/*
** Call sqlite3WalOpen() to open the WAL handle. If the pager is in
** exclusive-locking mode when this function is called, take an EXCLUSIVE
** lock on the database file and use heap-memory to store the wal-index
** in. Otherwise, use the normal shared-memory.
*/
static int pagerOpenWal(Pager *pPager)
{
int rc = SQLITE.OK;
assert( pPager.pWal==0 && pPager.tempFile==0 );
assert( pPager.eLock==SHARED_LOCK || pPager.eLock==EXCLUSIVE_LOCK || pPager.noReadlock);
/* If the pager is already in exclusive-mode, the WAL module will use
** heap-memory for the wal-index instead of the VFS shared-memory
** implementation. Take the exclusive lock now, before opening the WAL
** file, to make sure this is safe.
*/
if( pPager.exclusiveMode ){
rc = pagerExclusiveLock(pPager);
}
/* Open the connection to the log file. If this operation fails,
** (e.g. due to malloc() failure), return an error code.
*/
if( rc==SQLITE.OK ){
rc = sqlite3WalOpen(pPager.pVfs,
pPager.fd, pPager.zWal, pPager.exclusiveMode, &pPager.pWal
pPager.journalSizeLimit, &pPager.pWal
);
}
return rc;
}
/*
** Attempt to take an exclusive lock on the database file. If a PENDING lock
** is obtained instead, immediately release it.
*/
static int pagerExclusiveLock(Pager *pPager)
{
int rc; /* Return code */
assert( pPager.eLock==SHARED_LOCK || pPager.eLock==EXCLUSIVE_LOCK );
rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
if( rc!=SQLITE.OK ){
/* If the attempt to grab the exclusive lock failed, release the
** pending lock that may have been obtained instead. */
pagerUnlockDb(pPager, SHARED_LOCK);
}
return rc;
}
public void ClearState()
{
this.Data = null;
this.Extra = null;
this.Dirtys = null;
this.ID = 0;
this.Pager = null;
#if DEBUG
this.PageHash = 0;
#endif
this.Flags = 0;
this.Refs = 0;
this.CacheAllocated = false;
this.Cache = null;
this.DirtyNext = null;
this.DirtyPrev = null;
this.PgHdr1 = null;
}
/*
** Begin a read transaction on the WAL.
**
** This routine used to be called "pagerOpenSnapshot()" because it essentially
** makes a snapshot of the database at the current point in time and preserves
** that snapshot for use by the reader in spite of concurrently changes by
** other writers or checkpointers.
*/
static int pagerBeginReadTransaction(Pager* pPager)
{
int rc; /* Return code */
int changed = 0; /* True if cache must be reset */
assert(pagerUseWal(pPager));
assert(pPager.eState == PAGER_OPEN || pPager.eState == PAGER_READER);
/* sqlite3WalEndReadTransaction() was not called for the previous
** transaction in locking_mode=EXCLUSIVE. So call it now. If we
** are in locking_mode=NORMAL and EndRead() was previously called,
** the duplicate call is harmless.
*/
sqlite3WalEndReadTransaction(pPager.pWal);
rc = sqlite3WalBeginReadTransaction(pPager.pWal, &changed);
if (rc != SQLITE.OK || changed)
{
pager_reset(pPager);
}
return rc;
}
/*
** Return true if the underlying VFS for the given pager supports the
** primitives necessary for write-ahead logging.
*/
int sqlite3PagerWalSupported(Pager *pPager)
{
const sqlite3_io_methods *pMethods = pPager.fd->pMethods;
return pPager.exclusiveMode || (pMethods->iVersion>=2 && pMethods->xShmMap);
}
int sqlite3PagerWalCallback(Pager *pPager)
{
return sqlite3WalCallback(pPager.pWal);
