Drop archive.ar_text and ar_flags
[lhc/web/wiklou.git] / includes / Storage / RevisionStore.php
1 <?php
2 /**
3 * Service for looking up page revisions.
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 * Attribution notice: when this file was created, much of its content was taken
21 * from the Revision.php file as present in release 1.30. Refer to the history
22 * of that file for original authorship.
23 *
24 * @file
25 */
26
27 namespace MediaWiki\Storage;
28
29 use ActorMigration;
30 use CommentStore;
31 use CommentStoreComment;
32 use Content;
33 use ContentHandler;
34 use DBAccessObjectUtils;
35 use Hooks;
36 use IDBAccessObject;
37 use InvalidArgumentException;
38 use IP;
39 use LogicException;
40 use MediaWiki\Linker\LinkTarget;
41 use MediaWiki\User\UserIdentity;
42 use MediaWiki\User\UserIdentityValue;
43 use Message;
44 use MWException;
45 use MWUnknownContentModelException;
46 use Psr\Log\LoggerAwareInterface;
47 use Psr\Log\LoggerInterface;
48 use Psr\Log\NullLogger;
49 use RecentChange;
50 use stdClass;
51 use Title;
52 use User;
53 use WANObjectCache;
54 use Wikimedia\Assert\Assert;
55 use Wikimedia\Rdbms\Database;
56 use Wikimedia\Rdbms\DBConnRef;
57 use Wikimedia\Rdbms\IDatabase;
58 use Wikimedia\Rdbms\LoadBalancer;
59
60 /**
61 * Service for looking up page revisions.
62 *
63 * @since 1.31
64 *
65 * @note This was written to act as a drop-in replacement for the corresponding
66 * static methods in Revision.
67 */
68 class RevisionStore
69 implements IDBAccessObject, RevisionFactory, RevisionLookup, LoggerAwareInterface {
70
71 /**
72 * @var SqlBlobStore
73 */
74 private $blobStore;
75
76 /**
77 * @var bool|string
78 */
79 private $wikiId;
80
81 /**
82 * @var boolean
83 */
84 private $contentHandlerUseDB = true;
85
86 /**
87 * @var LoadBalancer
88 */
89 private $loadBalancer;
90
91 /**
92 * @var WANObjectCache
93 */
94 private $cache;
95
96 /**
97 * @var CommentStore
98 */
99 private $commentStore;
100
101 /**
102 * @var ActorMigration
103 */
104 private $actorMigration;
105
106 /**
107 * @var LoggerInterface
108 */
109 private $logger;
110
111 /**
112 * @todo $blobStore should be allowed to be any BlobStore!
113 *
114 * @param LoadBalancer $loadBalancer
115 * @param SqlBlobStore $blobStore
116 * @param WANObjectCache $cache
117 * @param CommentStore $commentStore
118 * @param ActorMigration $actorMigration
119 * @param bool|string $wikiId
120 */
121 public function __construct(
122 LoadBalancer $loadBalancer,
123 SqlBlobStore $blobStore,
124 WANObjectCache $cache,
125 CommentStore $commentStore,
126 ActorMigration $actorMigration,
127 $wikiId = false
128 ) {
129 Assert::parameterType( 'string|boolean', $wikiId, '$wikiId' );
130
131 $this->loadBalancer = $loadBalancer;
132 $this->blobStore = $blobStore;
133 $this->cache = $cache;
134 $this->commentStore = $commentStore;
135 $this->actorMigration = $actorMigration;
136 $this->wikiId = $wikiId;
137 $this->logger = new NullLogger();
138 }
139
140 public function setLogger( LoggerInterface $logger ) {
141 $this->logger = $logger;
142 }
143
144 /**
145 * @return bool Whether the store is read-only
146 */
147 public function isReadOnly() {
148 return $this->blobStore->isReadOnly();
149 }
150
151 /**
152 * @return bool
153 */
154 public function getContentHandlerUseDB() {
155 return $this->contentHandlerUseDB;
156 }
157
158 /**
159 * @param bool $contentHandlerUseDB
160 */
161 public function setContentHandlerUseDB( $contentHandlerUseDB ) {
162 $this->contentHandlerUseDB = $contentHandlerUseDB;
163 }
164
165 /**
166 * @return LoadBalancer
167 */
168 private function getDBLoadBalancer() {
169 return $this->loadBalancer;
170 }
171
172 /**
173 * @param int $mode DB_MASTER or DB_REPLICA
174 *
175 * @return IDatabase
176 */
177 private function getDBConnection( $mode ) {
178 $lb = $this->getDBLoadBalancer();
179 return $lb->getConnection( $mode, [], $this->wikiId );
180 }
181
182 /**
183 * @param IDatabase $connection
184 */
185 private function releaseDBConnection( IDatabase $connection ) {
186 $lb = $this->getDBLoadBalancer();
187 $lb->reuseConnection( $connection );
188 }
189
190 /**
191 * @param int $mode DB_MASTER or DB_REPLICA
192 *
193 * @return DBConnRef
194 */
195 private function getDBConnectionRef( $mode ) {
196 $lb = $this->getDBLoadBalancer();
197 return $lb->getConnectionRef( $mode, [], $this->wikiId );
198 }
199
200 /**
201 * Determines the page Title based on the available information.
202 *
203 * MCR migration note: this corresponds to Revision::getTitle
204 *
205 * @note this method should be private, external use should be avoided!
206 *
207 * @param int|null $pageId
208 * @param int|null $revId
209 * @param int $queryFlags
210 *
211 * @return Title
212 * @throws RevisionAccessException
213 */
214 public function getTitle( $pageId, $revId, $queryFlags = self::READ_NORMAL ) {
215 if ( !$pageId && !$revId ) {
216 throw new InvalidArgumentException( '$pageId and $revId cannot both be 0 or null' );
217 }
218
219 // This method recalls itself with READ_LATEST if READ_NORMAL doesn't get us a Title
220 // So ignore READ_LATEST_IMMUTABLE flags and handle the fallback logic in this method
221 if ( DBAccessObjectUtils::hasFlags( $queryFlags, self::READ_LATEST_IMMUTABLE ) ) {
222 $queryFlags = self::READ_NORMAL;
223 }
224
225 $canUseTitleNewFromId = ( $pageId !== null && $pageId > 0 && $this->wikiId === false );
226 list( $dbMode, $dbOptions ) = DBAccessObjectUtils::getDBOptions( $queryFlags );
227 $titleFlags = ( $dbMode == DB_MASTER ? Title::GAID_FOR_UPDATE : 0 );
228
229 // Loading by ID is best, but Title::newFromID does not support that for foreign IDs.
230 if ( $canUseTitleNewFromId ) {
231 // TODO: better foreign title handling (introduce TitleFactory)
232 $title = Title::newFromID( $pageId, $titleFlags );
233 if ( $title ) {
234 return $title;
235 }
236 }
237
238 // rev_id is defined as NOT NULL, but this revision may not yet have been inserted.
239 $canUseRevId = ( $revId !== null && $revId > 0 );
240
241 if ( $canUseRevId ) {
242 $dbr = $this->getDBConnectionRef( $dbMode );
243 // @todo: Title::getSelectFields(), or Title::getQueryInfo(), or something like that
244 $row = $dbr->selectRow(
245 [ 'revision', 'page' ],
246 [
247 'page_namespace',
248 'page_title',
249 'page_id',
250 'page_latest',
251 'page_is_redirect',
252 'page_len',
253 ],
254 [ 'rev_id' => $revId ],
255 __METHOD__,
256 $dbOptions,
257 [ 'page' => [ 'JOIN', 'page_id=rev_page' ] ]
258 );
259 if ( $row ) {
260 // TODO: better foreign title handling (introduce TitleFactory)
261 return Title::newFromRow( $row );
262 }
263 }
264
265 // If we still don't have a title, fallback to master if that wasn't already happening.
266 if ( $dbMode !== DB_MASTER ) {
267 $title = $this->getTitle( $pageId, $revId, self::READ_LATEST );
268 if ( $title ) {
269 $this->logger->info(
270 __METHOD__ . ' fell back to READ_LATEST and got a Title.',
271 [ 'trace' => wfBacktrace() ]
272 );
273 return $title;
274 }
275 }
276
277 throw new RevisionAccessException(
278 "Could not determine title for page ID $pageId and revision ID $revId"
279 );
280 }
281
282 /**
283 * @param mixed $value
284 * @param string $name
285 *
286 * @throw IncompleteRevisionException if $value is null
287 * @return mixed $value, if $value is not null
288 */
289 private function failOnNull( $value, $name ) {
290 if ( $value === null ) {
291 throw new IncompleteRevisionException(
292 "$name must not be " . var_export( $value, true ) . "!"
293 );
294 }
295
296 return $value;
297 }
298
299 /**
300 * @param mixed $value
301 * @param string $name
302 *
303 * @throw IncompleteRevisionException if $value is empty
304 * @return mixed $value, if $value is not null
305 */
306 private function failOnEmpty( $value, $name ) {
307 if ( $value === null || $value === 0 || $value === '' ) {
308 throw new IncompleteRevisionException(
309 "$name must not be " . var_export( $value, true ) . "!"
