btree.c

sqlite数据库管理系统开放源码

  pP1 = pBt->page1;
  rc = sqlitepager_write(pBt->page1);
  if( rc ) return rc;
  rc = sqlitepager_get(pBt->pPager, 2, (void**)&pRoot);
  if( rc ) return rc;
  rc = sqlitepager_write(pRoot);
  if( rc ){
    sqlitepager_unref(pRoot);
    return rc;
  }
  strcpy(pP1->zMagic, zMagicHeader);
  if( btree_native_byte_order ){
    pP1->iMagic = MAGIC;
    pBt->needSwab = 0;
  }else{
    pP1->iMagic = swab32(MAGIC);
    pBt->needSwab = 1;
  }
  zeroPage(pBt, pRoot);
  sqlitepager_unref(pRoot);
  return SQLITE_OK;
}

/*
** Attempt to start a new transaction.
**
** A transaction must be started before attempting any changes
** to the database.  None of the following routines will work
** unless a transaction is started first:
**
**      sqliteBtreeCreateTable()
**      sqliteBtreeCreateIndex()
**      sqliteBtreeClearTable()
**      sqliteBtreeDropTable()
**      sqliteBtreeInsert()
**      sqliteBtreeDelete()
**      sqliteBtreeUpdateMeta()
*/
static int fileBtreeBeginTrans(Btree *pBt){
  int rc;
  if( pBt->inTrans ) return SQLITE_ERROR;
  if( pBt->readOnly ) return SQLITE_READONLY;
  if( pBt->page1==0 ){
    rc = lockBtree(pBt);
    if( rc!=SQLITE_OK ){
      return rc;
    }
  }
  rc = sqlitepager_begin(pBt->page1);
  if( rc==SQLITE_OK ){
    rc = newDatabase(pBt);
  }
  if( rc==SQLITE_OK ){
    pBt->inTrans = 1;
    pBt->inCkpt = 0;
  }else{
    unlockBtreeIfUnused(pBt);
  }
  return rc;
}

/*
** Commit the transaction currently in progress.
**
** This will release the write lock on the database file.  If there
** are no active cursors, it also releases the read lock.
*/
static int fileBtreeCommit(Btree *pBt){
  int rc;
  rc = pBt->readOnly ? SQLITE_OK : sqlitepager_commit(pBt->pPager);
  pBt->inTrans = 0;
  pBt->inCkpt = 0;
  unlockBtreeIfUnused(pBt);
  return rc;
}

/*
** Rollback the transaction in progress.  All cursors will be
** invalided by this operation.  Any attempt to use a cursor
** that was open at the beginning of this operation will result
** in an error.
**
** This will release the write lock on the database file.  If there
** are no active cursors, it also releases the read lock.
*/
static int fileBtreeRollback(Btree *pBt){
  int rc;
  BtCursor *pCur;
  if( pBt->inTrans==0 ) return SQLITE_OK;
  pBt->inTrans = 0;
  pBt->inCkpt = 0;
  rc = pBt->readOnly ? SQLITE_OK : sqlitepager_rollback(pBt->pPager);
  for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
    if( pCur->pPage && pCur->pPage->isInit==0 ){
      sqlitepager_unref(pCur->pPage);
      pCur->pPage = 0;
    }
  }
  unlockBtreeIfUnused(pBt);
  return rc;
}

/*
** Set the checkpoint for the current transaction.  The checkpoint serves
** as a sub-transaction that can be rolled back independently of the
** main transaction.  You must start a transaction before starting a
** checkpoint.  The checkpoint is ended automatically if the transaction
** commits or rolls back.
**
** Only one checkpoint may be active at a time.  It is an error to try
** to start a new checkpoint if another checkpoint is already active.
*/
static int fileBtreeBeginCkpt(Btree *pBt){
  int rc;
  if( !pBt->inTrans || pBt->inCkpt ){
    return pBt->readOnly ? SQLITE_READONLY : SQLITE_ERROR;
  }
  rc = pBt->readOnly ? SQLITE_OK : sqlitepager_ckpt_begin(pBt->pPager);
  pBt->inCkpt = 1;
  return rc;
}


/*
** Commit a checkpoint to transaction currently in progress.  If no
** checkpoint is active, this is a no-op.
*/
static int fileBtreeCommitCkpt(Btree *pBt){
  int rc;
  if( pBt->inCkpt && !pBt->readOnly ){
    rc = sqlitepager_ckpt_commit(pBt->pPager);
  }else{
    rc = SQLITE_OK;
  }
  pBt->inCkpt = 0;
  return rc;
}

