2021SC@SDUSC SQLite源码分析（六）————b树增删改查API

---

一、sqlite3BtreeInsert

在b-树的适当位置插入一个新记录。新记录的内容由pX对象描述。pCur游标仅用于定义记录应该插入到哪个表中，以及被留在哪个表中，之后将指向一个随机位置。

源码：

int sqlite3BtreeInsert(
  BtCursor *pCur,                /* Insert data into the table of this cursor */
  const BtreePayload *pX,        /* Content of the row to be inserted */
  int flags,                     /* True if this is likely an append */
  int seekResult                 /* Result of prior MovetoUnpacked() call */
){
  int rc;
  int loc = seekResult;          /* -1: before desired location  +1: after */
  int szNew = 0;
  int idx;
  MemPage *pPage;
  Btree *p = pCur->pBtree;
  BtShared *pBt = p->pBt;
  unsigned char *oldCell;
  unsigned char *newCell = 0;

  assert( (flags & (BTREE_SAVEPOSITION|BTREE_APPEND|BTREE_PREFORMAT))==flags );
  assert( (flags & BTREE_PREFORMAT)==0 || seekResult || pCur->pKeyInfo==0 );

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

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

  /* Assert that the caller has been consistent. If this cursor was opened
  ** expecting an index b-tree, then the caller should be inserting blob
  ** keys with no associated data. If the cursor was opened expecting an
  ** intkey table, the caller should be inserting integer keys with a
  ** blob of associated data.  */
  assert( (flags & BTREE_PREFORMAT) || (pX->pKey==0)==(pCur->pKeyInfo==0) );

  /* Save the positions of any other cursors open on this table.
  **
  ** In some cases, the call to btreeMoveto() below is a no-op. For
  ** example, when inserting data into a table with auto-generated integer
  ** keys, the VDBE layer invokes sqlite3BtreeLast() to figure out the 
  ** integer key to use. It then calls this function to actually insert the 
  ** data into the intkey B-Tree. In this case btreeMoveto() recognizes
  ** that the cursor is already where it needs to be and returns without
  ** doing any work. To avoid thwarting these optimizations, it is important
  ** not to clear the cursor here.
  */
  if( pCur->curFlags & BTCF_Multiple ){
    rc = saveAllCursors(pBt, pCur->pgnoRoot, pCur);
    if( rc ) return rc;
    if( loc && pCur->iPage<0 ){
      /* This can only happen if the schema is corrupt such that there is more
      ** than one table or index with the same root page as used by the cursor.
      ** Which can only happen if the SQLITE_NoSchemaError flag was set when
      ** the schema was loaded. This cannot be asserted though, as a user might
      ** set the flag, load the schema, and then unset the flag.  */
      return SQLITE_CORRUPT_BKPT;
    }
  }

  if( pCur->pKeyInfo==0 ){
    assert( pX->pKey==0 );
    /* If this is an insert into a table b-tree, invalidate any incrblob 
    ** cursors open on the row being replaced */
    if( p->hasIncrblobCur ){
      invalidateIncrblobCursors(p, pCur->pgnoRoot, pX->nKey, 0);
    }

    /* If BTREE_SAVEPOSITION is set, the cursor must already be pointing 
    ** to a row with the same key as the new entry being inserted.
    */
#ifdef SQLITE_DEBUG
    if( flags & BTREE_SAVEPOSITION ){
      assert( pCur->curFlags & BTCF_ValidNKey );
      assert( pX->nKey==pCur->info.nKey );
      assert( loc==0 );
    }
#endif

