Merge "rcfilters: Stop the "Live updates" button animation from affecting text rendering"
[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 * @throws 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 * @throws 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 = SqlBlobStore::makeAddressFromTextId( $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 = SqlBlobStore::makeAddressFromTextId(
768 $row->rev_text_id
769 );
770 }
771
772 // This is used by null-revisions
773 $mainSlotRow->slot_origin = isset( $row->slot_origin )
774 ? intval( $row->slot_origin )
775 : null;
776
777 if ( isset( $row->old_text ) ) {
778 // this happens when the text-table gets joined directly, in the pre-1.30 schema
779 $blobData = isset( $row->old_text ) ? strval( $row->old_text ) : null;
780 // Check against selects that might have not included old_flags
781 if ( !property_exists( $row, 'old_flags' ) ) {
782 throw new InvalidArgumentException( 'old_flags was not set in $row' );
783 }
784 $blobFlags = ( $row->old_flags === null ) ? '' : $row->old_flags;
785 }
786
787 $mainSlotRow->slot_revision_id = intval( $row->rev_id );
788
789 $mainSlotRow->content_size = isset( $row->rev_len ) ? intval( $row->rev_len ) : null;
790 $mainSlotRow->content_sha1 = isset( $row->rev_sha1 ) ? strval( $row->rev_sha1 ) : null;
791 $mainSlotRow->model_name = isset( $row->rev_content_model )
792 ? strval( $row->rev_content_model )
793 : null;
794 // XXX: in the future, we'll probably always use the default format, and drop content_format
795 $mainSlotRow->format_name = isset( $row->rev_content_format )
796 ? strval( $row->rev_content_format )
797 : null;
798 } elseif ( is_array( $row ) ) {
799 $mainSlotRow->slot_revision_id = isset( $row['id'] ) ? intval( $row['id'] ) : null;
800
801 $mainSlotRow->slot_content_id = isset( $row['text_id'] )
802 ? intval( $row['text_id'] )
803 : null;
804 $mainSlotRow->slot_origin = isset( $row['slot_origin'] )
805 ? intval( $row['slot_origin'] )
806 : null;
807 $mainSlotRow->content_address = isset( $row['text_id'] )
808 ? SqlBlobStore::makeAddressFromTextId( intval( $row['text_id'] ) )
809 : null;
810 $mainSlotRow->content_size = isset( $row['len'] ) ? intval( $row['len'] ) : null;
811 $mainSlotRow->content_sha1 = isset( $row['sha1'] ) ? strval( $row['sha1'] ) : null;
812
813 $mainSlotRow->model_name = isset( $row['content_model'] )
814 ? strval( $row['content_model'] ) : null; // XXX: must be a string!
815 // XXX: in the future, we'll probably always use the default format, and drop content_format
816 $mainSlotRow->format_name = isset( $row['content_format'] )
817 ? strval( $row['content_format'] ) : null;
818 $blobData = isset( $row['text'] ) ? rtrim( strval( $row['text'] ) ) : null;
819 // XXX: If the flags field is not set then $blobFlags should be null so that no
820 // decoding will happen. An empty string will result in default decodings.
821 $blobFlags = isset( $row['flags'] ) ? trim( strval( $row['flags'] ) ) : null;
822
823 // if we have a Content object, override mText and mContentModel
824 if ( !empty( $row['content'] ) ) {
825 if ( !( $row['content'] instanceof Content ) ) {
826 throw new MWException( 'content field must contain a Content object.' );
827 }
828
829 /** @var Content $content */
830 $content = $row['content'];
831 $handler = $content->getContentHandler();
832
833 $mainSlotRow->model_name = $content->getModel();
834
835 // XXX: in the future, we'll probably always use the default format.
836 if ( $mainSlotRow->format_name === null ) {
837 $mainSlotRow->format_name = $handler->getDefaultFormat();
838 }
839 }
840 } else {
841 throw new MWException( 'Revision constructor passed invalid row format.' );
842 }
843
844 // With the old schema, the content changes with every revision,
845 // except for null-revisions.
