*
* This class defines the methods as abstract that subclasses must implement.
* Outside callers should *not* use functions with "Internal" in the name.
- *
+ *
* The FileBackend operations are implemented using basic functions
* such as storeInternal(), copyInternal(), deleteInternal() and the like.
* This class is also responsible for path resolution and sanitization.
- *
+ *
* @ingroup FileBackend
* @since 1.19
*/
abstract class FileBackendStore extends FileBackend {
+ /** @var BagOStuff */
+ protected $memCache;
+
/** @var Array Map of paths to small (RAM/disk) cache items */
protected $cache = array(); // (storage path => key => value)
- protected $maxCacheSize = 100; // integer; max paths with entries
+ protected $maxCacheSize = 300; // integer; max paths with entries
/** @var Array Map of paths to large (RAM/disk) cache items */
protected $expensiveCache = array(); // (storage path => key => value)
- protected $maxExpensiveCacheSize = 10; // integer; max paths with entries
+ protected $maxExpensiveCacheSize = 5; // integer; max paths with entries
/** @var Array Map of container names to sharding settings */
protected $shardViaHashLevels = array(); // (container name => config array)
protected $maxFileSize = 4294967296; // integer bytes (4GiB)
+ /**
+ * @see FileBackend::__construct()
+ *
+ * @param $config Array
+ */
+ public function __construct( array $config ) {
+ parent::__construct( $config );
+ $this->memCache = new EmptyBagOStuff(); // disabled by default
+ }
+
/**
* Get the maximum allowable file size given backend
* medium restrictions and basic performance constraints.
* Do not call this function from places outside FileBackend and FileOp.
- *
- * @return integer Bytes
+ *
+ * @return integer Bytes
*/
final public function maxFileSizeInternal() {
return $this->maxFileSize;
/**
* Create a file in the backend with the given contents.
* Do not call this function from places outside FileBackend and FileOp.
- *
+ *
* $params include:
* content : the raw file contents
* dst : destination storage path
* overwrite : overwrite any file that exists at the destination
- *
+ *
* @param $params Array
* @return Status
*/
final public function createInternal( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
if ( strlen( $params['content'] ) > $this->maxFileSizeInternal() ) {
$status = Status::newFatal( 'backend-fail-maxsize',
$params['dst'], $this->maxFileSizeInternal() );
$status = $this->doCreateInternal( $params );
$this->clearCache( array( $params['dst'] ) );
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
/**
* Store a file into the backend from a file on disk.
* Do not call this function from places outside FileBackend and FileOp.
- *
+ *
* $params include:
* src : source path on disk
* dst : destination storage path
* overwrite : overwrite any file that exists at the destination
- *
+ *
* @param $params Array
* @return Status
*/
final public function storeInternal( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
if ( filesize( $params['src'] ) > $this->maxFileSizeInternal() ) {
$status = Status::newFatal( 'backend-fail-store', $params['dst'] );
} else {
$status = $this->doStoreInternal( $params );
$this->clearCache( array( $params['dst'] ) );
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
/**
* Copy a file from one storage path to another in the backend.
* Do not call this function from places outside FileBackend and FileOp.
- *
+ *
* $params include:
* src : source storage path
* dst : destination storage path
* overwrite : overwrite any file that exists at the destination
- *
+ *
* @param $params Array
* @return Status
*/
final public function copyInternal( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = $this->doCopyInternal( $params );
$this->clearCache( array( $params['dst'] ) );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
/**
* Delete a file at the storage path.
* Do not call this function from places outside FileBackend and FileOp.
- *
+ *
* $params include:
* src : source storage path
* ignoreMissingSource : do nothing if the source file does not exist
- *
+ *
* @param $params Array
* @return Status
*/
final public function deleteInternal( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = $this->doDeleteInternal( $params );
$this->clearCache( array( $params['src'] ) );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
/**
* Move a file from one storage path to another in the backend.
* Do not call this function from places outside FileBackend and FileOp.