310 );
311 }
312
313 return $value;
314 }
315
316 /**
317 * Insert a new revision into the database, returning the new revision record
318 * on success and dies horribly on failure.
319 *
320 * MCR migration note: this replaces Revision::insertOn
321 *
322 * @param RevisionRecord $rev
323 * @param IDatabase $dbw (master connection)
324 *
325 * @throws InvalidArgumentException
326 * @return RevisionRecord the new revision record.
327 */
328 public function insertRevisionOn( RevisionRecord $rev, IDatabase $dbw ) {
329 // TODO: pass in a DBTransactionContext instead of a database connection.
330 $this->checkDatabaseWikiId( $dbw );
331
332 if ( !$rev->getSlotRoles() ) {
333 throw new InvalidArgumentException( 'At least one slot needs to be defined!' );
334 }
335
336 if ( $rev->getSlotRoles() !== [ 'main' ] ) {
337 throw new InvalidArgumentException( 'Only the main slot is supported for now!' );
338 }
339
340 // TODO: we shouldn't need an actual Title here.
341 $title = Title::newFromLinkTarget( $rev->getPageAsLinkTarget() );
342 $pageId = $this->failOnEmpty( $rev->getPageId(), 'rev_page field' ); // check this early
343
344 $parentId = $rev->getParentId() === null
345 ? $this->getPreviousRevisionId( $dbw, $rev )
346 : $rev->getParentId();
347
348 // Record the text (or external storage URL) to the blob store
349 $slot = $rev->getSlot( 'main', RevisionRecord::RAW );
350
351 $size = $this->failOnNull( $rev->getSize(), 'size field' );
352 $sha1 = $this->failOnEmpty( $rev->getSha1(), 'sha1 field' );
353
354 if ( !$slot->hasAddress() ) {
355 $content = $slot->getContent();
356 $format = $content->getDefaultFormat();
357 $model = $content->getModel();
358
359 $this->checkContentModel( $content, $title );
360
361 $data = $content->serialize( $format );
362
363 // Hints allow the blob store to optimize by "leaking" application level information to it.
364 // TODO: with the new MCR storage schema, we rev_id have this before storing the blobs.
365 // When we have it, add rev_id as a hint. Can be used with rev_parent_id for
366 // differential storage or compression of subsequent revisions.
367 $blobHints = [
368 BlobStore::DESIGNATION_HINT => 'page-content', // BlobStore may be used for other things too.
369 BlobStore::PAGE_HINT => $pageId,
370 BlobStore::ROLE_HINT => $slot->getRole(),
371 BlobStore::PARENT_HINT => $parentId,
372 BlobStore::SHA1_HINT => $slot->getSha1(),
373 BlobStore::MODEL_HINT => $model,
374 BlobStore::FORMAT_HINT => $format,
375 ];
376
377 $blobAddress = $this->blobStore->storeBlob( $data, $blobHints );
378 } else {
379 $blobAddress = $slot->getAddress();
380 $model = $slot->getModel();
381 $format = $slot->getFormat();
382 }
383
384 $textId = $this->blobStore->getTextIdFromAddress( $blobAddress );
385
386 if ( !$textId ) {
387 throw new LogicException(
388 'Blob address not supported in 1.29 database schema: ' . $blobAddress
389 );
390 }
391
392 // getTextIdFromAddress() is free to insert something into the text table, so $textId
393 // may be a new value, not anything already contained in $blobAddress.
394 $blobAddress = 'tt:' . $textId;
395
396 $comment = $this->failOnNull( $rev->getComment( RevisionRecord::RAW ), 'comment' );
397 $user = $this->failOnNull( $rev->getUser( RevisionRecord::RAW ), 'user' );
398 $timestamp = $this->failOnEmpty( $rev->getTimestamp(), 'timestamp field' );
399
400 // Checks.
401 $this->failOnNull( $user->getId(), 'user field' );
402 $this->failOnEmpty( $user->getName(), 'user_text field' );
403
404 # Record the edit in revisions
405 $row = [
406 'rev_page' => $pageId,
407 'rev_parent_id' => $parentId,
408 'rev_text_id' => $textId,
409 'rev_minor_edit' => $rev->isMinor() ? 1 : 0,
410 'rev_timestamp' => $dbw->timestamp( $timestamp ),
411 'rev_deleted' => $rev->getVisibility(),
412 'rev_len' => $size,
413 'rev_sha1' => $sha1,
414 ];
415
416 if ( $rev->getId() !== null ) {
417 // Needed to restore revisions with their original ID
418 $row['rev_id'] = $rev->getId();
419 }
420
421 list( $commentFields, $commentCallback ) =
422 $this->commentStore->insertWithTempTable( $dbw, 'rev_comment', $comment );
423 $row += $commentFields;
424
425 list( $actorFields, $actorCallback ) =
426 $this->actorMigration->getInsertValuesWithTempTable( $dbw, 'rev_user', $user );
427 $row += $actorFields;
428
429 if ( $this->contentHandlerUseDB ) {
430 // MCR migration note: rev_content_model and rev_content_format will go away
431
432 $defaultModel = ContentHandler::getDefaultModelFor( $title );
433 $defaultFormat = ContentHandler::getForModelID( $defaultModel )->getDefaultFormat();
434
435 $row['rev_content_model'] = ( $model === $defaultModel ) ? null : $model;
436 $row['rev_content_format'] = ( $format === $defaultFormat ) ? null : $format;
437 }
438
439 $dbw->insert( 'revision', $row, __METHOD__ );
440
441 if ( !isset( $row['rev_id'] ) ) {
442 // only if auto-increment was used
443 $row['rev_id'] = intval( $dbw->insertId() );
444 }
445 $commentCallback( $row['rev_id'] );
446 $actorCallback( $row['rev_id'], $row );
447
448 // Insert IP revision into ip_changes for use when querying for a range.
449 if ( $user->getId() === 0 && IP::isValid( $user->getName() ) ) {
450 $ipcRow = [
451 'ipc_rev_id' => $row['rev_id'],
452 'ipc_rev_timestamp' => $row['rev_timestamp'],
453 'ipc_hex' => IP::toHex( $user->getName() ),
454 ];
455 $dbw->insert( 'ip_changes', $ipcRow, __METHOD__ );
456 }
457
458 $newSlot = SlotRecord::newSaved( $row['rev_id'], $textId, $blobAddress, $slot );
459 $slots = new RevisionSlots( [ 'main' => $newSlot ] );
460
461 $rev = new RevisionStoreRecord(
462 $title,
463 $user,
464 $comment,
465 (object)$row,
466 $slots,
467 $this->wikiId
468 );
469
470 $newSlot = $rev->getSlot( 'main', RevisionRecord::RAW );
471
472 // sanity checks
473 Assert::postcondition( $rev->getId() > 0, 'revision must have an ID' );
474 Assert::postcondition( $rev->getPageId() > 0, 'revision must have a page ID' );
475 Assert::postcondition(
476 $rev->getComment( RevisionRecord::RAW ) !== null,
477 'revision must have a comment'
478 );
479 Assert::postcondition(
480 $rev->getUser( RevisionRecord::RAW ) !== null,
481 'revision must have a user'
482 );
483
484 Assert::postcondition( $newSlot !== null, 'revision must have a main slot' );
485 Assert::postcondition(
486 $newSlot->getAddress() !== null,
487 'main slot must have an addess'
488 );
489
490 Hooks::run( 'RevisionRecordInserted', [ $rev ] );
491
492 return $rev;
493 }
494
495 /**
496 * MCR migration note: this corresponds to Revision::checkContentModel
497 *
498 * @param Content $content
499 * @param Title $title
500 *
501 * @throws MWException
502 * @throws MWUnknownContentModelException
503 */
504 private function checkContentModel( Content $content, Title $title ) {
505 // Note: may return null for revisions that have not yet been inserted
506
507 $model = $content->getModel();
508 $format = $content->getDefaultFormat();
509 $handler = $content->getContentHandler();
510
511 $name = "$title";
512
513 if ( !$handler->isSupportedFormat( $format ) ) {
514 throw new MWException( "Can't use format $format with content model $model on $name" );
515 }
516
517 if ( !$this->contentHandlerUseDB ) {
518 // if $wgContentHandlerUseDB is not set,
519 // all revisions must use the default content model and format.
520
521 $defaultModel = ContentHandler::getDefaultModelFor( $title );
522 $defaultHandler = ContentHandler::getForModelID( $defaultModel );
523 $defaultFormat = $defaultHandler->getDefaultFormat();
524
525 if ( $model != $defaultModel ) {
526 throw new MWException( "Can't save non-default content model with "
527 . "\$wgContentHandlerUseDB disabled: model is $model, "
528 . "default for $name is $defaultModel"
529 );
530 }
531
532 if ( $format != $defaultFormat ) {
533 throw new MWException( "Can't use non-default content format with "
534 . "\$wgContentHandlerUseDB disabled: format is $format, "
535 . "default for $name is $defaultFormat"
536 );
537 }
538 }
539
540 if ( !$content->isValid() ) {
541 throw new MWException(
542 "New content for $name is not valid! Content model is $model"
543 );
544 }
545 }
546
547 /**
548 * Create a new null-revision for insertion into a page's
549 * history. This will not re-save the text, but simply refer
550 * to the text from the previous version.
