3 * UploadStash is intended to accomplish a few things:
4 * - enable applications to temporarily stash files without publishing them to the wiki.
5 * - Several parts of MediaWiki do this in similar ways: UploadBase, UploadWizard, and FirefoggChunkedExtension
6 * And there are several that reimplement stashing from scratch, in idiosyncratic ways. The idea is to unify them all here.
7 * Mostly all of them are the same except for storing some custom fields, which we subsume into the data array.
8 * - enable applications to find said files later, as long as the db table or temp files haven't been purged.
9 * - enable the uploading user (and *ONLY* the uploading user) to access said files, and thumbnails of said files, via a URL.
10 * We accomplish this using a database table, with ownership checking as you might expect. See SpecialUploadStash, which
11 * implements a web interface to some files stored this way.
13 * UploadStash represents the entire stash of temporary files.
14 * UploadStashFile is a filestore for the actual physical disk files.
15 * UploadFromStash extends UploadBase, and represents a single stashed file as it is moved from the stash to the regular file repository
19 // Format of the key for files -- has to be suitable as a filename itself (e.g. ab12cd34ef.jpg)
20 const KEY_FORMAT_REGEX
= '/^[\w-]+\.\w*$/';
22 // When a given stashed file can't be loaded, wait for the slaves to catch up. If they're more than MAX_LAG
23 // behind, throw an exception instead. (at what point is broken better than slow?)
26 // Age of the repository in hours. That is, after how long will files be assumed abandoned and deleted?
30 * repository that this uses to store temp files
31 * public because we sometimes need to get a LocalFile within the same repo.
37 // array of initialized repo objects
38 protected $files = array();
40 // cache of the file metadata that's stored in the database
41 protected $fileMetadata = array();
44 protected $fileProps = array();
47 protected $user, $userId, $isLoggedIn;
50 * Represents a temporary filestore, with metadata in the database.
51 * Designed to be compatible with the session stashing code in UploadBase (should replace it eventually)
53 * @param $repo FileRepo
55 public function __construct( $repo, $user = null ) {
56 // this might change based on wiki's configuration.
59 // if a user was passed, use it. otherwise, attempt to use the global.
60 // this keeps FileRepo from breaking when it creates an UploadStash object
65 $this->user
= $wgUser;
68 if( is_object($this->user
) ) {
69 $this->userId
= $this->user
->getId();
70 $this->isLoggedIn
= $this->user
->isLoggedIn();
75 * Get a file and its metadata from the stash.
77 * @param $key String: key under which file information is stored
78 * @param $noAuth Boolean (optional) Don't check authentication. Used by maintenance scripts.
79 * @throws UploadStashFileNotFoundException
80 * @throws UploadStashNotLoggedInException
81 * @throws UploadStashWrongOwnerException
82 * @throws UploadStashBadPathException
83 * @return UploadStashFile
85 public function getFile( $key, $noAuth = false ) {
87 if ( ! preg_match( self
::KEY_FORMAT_REGEX
, $key ) ) {
88 throw new UploadStashBadPathException( "key '$key' is not in a proper format" );
92 if( !$this->isLoggedIn
) {
93 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
97 $dbr = $this->repo
->getSlaveDb();
99 if ( !isset( $this->fileMetadata
[$key] ) ) {
100 // try this first. if it fails to find the row, check for lag, wait, try again. if its still missing, throw an exception.
101 // this more complex solution keeps things moving for page loads with many requests
102 // (ie. validating image ownership) when replag is high
103 if( !$this->fetchFileMetadata($key) ) {
104 $lag = $dbr->getLag();
105 if( $lag > 0 && $lag <= self
::MAX_LAG
) {
106 // if there's not too much replication lag, just wait for the slave to catch up to our last insert.
107 sleep( ceil( $lag ) );
108 } elseif($lag > self
::MAX_LAG
) {
109 // that's a lot of lag to introduce into the middle of the UI.
