Merge "Fixed E_STRICT notices in PostgresUpdater::renameIndex()."
[lhc/web/wiklou.git] / includes / upload / UploadStash.php
1 <?php
2 /**
3 * Temporary storage for uploaded files.
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
19 *
20 * @file
21 * @ingroup Upload
22 */
23
24 /**
25 * UploadStash is intended to accomplish a few things:
26 * - enable applications to temporarily stash files without publishing them to the wiki.
27 * - Several parts of MediaWiki do this in similar ways: UploadBase, UploadWizard, and FirefoggChunkedExtension
28 * And there are several that reimplement stashing from scratch, in idiosyncratic ways. The idea is to unify them all here.
29 * Mostly all of them are the same except for storing some custom fields, which we subsume into the data array.
30 * - enable applications to find said files later, as long as the db table or temp files haven't been purged.
31 * - enable the uploading user (and *ONLY* the uploading user) to access said files, and thumbnails of said files, via a URL.
32 * We accomplish this using a database table, with ownership checking as you might expect. See SpecialUploadStash, which
33 * implements a web interface to some files stored this way.
34 *
35 * UploadStash right now is *mostly* intended to show you one user's slice of the entire stash. The user parameter is only optional
36 * because there are few cases where we clean out the stash from an automated script. In the future we might refactor this.
37 *
38 * UploadStash represents the entire stash of temporary files.
39 * UploadStashFile is a filestore for the actual physical disk files.
40 * UploadFromStash extends UploadBase, and represents a single stashed file as it is moved from the stash to the regular file repository
41 *
42 * @ingroup Upload
43 */
44 class UploadStash {
45
46 // Format of the key for files -- has to be suitable as a filename itself (e.g. ab12cd34ef.jpg)
47 const KEY_FORMAT_REGEX = '/^[\w-\.]+\.\w*$/';
48
49 /**
50 * repository that this uses to store temp files
51 * public because we sometimes need to get a LocalFile within the same repo.
52 *
53 * @var LocalRepo
54 */
55 public $repo;
56
57 // array of initialized repo objects
58 protected $files = array();
59
60 // cache of the file metadata that's stored in the database
61 protected $fileMetadata = array();
62
63 // fileprops cache
64 protected $fileProps = array();
65
66 // current user
67 protected $user, $userId, $isLoggedIn;
68
69 /**
70 * Represents a temporary filestore, with metadata in the database.
71 * Designed to be compatible with the session stashing code in UploadBase
72 * (should replace it eventually).
73 *
74 * @param $repo FileRepo
75 * @param $user User (default null)
76 */
77 public function __construct( FileRepo $repo, $user = null ) {
78 // this might change based on wiki's configuration.
79 $this->repo = $repo;
80
81 // if a user was passed, use it. otherwise, attempt to use the global.
82 // this keeps FileRepo from breaking when it creates an UploadStash object
83 if ( $user ) {
84 $this->user = $user;
85 } else {
86 global $wgUser;
87 $this->user = $wgUser;
88 }
89
90 if ( is_object( $this->user ) ) {
91 $this->userId = $this->user->getId();
92 $this->isLoggedIn = $this->user->isLoggedIn();
93 }
94 }
95
96 /**
97 * Get a file and its metadata from the stash.
98 * The noAuth param is a bit janky but is required for automated scripts which clean out the stash.
99 *
100 * @param $key String: key under which file information is stored
101 * @param $noAuth Boolean (optional) Don't check authentication. Used by maintenance scripts.
102 * @throws UploadStashFileNotFoundException
103 * @throws UploadStashNotLoggedInException
104 * @throws UploadStashWrongOwnerException
105 * @throws UploadStashBadPathException
106 * @return UploadStashFile
107 */
108 public function getFile( $key, $noAuth = false ) {
109 if ( ! preg_match( self::KEY_FORMAT_REGEX, $key ) ) {
110 throw new UploadStashBadPathException( "key '$key' is not in a proper format" );
111 }
112
113 if ( !$noAuth && !$this->isLoggedIn ) {
114 throw new UploadStashNotLoggedInException( __METHOD__ .
