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 );
134 if ( ! $this->files
[$key]->exists() ) {
135 wfDebug( __METHOD__
. " tried to get file at $key, but it doesn't exist\n" );
136 throw new UploadStashBadPathException( "path doesn't exist" );
140 if ( $this->fileMetadata
[$key]['us_user'] != $this->userId
) {
141 throw new UploadStashWrongOwnerException( "This file ($key) doesn't belong to the current user." );
145 return $this->files
[$key];
149 * Getter for file metadata.
151 * @param key String: key under which file information is stored
154 public function getMetadata ( $key ) {
155 $this->getFile( $key );
156 return $this->fileMetadata
[$key];
160 * Getter for fileProps
162 * @param key String: key under which file information is stored
165 public function getFileProps ( $key ) {
166 $this->getFile( $key );
167 return $this->fileProps
[$key];
171 * Stash a file in a temp directory and record that we did this in the database, along with other metadata.
173 * @param $path String: path to file you want stashed
174 * @param $sourceType String: the type of upload that generated this file (currently, I believe, 'file' or null)
175 * @param $key String: optional, unique key for this file. Used for directory hashing when storing, otherwise not important
176 * @throws UploadStashBadPathException
177 * @throws UploadStashFileException
178 * @throws UploadStashNotLoggedInException
179 * @return UploadStashFile: file, or null on failure
181 public function stashFile( $path, $sourceType = null, $key = null ) {
182 if ( ! file_exists( $path ) ) {
183 wfDebug( __METHOD__
. " tried to stash file at '$path', but it doesn't exist\n" );
184 throw new UploadStashBadPathException( "path doesn't exist" );
186 $fileProps = File
::getPropsFromPath( $path );
187 wfDebug( __METHOD__
. " stashing file at '$path'\n" );
189 // we will be initializing from some tmpnam files that don't have extensions.
190 // most of MediaWiki assumes all uploaded files have good extensions. So, we fix this.
191 $extension = self
::getExtensionForPath( $path );
192 if ( ! preg_match( "/\\.\\Q$extension\\E$/", $path ) ) {
193 $pathWithGoodExtension = "$path.$extension";
194 if ( ! rename( $path, $pathWithGoodExtension ) ) {
195 throw new UploadStashFileException( "couldn't rename $path to have a better extension at $pathWithGoodExtension" );
197 $path = $pathWithGoodExtension;
200 // If no key was supplied, make one. a mysql insertid would be totally reasonable here, except
201 // that some users of this function might expect to supply the key instead of using the generated one.
202 if ( is_null( $key ) ) {
203 // some things that when combined will make a suitably unique key.
204 // see: http://www.jwz.org/doc/mid.html
205 list ($usec, $sec) = explode( ' ', microtime() );
206 $usec = substr($usec, 2);
207 $key = wfBaseConvert( $sec . $usec, 10, 36 ) . '.' .
208 wfBaseConvert( mt_rand(), 10, 36 ) . '.'.
209 $this->userId
. '.' .
213 $this->fileProps
[$key] = $fileProps;
215 if ( ! preg_match( self
::KEY_FORMAT_REGEX
, $key ) ) {
216 throw new UploadStashBadPathException( "key '$key' is not in a proper format" );
219 wfDebug( __METHOD__
. " key for '$path': $key\n" );
221 // if not already in a temporary area, put it there
222 $storeResult = $this->repo
->storeTemp( basename( $path ), $path );
224 if ( ! $storeResult->isOK() ) {
225 // It is a convention in MediaWiki to only return one error per API exception, even if multiple errors
226 // are available. We use reset() to pick the "first" thing that was wrong, preferring errors to warnings.
227 // 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
228 // redesigning API errors significantly.
229 // $storeResult->value just contains the virtual URL (if anything) which is probably useless to the caller
230 $error = $storeResult->getErrorsArray();
231 $error = reset( $error );
232 if ( ! count( $error ) ) {
233 $error = $storeResult->getWarningsArray();
234 $error = reset( $error );
235 if ( ! count( $error ) ) {
236 $error = array( 'unknown', 'no error recorded' );
239 throw new UploadStashFileException( "error storing file in '$path': " . implode( '; ', $error ) );
241 $stashPath = $storeResult->value
;
243 // fetch the current user ID
244 if ( !$this->isLoggedIn
) {
246 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
249 // insert the file metadata into the db.