110 throw new UploadStashMaxLagExceededException(
111 'Couldn\'t load stashed file metadata, and replication lag is above threshold: (MAX_LAG=' . self
::MAX_LAG
. ')'
115 // now that the waiting has happened, try again
116 $this->fetchFileMetadata($key);
119 if ( !isset( $this->fileMetadata
[$key] ) ) {
120 throw new UploadStashFileNotFoundException( "key '$key' not found in stash" );
123 // create $this->files[$key]
124 $this->initFile( $key );
127 $path = $this->fileMetadata
[$key]['us_path'];
128 if ( $this->repo
->isVirtualUrl( $path ) ) {
129 $path = $this->repo
->resolveVirtualUrl( $path );
131 $this->fileProps
[$key] = File
::getPropsFromPath( $path );
135 if( $this->fileMetadata
[$key]['us_user'] != $this->userId
) {
136 throw new UploadStashWrongOwnerException( "This file ($key) doesn't belong to the current user." );
140 return $this->files
[$key];
144 * Getter for file metadata.
146 * @param key String: key under which file information is stored
149 public function getMetadata ( $key ) {
150 $this->getFile( $key );
151 return $this->fileMetadata
[$key];
155 * Getter for fileProps
157 * @param key String: key under which file information is stored
160 public function getFileProps ( $key ) {
161 $this->getFile( $key );
162 return $this->fileProps
[$key];
166 * Stash a file in a temp directory and record that we did this in the database, along with other metadata.
168 * @param $path String: path to file you want stashed
169 * @param $sourceType String: the type of upload that generated this file (currently, I believe, 'file' or null)
170 * @param $key String: optional, unique key for this file. Used for directory hashing when storing, otherwise not important
171 * @throws UploadStashBadPathException
172 * @throws UploadStashFileException
173 * @throws UploadStashNotLoggedInException
174 * @return UploadStashFile: file, or null on failure
176 public function stashFile( $path, $sourceType = null, $key = null ) {
177 if ( ! file_exists( $path ) ) {
178 wfDebug( __METHOD__
. " tried to stash file at '$path', but it doesn't exist\n" );
179 throw new UploadStashBadPathException( "path doesn't exist" );
181 $fileProps = File
::getPropsFromPath( $path );
183 wfDebug( __METHOD__
. " stashing file at '$path'\n" );
185 // we will be initializing from some tmpnam files that don't have extensions.
186 // most of MediaWiki assumes all uploaded files have good extensions. So, we fix this.
187 $extension = self
::getExtensionForPath( $path );
188 if ( ! preg_match( "/\\.\\Q$extension\\E$/", $path ) ) {
189 $pathWithGoodExtension = "$path.$extension";
190 if ( ! rename( $path, $pathWithGoodExtension ) ) {
191 throw new UploadStashFileException( "couldn't rename $path to have a better extension at $pathWithGoodExtension" );
193 $path = $pathWithGoodExtension;
196 // If no key was supplied, use content hash. Also has the nice property of collapsing multiple identical files
197 // uploaded this session, which could happen if uploads had failed.
198 if ( is_null( $key ) ) {
199 $key = $fileProps['sha1'] . "." . $extension;
202 $this->fileProps
[$key] = $fileProps;
204 if ( ! preg_match( self
::KEY_FORMAT_REGEX
, $key ) ) {
205 throw new UploadStashBadPathException( "key '$key' is not in a proper format" );
208 wfDebug( __METHOD__
. " key for '$path': $key\n" );
210 // if not already in a temporary area, put it there
211 $storeResult = $this->repo
->storeTemp( basename( $path ), $path );
213 if( ! $storeResult->isOK() ) {
214 // It is a convention in MediaWiki to only return one error per API exception, even if multiple errors
215 // are available. We use reset() to pick the "first" thing that was wrong, preferring errors to warnings.
216 // This is a bit lame, as we may have more info in the $storeResult and we're throwing it away, but to fix it means
217 // redesigning API errors significantly.
218 // $storeResult->value just contains the virtual URL (if anything) which is probably useless to the caller
219 $error = $storeResult->getErrorsArray();
220 $error = reset( $error );
221 if ( ! count( $error ) ) {
222 $error = $storeResult->getWarningsArray();
223 $error = reset( $error );
224 if ( ! count( $error ) ) {
225 $error = array( 'unknown', 'no error recorded' );
228 throw new UploadStashFileException( "error storing file in '$path': " . implode( '; ', $error ) );
230 $stashPath = $storeResult->value
;
232 // fetch the current user ID
233 if( !$this->isLoggedIn
) {
235 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
238 // insert the file metadata into the db.