}
/*
** The caller must be holding a SHARED lock on the database file to call
** this function.
**
** If the pager passed as the first argument is open on a real database
** file (not a temp file or an in-memory database), and the WAL file
** is not already open, make an attempt to open it now. If successful,
** return SQLITE.OK. If an error occurs or the VFS used by the pager does
** not support the xShmXXX() methods, return an error code. *pbOpen is
** not modified in either case.
**
** If the pager is open on a temp-file (or in-memory database), or if
** the WAL file is already open, set *pbOpen to 1 and return SQLITE.OK
** without doing anything.
*/
int sqlite3PagerOpenWal(
Pager *pPager, /* Pager object */
int *pbOpen /* OUT: Set to true if call is a no-op */
)
{
int rc = SQLITE.OK; /* Return code */
assert( assert_pager_state(pPager) );
assert( pPager.eState==PAGER_OPEN || pbOpen );
assert( pPager.eState==PAGER_READER || !pbOpen );
assert( pbOpen==0 || *pbOpen==0 );
assert( pbOpen!=0 || (!pPager.tempFile && !pPager.pWal) );
if( !pPager.tempFile && !pPager.pWal ){
if( !sqlite3PagerWalSupported(pPager) ) return SQLITE_CANTOPEN;
/* Close any rollback journal previously open */
sqlite3OsClose(pPager.jfd);
rc = pagerOpenWal(pPager);
if( rc==SQLITE.OK ){
pPager.journalMode = PAGER_JOURNALMODE_WAL;
pPager.eState = PAGER_OPEN;
}
}else{
*pbOpen = 1;
}
return rc;
}
/*
** This function is called to close the connection to the log file prior
** to switching from WAL to rollback mode.
**
** Before closing the log file, this function attempts to take an
** EXCLUSIVE lock on the database file. If this cannot be obtained, an
** error (SQLITE_BUSY) is returned and the log connection is not closed.
** If successful, the EXCLUSIVE lock is not released before returning.
*/
int sqlite3PagerCloseWal(Pager *pPager)
{
int rc = SQLITE.OK;
assert( pPager.journalMode==PAGER_JOURNALMODE_WAL );
/* If the log file is not already open, but does exist in the file-system,
** it may need to be checkpointed before the connection can switch to
** rollback mode. Open it now so this can happen.
*/
if( !pPager.pWal ){
int logexists = 0;
rc = pagerLockDb(pPager, SHARED_LOCK);
if( rc==SQLITE.OK ){
rc = sqlite3OsAccess(
pPager.pVfs, pPager.zWal, SQLITE_ACCESS_EXISTS, &logexists
);
}
if( rc==SQLITE.OK && logexists ){
rc = pagerOpenWal(pPager);
}
}
/* Checkpoint and close the log. Because an EXCLUSIVE lock is held on
** the database file, the log and log-summary files will be deleted.
*/
if( rc==SQLITE.OK && pPager.pWal ){
rc = pagerExclusiveLock(pPager);
if( rc==SQLITE.OK ){
rc = sqlite3WalClose(pPager.pWal, pPager.ckptSyncFlags,
pPager.pageSize, (u8*)pPager.pTmpSpace);
pPager.pWal = 0;
}
}
return rc;
}
/*
** This function is called when the user invokes "PRAGMA wal_checkpoint",
** "PRAGMA wal_blocking_checkpoint" or calls the sqlite3_wal_checkpoint()
** or wal_blocking_checkpoint() API functions.
**
** Parameter eMode is one of SQLITE_CHECKPOINT_PASSIVE, FULL or RESTART.
*/
int sqlite3PagerCheckpoint(Pager *pPager, int eMode, int *pnLog, int *pnCkpt)
{
int rc = SQLITE.OK;
if( pPager.pWal ){
rc = sqlite3WalCheckpoint(pPager.pWal, eMode,
pPager.xBusyHandler, pPager.pBusyHandlerArg,
pPager.ckptSyncFlags, pPager.pageSize, (u8 *)pPager.pTmpSpace,
pnLog, pnCkpt
);
}
return rc;
}
// was:sqlite3BtreeDelete
public RC Delete()
{
MemPage pPage; // Page to delete cell from
int pCell; // Pointer to cell to delete
int iCellIdx; // Index of cell to delete
int iCellDepth; // Depth of node containing pCell
var p = this.Tree;
var pBt = p.Shared;
Debug.Assert(HoldsMutex());
Debug.Assert(pBt.InTransaction == TRANS.WRITE);
Debug.Assert(!pBt.ReadOnly);
Debug.Assert(this.Writeable);
Debug.Assert(p.hasSharedCacheTableLock(this.RootID, (this.KeyInfo != null), LOCK.WRITE));
Debug.Assert(!p.hasReadConflicts(this.RootID));
if (Check.NEVER(this.PagesIndexs[this.PageID] >= this.Pages[this.PageID].Cells) || Check.NEVER(this.State != CursorState.VALID))
{
return(RC.ERROR);
}
// 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 (this.KeyInfo == null)
{
Btree.invalidateIncrblobCursors(p, this.Info.nKey, false);
}
iCellDepth = this.PageID;
iCellIdx = this.PagesIndexs[iCellDepth];
pPage = this.Pages[iCellDepth];
pCell = pPage.FindCell(iCellIdx);
// 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.