    /* On the other hand, BTREE_SAVEPOSITION==0 does not imply
    ** that the cursor is not pointing to a row to be overwritten.
    ** So do a complete check.
    */
    if( (pCur->curFlags&BTCF_ValidNKey)!=0 && pX->nKey==pCur->info.nKey ){
      /* The cursor is pointing to the entry that is to be
      ** overwritten */
      assert( pX->nData>=0 && pX->nZero>=0 );
      if( pCur->info.nSize!=0
       && pCur->info.nPayload==(u32)pX->nData+pX->nZero
      ){
        /* New entry is the same size as the old.  Do an overwrite */
        return btreeOverwriteCell(pCur, pX);
      }
      assert( loc==0 );
    }else if( loc==0 ){
      /* The cursor is *not* pointing to the cell to be overwritten, nor
      ** to an adjacent cell.  Move the cursor so that it is pointing either
      ** to the cell to be overwritten or an adjacent cell.
      */
      rc = sqlite3BtreeMovetoUnpacked(pCur, 0, pX->nKey, flags!=0, &loc);
      if( rc ) return rc;
    }
  }else{
    /* This is an index or a WITHOUT ROWID table */

    /* If BTREE_SAVEPOSITION is set, the cursor must already be pointing 
    ** to a row with the same key as the new entry being inserted.
    */
    assert( (flags & BTREE_SAVEPOSITION)==0 || loc==0 );

    /* If the cursor is not already pointing either to the cell to be
    ** overwritten, or if a new cell is being inserted, if the cursor is
    ** not pointing to an immediately adjacent cell, then move the cursor
    ** so that it does.
    */
    if( loc==0 && (flags & BTREE_SAVEPOSITION)==0 ){
      if( pX->nMem ){
        UnpackedRecord r;
        r.pKeyInfo = pCur->pKeyInfo;
        r.aMem = pX->aMem;
        r.nField = pX->nMem;
        r.default_rc = 0;
        r.errCode = 0;
        r.r1 = 0;
        r.r2 = 0;
        r.eqSeen = 0;
        rc = sqlite3BtreeMovetoUnpacked(pCur, &r, 0, flags!=0, &loc);
      }else{
        rc = btreeMoveto(pCur, pX->pKey, pX->nKey, flags!=0, &loc);
      }
      if( rc ) return rc;
    }

    /* If the cursor is currently pointing to an entry to be overwritten
    ** and the new content is the same as as the old, then use the
    ** overwrite optimization.
    */
    if( loc==0 ){
      getCellInfo(pCur);
      if( pCur->info.nKey==pX->nKey ){
        BtreePayload x2;
        x2.pData = pX->pKey;
        x2.nData = pX->nKey;
        x2.nZero = 0;
        return btreeOverwriteCell(pCur, &x2);
      }
    }
  }
  assert( pCur->eState==CURSOR_VALID 
       || (pCur->eState==CURSOR_INVALID && loc)
       || CORRUPT_DB );

  pPage = pCur->pPage;
  assert( pPage->intKey || pX->nKey>=0 || (flags & BTREE_PREFORMAT) );
  assert( pPage->leaf || !pPage->intKey );
  if( pPage->nFree<0 ){
    if( NEVER(pCur->eState>CURSOR_INVALID) ){
      rc = SQLITE_CORRUPT_BKPT;
    }else{
      rc = btreeComputeFreeSpace(pPage);
    }
    if( rc ) return rc;
  }