115 ' No user is logged in, files must belong to users' );
116 }
117
118 if ( !isset( $this->fileMetadata[$key] ) ) {
119 if ( !$this->fetchFileMetadata( $key ) ) {
120 // If nothing was received, it's likely due to replication lag. Check the master to see if the record is there.
121 $this->fetchFileMetadata( $key, DB_MASTER );
122 }
123
124 if ( !isset( $this->fileMetadata[$key] ) ) {
125 throw new UploadStashFileNotFoundException( "key '$key' not found in stash" );
126 }
127
128 // create $this->files[$key]
129 $this->initFile( $key );
130
131 // fetch fileprops
132 if ( strlen( $this->fileMetadata[$key]['us_props'] ) ) {
133 $this->fileProps[$key] = unserialize( $this->fileMetadata[$key]['us_props'] );
134 } else { // b/c for rows with no us_props
135 wfDebug( __METHOD__ . " fetched props for $key from file\n" );
136 $path = $this->fileMetadata[$key]['us_path'];
137 $this->fileProps[$key] = $this->repo->getFileProps( $path );
138 }
139 }
140
141 if ( ! $this->files[$key]->exists() ) {
142 wfDebug( __METHOD__ . " tried to get file at $key, but it doesn't exist\n" );
143 throw new UploadStashBadPathException( "path doesn't exist" );
144 }
145
146 if ( !$noAuth ) {
147 if ( $this->fileMetadata[$key]['us_user'] != $this->userId ) {
148 throw new UploadStashWrongOwnerException( "This file ($key) doesn't belong to the current user." );
149 }
150 }
151
152 return $this->files[$key];
153 }
154
155 /**
156 * Getter for file metadata.
157 *
158 * @param $key String: key under which file information is stored
159 * @return Array
160 */
161 public function getMetadata ( $key ) {
162 $this->getFile( $key );
163 return $this->fileMetadata[$key];
164 }
165
166 /**
167 * Getter for fileProps
168 *
169 * @param $key String: key under which file information is stored
170 * @return Array
171 */
172 public function getFileProps ( $key ) {
173 $this->getFile( $key );
174 return $this->fileProps[$key];
175 }
176
177 /**
178 * Stash a file in a temp directory and record that we did this in the database, along with other metadata.
179 *
180 * @param $path String: path to file you want stashed
181 * @param $sourceType String: the type of upload that generated this file (currently, I believe, 'file' or null)
182 * @throws UploadStashBadPathException
183 * @throws UploadStashFileException
184 * @throws UploadStashNotLoggedInException
185 * @return UploadStashFile: file, or null on failure
186 */
187 public function stashFile( $path, $sourceType = null ) {
188 if ( !is_file( $path ) ) {
189 wfDebug( __METHOD__ . " tried to stash file at '$path', but it doesn't exist\n" );
190 throw new UploadStashBadPathException( "path doesn't exist" );
191 }
192 $fileProps = FSFile::getPropsFromPath( $path );
193 wfDebug( __METHOD__ . " stashing file at '$path'\n" );
194
195 // we will be initializing from some tmpnam files that don't have extensions.
196 // most of MediaWiki assumes all uploaded files have good extensions. So, we fix this.
197 $extension = self::getExtensionForPath( $path );
198 if ( !preg_match( "/\\.\\Q$extension\\E$/", $path ) ) {
199 $pathWithGoodExtension = "$path.$extension";
200 } else {
201 $pathWithGoodExtension = $path;
202 }
203
204 // If no key was supplied, make one. a mysql insertid would be totally reasonable here, except
205 // that for historical reasons, the key is this random thing instead. At least it's not guessable.
206 //
207 // some things that when combined will make a suitably unique key.
208 // see: http://www.jwz.org/doc/mid.html
209 list( $usec, $sec ) = explode( ' ', microtime() );
210 $usec = substr( $usec, 2 );
211 $key = wfBaseConvert( $sec . $usec, 10, 36 ) . '.' .
212 wfBaseConvert( mt_rand(), 10, 36 ) . '.'.
213 $this->userId . '.' .