- *
+ *
* $params include:
* src : source storage path
* dst : destination storage path
* overwrite : overwrite any file that exists at the destination
- *
+ *
* @param $params Array
* @return Status
*/
final public function moveInternal( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = $this->doMoveInternal( $params );
$this->clearCache( array( $params['src'], $params['dst'] ) );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
*/
final public function concatenate( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = Status::newGood();
// Try to lock the source files for the scope of this function
$status->merge( $this->doConcatenate( $params ) );
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
*/
final protected function doPrepare( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = Status::newGood();
list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
if ( $dir === null ) {
$status->fatal( 'backend-fail-invalidpath', $params['dir'] );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status; // invalid storage path
}
}
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
*/
final protected function doSecure( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = Status::newGood();
list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
if ( $dir === null ) {
$status->fatal( 'backend-fail-invalidpath', $params['dir'] );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status; // invalid storage path
}
}
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
*/
final protected function doClean( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = Status::newGood();
+ // Recursive: first delete all empty subdirs recursively
+ if ( !empty( $params['recursive'] ) && !$this->directoriesAreVirtual() ) {
+ $subDirsRel = $this->getTopDirectoryList( array( 'dir' => $params['dir'] ) );
+ if ( $subDirsRel !== null ) { // no errors
+ foreach ( $subDirsRel as $subDirRel ) {
+ $subDir = $params['dir'] . "/{$subDirRel}"; // full path
+ $status->merge( $this->doClean( array( 'dir' => $subDir ) + $params ) );
+ }
+ }
+ }
+
list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
if ( $dir === null ) {
$status->fatal( 'backend-fail-invalidpath', $params['dir'] );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status; // invalid storage path
}
$filesLockEx = array( $params['dir'] );
$scopedLockE = $this->getScopedFileLocks( $filesLockEx, LockManager::LOCK_EX, $status );
if ( !$status->isOK() ) {
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status; // abort
}
if ( $shard !== null ) { // confined to a single container/shard
$status->merge( $this->doCleanInternal( $fullCont, $dir, $params ) );
+ $this->deleteContainerCache( $fullCont ); // purge cache
} else { // directory is on several shards
wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
list( $b, $shortCont, $r ) = self::splitStoragePath( $params['dir'] );
foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
$status->merge( $this->doCleanInternal( "{$fullCont}{$suffix}", $dir, $params ) );
+ $this->deleteContainerCache( "{$fullCont}{$suffix}" ); // purge cache
}
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
*/
final public function fileExists( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$stat = $this->getFileStat( $params );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return ( $stat === null ) ? null : (bool)$stat; // null => failure
}
*/
final public function getFileTimestamp( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$stat = $this->getFileStat( $params );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $stat ? $stat['mtime'] : false;
}
*/
final public function getFileSize( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$stat = $this->getFileStat( $params );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $stat ? $stat['size'] : false;
}
*/
final public function getFileStat( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$path = self::normalizeStoragePath( $params['src'] );
if ( $path === null ) {
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return false; // invalid storage path
}
// If we want the latest data, check that this cached
// value was in fact fetched with the latest available data.