  TRACE(("INSERT: table=%d nkey=%lld ndata=%d page=%d %s\n",
          pCur->pgnoRoot, pX->nKey, pX->nData, pPage->pgno,
          loc==0 ? "overwrite" : "new entry"));
  assert( pPage->isInit );
  newCell = pBt->pTmpSpace;
  assert( newCell!=0 );
  if( flags & BTREE_PREFORMAT ){
    rc = SQLITE_OK;
    szNew = pBt->nPreformatSize;
    if( szNew<4 ) szNew = 4;
    if( ISAUTOVACUUM && szNew>pPage->maxLocal ){
      CellInfo info;
      pPage->xParseCell(pPage, newCell, &info);
      if( info.nPayload!=info.nLocal ){
        Pgno ovfl = get4byte(&newCell[szNew-4]);
        ptrmapPut(pBt, ovfl, PTRMAP_OVERFLOW1, pPage->pgno, &rc);
      }
    }
  }else{
    rc = fillInCell(pPage, newCell, pX, &szNew);
  }
  if( rc ) goto end_insert;
  assert( szNew==pPage->xCellSize(pPage, newCell) );
  assert( szNew <= MX_CELL_SIZE(pBt) );
  idx = pCur->ix;
  if( loc==0 ){
    CellInfo info;
    assert( idx<pPage->nCell );
    rc = sqlite3PagerWrite(pPage->pDbPage);
    if( rc ){
      goto end_insert;
    }
    oldCell = findCell(pPage, idx);
    if( !pPage->leaf ){
      memcpy(newCell, oldCell, 4);
    }
    BTREE_CLEAR_CELL(rc, pPage, oldCell, info);
    testcase( pCur->curFlags & BTCF_ValidOvfl );
    invalidateOverflowCache(pCur);
    if( info.nSize==szNew && info.nLocal==info.nPayload 
     && (!ISAUTOVACUUM || szNew<pPage->minLocal)
    ){
      /* Overwrite the old cell with the new if they are the same size.
      ** We could also try to do this if the old cell is smaller, then add
      ** the leftover space to the free list.  But experiments show that
      ** doing that is no faster then skipping this optimization and just
      ** calling dropCell() and insertCell(). 
      **
      ** This optimization cannot be used on an autovacuum database if the
      ** new entry uses overflow pages, as the insertCell() call below is
      ** necessary to add the PTRMAP_OVERFLOW1 pointer-map entry.  */
      assert( rc==SQLITE_OK ); /* clearCell never fails when nLocal==nPayload */
      if( oldCell < pPage->aData+pPage->hdrOffset+10 ){
        return SQLITE_CORRUPT_BKPT;
      }
      if( oldCell+szNew > pPage->aDataEnd ){
        return SQLITE_CORRUPT_BKPT;
      }
      memcpy(oldCell, newCell, szNew);
      return SQLITE_OK;
    }
    dropCell(pPage, idx, info.nSize, &rc);
    if( rc ) goto end_insert;
  }else if( loc<0 && pPage->nCell>0 ){
    assert( pPage->leaf );
    idx = ++pCur->ix;
    pCur->curFlags &= ~BTCF_ValidNKey;
  }else{
    assert( pPage->leaf );
  }
  insertCell(pPage, idx, newCell, szNew, 0, 0, &rc);
  assert( pPage->nOverflow==0 || rc==SQLITE_OK );
  assert( rc!=SQLITE_OK || pPage->nCell>0 || pPage->nOverflow>0 );

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

    /* Must make sure nOverflow is reset to zero even if the balance()
    ** fails. Internal data structure corruption will result otherwise. 
    ** Also, set the cursor state to invalid. This stops saveCursorPosition()
    ** from trying to save the current position of the cursor.  */
    pCur->pPage->nOverflow = 0;
    pCur->eState = CURSOR_INVALID;
    if( (flags & BTREE_SAVEPOSITION) && rc==SQLITE_OK ){
      btreeReleaseAllCursorPages(pCur);
      if( pCur->pKeyInfo ){
        assert( pCur->pKey==0 );
        pCur->pKey = sqlite3Malloc( pX->nKey );
        if( pCur->pKey==0 ){
          rc = SQLITE_NOMEM;
        }else{
          memcpy(pCur->pKey, pX->pKey, pX->nKey);
        }
      }
      pCur->eState = CURSOR_REQUIRESEEK;
      pCur->nKey = pX->nKey;
    }
  }
  assert( pCur->iPage<0 || pCur->pPage->nOverflow==0 );

end_insert:
  return rc;
}

二、sqlite3BtreeDelete

删除游标所指向的条目。
如果flags参数的BTREE_SAVEPOSITION位为零，则游标在删除后将被留在任意位置。
源码：

int sqlite3BtreeDelete(BtCursor *pCur, u8 flags){
  Btree *p = pCur->pBtree;
  BtShared *pBt = p->pBt;              
  int rc;                              /* Return code */
  MemPage *pPage;                      /* Page to delete cell from */
  unsigned char *pCell;                /* Pointer to cell to delete */
  int iCellIdx;                        /* Index of cell to delete */
  int iCellDepth;                      /* Depth of node containing pCell */ 
  CellInfo info;                       /* Size of the cell being deleted */
  int bSkipnext = 0;                   /* Leaf cursor in SKIPNEXT state */
  u8 bPreserve = flags & BTREE_SAVEPOSITION;  /* Keep cursor valid */