214 $extension;
215
216 $this->fileProps[$key] = $fileProps;
217
218 if ( ! preg_match( self::KEY_FORMAT_REGEX, $key ) ) {
219 throw new UploadStashBadPathException( "key '$key' is not in a proper format" );
220 }
221
222 wfDebug( __METHOD__ . " key for '$path': $key\n" );
223
224 // if not already in a temporary area, put it there
225 $storeStatus = $this->repo->storeTemp( basename( $pathWithGoodExtension ), $path );
226
227 if ( ! $storeStatus->isOK() ) {
228 // It is a convention in MediaWiki to only return one error per API exception, even if multiple errors
229 // are available. We use reset() to pick the "first" thing that was wrong, preferring errors to warnings.
230 // This is a bit lame, as we may have more info in the $storeStatus and we're throwing it away, but to fix it means
231 // redesigning API errors significantly.
232 // $storeStatus->value just contains the virtual URL (if anything) which is probably useless to the caller
233 $error = $storeStatus->getErrorsArray();
234 $error = reset( $error );
235 if ( ! count( $error ) ) {
236 $error = $storeStatus->getWarningsArray();
237 $error = reset( $error );
238 if ( ! count( $error ) ) {
239 $error = array( 'unknown', 'no error recorded' );
240 }
241 }
242 // at this point, $error should contain the single "most important" error, plus any parameters.
243 $errorMsg = array_shift( $error );
244 throw new UploadStashFileException( "Error storing file in '$path': " . wfMessage( $errorMsg, $error )->text() );
245 }
246 $stashPath = $storeStatus->value;
247
248 // fetch the current user ID
249 if ( !$this->isLoggedIn ) {
250 throw new UploadStashNotLoggedInException( __METHOD__ . ' No user is logged in, files must belong to users' );
251 }
252
253 // insert the file metadata into the db.
254 wfDebug( __METHOD__ . " inserting $stashPath under $key\n" );
255 $dbw = $this->repo->getMasterDb();
256
257 $this->fileMetadata[$key] = array(
258 'us_id' => $dbw->nextSequenceValue( 'uploadstash_us_id_seq' ),
259 'us_user' => $this->userId,
260 'us_key' => $key,
261 'us_orig_path' => $path,
262 'us_path' => $stashPath, // virtual URL
263 'us_props' => serialize( $fileProps ),
264 'us_size' => $fileProps['size'],
265 'us_sha1' => $fileProps['sha1'],
266 'us_mime' => $fileProps['mime'],
267 'us_media_type' => $fileProps['media_type'],
268 'us_image_width' => $fileProps['width'],
269 'us_image_height' => $fileProps['height'],
270 'us_image_bits' => $fileProps['bits'],
271 'us_source_type' => $sourceType,
272 'us_timestamp' => $dbw->timestamp(),
273 'us_status' => 'finished'
274 );
275
276 $dbw->insert(
277 'uploadstash',
278 $this->fileMetadata[$key],
279 __METHOD__
280 );
281
282 // store the insertid in the class variable so immediate retrieval (possibly laggy) isn't necesary.
283 $this->fileMetadata[$key]['us_id'] = $dbw->insertId();
284
285 # create the UploadStashFile object for this file.
286 $this->initFile( $key );
287
288 return $this->getFile( $key );
289 }
290
291 /**
292 * Remove all files from the stash.
293 * Does not clean up files in the repo, just the record of them.
294 *
295 * @throws UploadStashNotLoggedInException
296 * @return boolean: success
297 */
298 public function clear() {
299 if ( !$this->isLoggedIn ) {
300 throw new UploadStashNotLoggedInException( __METHOD__ . ' No user is logged in, files must belong to users' );
301 }
302
303 wfDebug( __METHOD__ . ' clearing all rows for user ' . $this->userId . "\n" );
304 $dbw = $this->repo->getMasterDb();
305 $dbw->delete(
306 'uploadstash',
307 array( 'us_user' => $this->userId ),
308 __METHOD__
309 );
310
311 # destroy objects.
312 $this->files = array();
313 $this->fileMetadata = array();
314
315 return true;
316 }
317
318 /**
319 * Remove a particular file from the stash. Also removes it from the repo.