RC rc;
if (pPage.Leaf == 0)
{
var notUsed = 0;
rc = MovePrevious(ref notUsed);
if (rc != RC.OK)
{
return(rc);
}
}
// Save the positions of any other cursors open on this table before making any modifications. 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 = pBt.saveAllCursors(this.RootID, this);
if (rc != RC.OK)
{
return(rc);
}
rc = Pager.Write(pPage.DbPage);
if (rc != RC.OK)
{
return(rc);
}
rc = pPage.clearCell(pCell);
pPage.dropCell(iCellIdx, pPage.cellSizePtr(pCell), ref rc);
if (rc != RC.OK)
{
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 == 0)
{
var pLeaf = this.Pages[this.PageID];
int nCell;
var n = this.Pages[iCellDepth + 1].ID;
pCell = pLeaf.FindCell(pLeaf.Cells - 1);
nCell = pLeaf.cellSizePtr(pCell);
Debug.Assert(Btree.MX_CELL_SIZE(pBt) >= nCell);
rc = Pager.Write(pLeaf.DbPage);
var pNext_4 = MallocEx.sqlite3Malloc(nCell + 4);
Buffer.BlockCopy(pLeaf.Data, pCell - 4, pNext_4, 0, nCell + 4);
pPage.insertCell(iCellIdx, pNext_4, nCell + 4, null, n, ref rc);
pLeaf.dropCell(pLeaf.Cells - 1, nCell, ref rc);
if (rc != RC.OK)
{
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();
if (rc == RC.OK && this.PageID > iCellDepth)
{
while (this.PageID > iCellDepth)
{
this.Pages[this.PageID--].releasePage();
}
rc = Balance();
}
if (rc == RC.OK)
{
MoveToRoot();
}
return(rc);
}
private static int PAGERID(Pager p)
{
return p.GetHashCode();
}
// This function is called to rollback a transaction on a WAL database.
static int pagerRollbackWal(Pager pPager)
{
int rc; /* Return Code */
PgHdr pList; /* List of dirty pages to revert */
/* For all pages in the cache that are currently dirty or have already been written (but not committed) to the log file, do one of the
** following:
** + Discard the cached page (if refcount==0), or
** + Reload page content from the database (if refcount>0).
*/
pPager.dbSize = pPager.dbOrigSize;
rc = sqlite3WalUndo(pPager.pWal, pagerUndoCallback, pPager);
pList = sqlite3PcacheDirtyList(pPager.pPCache);
while (pList && rc == SQLITE.OK)
{
PgHdr pNext = pList.pDirty;
rc = pagerUndoCallback(pPager, pList.pgno);
pList = pNext;
}
return rc;