  assert( cursorOwnsBtShared(pCur) );
  assert( pBt->inTransaction==TRANS_WRITE );
  assert( (pBt->btsFlags & BTS_READ_ONLY)==0 );
  assert( pCur->curFlags & BTCF_WriteFlag );
  assert( hasSharedCacheTableLock(p, pCur->pgnoRoot, pCur->pKeyInfo!=0, 2) );
  assert( !hasReadConflicts(p, pCur->pgnoRoot) );
  assert( (flags & ~(BTREE_SAVEPOSITION | BTREE_AUXDELETE))==0 );
  if( pCur->eState==CURSOR_REQUIRESEEK ){
    rc = btreeRestoreCursorPosition(pCur);
    assert( rc!=SQLITE_OK || CORRUPT_DB || pCur->eState==CURSOR_VALID );
    if( rc || pCur->eState!=CURSOR_VALID ) return rc;
  }
  assert( CORRUPT_DB || pCur->eState==CURSOR_VALID );

  iCellDepth = pCur->iPage;
  iCellIdx = pCur->ix;
  pPage = pCur->pPage;
  pCell = findCell(pPage, iCellIdx);
  if( pPage->nFree<0 && btreeComputeFreeSpace(pPage) ) return SQLITE_CORRUPT;

  /* If the bPreserve flag is set to true, then the cursor position must
  ** be preserved following this delete operation. If the current delete
  ** will cause a b-tree rebalance, then this is done by saving the cursor
  ** key and leaving the cursor in CURSOR_REQUIRESEEK state before 
  ** returning. 
  **
  ** Or, if the current delete will not cause a rebalance, then the cursor
  ** will be left in CURSOR_SKIPNEXT state pointing to the entry immediately
  ** before or after the deleted entry. In this case set bSkipnext to true.  */
  if( bPreserve ){
    if( !pPage->leaf 
     || (pPage->nFree+cellSizePtr(pPage,pCell)+2)>(int)(pBt->usableSize*2/3)
     || pPage->nCell==1  /* See dbfuzz001.test for a test case */
    ){
      /* A b-tree rebalance will be required after deleting this entry.
      ** Save the cursor key.  */
      rc = saveCursorKey(pCur);
      if( rc ) return rc;
    }else{
      bSkipnext = 1;
    }
  }

  /* If the page containing the entry to delete is not a leaf page, move
  ** the cursor to the largest entry in the tree that is smaller than
  ** the entry being deleted. This cell will replace the cell being deleted
  ** from the internal node. The 'previous' entry is used for this instead
  ** of the 'next' entry, as the previous entry is always a part of the
  ** sub-tree headed by the child page of the cell being deleted. This makes
  ** balancing the tree following the delete operation easier.  */
  if( !pPage->leaf ){
    rc = sqlite3BtreePrevious(pCur, 0);
    assert( rc!=SQLITE_DONE );
    if( rc ) return rc;
  }

  /* Save the positions of any other cursors open on this table before
  ** making any modifications.  */
  if( pCur->curFlags & BTCF_Multiple ){
    rc = saveAllCursors(pBt, pCur->pgnoRoot, pCur);
    if( rc ) return rc;
  }

  /* If this is a delete operation to remove a row from a table b-tree,
  ** invalidate any incrblob cursors open on the row being deleted.  */
  if( pCur->pKeyInfo==0 && p->hasIncrblobCur ){
    invalidateIncrblobCursors(p, pCur->pgnoRoot, pCur->info.nKey, 0);
  }