320 *
321 * @param $key
322 * @throws UploadStashNoSuchKeyException|UploadStashNotLoggedInException|UploadStashWrongOwnerException
323 * @return boolean: success
324 */
325 public function removeFile( $key ) {
326 if ( !$this->isLoggedIn ) {
327 throw new UploadStashNotLoggedInException( __METHOD__ . ' No user is logged in, files must belong to users' );
328 }
329
330 $dbw = $this->repo->getMasterDb();
331
332 // this is a cheap query. it runs on the master so that this function still works when there's lag.
333 // it won't be called all that often.
334 $row = $dbw->selectRow(
335 'uploadstash',
336 'us_user',
337 array( 'us_key' => $key ),
338 __METHOD__
339 );
340
341 if( !$row ) {
342 throw new UploadStashNoSuchKeyException( "No such key ($key), cannot remove" );
343 }
344
345 if ( $row->us_user != $this->userId ) {
346 throw new UploadStashWrongOwnerException( "Can't delete: the file ($key) doesn't belong to this user." );
347 }
348
349 return $this->removeFileNoAuth( $key );
350 }
351
352
353 /**
354 * Remove a file (see removeFile), but doesn't check ownership first.
355 *
356 * @return boolean: success
357 */
358 public function removeFileNoAuth( $key ) {
359 wfDebug( __METHOD__ . " clearing row $key\n" );
360
361 $dbw = $this->repo->getMasterDb();
362
363 $dbw->delete(
364 'uploadstash',
365 array( 'us_key' => $key ),
366 __METHOD__
367 );
368
369 // TODO: look into UnregisteredLocalFile and find out why the rv here is sometimes wrong (false when file was removed)
370 // for now, ignore.
371 $this->files[$key]->remove();
372
373 unset( $this->files[$key] );
374 unset( $this->fileMetadata[$key] );
375
376 return true;
377 }
378
379 /**
380 * List all files in the stash.
381 *
382 * @throws UploadStashNotLoggedInException
383 * @return Array
384 */
385 public function listFiles() {
386 if ( !$this->isLoggedIn ) {
387 throw new UploadStashNotLoggedInException( __METHOD__ . ' No user is logged in, files must belong to users' );
388 }
389
390 $dbr = $this->repo->getSlaveDb();
391 $res = $dbr->select(
392 'uploadstash',
393 'us_key',
394 array( 'us_user' => $this->userId ),
395 __METHOD__
396 );
397
398 if ( !is_object( $res ) || $res->numRows() == 0 ) {
399 // nothing to do.
400 return false;
401 }
402
403 // finish the read before starting writes.
404 $keys = array();
405 foreach ( $res as $row ) {
406 array_push( $keys, $row->us_key );
407 }
408
409 return $keys;
410 }
411
412 /**
413 * Find or guess extension -- ensuring that our extension matches our mime type.
414 * Since these files are constructed from php tempnames they may not start off
415 * with an extension.
416 * XXX this is somewhat redundant with the checks that ApiUpload.php does with incoming
417 * uploads versus the desired filename. Maybe we can get that passed to us...
418 * @param $path
419 * @throws UploadStashFileException
420 * @return string
421 */
422 public static function getExtensionForPath( $path ) {
423 // Does this have an extension?
424 $n = strrpos( $path, '.' );
425 $extension = null;
426 if ( $n !== false ) {
427 $extension = $n ? substr( $path, $n + 1 ) : '';
428 } else {
429 // If not, assume that it should be related to the mime type of the original file.
430 $magic = MimeMagic::singleton();
431 $mimeType = $magic->guessMimeType( $path );
432 $extensions = explode( ' ', MimeMagic::singleton()->getExtensionsForType( $mimeType ) );
433 if ( count( $extensions ) ) {
434 $extension = $extensions[0];
435 }
436 }
437
438 if ( is_null( $extension ) ) {
439 throw new UploadStashFileException( "extension is null" );
440 }
441
442 return File::normalizeExtension( $extension );
443 }
444
445 /**
446 * Helper function: do the actual database query to fetch file metadata.