}
/*
** Check if the *-wal file that corresponds to the database opened by pPager
** exists if the database is not empy, or verify that the *-wal file does
** not exist (by deleting it) if the database file is empty.
**
** If the database is not empty and the *-wal file exists, open the pager
** in WAL mode. If the database is empty or if no *-wal file exists and
** if no error occurs, make sure Pager.journalMode is not set to
** PAGER_JOURNALMODE_WAL.
**
** Return SQLITE.OK or an error code.
**
** The caller must hold a SHARED lock on the database file to call this
** function. Because an EXCLUSIVE lock on the db file is required to delete
** a WAL on a none-empty database, this ensures there is no race condition
** between the xAccess() below and an xDelete() being executed by some
** other connection.
*/
static int pagerOpenWalIfPresent(Pager* pPager)
{
int rc = SQLITE.OK;
Debug.Assert(pPager.eState == PAGER_OPEN);
Debug.Assert(pPager.eLock >= SHARED_LOCK || pPager.noReadlock);
if (!pPager.tempFile)
{
int isWal; /* True if WAL file exists */
Pgno nPage; /* Size of the database file */
rc = pagerPagecount(pPager, &nPage);
if (rc) return rc;
if (nPage == 0)
{
rc = sqlite3OsDelete(pPager.pVfs, pPager.zWal, 0);
isWal = 0;
}
else
{
rc = sqlite3OsAccess(
pPager.pVfs, pPager.zWal, SQLITE_ACCESS_EXISTS, &isWal
);
}
if (rc == SQLITE.OK)
{
if (isWal)
{
testcase(sqlite3PcachePagecount(pPager.pPCache) == 0);
rc = sqlite3PagerOpenWal(pPager, 0);
}
else if (pPager.journalMode == PAGER_JOURNALMODE_WAL)
{
pPager.journalMode = PAGER_JOURNALMODE_DELETE;
}
}
}
return rc;
}
private static uint JOURNAL_PG_SZ(Pager pPager)
{
return (uint)pPager.pageSize + 8;
}
private static uint JOURNAL_HDR_SZ(Pager pPager)
{
return pPager.sectorSize;
}
// was:sqlite3BtreeInsert
public RC Insert(byte[] key, long nKey, byte[] data, int nData, int nZero, bool appendBiasRight, int seekResult)
{
var loc = seekResult; // -1: before desired location +1: after
var szNew = 0;
int idx;
var p = this.Tree;
var pBt = p.Shared;
int oldCell;
byte[] newCell = null;
if (this.State == CursorState.FAULT)
{
Debug.Assert(this.SkipNext != 0);
return((RC)this.SkipNext);
}
Debug.Assert(HoldsMutex());
Debug.Assert(this.Writeable && pBt.InTransaction == TRANS.WRITE && !pBt.ReadOnly);
Debug.Assert(p.hasSharedCacheTableLock(this.RootID, (this.KeyInfo != null), LOCK.WRITE));
// 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.
Debug.Assert((key == null) == (this.KeyInfo == null));
// If this is an insert into a table b-tree, invalidate any incrblob cursors open on the row being replaced (assuming this is a replace
// operation - if it is not, the following is a no-op). */
if (this.KeyInfo == null)
{
Btree.invalidateIncrblobCursors(p, nKey, false);
}
// 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.
var rc = pBt.saveAllCursors(this.RootID, this);
if (rc != RC.OK)
{
return(rc);
}
if (loc == 0)
{
rc = BtreeMoveTo(key, nKey, appendBiasRight, ref loc);
if (rc != RC.OK)
{
return(rc);
}
}
Debug.Assert(this.State == CursorState.VALID || (this.State == CursorState.INVALID && loc != 0));
var pPage = this.Pages[this.PageID];
Debug.Assert(pPage.HasIntKey || nKey >= 0);
Debug.Assert(pPage.Leaf != 0 || !pPage.HasIntKey);
Btree.TRACE("INSERT: table=%d nkey=%lld ndata=%d page=%d %s\n", this.RootID, nKey, nData, pPage.ID, (loc == 0 ? "overwrite" : "new entry"));
Debug.Assert(pPage.HasInit);
pBt.allocateTempSpace();
newCell = pBt.pTmpSpace;
rc = pPage.fillInCell(newCell, key, nKey, data, nData, nZero, ref szNew);
if (rc != RC.OK)
{
goto end_insert;
}
Debug.Assert(szNew == pPage.cellSizePtr(newCell));
Debug.Assert(szNew <= Btree.MX_CELL_SIZE(pBt));
idx = this.PagesIndexs[this.PageID];
if (loc == 0)
{
ushort szOld;
Debug.Assert(idx < pPage.Cells);
rc = Pager.Write(pPage.DbPage);
if (rc != RC.OK)
{
goto end_insert;
}
oldCell = pPage.FindCell(idx);
if (0 == pPage.Leaf)
{
newCell[0] = pPage.Data[oldCell + 0];
newCell[1] = pPage.Data[oldCell + 1];
newCell[2] = pPage.Data[oldCell + 2];
newCell[3] = pPage.Data[oldCell + 3];
}
szOld = pPage.cellSizePtr(oldCell);
rc = pPage.clearCell(oldCell);
pPage.dropCell(idx, szOld, ref rc);
if (rc != RC.OK)
{
goto end_insert;
}
}
else if (loc < 0 && pPage.Cells > 0)
{
Debug.Assert(pPage.Leaf != 0);
idx = ++this.PagesIndexs[this.PageID];
}
else
{
Debug.Assert(pPage.Leaf != 0);
}
pPage.insertCell(idx, newCell, szNew, null, 0, ref rc);
Debug.Assert(rc != RC.OK || pPage.Cells > 0 || pPage.NOverflows > 0);
// If no error has occured 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 BtCursor.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.