551 *
552 * Such revisions can for instance identify page rename
553 * operations and other such meta-modifications.
554 *
555 * MCR migration note: this replaces Revision::newNullRevision
556 *
557 * @todo Introduce newFromParentRevision(). newNullRevision can then be based on that
558 * (or go away).
559 *
560 * @param IDatabase $dbw
561 * @param Title $title Title of the page to read from
562 * @param CommentStoreComment $comment RevisionRecord's summary
563 * @param bool $minor Whether the revision should be considered as minor
564 * @param User $user The user to attribute the revision to
565 * @return RevisionRecord|null RevisionRecord or null on error
566 */
567 public function newNullRevision(
568 IDatabase $dbw,
569 Title $title,
570 CommentStoreComment $comment,
571 $minor,
572 User $user
573 ) {
574 $this->checkDatabaseWikiId( $dbw );
575
576 $fields = [ 'page_latest', 'page_namespace', 'page_title',
577 'rev_id', 'rev_text_id', 'rev_len', 'rev_sha1' ];
578
579 if ( $this->contentHandlerUseDB ) {
580 $fields[] = 'rev_content_model';
581 $fields[] = 'rev_content_format';
582 }
583
584 $current = $dbw->selectRow(
585 [ 'page', 'revision' ],
586 $fields,
587 [
588 'page_id' => $title->getArticleID(),
589 'page_latest=rev_id',
590 ],
591 __METHOD__,
592 [ 'FOR UPDATE' ] // T51581
593 );
594
595 if ( $current ) {
596 $fields = [
597 'page' => $title->getArticleID(),
598 'user_text' => $user->getName(),
599 'user' => $user->getId(),
600 'actor' => $user->getActorId(),
601 'comment' => $comment,
602 'minor_edit' => $minor,
603 'text_id' => $current->rev_text_id,
604 'parent_id' => $current->page_latest,
605 'slot_origin' => $current->page_latest,
606 'len' => $current->rev_len,
607 'sha1' => $current->rev_sha1
608 ];
609
610 if ( $this->contentHandlerUseDB ) {
611 $fields['content_model'] = $current->rev_content_model;
612 $fields['content_format'] = $current->rev_content_format;
613 }
614
615 $fields['title'] = Title::makeTitle( $current->page_namespace, $current->page_title );
616
617 $mainSlot = $this->emulateMainSlot_1_29( $fields, self::READ_LATEST, $title );
618 $revision = new MutableRevisionRecord( $title, $this->wikiId );
619 $this->initializeMutableRevisionFromArray( $revision, $fields );
620 $revision->setSlot( $mainSlot );
621 } else {
622 $revision = null;
623 }
624
625 return $revision;
626 }
627
628 /**
629 * MCR migration note: this replaces Revision::isUnpatrolled
630 *
631 * @todo This is overly specific, so move or kill this method.
632 *
633 * @param RevisionRecord $rev
634 *
635 * @return int Rcid of the unpatrolled row, zero if there isn't one
636 */
637 public function getRcIdIfUnpatrolled( RevisionRecord $rev ) {
638 $rc = $this->getRecentChange( $rev );
639 if ( $rc && $rc->getAttribute( 'rc_patrolled' ) == RecentChange::PRC_UNPATROLLED ) {
640 return $rc->getAttribute( 'rc_id' );
641 } else {
642 return 0;
643 }
644 }
645
646 /**
647 * Get the RC object belonging to the current revision, if there's one
648 *
649 * MCR migration note: this replaces Revision::getRecentChange
650 *
651 * @todo move this somewhere else?
652 *
653 * @param RevisionRecord $rev
654 * @param int $flags (optional) $flags include:
655 * IDBAccessObject::READ_LATEST: Select the data from the master
656 *
657 * @return null|RecentChange
658 */
659 public function getRecentChange( RevisionRecord $rev, $flags = 0 ) {
660 $dbr = $this->getDBConnection( DB_REPLICA );
661
662 list( $dbType, ) = DBAccessObjectUtils::getDBOptions( $flags );
663
664 $userIdentity = $rev->getUser( RevisionRecord::RAW );
665
666 if ( !$userIdentity ) {
667 // If the revision has no user identity, chances are it never went
668 // into the database, and doesn't have an RC entry.
669 return null;
670 }
671
672 // TODO: Select by rc_this_oldid alone - but as of Nov 2017, there is no index on that!
673 $actorWhere = $this->actorMigration->getWhere( $dbr, 'rc_user', $rev->getUser(), false );
674 $rc = RecentChange::newFromConds(
675 [
676 $actorWhere['conds'],
677 'rc_timestamp' => $dbr->timestamp( $rev->getTimestamp() ),
678 'rc_this_oldid' => $rev->getId()
679 ],
680 __METHOD__,
681 $dbType
682 );
683
684 $this->releaseDBConnection( $dbr );
685
686 // XXX: cache this locally? Glue it to the RevisionRecord?
687 return $rc;
688 }
689
690 /**
691 * Maps fields of the archive row to corresponding revision rows.
692 *
693 * @param object $archiveRow
694 *
695 * @return object a revision row object, corresponding to $archiveRow.
696 */
697 private static function mapArchiveFields( $archiveRow ) {
698 $fieldMap = [
699 // keep with ar prefix:
700 'ar_id' => 'ar_id',
701
702 // not the same suffix:
703 'ar_page_id' => 'rev_page',
704 'ar_rev_id' => 'rev_id',
705
706 // same suffix:
707 'ar_text_id' => 'rev_text_id',
708 'ar_timestamp' => 'rev_timestamp',
709 'ar_user_text' => 'rev_user_text',
710 'ar_user' => 'rev_user',
711 'ar_actor' => 'rev_actor',
712 'ar_minor_edit' => 'rev_minor_edit',
713 'ar_deleted' => 'rev_deleted',
714 'ar_len' => 'rev_len',
715 'ar_parent_id' => 'rev_parent_id',
716 'ar_sha1' => 'rev_sha1',
717 'ar_comment' => 'rev_comment',
718 'ar_comment_cid' => 'rev_comment_cid',
719 'ar_comment_id' => 'rev_comment_id',
720 'ar_comment_text' => 'rev_comment_text',
721 'ar_comment_data' => 'rev_comment_data',
722 'ar_comment_old' => 'rev_comment_old',
723 'ar_content_format' => 'rev_content_format',
724 'ar_content_model' => 'rev_content_model',
725 ];
726
727 $revRow = new stdClass();
728 foreach ( $fieldMap as $arKey => $revKey ) {
729 if ( property_exists( $archiveRow, $arKey ) ) {
730 $revRow->$revKey = $archiveRow->$arKey;
731 }
732 }
733
734 return $revRow;
735 }
736
737 /**
738 * Constructs a RevisionRecord for the revisions main slot, based on the MW1.29 schema.
739 *
740 * @param object|array $row Either a database row or an array
741 * @param int $queryFlags for callbacks
742 * @param Title $title
743 *
744 * @return SlotRecord The main slot, extracted from the MW 1.29 style row.
745 * @throws MWException
746 */
747 private function emulateMainSlot_1_29( $row, $queryFlags, Title $title ) {
748 $mainSlotRow = new stdClass();
749 $mainSlotRow->role_name = 'main';
750 $mainSlotRow->model_name = null;
751 $mainSlotRow->slot_revision_id = null;
752 $mainSlotRow->content_address = null;
753 $mainSlotRow->slot_content_id = null;
754
755 $content = null;
756 $blobData = null;
757 $blobFlags = null;
758
759 if ( is_object( $row ) ) {
760 // archive row
761 if ( !isset( $row->rev_id ) && ( isset( $row->ar_user ) || isset( $row->ar_actor ) ) ) {
762 $row = $this->mapArchiveFields( $row );
763 }
764
765 if ( isset( $row->rev_text_id ) && $row->rev_text_id > 0 ) {
766 $mainSlotRow->slot_content_id = $row->rev_text_id;
767 $mainSlotRow->content_address = 'tt:' . $row->rev_text_id;
768 }
769
770 // This is used by null-revisions
771 $mainSlotRow->slot_origin = isset( $row->slot_origin )
772 ? intval( $row->slot_origin )
773 : null;
774
775 if ( isset( $row->old_text ) ) {
776 // this happens when the text-table gets joined directly, in the pre-1.30 schema
777 $blobData = isset( $row->old_text ) ? strval( $row->old_text ) : null;
778 // Check against selects that might have not included old_flags
779 if ( !property_exists( $row, 'old_flags' ) ) {
780 throw new InvalidArgumentException( 'old_flags was not set in $row' );
781 }
782 $blobFlags = ( $row->old_flags === null ) ? '' : $row->old_flags;
783 }
784
785 $mainSlotRow->slot_revision_id = intval( $row->rev_id );
786
787 $mainSlotRow->content_size = isset( $row->rev_len ) ? intval( $row->rev_len ) : null;
788 $mainSlotRow->content_sha1 = isset( $row->rev_sha1 ) ? strval( $row->rev_sha1 ) : null;
789 $mainSlotRow->model_name = isset( $row->rev_content_model )
790 ? strval( $row->rev_content_model )
791 : null;
792 // XXX: in the future, we'll probably always use the default format, and drop content_format
793 $mainSlotRow->format_name = isset( $row->rev_content_format )
794 ? strval( $row->rev_content_format )
795 : null;
796 } elseif ( is_array( $row ) ) {
797 $mainSlotRow->slot_revision_id = isset( $row['id'] ) ? intval( $row['id'] ) : null;
798
799 $mainSlotRow->slot_content_id = isset( $row['text_id'] )
800 ? intval( $row['text_id'] )
801 : null;
802 $mainSlotRow->slot_origin = isset( $row['slot_origin'] )
803 ? intval( $row['slot_origin'] )
804 : null;
805 $mainSlotRow->content_address = isset( $row['text_id'] )
806 ? 'tt:' . intval( $row['text_id'] )
807 : null;
808 $mainSlotRow->content_size = isset( $row['len'] ) ? intval( $row['len'] ) : null;
809 $mainSlotRow->content_sha1 = isset( $row['sha1'] ) ? strval( $row['sha1'] ) : null;
810
811 $mainSlotRow->model_name = isset( $row['content_model'] )
812 ? strval( $row['content_model'] ) : null; // XXX: must be a string!