447 *
448 * @param $key String: key
449 * @param $readFromDB: constant (default: DB_SLAVE)
450 * @return boolean
451 */
452 protected function fetchFileMetadata( $key, $readFromDB = DB_SLAVE ) {
453 // populate $fileMetadata[$key]
454 $dbr = null;
455 if( $readFromDB === DB_MASTER ) {
456 // sometimes reading from the master is necessary, if there's replication lag.
457 $dbr = $this->repo->getMasterDb();
458 } else {
459 $dbr = $this->repo->getSlaveDb();
460 }
461
462 $row = $dbr->selectRow(
463 'uploadstash',
464 '*',
465 array( 'us_key' => $key ),
466 __METHOD__
467 );
468
469 if ( !is_object( $row ) ) {
470 // key wasn't present in the database. this will happen sometimes.
471 return false;
472 }
473
474 $this->fileMetadata[$key] = (array)$row;
475
476 return true;
477 }
478
479 /**
480 * Helper function: Initialize the UploadStashFile for a given file.
481 *
482 * @param $key String: key under which to store the object
483 * @throws UploadStashZeroLengthFileException
484 * @return bool
485 */
486 protected function initFile( $key ) {
487 $file = new UploadStashFile( $this->repo, $this->fileMetadata[$key]['us_path'], $key );
488 if ( $file->getSize() === 0 ) {
489 throw new UploadStashZeroLengthFileException( "File is zero length" );
490 }
491 $this->files[$key] = $file;
492 return true;
493 }
494 }
495
496 class UploadStashFile extends UnregisteredLocalFile {
497 private $fileKey;
498 private $urlName;
499 protected $url;
500
501 /**
502 * A LocalFile wrapper around a file that has been temporarily stashed, so we can do things like create thumbnails for it
503 * Arguably UnregisteredLocalFile should be handling its own file repo but that class is a bit retarded currently
504 *
505 * @param $repo FileRepo: repository where we should find the path
506 * @param $path String: path to file
507 * @param $key String: key to store the path and any stashed data under
508 * @throws UploadStashBadPathException
509 * @throws UploadStashFileNotFoundException
510 */
511 public function __construct( $repo, $path, $key ) {
512 $this->fileKey = $key;
513
514 // resolve mwrepo:// urls
515 if ( $repo->isVirtualUrl( $path ) ) {
516 $path = $repo->resolveVirtualUrl( $path );
517 } else {
518
519 // check if path appears to be sane, no parent traversals, and is in this repo's temp zone.
520 $repoTempPath = $repo->getZonePath( 'temp' );
521 if ( ( ! $repo->validateFilename( $path ) ) ||
522 ( strpos( $path, $repoTempPath ) !== 0 ) ) {
523 wfDebug( "UploadStash: tried to construct an UploadStashFile from a file that should already exist at '$path', but path is not valid\n" );
524 throw new UploadStashBadPathException( 'path is not valid' );
525 }
526
527 // check if path exists! and is a plain file.
528 if ( ! $repo->fileExists( $path ) ) {
529 wfDebug( "UploadStash: tried to construct an UploadStashFile from a file that should already exist at '$path', but path is not found\n" );
530 throw new UploadStashFileNotFoundException( 'cannot find path, or not a plain file' );
531 }
532 }
533
534 parent::__construct( false, $repo, $path, false );
535
536 $this->name = basename( $this->path );
537 }
538
539 /**
540 * A method needed by the file transforming and scaling routines in File.php
541 * We do not necessarily care about doing the description at this point
542 * However, we also can't return the empty string, as the rest of MediaWiki demands this (and calls to imagemagick
543 * convert require it to be there)
544 *
545 * @return String: dummy value
546 */
547 public function getDescriptionUrl() {
548 return $this->getUrl();
549 }
550
551 /**
552 * Get the path for the thumbnail (actually any transformation of this file)
553 * The actual argument is the result of thumbName although we seem to have
554 * buggy code elsewhere that expects a boolean 'suffix'
555 *
556 * @param $thumbName String: name of thumbnail (e.g. "120px-123456.jpg" ), or false to just get the path
557 * @return String: path thumbnail should take on filesystem, or containing directory if thumbname is false
558 */
559 public function getThumbPath( $thumbName = false ) {
560 $path = dirname( $this->path );
561 if ( $thumbName !== false ) {
562 $path .= "/$thumbName";
563 }
564 return $path;
565 }
566
567 /**
568 * Return the file/url base name of a thumbnail with the specified parameters.