239 wfDebug( __METHOD__
. " inserting $stashPath under $key\n" );
240 $dbw = $this->repo
->getMasterDb();
242 // select happens on the master so this can all be in a transaction, which
243 // avoids a race condition that's likely with multiple people uploading from the same
246 // first, check to see if it's already there.
247 $row = $dbw->selectRow(
249 'us_user, us_timestamp',
250 array('us_key' => $key),
254 // The current user can't have this key if:
255 // - the key is owned by someone else and
256 // - the age of the key is less than REPO_AGE
257 if( is_object( $row ) ) {
258 if( $row->us_user
!= $this->userId
&&
259 $row->wfTimestamp( TS_UNIX
, $row->us_timestamp
) > time() - UploadStash
::REPO_AGE
* 3600
262 throw new UploadStashWrongOwnerException( "Attempting to upload a duplicate of a file that someone else has stashed" );
266 $this->fileMetadata
[$key] = array(
267 'us_user' => $this->userId
,
269 'us_orig_path' => $path,
270 'us_path' => $stashPath,
271 'us_size' => $fileProps['size'],
272 'us_sha1' => $fileProps['sha1'],
273 'us_mime' => $fileProps['mime'],
274 'us_media_type' => $fileProps['media_type'],
275 'us_image_width' => $fileProps['width'],
276 'us_image_height' => $fileProps['height'],
277 'us_image_bits' => $fileProps['bits'],
278 'us_source_type' => $sourceType,
279 'us_timestamp' => $dbw->timestamp(),
280 'us_status' => 'finished'
283 // if a row exists but previous checks on it passed, let the current user take over this key.
287 $this->fileMetadata
[$key],
292 // store the insertid in the class variable so immediate retrieval (possibly laggy) isn't necesary.
293 $this->fileMetadata
[$key]['us_id'] = $dbw->insertId();
295 # create the UploadStashFile object for this file.
296 $this->initFile( $key );
298 return $this->getFile( $key );
302 * Remove all files from the stash.
303 * Does not clean up files in the repo, just the record of them.
305 * @throws UploadStashNotLoggedInException
306 * @return boolean: success
308 public function clear() {
309 if( !$this->isLoggedIn
) {
310 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
313 wfDebug( __METHOD__
. " clearing all rows for user $userId\n" );
314 $dbw = $this->repo
->getMasterDb();
317 array( 'us_user' => $this->userId
),
322 $this->files
= array();
323 $this->fileMetadata
= array();
329 * Remove a particular file from the stash. Also removes it from the repo.
331 * @throws UploadStashNotLoggedInException
332 * @throws UploadStashWrongOwnerException
333 * @return boolean: success
335 public function removeFile( $key ){
336 if( !$this->isLoggedIn
) {
337 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
340 $dbw = $this->repo
->getMasterDb();
342 // this is a cheap query. it runs on the master so that this function still works when there's lag.
343 // it won't be called all that often.
344 $row = $dbw->selectRow(
347 array('us_key' => $key),
351 if( $row->us_user
!= $this->userId
) {
352 throw new UploadStashWrongOwnerException( "Can't delete: the file ($key) doesn't belong to this user." );
355 return $this->removeFileNoAuth( $key );
360 * Remove a file (see removeFile), but doesn't check ownership first.
362 * @return boolean: success
364 public function removeFileNoAuth( $key ) {
365 wfDebug( __METHOD__
. " clearing row $key\n" );
367 $dbw = $this->repo
->getMasterDb();
369 // this gets its own transaction since it's called serially by the cleanupUploadStash maintenance script
373 array( 'us_key' => $key),
378 // TODO: look into UnregisteredLocalFile and find out why the rv here is sometimes wrong (false when file was removed)
380 $this->files
[$key]->remove();
382 unset( $this->files
[$key] );
383 unset( $this->fileMetadata
[$key] );
389 * List all files in the stash.
391 * @throws UploadStashNotLoggedInException
394 public function listFiles() {
395 if( !$this->isLoggedIn
) {
396 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
399 $dbr = $this->repo
->getSlaveDb();
403 array('us_key' => $key),
407 if( !is_object( $res ) ||
$res->numRows() == 0 ) {
412 // finish the read before starting writes.
414 foreach($res as $row) {
415 array_push( $keys, $row->us_key
);
422 * Find or guess extension -- ensuring that our extension matches our mime type.