this.Info.nSize = 0;
this.ValidNKey = false;
if (rc == RC.OK && pPage.NOverflows != 0)
{
rc = Balance();
// 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.
this.Pages[this.PageID].NOverflows = 0;
this.State = CursorState.INVALID;
}
Debug.Assert(this.Pages[this.PageID].NOverflows == 0);
end_insert:
return(rc);
}
// This function is invoked once for each page that has already been written into the log file when a WAL transaction is rolled back.
// Parameter iPg is the page number of said page. The pCtx argument is actually a pointer to the Pager structure.
//
// If page iPg is present in the cache, and has no outstanding references, it is discarded. Otherwise, if there are one or more outstanding
// references, the page content is reloaded from the database. If the attempt to reload content from the database is required and fails,
// return an SQLite error code. Otherwise, SQLITE.OK.
static int pagerUndoCallback(Pager pCtx, Pgno iPg)
{
var rc = SQLITE.OK;
Pager pPager = (Pager)pCtx;
PgHdr pPg;
pPg = sqlite3PagerLookup(pPager, iPg);
if (pPg)
{
if (sqlite3PcachePageRefcount(pPg) == 1)
{
sqlite3PcacheDrop(pPg);
}
else
{
rc = readDbPage(pPg);
if (rc == SQLITE.OK)
{
pPager.xReiniter(pPg);
}
sqlite3PagerUnref(pPg);
}
}
// Normally, if a transaction is rolled back, any backup processes are updated as data is copied out of the rollback journal and into the
// database. This is not generally possible with a WAL database, as rollback involves simply truncating the log file. Therefore, if one
// or more frames have already been written to the log (and therefore also copied into the backup databases) as part of this transaction,
// the backups must be restarted.
sqlite3BackupRestart(pPager.pBackup);
return rc;
}
/*
** This function is a wrapper around sqlite3WalFrames(). As well as logging
** the contents of the list of pages headed by pList (connected by pDirty),
** this function notifies any active backup processes that the pages have
** changed.
**
** The list of pages passed into this routine is always sorted by page number.
** Hence, if page 1 appears anywhere on the list, it will be the first page.
*/
static int pagerWalFrames(Pager pPager, PgHdr pList, Pgno nTruncate, int isCommit, int syncFlags)
{
int rc; /* Return code */
#if DEBUG || (SQLITE_CHECK_PAGES)
PgHdr p; /* For looping over pages */
#endif
Debug.Assert(pPager.pWal);
#if SQLITE_DEBUG
/* Verify that the page list is in accending order */
for(p=pList; p && p->pDirty; p=p->pDirty){
assert( p->pgno < p->pDirty->pgno );
}
#endif
if (isCommit)
