*
** Delete a value from the database. No error is returned if the specified
** key value does not exist in the database.
*/
int lsm_delete(lsm_db *, const void *pKey, int nKey);

/*
** Delete all database entries with keys that are greater than or equal to
** (pKey1/nKey1) and smaller than or equal to (pKey2/nKey2).



*/
int lsm_delete_range(lsm_db *, 
    const void *pKey1, int nKey1, const void *pKey2, int nKey2
);

/*
** The lsm_tree_size() function reports on the current state of the 







|
|
>
>
>







376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
/*
** Delete a value from the database. No error is returned if the specified
** key value does not exist in the database.
*/
int lsm_delete(lsm_db *, const void *pKey, int nKey);

/*
** Delete all database entries with keys that are greater than (pKey1/nKey1) 
** and smaller than (pKey2/nKey2). Note that keys (pKey1/nKey1) and
** (pKey2/nKey2) themselves, if they exist in the database, are not deleted.
**
** Return LSM_OK if successful, or an LSM error code otherwise.
*/
int lsm_delete_range(lsm_db *, 
    const void *pKey1, int nKey1, const void *pKey2, int nKey2
);

/*
** The lsm_tree_size() function reports on the current state of the 

Changes to src/lsm_tree.c.

96
97
98
99
100
101
102

































103
104
105
106
107
108
109
110
111
112

113
114
115
116
117
118
119
....
1340
1341
1342
1343
1344
1345
1346

























































































































































































































































1347
1348
1349
1350
1351
1352
1353

struct TreeOld {
  u32 iShmid;                     /* Last shared-memory chunk in use by old */
  u32 iRoot;                      /* Offset of root node in shm file */
  u32 nHeight;                    /* Height of tree structure */
};


































/*
** Container for a key-value pair. Within the *-shm file, each key/value
** pair is stored in a single allocation (which may not actually be 
** contiguous in memory). Layout is the TreeKey structure, followed by
** the nKey bytes of key blob, followed by the nValue bytes of value blob
** (if nValue is non-negative).
*/
struct TreeKey {
  int nKey;                       /* Size of pKey in bytes */
  int nValue;                     /* Size of pValue. Or negative. */

};

#define TK_KEY(p) ((void *)&(p)[1])
#define TK_VAL(p) ((void *)(((u8 *)&(p)[1]) + (p)->nKey))

/*
** A single tree node. A node structure may contain up to 3 key/value
................................................................................

#if 0
  dump_tree_contents(pDb, "after");
#endif
  assert_tree_looks_ok(rc, pTree);
  return rc;
}


























































































































































































































































/*
** Return, in bytes, the amount of memory currently used by the tree 
** structure.
*/
int lsmTreeSize(lsm_db *pDb){
  return pDb->treehdr.nByte;







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










>







 







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







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
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
....
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
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
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
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
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
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

struct TreeOld {
  u32 iShmid;                     /* Last shared-memory chunk in use by old */
  u32 iRoot;                      /* Offset of root node in shm file */
  u32 nHeight;                    /* Height of tree structure */
};

/*
** Flags for the TreeKey.flags variable.
*/
#define LSM_START_DELETE 0x01     /* Start of delete range */
#define LSM_END_DELETE   0x02     /* End of delete range */
#define LSM_POINT_DELETE 0x04     /* Delete this key */
#define LSM_INSERT       0x08     /* Insert this key and value */

#ifndef NDEBUG
/*
** assert() that a TreeKey.flags value is sane. Usage:
**
**   assert( assertFlagsOk(pTreeKey->flags) );
*/
static int assertFlagsOk(u8 keyflags){
  /* At least one flag must be set. Otherwise, what is this key doing? */
  assert( keyflags!=0 );

  /* The POINT_DELETE and INSERT flags cannot both be set. */
  assert( (keyflags & LSM_POINT_DELETE)==0 || (keyflags & LSM_INSERT)==0 );