813 // XXX: in the future, we'll probably always use the default format, and drop content_format
814 $mainSlotRow->format_name = isset( $row['content_format'] )
815 ? strval( $row['content_format'] ) : null;
816 $blobData = isset( $row['text'] ) ? rtrim( strval( $row['text'] ) ) : null;
817 // XXX: If the flags field is not set then $blobFlags should be null so that no
818 // decoding will happen. An empty string will result in default decodings.
819 $blobFlags = isset( $row['flags'] ) ? trim( strval( $row['flags'] ) ) : null;
820
821 // if we have a Content object, override mText and mContentModel
822 if ( !empty( $row['content'] ) ) {
823 if ( !( $row['content'] instanceof Content ) ) {
824 throw new MWException( 'content field must contain a Content object.' );
825 }
826
827 /** @var Content $content */
828 $content = $row['content'];
829 $handler = $content->getContentHandler();
830
831 $mainSlotRow->model_name = $content->getModel();
832
833 // XXX: in the future, we'll probably always use the default format.
834 if ( $mainSlotRow->format_name === null ) {
835 $mainSlotRow->format_name = $handler->getDefaultFormat();
836 }
837 }
838 } else {
839 throw new MWException( 'Revision constructor passed invalid row format.' );
840 }
841
842 // With the old schema, the content changes with every revision,
843 // except for null-revisions.
844 if ( !isset( $mainSlotRow->slot_origin ) ) {
845 $mainSlotRow->slot_origin = $mainSlotRow->slot_revision_id;
846 }
847
848 if ( $mainSlotRow->model_name === null ) {
849 $mainSlotRow->model_name = function ( SlotRecord $slot ) use ( $title ) {
850 // TODO: MCR: consider slot role in getDefaultModelFor()! Use LinkTarget!
851 // TODO: MCR: deprecate $title->getModel().
852 return ContentHandler::getDefaultModelFor( $title );
853 };
854 }
855
856 if ( !$content ) {
857 $content = function ( SlotRecord $slot )
858 use ( $blobData, $blobFlags, $queryFlags, $mainSlotRow )
859 {
860 return $this->loadSlotContent(
861 $slot,
862 $blobData,
863 $blobFlags,
864 $mainSlotRow->format_name,
865 $queryFlags
866 );
867 };
868 }
869
870 $mainSlotRow->slot_id = $mainSlotRow->slot_revision_id;
871 return new SlotRecord( $mainSlotRow, $content );
872 }
873
874 /**
875 * Loads a Content object based on a slot row.
876 *
877 * This method does not call $slot->getContent(), and may be used as a callback
878 * called by $slot->getContent().
879 *
880 * MCR migration note: this roughly corresponds to Revision::getContentInternal
881 *
882 * @param SlotRecord $slot The SlotRecord to load content for
883 * @param string|null $blobData The content blob, in the form indicated by $blobFlags
884 * @param string|null $blobFlags Flags indicating how $blobData needs to be processed.
885 * Use null if no processing should happen. That is in constrast to the empty string,
886 * which causes the blob to be decoded according to the configured legacy encoding.
887 * @param string|null $blobFormat MIME type indicating how $dataBlob is encoded
888 * @param int $queryFlags
889 *
890 * @throw RevisionAccessException
891 * @return Content
892 */
893 private function loadSlotContent(
894 SlotRecord $slot,
895 $blobData = null,
896 $blobFlags = null,
897 $blobFormat = null,
898 $queryFlags = 0
899 ) {
900 if ( $blobData !== null ) {
901 Assert::parameterType( 'string', $blobData, '$blobData' );
902 Assert::parameterType( 'string|null', $blobFlags, '$blobFlags' );
903
904 $cacheKey = $slot->hasAddress() ? $slot->getAddress() : null;
905
906 if ( $blobFlags === null ) {
907 // No blob flags, so use the blob verbatim.
908 $data = $blobData;
909 } else {
910 $data = $this->blobStore->expandBlob( $blobData, $blobFlags, $cacheKey );
911 if ( $data === false ) {
912 throw new RevisionAccessException(
913 "Failed to expand blob data using flags $blobFlags (key: $cacheKey)"
914 );
915 }
916 }
917
918 } else {
919 $address = $slot->getAddress();
920 try {
921 $data = $this->blobStore->getBlob( $address, $queryFlags );
922 } catch ( BlobAccessException $e ) {
923 throw new RevisionAccessException(
924 "Failed to load data blob from $address: " . $e->getMessage(), 0, $e
925 );
926 }
927 }
928
929 // Unserialize content
930 $handler = ContentHandler::getForModelID( $slot->getModel() );
931
932 $content = $handler->unserializeContent( $data, $blobFormat );
933 return $content;
934 }
935
936 /**
937 * Load a page revision from a given revision ID number.
938 * Returns null if no such revision can be found.
939 *
940 * MCR migration note: this replaces Revision::newFromId
941 *
942 * $flags include:
943 * IDBAccessObject::READ_LATEST: Select the data from the master
944 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
945 *
946 * @param int $id
947 * @param int $flags (optional)
948 * @return RevisionRecord|null
949 */
950 public function getRevisionById( $id, $flags = 0 ) {
951 return $this->newRevisionFromConds( [ 'rev_id' => intval( $id ) ], $flags );
952 }
953
954 /**
955 * Load either the current, or a specified, revision
956 * that's attached to a given link target. If not attached
957 * to that link target, will return null.
958 *
959 * MCR migration note: this replaces Revision::newFromTitle
960 *
961 * $flags include:
962 * IDBAccessObject::READ_LATEST: Select the data from the master
963 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
964 *
965 * @param LinkTarget $linkTarget
966 * @param int $revId (optional)
967 * @param int $flags Bitfield (optional)
968 * @return RevisionRecord|null
969 */
970 public function getRevisionByTitle( LinkTarget $linkTarget, $revId = 0, $flags = 0 ) {
971 $conds = [
972 'page_namespace' => $linkTarget->getNamespace(),
973 'page_title' => $linkTarget->getDBkey()
974 ];
975 if ( $revId ) {
976 // Use the specified revision ID.
977 // Note that we use newRevisionFromConds here because we want to retry
978 // and fall back to master if the page is not found on a replica.
979 // Since the caller supplied a revision ID, we are pretty sure the revision is
980 // supposed to exist, so we should try hard to find it.
981 $conds['rev_id'] = $revId;
982 return $this->newRevisionFromConds( $conds, $flags );
983 } else {
984 // Use a join to get the latest revision.
985 // Note that we don't use newRevisionFromConds here because we don't want to retry
986 // and fall back to master. The assumption is that we only want to force the fallback
987 // if we are quite sure the revision exists because the caller supplied a revision ID.
988 // If the page isn't found at all on a replica, it probably simply does not exist.
989 $db = $this->getDBConnection( ( $flags & self::READ_LATEST ) ? DB_MASTER : DB_REPLICA );
990
991 $conds[] = 'rev_id=page_latest';
992 $rev = $this->loadRevisionFromConds( $db, $conds, $flags );
993
994 $this->releaseDBConnection( $db );
995 return $rev;
996 }
997 }
998
999 /**
1000 * Load either the current, or a specified, revision
1001 * that's attached to a given page ID.
1002 * Returns null if no such revision can be found.
1003 *
1004 * MCR migration note: this replaces Revision::newFromPageId
1005 *
1006 * $flags include:
1007 * IDBAccessObject::READ_LATEST: Select the data from the master (since 1.20)
1008 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
1009 *
1010 * @param int $pageId
1011 * @param int $revId (optional)
1012 * @param int $flags Bitfield (optional)
1013 * @return RevisionRecord|null
1014 */
1015 public function getRevisionByPageId( $pageId, $revId = 0, $flags = 0 ) {
1016 $conds = [ 'page_id' => $pageId ];
1017 if ( $revId ) {
1018 // Use the specified revision ID.