569 * We override this because we want to use the pretty url name instead of the
570 * ugly file name.
571 *
572 * @param $params Array: handler-specific parameters
573 * @param $flags integer Bitfield that supports THUMB_* constants
574 * @return String: base name for URL, like '120px-12345.jpg', or null if there is no handler
575 */
576 function thumbName( $params, $flags = 0 ) {
577 return $this->generateThumbName( $this->getUrlName(), $params );
578 }
579
580 /**
581 * Helper function -- given a 'subpage', return the local URL e.g. /wiki/Special:UploadStash/subpage
582 * @param $subPage String
583 * @return String: local URL for this subpage in the Special:UploadStash space.
584 */
585 private function getSpecialUrl( $subPage ) {
586 return SpecialPage::getTitleFor( 'UploadStash', $subPage )->getLocalURL();
587 }
588
589 /**
590 * Get a URL to access the thumbnail
591 * This is required because the model of how files work requires that
592 * the thumbnail urls be predictable. However, in our model the URL is not based on the filename
593 * (that's hidden in the db)
594 *
595 * @param $thumbName String: basename of thumbnail file -- however, we don't want to use the file exactly
596 * @return String: URL to access thumbnail, or URL with partial path
597 */
598 public function getThumbUrl( $thumbName = false ) {
599 wfDebug( __METHOD__ . " getting for $thumbName \n" );
600 return $this->getSpecialUrl( 'thumb/' . $this->getUrlName() . '/' . $thumbName );
601 }
602
603 /**
604 * The basename for the URL, which we want to not be related to the filename.
605 * Will also be used as the lookup key for a thumbnail file.
606 *
607 * @return String: base url name, like '120px-123456.jpg'
608 */
609 public function getUrlName() {
610 if ( ! $this->urlName ) {
611 $this->urlName = $this->fileKey;
612 }
613 return $this->urlName;
614 }
615
616 /**
617 * Return the URL of the file, if for some reason we wanted to download it
618 * We tend not to do this for the original file, but we do want thumb icons
619 *
620 * @return String: url
621 */
622 public function getUrl() {
623 if ( !isset( $this->url ) ) {
624 $this->url = $this->getSpecialUrl( 'file/' . $this->getUrlName() );
625 }
626 return $this->url;
627 }
628
629 /**
630 * Parent classes use this method, for no obvious reason, to return the path (relative to wiki root, I assume).
631 * But with this class, the URL is unrelated to the path.
632 *
633 * @return String: url
634 */
635 public function getFullUrl() {
636 return $this->getUrl();
637 }
638
639 /**
640 * Getter for file key (the unique id by which this file's location & metadata is stored in the db)
641 *
642 * @return String: file key
643 */
644 public function getFileKey() {
645 return $this->fileKey;
646 }
647
648 /**
649 * Remove the associated temporary file
650 * @return Status: success
651 */
652 public function remove() {
653 if ( !$this->repo->fileExists( $this->path ) ) {
654 // Maybe the file's already been removed? This could totally happen in UploadBase.
655 return true;
656 }
657
658 return $this->repo->freeTemp( $this->path );
659 }
660
661 public function exists() {
662 return $this->repo->fileExists( $this->path );
663 }
664
665 }
666
667 class UploadStashNotAvailableException extends MWException {};
668 class UploadStashFileNotFoundException extends MWException {};
669 class UploadStashBadPathException extends MWException {};
670 class UploadStashFileException extends MWException {};
671 class UploadStashZeroLengthFileException extends MWException {};
672 class UploadStashNotLoggedInException extends MWException {};
673 class UploadStashWrongOwnerException extends MWException {};
674 class UploadStashNoSuchKeyException extends MWException {};