{
/* If a WAL transaction is being committed, there is no point in writing
** any pages with page numbers greater than nTruncate into the WAL file.
** They will never be read by any client. So remove them from the pDirty
** list here. */
PgHdr* p;
PgHdr** ppNext = &pList;
for (p = pList; (*ppNext = p); p = p->pDirty)
{
if (p->pgno <= nTruncate) ppNext = &p->pDirty;
}
assert(pList);
}
if (pList->pgno == 1) pager_write_changecounter(pList);
rc = sqlite3WalFrames(pPager.pWal,
pPager.pageSize, pList, nTruncate, isCommit, syncFlags
);
if (rc == SQLITE.OK && pPager.pBackup)
{
PgHdr* p;
for (p = pList; p; p = p->pDirty)
{
sqlite3BackupUpdate(pPager.pBackup, p->pgno, (u8*)p->pData);
}
}
#if SQLITE_CHECK_PAGES
pList = sqlite3PcacheDirtyList(pPager.pPCache);
for(p=pList; p; p=p->pDirty){
pager_set_pagehash(p);
}
#endif
return rc;
}
// was:sqlite3PagerOpen
public static RC Open(VirtualFileSystem pVfs, out Pager ppPager, string zFilename, int nExtra, PAGEROPEN flags, VFSOPEN vfsFlags, Action<PgHdr> xReinit, Func<object> memPageBuilder)
{
Pager pPager = null; // Pager object to allocate and return
byte memDb = 0; // True if this is an in-memory file
bool readOnly = false; // True if this is a read-only file
string zPathname = null; // Full path to database file
//int nPathname = 0; // Number of bytes in zPathname
bool useJournal = (flags & PAGEROPEN.OMIT_JOURNAL) == 0; // False to omit journal
bool noReadlock = (flags & PAGEROPEN.NO_READLOCK) != 0; // True to omit read-lock
int pcacheSize = PCache.sqlite3PcacheSize(); // Bytes to allocate for PCache
uint szPageDflt = SQLITE_DEFAULT_PAGE_SIZE; // Default page size
string zUri = null; // URI args to copy
int nUri = 0; // Number of bytes of URI args at *zUri
// Figure out how much space is required for each journal file-handle (there are two of them, the main journal and the sub-journal). This
// is the maximum space required for an in-memory journal file handle and a regular journal file-handle. Note that a "regular journal-handle"
// may be a wrapper capable of caching the first portion of the journal file in memory to implement the atomic-write optimization (see
// source file journal.c).
int journalFileSize = SysEx.ROUND8(sqlite3JournalSize(pVfs) > MemJournalFile.sqlite3MemJournalSize() ? sqlite3JournalSize(pVfs) : MemJournalFile.sqlite3MemJournalSize()); // Bytes to allocate for each journal fd
// Set the output variable to NULL in case an error occurs.
ppPager = null;
#if !SQLITE_OMIT_MEMORYDB
if ((flags & PAGEROPEN.MEMORY) != 0)
{
memDb = 1;
zFilename = null;
}
#endif
// Compute and store the full pathname in an allocated buffer pointed to by zPathname, length nPathname. Or, if this is a temporary file,
// leave both nPathname and zPathname set to 0.
var rc = RC.OK;
if (!string.IsNullOrEmpty(zFilename))
{
rc = pVfs.xFullPathname(zFilename, out zPathname);
var z = zUri = zFilename;
nUri = zUri.Length;
if (rc == RC.OK && zPathname.Length + 8 > pVfs.mxPathname)
// This branch is taken when the journal path required by the database being opened will be more than pVfs.mxPathname
// bytes in length. This means the database cannot be opened, as it will not be possible to open the journal file or even
// check for a hot-journal before reading.
rc = SysEx.SQLITE_CANTOPEN_BKPT();
if (rc != RC.OK)
return rc;
}
// Allocate memory for the Pager structure, PCache object, the three file descriptors, the database file name and the journal
// file name. The layout in memory is as follows:
// Pager object (sizeof(Pager) bytes)
// PCache object (sqlite3PcacheSize() bytes)
// Database file handle (pVfs.szOsFile bytes)
// Sub-journal file handle (journalFileSize bytes)
// Main journal file handle (journalFileSize bytes)
// Database file name (nPathname+1 bytes)
// Journal file name (nPathname+8+1 bytes)
pPager = new Pager(memPageBuilder);
pPager.pPCache = new PCache();
pPager.fd = new VirtualFile();
pPager.sjfd = new VirtualFile();
pPager.jfd = new VirtualFile();
// Fill in the Pager.zFilename and Pager.zJournal buffers, if required.
if (zPathname != null)
{
Debug.Assert(zPathname.Length > 0);
pPager.zFilename = zPathname.ToString();
zUri = pPager.zFilename;
pPager.zJournal = pPager.zFilename + "-journal";
#if !SQLITE_OMIT_WAL
pPager.zWal = &pPager.zJournal[nPathname + 8 + 1];
memcpy(pPager.zWal, zPathname, nPathname);
memcpy(&pPager.zWal[nPathname], "-wal", 4);
#endif
}
else
pPager.zFilename = string.Empty;
pPager.pVfs = pVfs;
pPager.vfsFlags = vfsFlags;
// Open the pager file.