1019 // Note that we use newRevisionFromConds here because we want to retry
1020 // and fall back to master if the page is not found on a replica.
1021 // Since the caller supplied a revision ID, we are pretty sure the revision is
1022 // supposed to exist, so we should try hard to find it.
1023 $conds['rev_id'] = $revId;
1024 return $this->newRevisionFromConds( $conds, $flags );
1025 } else {
1026 // Use a join to get the latest revision.
1027 // Note that we don't use newRevisionFromConds here because we don't want to retry
1028 // and fall back to master. The assumption is that we only want to force the fallback
1029 // if we are quite sure the revision exists because the caller supplied a revision ID.
1030 // If the page isn't found at all on a replica, it probably simply does not exist.
1031 $db = $this->getDBConnection( ( $flags & self::READ_LATEST ) ? DB_MASTER : DB_REPLICA );
1032
1033 $conds[] = 'rev_id=page_latest';
1034 $rev = $this->loadRevisionFromConds( $db, $conds, $flags );
1035
1036 $this->releaseDBConnection( $db );
1037 return $rev;
1038 }
1039 }
1040
1041 /**
1042 * Load the revision for the given title with the given timestamp.
1043 * WARNING: Timestamps may in some circumstances not be unique,
1044 * so this isn't the best key to use.
1045 *
1046 * MCR migration note: this replaces Revision::loadFromTimestamp
1047 *
1048 * @param Title $title
1049 * @param string $timestamp
1050 * @return RevisionRecord|null
1051 */
1052 public function getRevisionByTimestamp( $title, $timestamp ) {
1053 $db = $this->getDBConnection( DB_REPLICA );
1054 return $this->newRevisionFromConds(
1055 [
1056 'rev_timestamp' => $db->timestamp( $timestamp ),
1057 'page_namespace' => $title->getNamespace(),
1058 'page_title' => $title->getDBkey()
1059 ],
1060 0,
1061 $title
1062 );
1063 }
1064
1065 /**
1066 * Make a fake revision object from an archive table row. This is queried
1067 * for permissions or even inserted (as in Special:Undelete)
1068 *
1069 * MCR migration note: this replaces Revision::newFromArchiveRow
1070 *
1071 * @param object $row
1072 * @param int $queryFlags
1073 * @param Title|null $title
1074 * @param array $overrides associative array with fields of $row to override. This may be
1075 * used e.g. to force the parent revision ID or page ID. Keys in the array are fields
1076 * names from the archive table without the 'ar_' prefix, i.e. use 'parent_id' to
1077 * override ar_parent_id.
1078 *
1079 * @return RevisionRecord
1080 * @throws MWException
1081 */
1082 public function newRevisionFromArchiveRow(
1083 $row,
1084 $queryFlags = 0,
1085 Title $title = null,
1086 array $overrides = []
1087 ) {
1088 Assert::parameterType( 'object', $row, '$row' );
1089
1090 // check second argument, since Revision::newFromArchiveRow had $overrides in that spot.
1091 Assert::parameterType( 'integer', $queryFlags, '$queryFlags' );
1092
1093 if ( !$title && isset( $overrides['title'] ) ) {
1094 if ( !( $overrides['title'] instanceof Title ) ) {
1095 throw new MWException( 'title field override must contain a Title object.' );
1096 }
1097
1098 $title = $overrides['title'];
1099 }
1100
1101 if ( !isset( $title ) ) {
1102 if ( isset( $row->ar_namespace ) && isset( $row->ar_title ) ) {
1103 $title = Title::makeTitle( $row->ar_namespace, $row->ar_title );
1104 } else {
1105 throw new InvalidArgumentException(
1106 'A Title or ar_namespace and ar_title must be given'
1107 );
1108 }
1109 }
1110
1111 foreach ( $overrides as $key => $value ) {
1112 $field = "ar_$key";
1113 $row->$field = $value;
1114 }
1115
1116 try {
1117 $user = User::newFromAnyId(
1118 isset( $row->ar_user ) ? $row->ar_user : null,
1119 isset( $row->ar_user_text ) ? $row->ar_user_text : null,
1120 isset( $row->ar_actor ) ? $row->ar_actor : null
1121 );
1122 } catch ( InvalidArgumentException $ex ) {
1123 wfWarn( __METHOD__ . ': ' . $ex->getMessage() );
1124 $user = new UserIdentityValue( 0, '', 0 );
1125 }
1126
1127 $comment = $this->commentStore
1128 // Legacy because $row may have come from self::selectFields()
1129 ->getCommentLegacy( $this->getDBConnection( DB_REPLICA ), 'ar_comment', $row, true );
1130
1131 $mainSlot = $this->emulateMainSlot_1_29( $row, $queryFlags, $title );
1132 $slots = new RevisionSlots( [ 'main' => $mainSlot ] );
1133
1134 return new RevisionArchiveRecord( $title, $user, $comment, $row, $slots, $this->wikiId );
1135 }
1136
1137 /**
1138 * @see RevisionFactory::newRevisionFromRow_1_29
1139 *
1140 * MCR migration note: this replaces Revision::newFromRow
1141 *
1142 * @param object $row
1143 * @param int $queryFlags
1144 * @param Title|null $title
1145 *
1146 * @return RevisionRecord
1147 * @throws MWException
1148 * @throws RevisionAccessException
1149 */
1150 private function newRevisionFromRow_1_29( $row, $queryFlags = 0, Title $title = null ) {
1151 Assert::parameterType( 'object', $row, '$row' );
1152
1153 if ( !$title ) {
1154 $pageId = isset( $row->rev_page ) ? $row->rev_page : 0; // XXX: also check page_id?
1155 $revId = isset( $row->rev_id ) ? $row->rev_id : 0;
1156
1157 $title = $this->getTitle( $pageId, $revId, $queryFlags );
1158 }
1159
1160 if ( !isset( $row->page_latest ) ) {
1161 $row->page_latest = $title->getLatestRevID();
1162 if ( $row->page_latest === 0 && $title->exists() ) {
1163 wfWarn( 'Encountered title object in limbo: ID ' . $title->getArticleID() );
1164 }
1165 }
1166
1167 try {
1168 $user = User::newFromAnyId(
1169 isset( $row->rev_user ) ? $row->rev_user : null,
1170 isset( $row->rev_user_text ) ? $row->rev_user_text : null,
1171 isset( $row->rev_actor ) ? $row->rev_actor : null
1172 );
1173 } catch ( InvalidArgumentException $ex ) {
1174 wfWarn( __METHOD__ . ': ' . $ex->getMessage() );
1175 $user = new UserIdentityValue( 0, '', 0 );
1176 }
1177
1178 $comment = $this->commentStore
1179 // Legacy because $row may have come from self::selectFields()
1180 ->getCommentLegacy( $this->getDBConnection( DB_REPLICA ), 'rev_comment', $row, true );
1181
1182 $mainSlot = $this->emulateMainSlot_1_29( $row, $queryFlags, $title );
1183 $slots = new RevisionSlots( [ 'main' => $mainSlot ] );
1184
1185 return new RevisionStoreRecord( $title, $user, $comment, $row, $slots, $this->wikiId );
1186 }
1187
1188 /**
1189 * @see RevisionFactory::newRevisionFromRow
1190 *
1191 * MCR migration note: this replaces Revision::newFromRow
1192 *
1193 * @param object $row
1194 * @param int $queryFlags
1195 * @param Title|null $title
1196 *
1197 * @return RevisionRecord
1198 */
1199 public function newRevisionFromRow( $row, $queryFlags = 0, Title $title = null ) {
1200 return $this->newRevisionFromRow_1_29( $row, $queryFlags, $title );
1201 }
1202
1203 /**
1204 * Constructs a new MutableRevisionRecord based on the given associative array following
1205 * the MW1.29 convention for the Revision constructor.
1206 *
1207 * MCR migration note: this replaces Revision::newFromRow
1208 *
1209 * @param array $fields
1210 * @param int $queryFlags
1211 * @param Title|null $title
1212 *
1213 * @return MutableRevisionRecord
1214 * @throws MWException
1215 * @throws RevisionAccessException
1216 */
1217 public function newMutableRevisionFromArray(
1218 array $fields,
1219 $queryFlags = 0,
1220 Title $title = null
1221 ) {
1222 if ( !$title && isset( $fields['title'] ) ) {
1223 if ( !( $fields['title'] instanceof Title ) ) {
1224 throw new MWException( 'title field must contain a Title object.' );
1225 }
1226
1227 $title = $fields['title'];
1228 }
1229
1230 if ( !$title ) {
1231 $pageId = isset( $fields['page'] ) ? $fields['page'] : 0;
1232 $revId = isset( $fields['id'] ) ? $fields['id'] : 0;
1233
1234 $title = $this->getTitle( $pageId, $revId, $queryFlags );
1235 }
1236
1237 if ( !isset( $fields['page'] ) ) {
1238 $fields['page'] = $title->getArticleID( $queryFlags );
1239 }
1240
1241 // if we have a content object, use it to set the model and type
1242 if ( !empty( $fields['content'] ) ) {
1243 if ( !( $fields['content'] instanceof Content ) ) {
1244 throw new MWException( 'content field must contain a Content object.' );
1245 }
1246
1247 if ( !empty( $fields['text_id'] ) ) {
1248 throw new MWException(
1249 "Text already stored in external store (id {$fields['text_id']}), " .