  /* If both the START_DELETE and END_DELETE flags are set, then the INSERT
  ** flag must also be set. In other words - the three DELETE flags cannot
  ** all be set */
  assert( (keyflags & LSM_END_DELETE)==0 
       || (keyflags & LSM_START_DELETE)==0 
       || (keyflags & LSM_POINT_DELETE)==0 
  );

  return 1;
}
#endif

/*
** Container for a key-value pair. Within the *-shm file, each key/value
** pair is stored in a single allocation (which may not actually be 
** contiguous in memory). Layout is the TreeKey structure, followed by
** the nKey bytes of key blob, followed by the nValue bytes of value blob
** (if nValue is non-negative).
*/
struct TreeKey {
  int nKey;                       /* Size of pKey in bytes */
  int nValue;                     /* Size of pValue. Or negative. */
  int flags;
};

#define TK_KEY(p) ((void *)&(p)[1])
#define TK_VAL(p) ((void *)(((u8 *)&(p)[1]) + (p)->nKey))

/*
** A single tree node. A node structure may contain up to 3 key/value
................................................................................

#if 0
  dump_tree_contents(pDb, "after");
#endif
  assert_tree_looks_ok(rc, pTree);
  return rc;
}

static int treeDeleteEntry(lsm_db *db, TreeCursor *pCsr, u32 iNewptr){
  TreeRoot *p = &pDb->treehdr.root;
  TreeNode *pNode = pCsr->apTreeNode[pCsr->iNode];
  int iSlot = pCsr->aiCell[pCsr->iNode];
  int bLeaf;
  int rc = LSM_OK;

  assert( pNode->aiKeyPtr[1] );
  assert( pNode->aiKeyPtr[iSlot] );
  assert( iSlot==0 || iSlot==1 || iSlot==2 );
  assert( (pCsr->iNode==(db->treehdr.root.nHeight-1))==(iPtr==0) );

  bLeaf = (pCsr->iNode==(p->nHeight-1) && p->nHeight>1);
  
  if( pNode->aiKeyPtr[0] || pNode->aiKeyPtr[2] ){
    /* There are currently at least 2 keys on this leaf. So just create
    ** a new copy of the leaf with one of the keys removed. If the leaf
    ** happens to be the root node of the tree, allocate an entire 
    ** TreeNode structure instead of just a TreeLeaf.  */
    TreeNode *pNew;
    u32 iNew;

    if( bLeaf ){
      pNew = (TreeNode *)newTreeLeaf(db, &iNew, &rc);
    }else{
      pNew = newTreeNode(db, &iNew, &rc);
    }
    if( pNew ){
      int i;
      int iOut = 1;
      for(i=0; i<4; i++){
        if( i==iSlot ){
          i++;
          if( bLeaf==0 ) pNew->aiChildPtr[iOut] = iNewptr;
          if( i<3 ) pNew->aiKeyPtr[iOut] = pNode->aiKeyPtr[i];
          iOut++;
        }else{
          if( pNode->aiChildPtr[i] ){
            if( bLeaf==0 ) pNew->aiChildPtr[iOut] = pNode->aiChildPtr[i];
            if( i<3 && iOut<3 ) pNew->aiKeyPtr[iOut] = pNode->aiKeyPtr[i];
            iOut++;
          }
        }
      }
      assert( iOut<=3 );
      pCsr->iNode--;
      rc = treeUpdatePtr(pDb, pCsr, iNew);
    }

  }else if( pCsr->iNode==0 ){
    /* Removing the only key in the root node. iNewptr is the new root. */
    assert( iSlot==1 );
    pDb->treehdr.root.iRoot = iNewptr;