if ( !$latest || $this->cache[$path]['stat']['latest'] ) {
+ $this->pingCache( $path ); // LRU
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $this->cache[$path]['stat'];
}
}
wfProfileIn( __METHOD__ . '-miss' );
+ wfProfileIn( __METHOD__ . '-miss-' . $this->name );
$stat = $this->doGetFileStat( $params );
+ wfProfileOut( __METHOD__ . '-miss-' . $this->name );
wfProfileOut( __METHOD__ . '-miss' );
if ( is_array( $stat ) ) { // don't cache negatives
$this->trimCache(); // limit memory
$this->cache[$path]['stat'] = $stat;
$this->cache[$path]['stat']['latest'] = $latest;
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $stat;
}
*/
public function getFileContents( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$tmpFile = $this->getLocalReference( $params );
if ( !$tmpFile ) {
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return false;
}
wfSuppressWarnings();
$data = file_get_contents( $tmpFile->getPath() );
wfRestoreWarnings();
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $data;
}
*/
final public function getFileSha1Base36( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$path = $params['src'];
if ( isset( $this->cache[$path]['sha1'] ) ) {
+ $this->pingCache( $path ); // LRU
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $this->cache[$path]['sha1'];
}
wfProfileIn( __METHOD__ . '-miss' );
+ wfProfileIn( __METHOD__ . '-miss-' . $this->name );
$hash = $this->doGetFileSha1Base36( $params );
+ wfProfileOut( __METHOD__ . '-miss-' . $this->name );
wfProfileOut( __METHOD__ . '-miss' );
if ( $hash ) { // don't cache negatives
$this->trimCache(); // limit memory
$this->cache[$path]['sha1'] = $hash;
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $hash;
}
*/
final public function getFileProps( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$fsFile = $this->getLocalReference( $params );
$props = $fsFile ? $fsFile->getProps() : FSFile::placeholderProps();
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $props;
}
*/
public function getLocalReference( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$path = $params['src'];
if ( isset( $this->expensiveCache[$path]['localRef'] ) ) {
+ $this->pingExpensiveCache( $path );
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $this->expensiveCache[$path]['localRef'];
}
$this->trimExpensiveCache(); // limit memory
$this->expensiveCache[$path]['localRef'] = $tmpFile;
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $tmpFile;
}
*/
final public function streamFile( array $params ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = Status::newGood();
$info = $this->getFileStat( $params );
// do nothing; client cache is up to date
} elseif ( $res == StreamFile::READY_STREAM ) {
wfProfileIn( __METHOD__ . '-send' );
+ wfProfileIn( __METHOD__ . '-send-' . $this->name );
$status = $this->doStreamFile( $params );
+ wfProfileOut( __METHOD__ . '-send-' . $this->name );
wfProfileOut( __METHOD__ . '-send' );
} else {
$status->fatal( 'backend-fail-stream', $params['src'] );
}
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
}
/**
- * @copydoc FileBackend::getFileList()
- * @return Array|null|Traversable
+ * @see FileBackend::directoryExists()
+ * @return bool|null
+ */
+ final public function directoryExists( array $params ) {
+ list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
+ if ( $dir === null ) {
+ return false; // invalid storage path
+ }
+ if ( $shard !== null ) { // confined to a single container/shard
+ return $this->doDirectoryExists( $fullCont, $dir, $params );
+ } else { // directory is on several shards
+ wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
+ list( $b, $shortCont, $r ) = self::splitStoragePath( $params['dir'] );
+ $res = false; // response
+ foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
+ $exists = $this->doDirectoryExists( "{$fullCont}{$suffix}", $dir, $params );
+ if ( $exists ) {
+ $res = true;
+ break; // found one!
+ } elseif ( $exists === null ) { // error?