1250 "can't serialize content object"
1251 );
1252 }
1253 }
1254
1255 if (
1256 isset( $fields['comment'] )
1257 && !( $fields['comment'] instanceof CommentStoreComment )
1258 ) {
1259 $commentData = isset( $fields['comment_data'] ) ? $fields['comment_data'] : null;
1260
1261 if ( $fields['comment'] instanceof Message ) {
1262 $fields['comment'] = CommentStoreComment::newUnsavedComment(
1263 $fields['comment'],
1264 $commentData
1265 );
1266 } else {
1267 $commentText = trim( strval( $fields['comment'] ) );
1268 $fields['comment'] = CommentStoreComment::newUnsavedComment(
1269 $commentText,
1270 $commentData
1271 );
1272 }
1273 }
1274
1275 $mainSlot = $this->emulateMainSlot_1_29( $fields, $queryFlags, $title );
1276
1277 $revision = new MutableRevisionRecord( $title, $this->wikiId );
1278 $this->initializeMutableRevisionFromArray( $revision, $fields );
1279 $revision->setSlot( $mainSlot );
1280
1281 return $revision;
1282 }
1283
1284 /**
1285 * @param MutableRevisionRecord $record
1286 * @param array $fields
1287 */
1288 private function initializeMutableRevisionFromArray(
1289 MutableRevisionRecord $record,
1290 array $fields
1291 ) {
1292 /** @var UserIdentity $user */
1293 $user = null;
1294
1295 if ( isset( $fields['user'] ) && ( $fields['user'] instanceof UserIdentity ) ) {
1296 $user = $fields['user'];
1297 } else {
1298 try {
1299 $user = User::newFromAnyId(
1300 isset( $fields['user'] ) ? $fields['user'] : null,
1301 isset( $fields['user_text'] ) ? $fields['user_text'] : null,
1302 isset( $fields['actor'] ) ? $fields['actor'] : null
1303 );
1304 } catch ( InvalidArgumentException $ex ) {
1305 $user = null;
1306 }
1307 }
1308
1309 if ( $user ) {
1310 $record->setUser( $user );
1311 }
1312
1313 $timestamp = isset( $fields['timestamp'] )
1314 ? strval( $fields['timestamp'] )
1315 : wfTimestampNow(); // TODO: use a callback, so we can override it for testing.
1316
1317 $record->setTimestamp( $timestamp );
1318
1319 if ( isset( $fields['page'] ) ) {
1320 $record->setPageId( intval( $fields['page'] ) );
1321 }
1322
1323 if ( isset( $fields['id'] ) ) {
1324 $record->setId( intval( $fields['id'] ) );
1325 }
1326 if ( isset( $fields['parent_id'] ) ) {
1327 $record->setParentId( intval( $fields['parent_id'] ) );
1328 }
1329
1330 if ( isset( $fields['sha1'] ) ) {
1331 $record->setSha1( $fields['sha1'] );
1332 }
1333 if ( isset( $fields['size'] ) ) {
1334 $record->setSize( intval( $fields['size'] ) );
1335 }
1336
1337 if ( isset( $fields['minor_edit'] ) ) {
1338 $record->setMinorEdit( intval( $fields['minor_edit'] ) !== 0 );
1339 }
1340 if ( isset( $fields['deleted'] ) ) {
1341 $record->setVisibility( intval( $fields['deleted'] ) );
1342 }
1343
1344 if ( isset( $fields['comment'] ) ) {
1345 Assert::parameterType(
1346 CommentStoreComment::class,
1347 $fields['comment'],
1348 '$row[\'comment\']'
1349 );
1350 $record->setComment( $fields['comment'] );
1351 }
1352 }
1353
1354 /**
1355 * Load a page revision from a given revision ID number.
1356 * Returns null if no such revision can be found.
1357 *
1358 * MCR migration note: this corresponds to Revision::loadFromId
1359 *
1360 * @note direct use is deprecated!
1361 * @todo remove when unused! there seem to be no callers of Revision::loadFromId
1362 *
1363 * @param IDatabase $db
1364 * @param int $id
1365 *
1366 * @return RevisionRecord|null
1367 */
1368 public function loadRevisionFromId( IDatabase $db, $id ) {
1369 return $this->loadRevisionFromConds( $db, [ 'rev_id' => intval( $id ) ] );
1370 }
1371
1372 /**
1373 * Load either the current, or a specified, revision
1374 * that's attached to a given page. If not attached
1375 * to that page, will return null.
1376 *
1377 * MCR migration note: this replaces Revision::loadFromPageId
1378 *
1379 * @note direct use is deprecated!
1380 * @todo remove when unused!
1381 *
1382 * @param IDatabase $db
1383 * @param int $pageid
1384 * @param int $id
1385 * @return RevisionRecord|null
1386 */
1387 public function loadRevisionFromPageId( IDatabase $db, $pageid, $id = 0 ) {
1388 $conds = [ 'rev_page' => intval( $pageid ), 'page_id' => intval( $pageid ) ];
1389 if ( $id ) {
1390 $conds['rev_id'] = intval( $id );
1391 } else {
1392 $conds[] = 'rev_id=page_latest';
1393 }
1394 return $this->loadRevisionFromConds( $db, $conds );
1395 }
1396
1397 /**
1398 * Load either the current, or a specified, revision
1399 * that's attached to a given page. If not attached
1400 * to that page, will return null.
1401 *
1402 * MCR migration note: this replaces Revision::loadFromTitle
1403 *
1404 * @note direct use is deprecated!
1405 * @todo remove when unused!
1406 *
1407 * @param IDatabase $db
1408 * @param Title $title
1409 * @param int $id
1410 *
1411 * @return RevisionRecord|null
1412 */
1413 public function loadRevisionFromTitle( IDatabase $db, $title, $id = 0 ) {
1414 if ( $id ) {
1415 $matchId = intval( $id );
1416 } else {
1417 $matchId = 'page_latest';
1418 }
1419
1420 return $this->loadRevisionFromConds(
1421 $db,
1422 [
1423 "rev_id=$matchId",
1424 'page_namespace' => $title->getNamespace(),
1425 'page_title' => $title->getDBkey()
1426 ],
1427 0,
1428 $title
1429 );
1430 }
1431
1432 /**
1433 * Load the revision for the given title with the given timestamp.
1434 * WARNING: Timestamps may in some circumstances not be unique,
1435 * so this isn't the best key to use.
1436 *
1437 * MCR migration note: this replaces Revision::loadFromTimestamp
1438 *
1439 * @note direct use is deprecated! Use getRevisionFromTimestamp instead!
1440 * @todo remove when unused!
1441 *
1442 * @param IDatabase $db
1443 * @param Title $title
1444 * @param string $timestamp
1445 * @return RevisionRecord|null
1446 */
1447 public function loadRevisionFromTimestamp( IDatabase $db, $title, $timestamp ) {
1448 return $this->loadRevisionFromConds( $db,
1449 [
1450 'rev_timestamp' => $db->timestamp( $timestamp ),
1451 'page_namespace' => $title->getNamespace(),
1452 'page_title' => $title->getDBkey()
1453 ],
1454 0,
1455 $title
1456 );
1457 }
1458
1459 /**
1460 * Given a set of conditions, fetch a revision
1461 *
1462 * This method should be used if we are pretty sure the revision exists.
1463 * Unless $flags has READ_LATEST set, this method will first try to find the revision
1464 * on a replica before hitting the master database.
1465 *
1466 * MCR migration note: this corresponds to Revision::newFromConds
1467 *
1468 * @param array $conditions
1469 * @param int $flags (optional)
1470 * @param Title $title
1471 *
1472 * @return RevisionRecord|null
1473 */
1474 private function newRevisionFromConds( $conditions, $flags = 0, Title $title = null ) {
1475 $db = $this->getDBConnection( ( $flags & self::READ_LATEST ) ? DB_MASTER : DB_REPLICA );
1476 $rev = $this->loadRevisionFromConds( $db, $conditions, $flags, $title );
1477 $this->releaseDBConnection( $db );
1478
1479 $lb = $this->getDBLoadBalancer();
1480
1481 // Make sure new pending/committed revision are visibile later on
1482 // within web requests to certain avoid bugs like T93866 and T94407.
1483 if ( !$rev
1484 && !( $flags & self::READ_LATEST )
1485 && $lb->getServerCount() > 1
1486 && $lb->hasOrMadeRecentMasterChanges()
1487 ) {
1488 $flags = self::READ_LATEST;
1489 $db = $this->getDBConnection( DB_MASTER );
1490 $rev = $this->loadRevisionFromConds( $db, $conditions, $flags, $title );
1491 $this->releaseDBConnection( $db );
1492 }
1493
1494 return $rev;
1495 }
1496
1497 /**
1498 * Given a set of conditions, fetch a revision from
1499 * the given database connection.