/*
** Rollback the checkpoint to the current transaction.  If there
** is no active checkpoint or transaction, this routine is a no-op.
**
** All cursors will be invalided by this operation.  Any attempt
** to use a cursor that was open at the beginning of this operation
** will result in an error.
*/
static int fileBtreeRollbackCkpt(Btree *pBt){
  int rc;
  BtCursor *pCur;
  if( pBt->inCkpt==0 || pBt->readOnly ) return SQLITE_OK;
  rc = sqlitepager_ckpt_rollback(pBt->pPager);
  for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
    if( pCur->pPage && pCur->pPage->isInit==0 ){
      sqlitepager_unref(pCur->pPage);
      pCur->pPage = 0;
    }
  }
  pBt->inCkpt = 0;
  return rc;
}

/*
** Create a new cursor for the BTree whose root is on the page
** iTable.  The act of acquiring a cursor gets a read lock on 
** the database file.
**
** If wrFlag==0, then the cursor can only be used for reading.
** If wrFlag==1, then the cursor can be used for reading or for
** writing if other conditions for writing are also met.  These
** are the conditions that must be met in order for writing to
** be allowed:
**
** 1:  The cursor must have been opened with wrFlag==1
**
** 2:  No other cursors may be open with wrFlag==0 on the same table
**
** 3:  The database must be writable (not on read-only media)
**
** 4:  There must be an active transaction.
**
** Condition 2 warrants further discussion.  If any cursor is opened
** on a table with wrFlag==0, that prevents all other cursors from
** writing to that table.  This is a kind of "read-lock".  When a cursor
** is opened with wrFlag==0 it is guaranteed that the table will not
** change as long as the cursor is open.  This allows the cursor to
** do a sequential scan of the table without having to worry about
** entries being inserted or deleted during the scan.  Cursors should
** be opened with wrFlag==0 only if this read-lock property is needed.
** That is to say, cursors should be opened with wrFlag==0 only if they
** intend to use the sqliteBtreeNext() system call.  All other cursors
** should be opened with wrFlag==1 even if they never really intend
** to write.
** 
** No checking is done to make sure that page iTable really is the
** root page of a b-tree.  If it is not, then the cursor acquired
** will not work correctly.
*/
static 
int fileBtreeCursor(Btree *pBt, int iTable, int wrFlag, BtCursor **ppCur){
  int rc;
  BtCursor *pCur, *pRing;

  if( pBt->readOnly && wrFlag ){
    *ppCur = 0;
    return SQLITE_READONLY;
  }
  if( pBt->page1==0 ){
    rc = lockBtree(pBt);
    if( rc!=SQLITE_OK ){
      *ppCur = 0;
      return rc;
    }
  }
  pCur = sqliteMalloc( sizeof(*pCur) );
  if( pCur==0 ){
    rc = SQLITE_NOMEM;
    goto create_cursor_exception;
  }
  pCur->pgnoRoot = (Pgno)iTable;
  rc = sqlitepager_get(pBt->pPager, pCur->pgnoRoot, (void**)&pCur->pPage);
  if( rc!=SQLITE_OK ){
    goto create_cursor_exception;
  }
  rc = initPage(pBt, pCur->pPage, pCur->pgnoRoot, 0);
  if( rc!=SQLITE_OK ){
    goto create_cursor_exception;
  }
  pCur->pOps = &sqliteBtreeCursorOps;
  pCur->pBt = pBt;
  pCur->wrFlag = wrFlag;
  pCur->idx = 0;
  pCur->eSkip = SKIP_INVALID;
  pCur->pNext = pBt->pCursor;
  if( pCur->pNext ){
    pCur->pNext->pPrev = pCur;
  }
  pCur->pPrev = 0;
  pRing = pBt->pCursor;
  while( pRing && pRing->pgnoRoot!=pCur->pgnoRoot ){ pRing = pRing->pNext; }
  if( pRing ){
    pCur->pShared = pRing->pShared;
    pRing->pShared = pCur;
  }else{
    pCur->pShared = pCur;
  }
  pBt->pCursor = pCur;
  *ppCur = pCur;
  return SQLITE_OK;

create_cursor_exception:
  *ppCur = 0;
  if( pCur ){
    if( pCur->pPage ) sqlitepager_unref(pCur->pPage);
    sqliteFree(pCur);
  }
  unlockBtreeIfUnused(pBt);
  return rc;
}

/*
** Close a cursor.  The read lock on the database file is released
** when the last cursor is closed.
*/
static int fileBtreeCloseCursor(BtCursor *pCur){
  Btree *pBt = pCur->pBt;
  if( pCur->pPrev ){
    pCur->pPrev->pNext = pCur->pNext;
  }else{
    pBt->pCursor = pCur->pNext;
  }
  if( pCur->pNext ){
    pCur->pNext->pPrev = pCur->pPrev;
  }
  if( pCur->pPage ){
    sqlitepager_unref(pCur->pPage);
  }
  if( pCur->pShared!=pCur ){
    BtCursor *pRing = pCur->pShared;
    while( pRing->pShared!=pCur ){ pRing = pRing->pShared; }
    pRing->pShared = pCur->pShared;
  }
  unlockBtreeIfUnused(pBt);
  sqliteFree(pCur);
  return SQLITE_OK;
}