  }else{
    /* There is only one key on this leaf and the leaf is not the root
    ** node. Find a peer for this leaf. Then redistribute the contents of
    ** the peer and the parent cell between the parent and either one or
    ** two new leaves.  */
    TreeNode *pParent;            /* Parent tree node */
    int iPSlot;
    u32 iPeer;                    /* Pointer to peer leaf node */
    int iDir;
    TreeNode *pPeer;              /* The peer leaf node */
    TreeNode *pNew1; u32 iNew1;   /* First new leaf node */

    assert( iSlot==1 );

    pParent = pCsr->apTreeNode[pCsr->iNode-1];
    iPSlot = pCsr->aiCell[pCsr->iNode-1];

    if( iPSlot>0 && pParent->aiChildPtr[iPSlot-1] ){
      iDir = -1;
    }else{
      iDir = +1;
    }
    iPeer = pParent->aiChildPtr[iPSlot+iDir];
    pPeer = (TreeLeaf *)treeShmptr(pDb, iPeer, &rc);
    assert( pLeaf==(TreeLeaf*)treeShmptr(pDb,pParent->aiChildPtr[iPSlot],&rc) );

    /* Allocate the first new leaf node. This is always required. */
    if( bLeaf ){
      pNew1 = (TreeNode *)newTreeLeaf(db, &iNew1, &rc);
    }else{
      pNew1 = (TreeNode *)newTreeNode(db, &iNew1, &rc);
    }

    if( pPeer->aiKeyPtr[0] && pPeer->aiKeyPtr[1] ){
      /* Peer node is completely full. This means that two new leaf nodes
      ** and a new parent node are required. */

      TreeNew *pNew2; u32 iNew2;  /* Second new leaf node */
      TreeNew *pNewP; u32 iNewP;  /* New parent node */

      if( bLeaf ){
        pNew2 = (TreeNode *)newTreeLeaf(db, &iNew2, &rc);
      }else{
        pNew2 = (TreeNode *)newTreeNode(db, &iNew2, &rc);
      }
      pNewP = copyTreeNode(pDb, pParent, &iNewP, &rc);

      if( iDir==-1 ){
        pNew1->aiKeyPtr[1] = pPeer->aiKeyPtr[0];
        if( bLeaf==0 ){
          pNew1->aiChildPtr[1] = pPeer->aiChildPtr[0];
          pNew1->aiChildPtr[2] = pPeer->aiChildPtr[1];
        }

        pNewP->aiChildPtr[iPSlot-1] = iNew1;
        pNewP->aiKeyPtr[iPSlot-1] = pPeer->aiKeyPtr[1];
        pNewP->aiChildPtr[iPSlot] = iNew2;

        pNew2->aiKeyPtr[0] = pPeer->aiKeyPtr[2];
        pNew2->aiKeyPtr[1] = pParent->aiKeyPtr[iPSlot-1];
        if( bLeaf==0 ){
          pNew2->aiChildPtr[0] = pPeer->aiChildPtr[2];
          pNew2->aiChildPtr[1] = pPeer->aiChildPtr[3];
          pNew2->aiChildPtr[2] = iNewptr;
        }
      }else{
        pNew1->aiKeyPtr[1] = pParent->aiKeyPtr[iPSlot];
        if( bLeaf==0 ){
          pNew1->aiChildPtr[1] = iNewptr;
          pNew1->aiChildPtr[2] = pPeer->aiChildPtr[0];
        }

        pNewP->aiChildPtr[iPSlot] = iNew1;
        pNewP->aiKeyPtr[iPSlot] = pPeer->aiKeyPtr[0];
        pNewP->aiChildPtr[iPSlot+1] = iNew2;

        pNew2->aiKeyPtr[0] = pPeer->aiKeyPtr[1];
        pNew2->aiKeyPtr[1] = pPeer->aiKeyPtr[2];
        if( bLeaf==0 ){
          pNew2->aiChildPtr[0] = pPeer->aiChildPtr[1];
          pNew2->aiChildPtr[1] = pPeer->aiChildPtr[2];
          pNew2->aiChildPtr[2] = pPeer->aiChildPtr[3];
        }
      }
      assert( pCsr->iNode>=1 );
      pCsr->iNode -= 2;
      if( rc==LSM_OK ){
        rc = treeUpdatePtr(pDb, pCsr, iNewP);
      }
    }else{
      int iKOut = 0;
      int iPOut = 0;
      int i;