1500 *
1501 * MCR migration note: this corresponds to Revision::loadFromConds
1502 *
1503 * @param IDatabase $db
1504 * @param array $conditions
1505 * @param int $flags (optional)
1506 * @param Title $title
1507 *
1508 * @return RevisionRecord|null
1509 */
1510 private function loadRevisionFromConds(
1511 IDatabase $db,
1512 $conditions,
1513 $flags = 0,
1514 Title $title = null
1515 ) {
1516 $row = $this->fetchRevisionRowFromConds( $db, $conditions, $flags );
1517 if ( $row ) {
1518 $rev = $this->newRevisionFromRow( $row, $flags, $title );
1519
1520 return $rev;
1521 }
1522
1523 return null;
1524 }
1525
1526 /**
1527 * Throws an exception if the given database connection does not belong to the wiki this
1528 * RevisionStore is bound to.
1529 *
1530 * @param IDatabase $db
1531 * @throws MWException
1532 */
1533 private function checkDatabaseWikiId( IDatabase $db ) {
1534 $storeWiki = $this->wikiId;
1535 $dbWiki = $db->getDomainID();
1536
1537 if ( $dbWiki === $storeWiki ) {
1538 return;
1539 }
1540
1541 // XXX: we really want the default database ID...
1542 $storeWiki = $storeWiki ?: wfWikiID();
1543 $dbWiki = $dbWiki ?: wfWikiID();
1544
1545 if ( $dbWiki === $storeWiki ) {
1546 return;
1547 }
1548
1549 // HACK: counteract encoding imposed by DatabaseDomain
1550 $storeWiki = str_replace( '?h', '-', $storeWiki );
1551 $dbWiki = str_replace( '?h', '-', $dbWiki );
1552
1553 if ( $dbWiki === $storeWiki ) {
1554 return;
1555 }
1556
1557 throw new MWException( "RevisionStore for $storeWiki "
1558 . "cannot be used with a DB connection for $dbWiki" );
1559 }
1560
1561 /**
1562 * Given a set of conditions, return a row with the
1563 * fields necessary to build RevisionRecord objects.
1564 *
1565 * MCR migration note: this corresponds to Revision::fetchFromConds
1566 *
1567 * @param IDatabase $db
1568 * @param array $conditions
1569 * @param int $flags (optional)
1570 *
1571 * @return object|false data row as a raw object
1572 */
1573 private function fetchRevisionRowFromConds( IDatabase $db, $conditions, $flags = 0 ) {
1574 $this->checkDatabaseWikiId( $db );
1575
1576 $revQuery = self::getQueryInfo( [ 'page', 'user' ] );
1577 $options = [];
1578 if ( ( $flags & self::READ_LOCKING ) == self::READ_LOCKING ) {
1579 $options[] = 'FOR UPDATE';
1580 }
1581 return $db->selectRow(
1582 $revQuery['tables'],
1583 $revQuery['fields'],
1584 $conditions,
1585 __METHOD__,
1586 $options,
1587 $revQuery['joins']
1588 );
1589 }
1590
1591 /**
1592 * Return the tables, fields, and join conditions to be selected to create
1593 * a new revision object.
1594 *
1595 * MCR migration note: this replaces Revision::getQueryInfo
1596 *
1597 * @since 1.31
1598 *
1599 * @param array $options Any combination of the following strings
1600 * - 'page': Join with the page table, and select fields to identify the page
1601 * - 'user': Join with the user table, and select the user name
1602 * - 'text': Join with the text table, and select fields to load page text
1603 *
1604 * @return array With three keys:
1605 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
1606 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
1607 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
1608 */
1609 public function getQueryInfo( $options = [] ) {
1610 $ret = [
1611 'tables' => [],
1612 'fields' => [],
1613 'joins' => [],
1614 ];
1615
1616 $ret['tables'][] = 'revision';
1617 $ret['fields'] = array_merge( $ret['fields'], [
1618 'rev_id',
1619 'rev_page',
1620 'rev_text_id',
1621 'rev_timestamp',
1622 'rev_minor_edit',
1623 'rev_deleted',
1624 'rev_len',
1625 'rev_parent_id',
1626 'rev_sha1',
1627 ] );
1628
1629 $commentQuery = $this->commentStore->getJoin( 'rev_comment' );
1630 $ret['tables'] = array_merge( $ret['tables'], $commentQuery['tables'] );
1631 $ret['fields'] = array_merge( $ret['fields'], $commentQuery['fields'] );
1632 $ret['joins'] = array_merge( $ret['joins'], $commentQuery['joins'] );
1633
1634 $actorQuery = $this->actorMigration->getJoin( 'rev_user' );
1635 $ret['tables'] = array_merge( $ret['tables'], $actorQuery['tables'] );
1636 $ret['fields'] = array_merge( $ret['fields'], $actorQuery['fields'] );
1637 $ret['joins'] = array_merge( $ret['joins'], $actorQuery['joins'] );
1638
1639 if ( $this->contentHandlerUseDB ) {
1640 $ret['fields'][] = 'rev_content_format';
1641 $ret['fields'][] = 'rev_content_model';
1642 }
1643
1644 if ( in_array( 'page', $options, true ) ) {
1645 $ret['tables'][] = 'page';
1646 $ret['fields'] = array_merge( $ret['fields'], [
1647 'page_namespace',
1648 'page_title',
1649 'page_id',
1650 'page_latest',
1651 'page_is_redirect',
1652 'page_len',
1653 ] );
1654 $ret['joins']['page'] = [ 'INNER JOIN', [ 'page_id = rev_page' ] ];
1655 }
1656
1657 if ( in_array( 'user', $options, true ) ) {
1658 $ret['tables'][] = 'user';
1659 $ret['fields'] = array_merge( $ret['fields'], [
1660 'user_name',
1661 ] );
1662 $u = $actorQuery['fields']['rev_user'];
1663 $ret['joins']['user'] = [ 'LEFT JOIN', [ "$u != 0", "user_id = $u" ] ];
1664 }
1665
1666 if ( in_array( 'text', $options, true ) ) {
1667 $ret['tables'][] = 'text';
1668 $ret['fields'] = array_merge( $ret['fields'], [
1669 'old_text',
1670 'old_flags'
1671 ] );
1672 $ret['joins']['text'] = [ 'INNER JOIN', [ 'rev_text_id=old_id' ] ];
1673 }
1674
1675 return $ret;
1676 }
1677
1678 /**
1679 * Return the tables, fields, and join conditions to be selected to create
1680 * a new archived revision object.
1681 *
1682 * MCR migration note: this replaces Revision::getArchiveQueryInfo
1683 *
1684 * @since 1.31
1685 *
1686 * @return array With three keys:
1687 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
1688 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
1689 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
1690 */
1691 public function getArchiveQueryInfo() {
1692 $commentQuery = $this->commentStore->getJoin( 'ar_comment' );
1693 $actorQuery = $this->actorMigration->getJoin( 'ar_user' );
1694 $ret = [
1695 'tables' => [ 'archive' ] + $commentQuery['tables'] + $actorQuery['tables'],
1696 'fields' => [
1697 'ar_id',
1698 'ar_page_id',
1699 'ar_namespace',
1700 'ar_title',
1701 'ar_rev_id',
1702 'ar_text_id',
1703 'ar_timestamp',
1704 'ar_minor_edit',
1705 'ar_deleted',
1706 'ar_len',
1707 'ar_parent_id',
1708 'ar_sha1',
1709 ] + $commentQuery['fields'] + $actorQuery['fields'],
1710 'joins' => $commentQuery['joins'] + $actorQuery['joins'],
1711 ];
1712
1713 if ( $this->contentHandlerUseDB ) {
1714 $ret['fields'][] = 'ar_content_format';
1715 $ret['fields'][] = 'ar_content_model';
1716 }
1717
1718 return $ret;
1719 }
1720
1721 /**
1722 * Do a batched query for the sizes of a set of revisions.
1723 *
1724 * MCR migration note: this replaces Revision::getParentLengths
1725 *
1726 * @param int[] $revIds
1727 * @return int[] associative array mapping revision IDs from $revIds to the nominal size
1728 * of the corresponding revision.
1729 */
1730 public function getRevisionSizes( array $revIds ) {
1731 return $this->listRevisionSizes( $this->getDBConnection( DB_REPLICA ), $revIds );
1732 }
1733
1734 /**
1735 * Do a batched query for the sizes of a set of revisions.
1736 *
1737 * MCR migration note: this replaces Revision::getParentLengths
1738 *
1739 * @deprecated use RevisionStore::getRevisionSizes instead.
1740 *
1741 * @param IDatabase $db
1742 * @param int[] $revIds
1743 * @return int[] associative array mapping revision IDs from $revIds to the nominal size
1744 * of the corresponding revision.