+ $res = null; // if we don't find anything, it is indeterminate
+ }
+ }
+ return $res;
+ }
+ }
+
+ /**
+ * @see FileBackendStore::directoryExists()
+ *
+ * @param $container string Resolved container name
+ * @param $dir string Resolved path relative to container
+ * @param $params Array
+ * @return bool|null
+ */
+ abstract protected function doDirectoryExists( $container, $dir, array $params );
+
+ /**
+ * @see FileBackend::getDirectoryList()
+ * @return Traversable|Array|null Returns null on failure
+ */
+ final public function getDirectoryList( array $params ) {
+ list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
+ if ( $dir === null ) { // invalid storage path
+ return null;
+ }
+ if ( $shard !== null ) {
+ // File listing is confined to a single container/shard
+ return $this->getDirectoryListInternal( $fullCont, $dir, $params );
+ } else {
+ wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
+ // File listing spans multiple containers/shards
+ list( $b, $shortCont, $r ) = self::splitStoragePath( $params['dir'] );
+ return new FileBackendStoreShardDirIterator( $this,
+ $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
+ }
+ }
+
+ /**
+ * Do not call this function from places outside FileBackend
+ *
+ * @see FileBackendStore::getDirectoryList()
+ *
+ * @param $container string Resolved container name
+ * @param $dir string Resolved path relative to container
+ * @param $params Array
+ * @return Traversable|Array|null Returns null on failure
+ */
+ abstract public function getDirectoryListInternal( $container, $dir, array $params );
+
+ /**
+ * @see FileBackend::getFileList()
+ * @return Traversable|Array|null Returns null on failure
*/
final public function getFileList( array $params ) {
list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
// File listing spans multiple containers/shards
list( $b, $shortCont, $r ) = self::splitStoragePath( $params['dir'] );
- return new FileBackendStoreShardListIterator( $this,
+ return new FileBackendStoreShardFileIterator( $this,
$fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
}
}
* Do not call this function from places outside FileBackend
*
* @see FileBackendStore::getFileList()
- *
+ *
* @param $container string Resolved container name
* @param $dir string Resolved path relative to container
* @param $params Array
- * @return Traversable|Array|null
+ * @return Traversable|Array|null Returns null on failure
*/
abstract public function getFileListInternal( $container, $dir, array $params );
/**
* Get the list of supported operations and their corresponding FileOp classes.
- *
+ *
* @return Array
*/
protected function supportedOperations() {
*
* The result must have the same number of items as the input.
* An exception is thrown if an unsupported operation is requested.
- *
+ *
* @param $ops Array Same format as doOperations()
* @return Array List of FileOp objects
* @throws MWException
*/
- final public function getOperations( array $ops ) {
+ final public function getOperationsInternal( array $ops ) {
$supportedOps = $this->supportedOperations();
$performOps = array(); // array of FileOp objects
// Append the FileOp class
$performOps[] = new $class( $this, $params );
} else {
- throw new MWException( "Operation `$opName` is not supported." );
+ throw new MWException( "Operation '$opName' is not supported." );
}
}
return $performOps;
}
+ /**
+ * Get a list of storage paths to lock for a list of operations
+ * Returns an array with 'sh' (shared) and 'ex' (exclusive) keys,
+ * each corresponding to a list of storage paths to be locked.
+ *
+ * @param $performOps Array List of FileOp objects
+ * @return Array ('sh' => list of paths, 'ex' => list of paths)
+ */
+ final public function getPathsToLockForOpsInternal( array $performOps ) {
+ // Build up a list of files to lock...
+ $paths = array( 'sh' => array(), 'ex' => array() );
+ foreach ( $performOps as $fileOp ) {
+ $paths['sh'] = array_merge( $paths['sh'], $fileOp->storagePathsRead() );
+ $paths['ex'] = array_merge( $paths['ex'], $fileOp->storagePathsChanged() );
+ }
+ // Optimization: if doing an EX lock anyway, don't also set an SH one
+ $paths['sh'] = array_diff( $paths['sh'], $paths['ex'] );
+ // Get a shared lock on the parent directory of each path changed
+ $paths['sh'] = array_merge( $paths['sh'], array_map( 'dirname', $paths['ex'] ) );
+
+ return $paths;
+ }
+
/**
* @see FileBackend::doOperationsInternal()
* @return Status
*/
protected function doOperationsInternal( array $ops, array $opts ) {
wfProfileIn( __METHOD__ );
+ wfProfileIn( __METHOD__ . '-' . $this->name );
$status = Status::newGood();
// Build up a list of FileOps...
- $performOps = $this->getOperations( $ops );
+ $performOps = $this->getOperationsInternal( $ops );
// Acquire any locks as needed...
if ( empty( $opts['nonLocking'] ) ) {
// Build up a list of files to lock...