846 if ( !isset( $mainSlotRow->slot_origin ) ) {
847 $mainSlotRow->slot_origin = $mainSlotRow->slot_revision_id;
848 }
849
850 if ( $mainSlotRow->model_name === null ) {
851 $mainSlotRow->model_name = function ( SlotRecord $slot ) use ( $title ) {
852 // TODO: MCR: consider slot role in getDefaultModelFor()! Use LinkTarget!
853 // TODO: MCR: deprecate $title->getModel().
854 return ContentHandler::getDefaultModelFor( $title );
855 };
856 }
857
858 if ( !$content ) {
859 $content = function ( SlotRecord $slot )
860 use ( $blobData, $blobFlags, $queryFlags, $mainSlotRow )
861 {
862 return $this->loadSlotContent(
863 $slot,
864 $blobData,
865 $blobFlags,
866 $mainSlotRow->format_name,
867 $queryFlags
868 );
869 };
870 }
871
872 $mainSlotRow->slot_id = $mainSlotRow->slot_revision_id;
873 return new SlotRecord( $mainSlotRow, $content );
874 }
875
876 /**
877 * Loads a Content object based on a slot row.
878 *
879 * This method does not call $slot->getContent(), and may be used as a callback
880 * called by $slot->getContent().
881 *
882 * MCR migration note: this roughly corresponds to Revision::getContentInternal
883 *
884 * @param SlotRecord $slot The SlotRecord to load content for
885 * @param string|null $blobData The content blob, in the form indicated by $blobFlags
886 * @param string|null $blobFlags Flags indicating how $blobData needs to be processed.
887 * Use null if no processing should happen. That is in constrast to the empty string,
888 * which causes the blob to be decoded according to the configured legacy encoding.
889 * @param string|null $blobFormat MIME type indicating how $dataBlob is encoded
890 * @param int $queryFlags
891 *
892 * @throws RevisionAccessException
893 * @return Content
894 */
895 private function loadSlotContent(
896 SlotRecord $slot,
897 $blobData = null,
898 $blobFlags = null,
899 $blobFormat = null,
900 $queryFlags = 0
901 ) {
902 if ( $blobData !== null ) {
903 Assert::parameterType( 'string', $blobData, '$blobData' );
904 Assert::parameterType( 'string|null', $blobFlags, '$blobFlags' );
905
906 $cacheKey = $slot->hasAddress() ? $slot->getAddress() : null;
907
908 if ( $blobFlags === null ) {
909 // No blob flags, so use the blob verbatim.
910 $data = $blobData;
911 } else {
912 $data = $this->blobStore->expandBlob( $blobData, $blobFlags, $cacheKey );
913 if ( $data === false ) {
914 throw new RevisionAccessException(
915 "Failed to expand blob data using flags $blobFlags (key: $cacheKey)"
916 );
917 }
918 }
919
920 } else {
921 $address = $slot->getAddress();
922 try {
923 $data = $this->blobStore->getBlob( $address, $queryFlags );
924 } catch ( BlobAccessException $e ) {
925 throw new RevisionAccessException(
926 "Failed to load data blob from $address: " . $e->getMessage(), 0, $e
927 );
928 }
929 }
930
931 // Unserialize content
932 $handler = ContentHandler::getForModelID( $slot->getModel() );
933
934 $content = $handler->unserializeContent( $data, $blobFormat );
935 return $content;
936 }
937
938 /**
939 * Load a page revision from a given revision ID number.
940 * Returns null if no such revision can be found.
941 *
942 * MCR migration note: this replaces Revision::newFromId
943 *
944 * $flags include:
945 * IDBAccessObject::READ_LATEST: Select the data from the master
946 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
947 *
948 * @param int $id
949 * @param int $flags (optional)
950 * @return RevisionRecord|null
951 */
952 public function getRevisionById( $id, $flags = 0 ) {
953 return $this->newRevisionFromConds( [ 'rev_id' => intval( $id ) ], $flags );
954 }
955
956 /**
957 * Load either the current, or a specified, revision
958 * that's attached to a given link target. If not attached
959 * to that link target, will return null.