250 wfDebug( __METHOD__
. " inserting $stashPath under $key\n" );
251 $dbw = $this->repo
->getMasterDb();
253 // select happens on the master so this can all be in a transaction, which
254 // avoids a race condition that's likely with multiple people uploading from the same
257 // first, check to see if it's already there.
258 $row = $dbw->selectRow(
260 'us_user, us_timestamp',
261 array( 'us_key' => $key ),
265 // The current user can't have this key if:
266 // - the key is owned by someone else and
267 // - the age of the key is less than REPO_AGE
268 if ( is_object( $row ) ) {
269 if ( $row->us_user
!= $this->userId
&&
270 $row->wfTimestamp( TS_UNIX
, $row->us_timestamp
) > time() - UploadStash
::REPO_AGE
* 3600
273 throw new UploadStashWrongOwnerException( "Attempting to upload a duplicate of a file that someone else has stashed" );
277 $this->fileMetadata
[$key] = array(
278 'us_user' => $this->userId
,
280 'us_orig_path' => $path,
281 'us_path' => $stashPath,
282 'us_size' => $fileProps['size'],
283 'us_sha1' => $fileProps['sha1'],
284 'us_mime' => $fileProps['mime'],
285 'us_media_type' => $fileProps['media_type'],
286 'us_image_width' => $fileProps['width'],
287 'us_image_height' => $fileProps['height'],
288 'us_image_bits' => $fileProps['bits'],
289 'us_source_type' => $sourceType,
290 'us_timestamp' => $dbw->timestamp(),
291 'us_status' => 'finished'
294 // if a row exists but previous checks on it passed, let the current user take over this key.
298 $this->fileMetadata
[$key],
303 // store the insertid in the class variable so immediate retrieval (possibly laggy) isn't necesary.
304 $this->fileMetadata
[$key]['us_id'] = $dbw->insertId();
306 # create the UploadStashFile object for this file.
307 $this->initFile( $key );
309 return $this->getFile( $key );
313 * Remove all files from the stash.
314 * Does not clean up files in the repo, just the record of them.
316 * @throws UploadStashNotLoggedInException
317 * @return boolean: success
319 public function clear() {
320 if ( !$this->isLoggedIn
) {
321 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
324 wfDebug( __METHOD__
. " clearing all rows for user $userId\n" );
325 $dbw = $this->repo
->getMasterDb();
328 array( 'us_user' => $this->userId
),
333 $this->files
= array();
334 $this->fileMetadata
= array();
340 * Remove a particular file from the stash. Also removes it from the repo.
342 * @throws UploadStashNotLoggedInException
343 * @throws UploadStashWrongOwnerException
344 * @return boolean: success
346 public function removeFile( $key ) {
347 if ( !$this->isLoggedIn
) {
348 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
351 $dbw = $this->repo
->getMasterDb();
353 // this is a cheap query. it runs on the master so that this function still works when there's lag.
354 // it won't be called all that often.
355 $row = $dbw->selectRow(
358 array( 'us_key' => $key ),
363 throw new UploadStashNoSuchKeyException( "No such key ($key), cannot remove" );
366 if ( $row->us_user
!= $this->userId
) {
367 throw new UploadStashWrongOwnerException( "Can't delete: the file ($key) doesn't belong to this user." );
370 return $this->removeFileNoAuth( $key );
375 * Remove a file (see removeFile), but doesn't check ownership first.
377 * @return boolean: success
379 public function removeFileNoAuth( $key ) {
380 wfDebug( __METHOD__
. " clearing row $key\n" );
382 $dbw = $this->repo
->getMasterDb();
384 // this gets its own transaction since it's called serially by the cleanupUploadStash maintenance script
388 array( 'us_key' => $key ),
393 // TODO: look into UnregisteredLocalFile and find out why the rv here is sometimes wrong (false when file was removed)
395 $this->files
[$key]->remove();
397 unset( $this->files
[$key] );
398 unset( $this->fileMetadata
[$key] );
404 * List all files in the stash.
406 * @throws UploadStashNotLoggedInException
409 public function listFiles() {
410 if ( !$this->isLoggedIn
) {
411 throw new UploadStashNotLoggedInException( __METHOD__
. ' No user is logged in, files must belong to users' );
414 $dbr = $this->repo
->getSlaveDb();
418 array( 'us_key' => $key ),
422 if ( !is_object( $res ) ||
$res->numRows() == 0 ) {
427 // finish the read before starting writes.