- $filesLockEx = $filesLockSh = array();
- foreach ( $performOps as $fileOp ) {
- $filesLockSh = array_merge( $filesLockSh, $fileOp->storagePathsRead() );
- $filesLockEx = array_merge( $filesLockEx, $fileOp->storagePathsChanged() );
- }
- // Optimization: if doing an EX lock anyway, don't also set an SH one
- $filesLockSh = array_diff( $filesLockSh, $filesLockEx );
- // Get a shared lock on the parent directory of each path changed
- $filesLockSh = array_merge( $filesLockSh, array_map( 'dirname', $filesLockEx ) );
+ $paths = $this->getPathsToLockForOpsInternal( $performOps );
// Try to lock those files for the scope of this function...
- $scopeLockS = $this->getScopedFileLocks( $filesLockSh, LockManager::LOCK_UW, $status );
- $scopeLockE = $this->getScopedFileLocks( $filesLockEx, LockManager::LOCK_EX, $status );
+ $scopeLockS = $this->getScopedFileLocks( $paths['sh'], LockManager::LOCK_UW, $status );
+ $scopeLockE = $this->getScopedFileLocks( $paths['ex'], LockManager::LOCK_EX, $status );
if ( !$status->isOK() ) {
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status; // abort
}
}
- // Clear any cache entries (after locks acquired)
+ // Clear any file cache entries (after locks acquired)
$this->clearCache();
+ // Load from the persistent container cache
+ $this->primeContainerCache( $performOps );
+
// Actually attempt the operation batch...
- $subStatus = FileOp::attemptBatch( $performOps, $opts );
+ $subStatus = FileOp::attemptBatch( $performOps, $opts, $this->fileJournal );
// Merge errors into status fields
$status->merge( $subStatus );
$status->success = $subStatus->success; // not done in merge()
+ wfProfileOut( __METHOD__ . '-' . $this->name );
wfProfileOut( __METHOD__ );
return $status;
}
/**
* Clears any additional stat caches for storage paths
- *
+ *
* @see FileBackend::clearCache()
- *
+ *
* @param $paths Array Storage paths (optional)
* @return void
*/
protected function doClearCache( array $paths = null ) {}
+ /**
+ * Is this a key/value store where directories are just virtual?
+ * Virtual directories exists in so much as files exists that are
+ * prefixed with the directory path followed by a forward slash.
+ *
+ * @return bool
+ */
+ abstract protected function directoriesAreVirtual();
+
+ /**
+ * Move a cache entry to the top (such as when accessed)
+ *
+ * @param $path string Storage path
+ * @return void
+ */
+ protected function pingCache( $path ) {
+ if ( isset( $this->cache[$path] ) ) {
+ $tmp = $this->cache[$path];
+ unset( $this->cache[$path] );
+ $this->cache[$path] = $tmp;
+ }
+ }
+
/**
* Prune the inexpensive cache if it is too big to add an item
- *
+ *
* @return void
*/
protected function trimCache() {
}
}
+ /**
+ * Move a cache entry to the top (such as when accessed)
+ *
+ * @param $path string Storage path
+ * @return void
+ */
+ protected function pingExpensiveCache( $path ) {
+ if ( isset( $this->expensiveCache[$path] ) ) {
+ $tmp = $this->expensiveCache[$path];
+ unset( $this->expensiveCache[$path] );
+ $this->expensiveCache[$path] = $tmp;
+ }
+ }
+
/**
* Prune the expensive cache if it is too big to add an item
- *
+ *
* @return void
*/
protected function trimExpensiveCache() {
/**
* Check if a container name is valid.
* This checks for for length and illegal characters.
- *
+ *
* @param $container string
* @return bool
*/
final protected static function isValidContainerName( $container ) {
- // This accounts for Swift and S3 restrictions while leaving room
+ // This accounts for Swift and S3 restrictions while leaving room
// for things like '.xxx' (hex shard chars) or '.seg' (segments).
// This disallows directory separators or traversal characters.
// Note that matching strings URL encode to the same string;
return ''; // no sharding
}
+ /**
+ * Check if a storage path maps to a single shard.