423 * Since these files are constructed from php tempnames they may not start off
425 * XXX this is somewhat redundant with the checks that ApiUpload.php does with incoming
426 * uploads versus the desired filename. Maybe we can get that passed to us...
428 public static function getExtensionForPath( $path ) {
429 // Does this have an extension?
430 $n = strrpos( $path, '.' );
432 if ( $n !== false ) {
433 $extension = $n ?
substr( $path, $n +
1 ) : '';
435 // If not, assume that it should be related to the mime type of the original file.
436 $magic = MimeMagic
::singleton();
437 $mimeType = $magic->guessMimeType( $path );
438 $extensions = explode( ' ', MimeMagic
::singleton()->getExtensionsForType( $mimeType ) );
439 if ( count( $extensions ) ) {
440 $extension = $extensions[0];
444 if ( is_null( $extension ) ) {
445 throw new UploadStashFileException( "extension is null" );
448 return File
::normalizeExtension( $extension );
452 * Helper function: do the actual database query to fetch file metadata.
454 * @param $key String: key
457 protected function fetchFileMetadata( $key ) {
458 // populate $fileMetadata[$key]
459 $dbr = $this->repo
->getSlaveDb();
460 $row = $dbr->selectRow(
463 array('us_key' => $key),
467 if( !is_object( $row ) ) {
468 // key wasn't present in the database. this will happen sometimes.
472 $this->fileMetadata
[$key] = array(
473 'us_user' => $row->us_user
,
474 'us_key' => $row->us_key
,
475 'us_orig_path' => $row->us_orig_path
,
476 'us_path' => $row->us_path
,
477 'us_size' => $row->us_size
,
478 'us_sha1' => $row->us_sha1
,
479 'us_mime' => $row->us_mime
,
480 'us_media_type' => $row->us_media_type
,
481 'us_image_width' => $row->us_image_width
,
482 'us_image_height' => $row->us_image_height
,
483 'us_image_bits' => $row->us_image_bits
,
484 'us_source_type' => $row->us_source_type
,
485 'us_timestamp' => $row->us_timestamp
,
486 'us_status' => $row->us_status
493 * Helper function: Initialize the UploadStashFile for a given file.
495 * @param $path String: path to file
496 * @param $key String: key under which to store the object
497 * @throws UploadStashZeroLengthFileException
500 protected function initFile( $key ) {
501 $file = new UploadStashFile( $this->repo
, $this->fileMetadata
[$key]['us_path'], $key );
502 if ( $file->getSize() === 0 ) {
503 throw new UploadStashZeroLengthFileException( "File is zero length" );
505 $this->files
[$key] = $file;
510 class UploadStashFile
extends UnregisteredLocalFile
{
516 * A LocalFile wrapper around a file that has been temporarily stashed, so we can do things like create thumbnails for it
517 * Arguably UnregisteredLocalFile should be handling its own file repo but that class is a bit retarded currently
519 * @param $repo FSRepo: repository where we should find the path
520 * @param $path String: path to file
521 * @param $key String: key to store the path and any stashed data under
522 * @throws UploadStashBadPathException
523 * @throws UploadStashFileNotFoundException
525 public function __construct( $repo, $path, $key ) {
526 $this->fileKey
= $key;
528 // resolve mwrepo:// urls
529 if ( $repo->isVirtualUrl( $path ) ) {
530 $path = $repo->resolveVirtualUrl( $path );
533 // check if path appears to be sane, no parent traversals, and is in this repo's temp zone.
534 $repoTempPath = $repo->getZonePath( 'temp' );
535 if ( ( ! $repo->validateFilename( $path ) ) ||
536 ( strpos( $path, $repoTempPath ) !== 0 ) ) {
537 wfDebug( "UploadStash: tried to construct an UploadStashFile from a file that should already exist at '$path', but path is not valid\n" );
538 throw new UploadStashBadPathException( 'path is not valid' );
541 // check if path exists! and is a plain file.