429 foreach ( $res as $row ) {
430 array_push( $keys, $row->us_key
);
437 * Find or guess extension -- ensuring that our extension matches our mime type.
438 * Since these files are constructed from php tempnames they may not start off
440 * XXX this is somewhat redundant with the checks that ApiUpload.php does with incoming
441 * uploads versus the desired filename. Maybe we can get that passed to us...
443 public static function getExtensionForPath( $path ) {
444 // Does this have an extension?
445 $n = strrpos( $path, '.' );
447 if ( $n !== false ) {
448 $extension = $n ?
substr( $path, $n +
1 ) : '';
450 // If not, assume that it should be related to the mime type of the original file.
451 $magic = MimeMagic
::singleton();
452 $mimeType = $magic->guessMimeType( $path );
453 $extensions = explode( ' ', MimeMagic
::singleton()->getExtensionsForType( $mimeType ) );
454 if ( count( $extensions ) ) {
455 $extension = $extensions[0];
459 if ( is_null( $extension ) ) {
460 throw new UploadStashFileException( "extension is null" );
463 return File
::normalizeExtension( $extension );
467 * Helper function: do the actual database query to fetch file metadata.
469 * @param $key String: key
472 protected function fetchFileMetadata( $key ) {
473 // populate $fileMetadata[$key]
474 $dbr = $this->repo
->getSlaveDb();
475 $row = $dbr->selectRow(
478 array( 'us_key' => $key ),
482 if ( !is_object( $row ) ) {
483 // key wasn't present in the database. this will happen sometimes.
487 $this->fileMetadata
[$key] = array(
488 'us_user' => $row->us_user
,
489 'us_key' => $row->us_key
,
490 'us_orig_path' => $row->us_orig_path
,
491 'us_path' => $row->us_path
,
492 'us_size' => $row->us_size
,
493 'us_sha1' => $row->us_sha1
,
494 'us_mime' => $row->us_mime
,
495 'us_media_type' => $row->us_media_type
,
496 'us_image_width' => $row->us_image_width
,
497 'us_image_height' => $row->us_image_height
,
498 'us_image_bits' => $row->us_image_bits
,
499 'us_source_type' => $row->us_source_type
,
500 'us_timestamp' => $row->us_timestamp
,
501 'us_status' => $row->us_status
508 * Helper function: Initialize the UploadStashFile for a given file.
510 * @param $path String: path to file
511 * @param $key String: key under which to store the object
512 * @throws UploadStashZeroLengthFileException
515 protected function initFile( $key ) {
516 $file = new UploadStashFile( $this->repo
, $this->fileMetadata
[$key]['us_path'], $key );
517 if ( $file->getSize() === 0 ) {
518 throw new UploadStashZeroLengthFileException( "File is zero length" );
520 $this->files
[$key] = $file;
525 class UploadStashFile
extends UnregisteredLocalFile
{
531 * A LocalFile wrapper around a file that has been temporarily stashed, so we can do things like create thumbnails for it
532 * Arguably UnregisteredLocalFile should be handling its own file repo but that class is a bit retarded currently
534 * @param $repo FSRepo: repository where we should find the path
535 * @param $path String: path to file
536 * @param $key String: key to store the path and any stashed data under
537 * @throws UploadStashBadPathException
538 * @throws UploadStashFileNotFoundException
540 public function __construct( $repo, $path, $key ) {
541 $this->fileKey
= $key;
543 // resolve mwrepo:// urls
544 if ( $repo->isVirtualUrl( $path ) ) {
545 $path = $repo->resolveVirtualUrl( $path );
548 // check if path appears to be sane, no parent traversals, and is in this repo's temp zone.
549 $repoTempPath = $repo->getZonePath( 'temp' );
550 if ( ( ! $repo->validateFilename( $path ) ) ||
551 ( strpos( $path, $repoTempPath ) !== 0 ) ) {
552 wfDebug( "UploadStash: tried to construct an UploadStashFile from a file that should already exist at '$path', but path is not valid\n" );
553 throw new UploadStashBadPathException( 'path is not valid' );
556 // check if path exists! and is a plain file.