var tempFile = LOCKINGMODE.NORMAL; // True for temp files (incl. in-memory files)
if (!string.IsNullOrEmpty(zFilename))
{
VFSOPEN fout = 0; // VFS flags returned by xOpen()
rc = FileEx.sqlite3OsOpen(pVfs, zFilename, pPager.fd, vfsFlags, ref fout);
Debug.Assert(0 == memDb);
readOnly = (fout & VFSOPEN.READONLY) != 0;
// If the file was successfully opened for read/write access, choose a default page size in case we have to create the
// database file. The default page size is the maximum of:
// + SQLITE_DEFAULT_PAGE_SIZE,
// + The value returned by sqlite3OsSectorSize()
// + The largest page size that can be written atomically.
if (rc == RC.OK && !readOnly)
{
pPager.setSectorSize();
Debug.Assert(SQLITE_DEFAULT_PAGE_SIZE <= SQLITE_MAX_DEFAULT_PAGE_SIZE);
if (szPageDflt < pPager.sectorSize)
szPageDflt = (pPager.sectorSize > SQLITE_MAX_DEFAULT_PAGE_SIZE ? SQLITE_MAX_DEFAULT_PAGE_SIZE : (uint)pPager.sectorSize);
#if SQLITE_ENABLE_ATOMIC_WRITE
int iDc = sqlite3OsDeviceCharacteristics(pPager.fd);
Debug.Assert(SQLITE_IOCAP_ATOMIC512 == (512 >> 8));
Debug.Assert(SQLITE_IOCAP_ATOMIC64K == (65536 >> 8));
Debug.Assert(SQLITE_MAX_DEFAULT_PAGE_SIZE <= 65536);
for (var ii = szPageDflt; ii <= SQLITE_MAX_DEFAULT_PAGE_SIZE; ii = ii * 2)
if (iDc & (SQLITE_IOCAP_ATOMIC | (ii >> 8)))
szPageDflt = ii;
#endif
}
}
else
{
// If a temporary file is requested, it is not opened immediately. In this case we accept the default page size and delay actually
// opening the file until the first call to OsWrite().
// This branch is also run for an in-memory database. An in-memory database is the same as a temp-file that is never written out to
// disk and uses an in-memory rollback journal.
tempFile = LOCKINGMODE.EXCLUSIVE;
pPager.eState = PAGER.READER;
pPager.eLock = VFSLOCK.EXCLUSIVE;
readOnly = (vfsFlags & VFSOPEN.READONLY) != 0;
}
// The following call to PagerSetPagesize() serves to set the value of Pager.pageSize and to allocate the Pager.pTmpSpace buffer.
if (rc == RC.OK)
{
Debug.Assert(pPager.memDb == 0);
rc = pPager.SetPageSize(ref szPageDflt, -1);
}
// If an error occurred in either of the blocks above, free the Pager structure and close the file.
if (rc != RC.OK)
{
Debug.Assert(null == pPager.pTmpSpace);
FileEx.sqlite3OsClose(pPager.fd);
return rc;
}
// Initialize the PCache object.
Debug.Assert(nExtra < 1000);
nExtra = SysEx.ROUND8(nExtra);
PCache.Open((int)szPageDflt, nExtra, (memDb == 0), (memDb == 0 ? (Func<object, PgHdr, RC>)pagerStress : null), pPager, pPager.pPCache);
PAGERTRACE("OPEN {0} {1}", FILEHANDLEID(pPager.fd), pPager.zFilename);
SysEx.IOTRACE("OPEN {0:x} {1}", pPager.GetHashCode(), pPager.zFilename);
pPager.useJournal = (byte)(useJournal ? 1 : 0);
pPager.noReadlock = (byte)(noReadlock && readOnly ? 1 : 0);
pPager.mxPgno = SQLITE_MAX_PAGE_COUNT;
#if false
Debug.Assert(pPager.state == (tempFile != 0 ? PAGER.EXCLUSIVE : PAGER.UNLOCK));
#endif
pPager.tempFile = tempFile != 0;
Debug.Assert(tempFile == LOCKINGMODE.NORMAL || tempFile == LOCKINGMODE.EXCLUSIVE);
pPager.exclusiveMode = tempFile != 0;
pPager.changeCountDone = pPager.tempFile;
pPager.memDb = memDb;
pPager.readOnly = readOnly;
Debug.Assert(useJournal || pPager.tempFile);
pPager.noSync = pPager.tempFile;
pPager.fullSync = pPager.noSync;
pPager.syncFlags = (pPager.noSync ? 0 : VirtualFile.SYNC.NORMAL);
pPager.ckptSyncFlags = pPager.syncFlags;
pPager.nExtra = (ushort)nExtra;
pPager.journalSizeLimit = SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT;
Debug.Assert(pPager.fd.IsOpen || tempFile != 0);