  /* Make the page containing the entry to be deleted writable. Then free any
  ** overflow pages associated with the entry and finally remove the cell
  ** itself from within the page.  */
  rc = sqlite3PagerWrite(pPage->pDbPage);
  if( rc ) return rc;
  BTREE_CLEAR_CELL(rc, pPage, pCell, info);
  dropCell(pPage, iCellIdx, info.nSize, &rc);
  if( rc ) return rc;

  /* If the cell deleted was not located on a leaf page, then the cursor
  ** is currently pointing to the largest entry in the sub-tree headed
  ** by the child-page of the cell that was just deleted from an internal
  ** node. The cell from the leaf node needs to be moved to the internal
  ** node to replace the deleted cell.  */
  if( !pPage->leaf ){
    MemPage *pLeaf = pCur->pPage;
    int nCell;
    Pgno n;
    unsigned char *pTmp;

    if( pLeaf->nFree<0 ){
      rc = btreeComputeFreeSpace(pLeaf);
      if( rc ) return rc;
    }
    if( iCellDepth<pCur->iPage-1 ){
      n = pCur->apPage[iCellDepth+1]->pgno;
    }else{
      n = pCur->pPage->pgno;
    }
    pCell = findCell(pLeaf, pLeaf->nCell-1);
    if( pCell<&pLeaf->aData[4] ) return SQLITE_CORRUPT_BKPT;
    nCell = pLeaf->xCellSize(pLeaf, pCell);
    assert( MX_CELL_SIZE(pBt) >= nCell );
    pTmp = pBt->pTmpSpace;
    assert( pTmp!=0 );
    rc = sqlite3PagerWrite(pLeaf->pDbPage);
    if( rc==SQLITE_OK ){
      insertCell(pPage, iCellIdx, pCell-4, nCell+4, pTmp, n, &rc);
    }
    dropCell(pLeaf, pLeaf->nCell-1, nCell, &rc);
    if( rc ) return rc;
  }

  /* Balance the tree. If the entry deleted was located on a leaf page,
  ** then the cursor still points to that page. In this case the first
  ** call to balance() repairs the tree, and the if(...) condition is
  ** never true.
  **
  ** Otherwise, if the entry deleted was on an internal node page, then
  ** pCur is pointing to the leaf page from which a cell was removed to
  ** replace the cell deleted from the internal node. This is slightly
  ** tricky as the leaf node may be underfull, and the internal node may
  ** be either under or overfull. In this case run the balancing algorithm
  ** on the leaf node first. If the balance proceeds far enough up the
  ** tree that we can be sure that any problem in the internal node has
  ** been corrected, so be it. Otherwise, after balancing the leaf node,
  ** walk the cursor up the tree to the internal node and balance it as 
  ** well.  */
  rc = balance(pCur);
  if( rc==SQLITE_OK && pCur->iPage>iCellDepth ){
    releasePageNotNull(pCur->pPage);
    pCur->iPage--;
    while( pCur->iPage>iCellDepth ){
      releasePage(pCur->apPage[pCur->iPage--]);
    }
    pCur->pPage = pCur->apPage[pCur->iPage];
    rc = balance(pCur);
  }

  if( rc==SQLITE_OK ){
    if( bSkipnext ){
      assert( bPreserve && (pCur->iPage==iCellDepth || CORRUPT_DB) );
      assert( pPage==pCur->pPage || CORRUPT_DB );
      assert( (pPage->nCell>0 || CORRUPT_DB) && iCellIdx<=pPage->nCell );
      pCur->eState = CURSOR_SKIPNEXT;
      if( iCellIdx>=pPage->nCell ){
        pCur->skipNext = -1;
        pCur->ix = pPage->nCell-1;
      }else{
        pCur->skipNext = 1;
      }
    }else{
      rc = moveToRoot(pCur);
      if( bPreserve ){
        btreeReleaseAllCursorPages(pCur);
        pCur->eState = CURSOR_REQUIRESEEK;
      }
      if( rc==SQLITE_EMPTY ) rc = SQLITE_OK;
    }
  }
  return rc;
}