+ * Container dirs like "a", where the container shards on "x/xy",
+ * can reside on several shards. Such paths are tricky to handle.
+ *
+ * @param $storagePath string Storage path
+ * @return bool
+ */
+ final public function isSingleShardPathInternal( $storagePath ) {
+ list( $c, $r, $shard ) = $this->resolveStoragePath( $storagePath );
+ return ( $shard !== null );
+ }
+
/**
* Get the sharding config for a container.
* If greater than 0, then all file storage paths within
/**
* Get a list of full container shard suffixes for a container
- *
+ *
* @param $container string
- * @return Array
+ * @return Array
*/
final protected function getContainerSuffixes( $container ) {
$shards = array();
/**
* Get the full container name, including the wiki ID prefix
- *
+ *
* @param $container string
- * @return string
+ * @return string
*/
final protected function fullContainerName( $container ) {
if ( $this->wikiId != '' ) {
* Resolve a container name, checking if it's allowed by the backend.
* This is intended for internal use, such as encoding illegal chars.
* Subclasses can override this to be more restrictive.
- *
+ *
* @param $container string
- * @return string|null
+ * @return string|null
*/
protected function resolveContainerName( $container ) {
return $container;
protected function resolveContainerPath( $container, $relStoragePath ) {
return $relStoragePath;
}
+
+ /**
+ * Get the cache key for a container
+ *
+ * @param $container Resolved container name
+ * @return string
+ */
+ private function containerCacheKey( $container ) {
+ return wfMemcKey( 'backend', $this->getName(), 'container', $container );
+ }
+
+ /**
+ * Set the cached info for a container
+ *
+ * @param $container Resolved container name
+ * @param $val mixed Information to cache
+ * @return void
+ */
+ final protected function setContainerCache( $container, $val ) {
+ $this->memCache->set( $this->containerCacheKey( $container ), $val, 7*86400 );
+ }
+
+ /**
+ * Delete the cached info for a container
+ *
+ * @param $container Resolved container name
+ * @return void
+ */
+ final protected function deleteContainerCache( $container ) {
+ $this->memCache->delete( $this->containerCacheKey( $container ) );
+ }
+
+ /**
+ * Do a batch lookup from cache for container stats for all containers
+ * used in a list of container names, storage paths, or FileOp objects.
+ *
+ * @param $items Array List of storage paths or FileOps
+ * @return void
+ */
+ final protected function primeContainerCache( array $items ) {
+ $paths = array(); // list of storage paths
+ $contNames = array(); // (cache key => resolved container name)
+ // Get all the paths/containers from the items...
+ foreach ( $items as $item ) {
+ if ( $item instanceof FileOp ) {
+ $paths = array_merge( $paths, $item->storagePathsRead() );
+ $paths = array_merge( $paths, $item->storagePathsChanged() );
+ } elseif ( self::isStoragePath( $item ) ) {
+ $paths[] = $item;
+ } elseif ( is_string( $item ) ) { // full container name
+ $contNames[$this->containerCacheKey( $item )] = $item;
+ }
+ }
+ // Get all the corresponding cache keys for paths...
+ foreach ( $paths as $path ) {
+ list( $fullCont, $r, $s ) = $this->resolveStoragePath( $path );
+ if ( $fullCont !== null ) { // valid path for this backend
+ $contNames[$this->containerCacheKey( $fullCont )] = $fullCont;
+ }
+ }
+
+ $contInfo = array(); // (resolved container name => cache value)
+ // Get all cache entries for these container cache keys...
+ $values = $this->memCache->getBatch( array_keys( $contNames ) );
+ foreach ( $values as $cacheKey => $val ) {
+ $contInfo[$contNames[$cacheKey]] = $val;
+ }
+
+ // Populate the container process cache for the backend...
+ $this->doPrimeContainerCache( array_filter( $contInfo, 'is_array' ) );
+ }
+
+ /**
+ * Fill the backend-specific process cache given an array of
+ * resolved container names and their corresponding cached info.