      pCsr->iNode--;

      if( iDir==1 ){
        pNew1->aiKeyPtr[iKOut++] = pParent->aiKeyPtr[iPSlot];
        if( bLeaf==0 ) pNew->aiChildPtr[iPOut] = iNewptr;
      }
      for(i=0; i<3; i++){
        if( pPeer->aiKeyPtr[i] ){
          pNew1->aiKeyPtr[iKOut++] = pPeer->aiKeyPtr[i];
        }
      }
      if( bLeaf==0 ){
        for(i=0; i<4; i++){
          if( pPeer->aiChildPtr[i] ){
            pNew1->aiChildPtr[iPOut++] = pPeer->aiChildPtr[i];
          }
        }
      }
      if( iDir==-1 ){
        iPSlot--;
        pNew1->aiKeyPtr[iKOut++] = pParent->aiKeyPtr[iPSlot];
        if( bLeaf==0 ) pNew1->aiChildPtr[iPOut++] = iNewptr;
        pCsr->aiCell[pCsr->iNode] = iPSlot;
      }

      rc = treeDeleteEntry(db, pCsr, iNew1);
    }
  }

  return rc;
}

/*
** Delete a range of keys from the tree structure.
**
** This is a two step process: 
**
**     1) Remove all entries currently stored in the tree that have keys
**        that fall into the deleted range.
**
**        TODO: There are surely good ways to optimize this step - removing 
**        a range of keys from a b-tree. But for now, this function removes
**        them one at a time using the usual approach.
**
**     2) Unless the largest key smaller than or equal to (pKey1/nKey1) is
**        already marked as START_DELETE, insert a START_DELETE key. 
**        Similarly, unless the smallest key greater than or equal to
**        (pKey2/nKey2) is already START_END, insert a START_END key.
*/
int lsmTreeDelete(
  lsm_db *db,
  void *pKey1, int nKey1,         /* Start of range */
  void *pKey2, int nKey2          /* End of range */
){
  int bDone = 0;
  TreeRoot *p = &pDb->treehdr.root;

  /* The range must be sensible - that (key1 < key2). */
  assert( db->xCmp(pKey1, nKey1, pKey2, nKey2)<0 );

  /* Step 1. This loop runs until the tree contains no keys within the
  ** range being deleted. Or until an error occurs. */
  while( bDone==0 ){
    TreeCursor csr;               /* Cursor to seek to first key in range */
    void *pDel; int nDel;         /* Key to (possibly) delete this iteration */

    /* Seek the cursor to the first entry in the tree greater than pKey1. */
    treeCursorInit(pDb, 0, &csr);
    lsmTreeCursorSeek(&csr, pKey1, nKey1, &res);
    if( res<=0 && lsmTreeCursorValid(&csr) ) lsmTreeCursorNext(&csr);

    /* If there is no such entry, or if it is greater than pKey2, then the
    ** tree now contains no keys in the range being deleted. In this case
    ** break out of the loop.  */
    bDone = 1;
    if( lsmTreeCursorValid(&csr) ){
      lsmTreeCursorKey(&csr, &pDel, &nDel);
      if( db->xCmp(pDel, nDel, pKey2, nKey2)<0 ) bDone = 0;
    }

    if( bDone==0 ){
      if( csr.iNode==(p->nHeight-1) ){
        /* The element to delete already lies on a leaf node */
        rc = treeDeleteEntry(db, &csr);
      }else{
        /* 1. Overwrite the current key with a previous key in the tree (P).
        **
        ** 2. Seek to key P (cursor will stop at the internal nodes copy of
        **    P). Move to the previous key (original copy of P). Delete
        **    this entry. 
        */


      }
    }

    /* Clean up any memory allocated by the cursor. */
    tblobFree(pDb, &csr.blob);
  }
}

/*
** Return, in bytes, the amount of memory currently used by the tree 
** structure.
*/
int lsmTreeSize(lsm_db *pDb){
  return pDb->treehdr.nByte;