542 if ( ! $repo->fileExists( $path, FileRepo
::FILES_ONLY
) ) {
543 wfDebug( "UploadStash: tried to construct an UploadStashFile from a file that should already exist at '$path', but path is not found\n" );
544 throw new UploadStashFileNotFoundException( 'cannot find path, or not a plain file' );
548 parent
::__construct( false, $repo, $path, false );
550 $this->name
= basename( $this->path
);
554 * A method needed by the file transforming and scaling routines in File.php
555 * We do not necessarily care about doing the description at this point
556 * However, we also can't return the empty string, as the rest of MediaWiki demands this (and calls to imagemagick
557 * convert require it to be there)
559 * @return String: dummy value
561 public function getDescriptionUrl() {
562 return $this->getUrl();
566 * Get the path for the thumbnail (actually any transformation of this file)
567 * The actual argument is the result of thumbName although we seem to have
568 * buggy code elsewhere that expects a boolean 'suffix'
570 * @param $thumbName String: name of thumbnail (e.g. "120px-123456.jpg" ), or false to just get the path
571 * @return String: path thumbnail should take on filesystem, or containing directory if thumbname is false
573 public function getThumbPath( $thumbName = false ) {
574 $path = dirname( $this->path
);
575 if ( $thumbName !== false ) {
576 $path .= "/$thumbName";
582 * Return the file/url base name of a thumbnail with the specified parameters.
583 * We override this because we want to use the pretty url name instead of the
586 * @param $params Array: handler-specific parameters
587 * @return String: base name for URL, like '120px-12345.jpg', or null if there is no handler
589 function thumbName( $params ) {
590 return $this->generateThumbName( $this->getUrlName(), $params );
594 * Helper function -- given a 'subpage', return the local URL e.g. /wiki/Special:UploadStash/subpage
595 * @param {String} $subPage
596 * @return {String} local URL for this subpage in the Special:UploadStash space.
598 private function getSpecialUrl( $subPage ) {
599 return SpecialPage
::getTitleFor( 'UploadStash', $subPage )->getLocalURL();
603 * Get a URL to access the thumbnail
604 * This is required because the model of how files work requires that
605 * the thumbnail urls be predictable. However, in our model the URL is not based on the filename
606 * (that's hidden in the db)
608 * @param $thumbName String: basename of thumbnail file -- however, we don't want to use the file exactly
609 * @return String: URL to access thumbnail, or URL with partial path
611 public function getThumbUrl( $thumbName = false ) {
612 wfDebug( __METHOD__
. " getting for $thumbName \n" );
613 return $this->getSpecialUrl( 'thumb/' . $this->getUrlName() . '/' . $thumbName );
617 * The basename for the URL, which we want to not be related to the filename.
618 * Will also be used as the lookup key for a thumbnail file.
620 * @return String: base url name, like '120px-123456.jpg'
622 public function getUrlName() {
623 if ( ! $this->urlName
) {
624 $this->urlName
= $this->fileKey
;
626 return $this->urlName
;
630 * Return the URL of the file, if for some reason we wanted to download it
631 * We tend not to do this for the original file, but we do want thumb icons
633 * @return String: url
635 public function getUrl() {
636 if ( !isset( $this->url
) ) {
637 $this->url
= $this->getSpecialUrl( 'file/' . $this->getUrlName() );
643 * Parent classes use this method, for no obvious reason, to return the path (relative to wiki root, I assume).
644 * But with this class, the URL is unrelated to the path.
646 * @return String: url
648 public function getFullUrl() {
649 return $this->getUrl();
653 * Getter for file key (the unique id by which this file's location & metadata is stored in the db)
655 * @return String: file key
657 public function getFileKey() {
658 return $this->fileKey
;
662 * Remove the associated temporary file
663 * @return Status: success
665 public function remove() {
666 if( !$this->repo
->fileExists( $this->path
, FileRepo
::FILES_ONLY
) ) {
667 // Maybe the file's already been removed? This could totally happen in UploadBase.
671 return $this->repo
->freeTemp( $this->path
);
676 class UploadStashNotAvailableException
extends MWException
{};
677 class UploadStashFileNotFoundException
extends MWException
{};
678 class UploadStashBadPathException
extends MWException
{};
679 class UploadStashFileException
extends MWException
{};
680 class UploadStashZeroLengthFileException
extends MWException
{};
681 class UploadStashNotLoggedInException
extends MWException
{};
682 class UploadStashWrongOwnerException
extends MWException
{};
683 class UploadStashMaxLagExceededException
extends MWException
{};