+ * Only containers that actually exist should appear in the map.
+ *
+ * @param $containerInfo Array Map of resolved container names to cached info
+ * @return void
+ */
+ protected function doPrimeContainerCache( array $containerInfo ) {}
}
/**
- * FileBackendStore helper function to handle file listings that span container shards.
+ * FileBackendStore helper function to handle listings that span container shards.
* Do not use this class from places outside of FileBackendStore.
*
* @ingroup FileBackend
*/
-class FileBackendStoreShardListIterator implements Iterator {
- /* @var FileBackendStore */
+abstract class FileBackendStoreShardListIterator implements Iterator {
+ /** @var FileBackendStore */
protected $backend;
- /* @var Array */
+ /** @var Array */
protected $params;
- /* @var Array */
+ /** @var Array */
protected $shardSuffixes;
- protected $container; // string
- protected $directory; // string
+ protected $container; // string; full container name
+ protected $directory; // string; resolved relative path
- /* @var Traversable */
+ /** @var Traversable */
protected $iter;
protected $curShard = 0; // integer
protected $pos = 0; // integer
+ /** @var Array */
+ protected $multiShardPaths = array(); // (rel path => 1)
+
/**
* @param $backend FileBackendStore
* @param $container string Full storage container name
} else {
$this->iter->next();
}
+ // Filter out items that we already listed
+ $this->filterViaNext();
// Find the next non-empty shard if no elements are left
$this->nextShardIteratorIfNotValid();
}
$this->pos = 0;
$this->curShard = 0;
$this->setIteratorFromCurrentShard();
+ // Filter out items that we already listed
+ $this->filterViaNext();
// Find the next non-empty shard if this one has no elements
$this->nextShardIteratorIfNotValid();
}
* @return bool
*/
public function valid() {
- if ( $this->iter == null ) {
+ if ( $this->iter === null ) {
return false; // some failure?
} elseif ( is_array( $this->iter ) ) {
return ( current( $this->iter ) !== false ); // no paths can have this value
}
}
+ /**
+ * Filter out duplicate items by advancing to the next ones
+ */
+ protected function filterViaNext() {
+ while ( $this->iter->valid() ) {
+ $rel = $this->iter->current(); // path relative to given directory
+ $path = $this->params['dir'] . "/{$rel}"; // full storage path
+ if ( !$this->backend->isSingleShardPathInternal( $path ) ) {
+ // Don't keep listing paths that are on multiple shards
+ if ( isset( $this->multiShardPaths[$rel] ) ) {
+ $this->iter->next(); // we already listed this path
+ } else {
+ $this->multiShardPaths[$rel] = 1;
+ break;
+ }
+ }
+ }
+ }
+
/**
* If the list iterator for this container shard is out of items,
* then move on to the next container that has items.
*/
protected function setIteratorFromCurrentShard() {
$suffix = $this->shardSuffixes[$this->curShard];
- $this->iter = $this->backend->getFileListInternal(
+ $this->iter = $this->listFromShard(
"{$this->container}{$suffix}", $this->directory, $this->params );
}
+
+ /**
+ * Get the list for a given container shard
+ *
+ * @param $container string Resolved container name
+ * @param $dir string Resolved path relative to container
+ * @param $params Array
+ * @return Traversable|Array|null
+ */
+ abstract protected function listFromShard( $container, $dir, array $params );
+}
+
+/**
+ * Iterator for listing directories
+ */
+class FileBackendStoreShardDirIterator extends FileBackendStoreShardListIterator {
+ protected function listFromShard( $container, $dir, array $params ) {
+ return $this->backend->getDirectoryListInternal( $container, $dir, $params );
+ }
+}
+
+/**
+ * Iterator for listing regular files
+ */
+class FileBackendStoreShardFileIterator extends FileBackendStoreShardListIterator {
+ protected function listFromShard( $container, $dir, array $params ) {
+ return $this->backend->getFileListInternal( $container, $dir, $params );
+ }
}