1745 */
1746 public function listRevisionSizes( IDatabase $db, array $revIds ) {
1747 $this->checkDatabaseWikiId( $db );
1748
1749 $revLens = [];
1750 if ( !$revIds ) {
1751 return $revLens; // empty
1752 }
1753
1754 $res = $db->select(
1755 'revision',
1756 [ 'rev_id', 'rev_len' ],
1757 [ 'rev_id' => $revIds ],
1758 __METHOD__
1759 );
1760
1761 foreach ( $res as $row ) {
1762 $revLens[$row->rev_id] = intval( $row->rev_len );
1763 }
1764
1765 return $revLens;
1766 }
1767
1768 /**
1769 * Get previous revision for this title
1770 *
1771 * MCR migration note: this replaces Revision::getPrevious
1772 *
1773 * @param RevisionRecord $rev
1774 * @param Title $title if known (optional)
1775 *
1776 * @return RevisionRecord|null
1777 */
1778 public function getPreviousRevision( RevisionRecord $rev, Title $title = null ) {
1779 if ( $title === null ) {
1780 $title = $this->getTitle( $rev->getPageId(), $rev->getId() );
1781 }
1782 $prev = $title->getPreviousRevisionID( $rev->getId() );
1783 if ( $prev ) {
1784 return $this->getRevisionByTitle( $title, $prev );
1785 }
1786 return null;
1787 }
1788
1789 /**
1790 * Get next revision for this title
1791 *
1792 * MCR migration note: this replaces Revision::getNext
1793 *
1794 * @param RevisionRecord $rev
1795 * @param Title $title if known (optional)
1796 *
1797 * @return RevisionRecord|null
1798 */
1799 public function getNextRevision( RevisionRecord $rev, Title $title = null ) {
1800 if ( $title === null ) {
1801 $title = $this->getTitle( $rev->getPageId(), $rev->getId() );
1802 }
1803 $next = $title->getNextRevisionID( $rev->getId() );
1804 if ( $next ) {
1805 return $this->getRevisionByTitle( $title, $next );
1806 }
1807 return null;
1808 }
1809
1810 /**
1811 * Get previous revision Id for this page_id
1812 * This is used to populate rev_parent_id on save
1813 *
1814 * MCR migration note: this corresponds to Revision::getPreviousRevisionId
1815 *
1816 * @param IDatabase $db
1817 * @param RevisionRecord $rev
1818 *
1819 * @return int
1820 */
1821 private function getPreviousRevisionId( IDatabase $db, RevisionRecord $rev ) {
1822 $this->checkDatabaseWikiId( $db );
1823
1824 if ( $rev->getPageId() === null ) {
1825 return 0;
1826 }
1827 # Use page_latest if ID is not given
1828 if ( !$rev->getId() ) {
1829 $prevId = $db->selectField(
1830 'page', 'page_latest',
1831 [ 'page_id' => $rev->getPageId() ],
1832 __METHOD__
1833 );
1834 } else {
1835 $prevId = $db->selectField(
1836 'revision', 'rev_id',
1837 [ 'rev_page' => $rev->getPageId(), 'rev_id < ' . $rev->getId() ],
1838 __METHOD__,
1839 [ 'ORDER BY' => 'rev_id DESC' ]
1840 );
1841 }
1842 return intval( $prevId );
1843 }
1844
1845 /**
1846 * Get rev_timestamp from rev_id, without loading the rest of the row
1847 *
1848 * MCR migration note: this replaces Revision::getTimestampFromId
1849 *
1850 * @param Title $title
1851 * @param int $id
1852 * @param int $flags
1853 * @return string|bool False if not found
1854 */
1855 public function getTimestampFromId( $title, $id, $flags = 0 ) {
1856 $db = $this->getDBConnection(
1857 ( $flags & IDBAccessObject::READ_LATEST ) ? DB_MASTER : DB_REPLICA
1858 );
1859
1860 $conds = [ 'rev_id' => $id ];
1861 $conds['rev_page'] = $title->getArticleID();
1862 $timestamp = $db->selectField( 'revision', 'rev_timestamp', $conds, __METHOD__ );
1863
1864 $this->releaseDBConnection( $db );
1865 return ( $timestamp !== false ) ? wfTimestamp( TS_MW, $timestamp ) : false;
1866 }
1867
1868 /**
1869 * Get count of revisions per page...not very efficient
1870 *
1871 * MCR migration note: this replaces Revision::countByPageId
1872 *
1873 * @param IDatabase $db
1874 * @param int $id Page id
1875 * @return int
1876 */
1877 public function countRevisionsByPageId( IDatabase $db, $id ) {
1878 $this->checkDatabaseWikiId( $db );
1879
1880 $row = $db->selectRow( 'revision',
1881 [ 'revCount' => 'COUNT(*)' ],
1882 [ 'rev_page' => $id ],
1883 __METHOD__
1884 );
1885 if ( $row ) {
1886 return intval( $row->revCount );
1887 }
1888 return 0;
1889 }
1890
1891 /**
1892 * Get count of revisions per page...not very efficient
1893 *
1894 * MCR migration note: this replaces Revision::countByTitle
1895 *
1896 * @param IDatabase $db
1897 * @param Title $title
1898 * @return int
1899 */
1900 public function countRevisionsByTitle( IDatabase $db, $title ) {
1901 $id = $title->getArticleID();
1902 if ( $id ) {
1903 return $this->countRevisionsByPageId( $db, $id );
1904 }
1905 return 0;
1906 }
1907
1908 /**
1909 * Check if no edits were made by other users since
1910 * the time a user started editing the page. Limit to
1911 * 50 revisions for the sake of performance.
1912 *
1913 * MCR migration note: this replaces Revision::userWasLastToEdit
1914 *
1915 * @deprecated since 1.31; Can possibly be removed, since the self-conflict suppression
1916 * logic in EditPage that uses this seems conceptually dubious. Revision::userWasLastToEdit
1917 * has been deprecated since 1.24.
1918 *
1919 * @param IDatabase $db The Database to perform the check on.
1920 * @param int $pageId The ID of the page in question
1921 * @param int $userId The ID of the user in question
1922 * @param string $since Look at edits since this time
1923 *
1924 * @return bool True if the given user was the only one to edit since the given timestamp
1925 */
1926 public function userWasLastToEdit( IDatabase $db, $pageId, $userId, $since ) {
1927 $this->checkDatabaseWikiId( $db );
1928
1929 if ( !$userId ) {
1930 return false;
1931 }
1932
1933 $revQuery = self::getQueryInfo();
1934 $res = $db->select(
1935 $revQuery['tables'],
1936 [
1937 'rev_user' => $revQuery['fields']['rev_user'],
1938 ],
1939 [
1940 'rev_page' => $pageId,
1941 'rev_timestamp > ' . $db->addQuotes( $db->timestamp( $since ) )
1942 ],
1943 __METHOD__,
1944 [ 'ORDER BY' => 'rev_timestamp ASC', 'LIMIT' => 50 ],
1945 $revQuery['joins']
1946 );
1947 foreach ( $res as $row ) {
1948 if ( $row->rev_user != $userId ) {
1949 return false;
1950 }
1951 }
1952 return true;
1953 }
1954
1955 /**
1956 * Load a revision based on a known page ID and current revision ID from the DB
1957 *
1958 * This method allows for the use of caching, though accessing anything that normally
1959 * requires permission checks (aside from the text) will trigger a small DB lookup.
1960 *
1961 * MCR migration note: this replaces Revision::newKnownCurrent
1962 *
1963 * @param Title $title the associated page title
1964 * @param int $revId current revision of this page. Defaults to $title->getLatestRevID().
1965 *
1966 * @return RevisionRecord|bool Returns false if missing
1967 */
1968 public function getKnownCurrentRevision( Title $title, $revId ) {
1969 $db = $this->getDBConnectionRef( DB_REPLICA );
1970
1971 $pageId = $title->getArticleID();
1972
1973 if ( !$pageId ) {
1974 return false;
1975 }
1976
1977 if ( !$revId ) {
1978 $revId = $title->getLatestRevID();
1979 }
1980
1981 if ( !$revId ) {
1982 wfWarn(
1983 'No latest revision known for page ' . $title->getPrefixedDBkey()
1984 . ' even though it exists with page ID ' . $pageId
1985 );
1986 return false;
1987 }
1988
1989 $row = $this->cache->getWithSetCallback(
1990 // Page/rev IDs passed in from DB to reflect history merges
1991 $this->cache->makeGlobalKey( 'revision-row-1.29', $db->getDomainID(), $pageId, $revId ),
1992 WANObjectCache::TTL_WEEK,
1993 function ( $curValue, &$ttl, array &$setOpts ) use ( $db, $pageId, $revId ) {
1994 $setOpts += Database::getCacheSetOptions( $db );
1995
1996 $conds = [
1997 'rev_page' => intval( $pageId ),
1998 'page_id' => intval( $pageId ),
1999 'rev_id' => intval( $revId ),
2000 ];
2001
2002 $row = $this->fetchRevisionRowFromConds( $db, $conds );
2003 return $row ?: false; // don't cache negatives
2004 }
2005 );
2006
2007 // Reflect revision deletion and user renames
2008 if ( $row ) {
2009 return $this->newRevisionFromRow( $row, 0, $title );
2010 } else {
2011 return false;
2012 }
2013 }
2014
2015 // TODO: move relevant methods from Title here, e.g. getFirstRevision, isBigDeletion, etc.
2016
2017 }