557 if ( ! $repo->fileExists( $path, FileRepo
::FILES_ONLY
) ) {
558 wfDebug( "UploadStash: tried to construct an UploadStashFile from a file that should already exist at '$path', but path is not found\n" );
559 throw new UploadStashFileNotFoundException( 'cannot find path, or not a plain file' );
563 parent
::__construct( false, $repo, $path, false );
565 $this->name
= basename( $this->path
);
569 * A method needed by the file transforming and scaling routines in File.php
570 * We do not necessarily care about doing the description at this point
571 * However, we also can't return the empty string, as the rest of MediaWiki demands this (and calls to imagemagick
572 * convert require it to be there)
574 * @return String: dummy value
576 public function getDescriptionUrl() {
577 return $this->getUrl();
581 * Get the path for the thumbnail (actually any transformation of this file)
582 * The actual argument is the result of thumbName although we seem to have
583 * buggy code elsewhere that expects a boolean 'suffix'
585 * @param $thumbName String: name of thumbnail (e.g. "120px-123456.jpg" ), or false to just get the path
586 * @return String: path thumbnail should take on filesystem, or containing directory if thumbname is false
588 public function getThumbPath( $thumbName = false ) {
589 $path = dirname( $this->path
);
590 if ( $thumbName !== false ) {
591 $path .= "/$thumbName";
597 * Return the file/url base name of a thumbnail with the specified parameters.
598 * We override this because we want to use the pretty url name instead of the
601 * @param $params Array: handler-specific parameters
602 * @return String: base name for URL, like '120px-12345.jpg', or null if there is no handler
604 function thumbName( $params ) {
605 return $this->generateThumbName( $this->getUrlName(), $params );
609 * Helper function -- given a 'subpage', return the local URL e.g. /wiki/Special:UploadStash/subpage
610 * @param {String} $subPage
611 * @return {String} local URL for this subpage in the Special:UploadStash space.
613 private function getSpecialUrl( $subPage ) {
614 return SpecialPage
::getTitleFor( 'UploadStash', $subPage )->getLocalURL();
618 * Get a URL to access the thumbnail
619 * This is required because the model of how files work requires that
620 * the thumbnail urls be predictable. However, in our model the URL is not based on the filename
621 * (that's hidden in the db)
623 * @param $thumbName String: basename of thumbnail file -- however, we don't want to use the file exactly
624 * @return String: URL to access thumbnail, or URL with partial path
626 public function getThumbUrl( $thumbName = false ) {
627 wfDebug( __METHOD__
. " getting for $thumbName \n" );
628 return $this->getSpecialUrl( 'thumb/' . $this->getUrlName() . '/' . $thumbName );
632 * The basename for the URL, which we want to not be related to the filename.
633 * Will also be used as the lookup key for a thumbnail file.
635 * @return String: base url name, like '120px-123456.jpg'
637 public function getUrlName() {
638 if ( ! $this->urlName
) {
639 $this->urlName
= $this->fileKey
;
641 return $this->urlName
;
645 * Return the URL of the file, if for some reason we wanted to download it
646 * We tend not to do this for the original file, but we do want thumb icons
648 * @return String: url
650 public function getUrl() {
651 if ( !isset( $this->url
) ) {
652 $this->url
= $this->getSpecialUrl( 'file/' . $this->getUrlName() );
658 * Parent classes use this method, for no obvious reason, to return the path (relative to wiki root, I assume).
659 * But with this class, the URL is unrelated to the path.
661 * @return String: url
663 public function getFullUrl() {
664 return $this->getUrl();
668 * Getter for file key (the unique id by which this file's location & metadata is stored in the db)
670 * @return String: file key
672 public function getFileKey() {
673 return $this->fileKey
;
677 * Remove the associated temporary file
678 * @return Status: success
680 public function remove() {
681 if ( !$this->repo
->fileExists( $this->path
, FileRepo
::FILES_ONLY
) ) {
682 // Maybe the file's already been removed? This could totally happen in UploadBase.
686 return $this->repo
->freeTemp( $this->path
);
689 public function exists() {
690 return $this->repo
->fileExists( $this->path
, FileRepo
::FILES_ONLY
);
695 class UploadStashNotAvailableException
extends MWException
{};
696 class UploadStashFileNotFoundException
extends MWException
{};
697 class UploadStashBadPathException
extends MWException
{};
698 class UploadStashFileException
extends MWException
{};
699 class UploadStashZeroLengthFileException
extends MWException
{};
700 class UploadStashNotLoggedInException
extends MWException
{};
701 class UploadStashWrongOwnerException
extends MWException
{};
702 class UploadStashMaxLagExceededException
extends MWException
{};
703 class UploadStashNoSuchKeyException
extends MWException
{};