pPager.setSectorSize();
if (!useJournal)
pPager.journalMode = JOURNALMODE.OFF;
else if (memDb != 0)
pPager.journalMode = JOURNALMODE.MEMORY;
pPager.xReiniter = xReinit;
ppPager = pPager;
return RC.OK;
}
// was:balance
private RC Balance()
{
var rc = RC.OK;
var nMin = (int)this.Shared.UsableSize * 2 / 3;
var balance_quick_called = 0;
var balance_deeper_called = 0;
do
{
var pageID = this.PageID;
var page = this.Pages[pageID];
if (pageID == 0)
{
if (page.NOverflows != 0)
{
// The root page of the b-tree is overfull. In this case call the balance_deeper() function to create a new child for the root-page
// and copy the current contents of the root-page to it. The next iteration of the do-loop will balance the child page.
Debug.Assert((balance_deeper_called++) == 0);
rc = MemPage.balance_deeper(page, ref this.Pages[1]);
if (rc == RC.OK)
{
this.PageID = 1;
this.PagesIndexs[0] = 0;
this.PagesIndexs[1] = 0;
Debug.Assert(this.Pages[1].NOverflows != 0);
}
}
else
{
break;
}
}
else if (page.NOverflows == 0 && page.FreeBytes <= nMin)
{
break;
}
else
{
var pParent = this.Pages[pageID - 1];
var iIdx = this.PagesIndexs[pageID - 1];
rc = Pager.Write(pParent.DbPage);
if (rc == RC.OK)
{
#if !SQLITE_OMIT_QUICKBALANCE
if (page.HasData != 0 && page.NOverflows == 1 && page.Overflows[0].Index == page.Cells && pParent.ID != 1 && pParent.Cells == iIdx)
{
// Call balance_quick() to create a new sibling of pPage on which to store the overflow cell. balance_quick() inserts a new cell
// into pParent, which may cause pParent overflow. If this happens, the next interation of the do-loop will balance pParent
// use either balance_nonroot() or balance_deeper(). Until this happens, the overflow cell is stored in the aBalanceQuickSpace[]
// buffer.
// The purpose of the following Debug.Assert() is to check that only a single call to balance_quick() is made for each call to this
// function. If this were not verified, a subtle bug involving reuse of the aBalanceQuickSpace[] might sneak in.
Debug.Assert((balance_quick_called++) == 0);
rc = MemPage.balance_quick(pParent, page, _balanceQuickSpaces);
}
else
#endif
{
// In this case, call balance_nonroot() to redistribute cells between pPage and up to 2 of its sibling pages. This involves
// modifying the contents of pParent, which may cause pParent to become overfull or underfull. The next iteration of the do-loop
// will balance the parent page to correct this.
// If the parent page becomes overfull, the overflow cell or cells are stored in the pSpace buffer allocated immediately below.
// A subsequent iteration of the do-loop will deal with this by calling balance_nonroot() (balance_deeper() may be called first,
// but it doesn't deal with overflow cells - just moves them to a different page). Once this subsequent call to balance_nonroot()
// has completed, it is safe to release the pSpace buffer used by the previous call, as the overflow cell data will have been
// copied either into the body of a database page or into the new pSpace buffer passed to the latter call to balance_nonroot().
var pSpace = new byte[this.Shared.PageSize];// u8 pSpace = sqlite3PageMalloc( pCur.pBt.pageSize );
rc = MemPage.balance_nonroot(pParent, iIdx, null, pageID == 1 ? 1 : 0);
// The pSpace buffer will be freed after the next call to balance_nonroot(), or just before this function returns, whichever comes first.
}
}
page.NOverflows = 0;
// The next iteration of the do-loop balances the parent page.
page.releasePage();
this.PageID--;
}
} while (rc == RC.OK);
return(rc);
}