/*
** Make a temporary cursor by filling in the fields of pTempCur.
** The temporary cursor is not on the cursor list for the Btree.
*/
static void getTempCursor(BtCursor *pCur, BtCursor *pTempCur){
  memcpy(pTempCur, pCur, sizeof(*pCur));
  pTempCur->pNext = 0;
  pTempCur->pPrev = 0;
  if( pTempCur->pPage ){
    sqlitepager_ref(pTempCur->pPage);
  }
}

/*
** Delete a temporary cursor such as was made by the CreateTemporaryCursor()
** function above.
*/
static void releaseTempCursor(BtCursor *pCur){
  if( pCur->pPage ){
    sqlitepager_unref(pCur->pPage);
  }
}

/*
** Set *pSize to the number of bytes of key in the entry the
** cursor currently points to.  Always return SQLITE_OK.
** Failure is not possible.  If the cursor is not currently
** pointing to an entry (which can happen, for example, if
** the database is empty) then *pSize is set to 0.
*/
static int fileBtreeKeySize(BtCursor *pCur, int *pSize){
  Cell *pCell;
  MemPage *pPage;

  pPage = pCur->pPage;
  assert( pPage!=0 );
  if( pCur->idx >= pPage->nCell ){
    *pSize = 0;
  }else{
    pCell = pPage->apCell[pCur->idx];
    *pSize = NKEY(pCur->pBt, pCell->h);
  }
  return SQLITE_OK;
}

/*
** Read payload information from the entry that the pCur cursor is
** pointing to.  Begin reading the payload at "offset" and read
** a total of "amt" bytes.  Put the result in zBuf.
**
** This routine does not make a distinction between key and data.
** It just reads bytes from the payload area.
*/
static int getPayload(BtCursor *pCur, int offset, int amt, char *zBuf){
  char *aPayload;
  Pgno nextPage;
  int rc;
  Btree *pBt = pCur->pBt;
  assert( pCur!=0 && pCur->pPage!=0 );
  assert( pCur->idx>=0 && pCur->idx<pCur->pPage->nCell );
  aPayload = pCur->pPage->apCell[pCur->idx]->aPayload;
  if( offset<MX_LOCAL_PAYLOAD ){
    int a = amt;
    if( a+offset>MX_LOCAL_PAYLOAD ){
      a = MX_LOCAL_PAYLOAD - offset;
    }
    memcpy(zBuf, &aPayload[offset], a);
    if( a==amt ){
      return SQLITE_OK;
    }
    offset = 0;
    zBuf += a;
    amt -= a;
  }else{
    offset -= MX_LOCAL_PAYLOAD;
  }
  if( amt>0 ){
    nextPage = SWAB32(pBt, pCur->pPage->apCell[pCur->idx]->ovfl);
  }
  while( amt>0 && nextPage ){
    OverflowPage *pOvfl;
    rc = sqlitepager_get(pBt->pPager, nextPage, (void**)&pOvfl);
    if( rc!=0 ){
      return rc;
    }
    nextPage = SWAB32(pBt, pOvfl->iNext);
    if( offset<OVERFLOW_SIZE ){
      int a = amt;
      if( a + offset > OVERFLOW_SIZE ){
        a = OVERFLOW_SIZE - offset;
      }
      memcpy(zBuf, &pOvfl->aPayload[offset], a);
      offset = 0;
      amt -= a;
      zBuf += a;
    }else{
      offset -= OVERFLOW_SIZE;
    }
    sqlitepager_unref(pOvfl);
  }
  if( amt>0 ){
    return SQLITE_CORRUPT;
  }
  return SQLITE_OK;
}

/*
** Read part of the key associated with cursor pCur.  A maximum
** of "amt" bytes will be transfered into zBuf[].  The transfer
** begins at "offset".  The number of bytes actually read is
** returned. 
**
** Change:  It used to be that the amount returned will be smaller
** than the amount requested if there are not enough bytes in the key
** to satisfy the request.  But now, it must be the case that there
** is enough data available to satisfy the request.  If not, an exception
** is raised.  The change was made in an effort to boost performance
** by eliminating unneeded tests.
*/
static int fileBtreeKey(BtCursor *pCur, int offset, int amt, char *zBuf){
  MemPage *pPage;

  assert( amt>=0 );
  assert( offset>=0 );
  assert( pCur->pPage!=0 );
  pPage = pCur->pPage;
  if( pCur->idx >= pPage->nCell ){
    return 0;
  }
  assert( amt+offset <= NKEY(pCur->pBt, pPage->apCell[pCur->idx]->h) );
  getPayload(pCur, offset, amt, zBuf);
  return amt;