960 *
961 * MCR migration note: this replaces Revision::newFromTitle
962 *
963 * $flags include:
964 * IDBAccessObject::READ_LATEST: Select the data from the master
965 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
966 *
967 * @param LinkTarget $linkTarget
968 * @param int $revId (optional)
969 * @param int $flags Bitfield (optional)
970 * @return RevisionRecord|null
971 */
972 public function getRevisionByTitle( LinkTarget $linkTarget, $revId = 0, $flags = 0 ) {
973 $conds = [
974 'page_namespace' => $linkTarget->getNamespace(),
975 'page_title' => $linkTarget->getDBkey()
976 ];
977 if ( $revId ) {
978 // Use the specified revision ID.
979 // Note that we use newRevisionFromConds here because we want to retry
980 // and fall back to master if the page is not found on a replica.
981 // Since the caller supplied a revision ID, we are pretty sure the revision is
982 // supposed to exist, so we should try hard to find it.
983 $conds['rev_id'] = $revId;
984 return $this->newRevisionFromConds( $conds, $flags );
985 } else {
986 // Use a join to get the latest revision.
987 // Note that we don't use newRevisionFromConds here because we don't want to retry
988 // and fall back to master. The assumption is that we only want to force the fallback
989 // if we are quite sure the revision exists because the caller supplied a revision ID.
990 // If the page isn't found at all on a replica, it probably simply does not exist.
991 $db = $this->getDBConnection( ( $flags & self::READ_LATEST ) ? DB_MASTER : DB_REPLICA );
992
993 $conds[] = 'rev_id=page_latest';
994 $rev = $this->loadRevisionFromConds( $db, $conds, $flags );
995
996 $this->releaseDBConnection( $db );
997 return $rev;
998 }
999 }
1000
1001 /**
1002 * Load either the current, or a specified, revision
1003 * that's attached to a given page ID.
1004 * Returns null if no such revision can be found.
1005 *
1006 * MCR migration note: this replaces Revision::newFromPageId
1007 *
1008 * $flags include:
1009 * IDBAccessObject::READ_LATEST: Select the data from the master (since 1.20)
1010 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
1011 *
1012 * @param int $pageId
1013 * @param int $revId (optional)
1014 * @param int $flags Bitfield (optional)
1015 * @return RevisionRecord|null
1016 */
1017 public function getRevisionByPageId( $pageId, $revId = 0, $flags = 0 ) {
1018 $conds = [ 'page_id' => $pageId ];
1019 if ( $revId ) {
1020 // Use the specified revision ID.
1021 // Note that we use newRevisionFromConds here because we want to retry
1022 // and fall back to master if the page is not found on a replica.
1023 // Since the caller supplied a revision ID, we are pretty sure the revision is
1024 // supposed to exist, so we should try hard to find it.
1025 $conds['rev_id'] = $revId;
1026 return $this->newRevisionFromConds( $conds, $flags );
1027 } else {
1028 // Use a join to get the latest revision.
1029 // Note that we don't use newRevisionFromConds here because we don't want to retry
1030 // and fall back to master. The assumption is that we only want to force the fallback
1031 // if we are quite sure the revision exists because the caller supplied a revision ID.
1032 // If the page isn't found at all on a replica, it probably simply does not exist.
1033 $db = $this->getDBConnection( ( $flags & self::READ_LATEST ) ? DB_MASTER : DB_REPLICA );
1034
1035 $conds[] = 'rev_id=page_latest';
1036 $rev = $this->loadRevisionFromConds( $db, $conds, $flags );
1037
1038 $this->releaseDBConnection( $db );
1039 return $rev;
1040 }
1041 }
1042
1043 /**
1044 * Load the revision for the given title with the given timestamp.
1045 * WARNING: Timestamps may in some circumstances not be unique,
1046 * so this isn't the best key to use.
1047 *
1048 * MCR migration note: this replaces Revision::loadFromTimestamp
1049 *
1050 * @param Title $title
1051 * @param string $timestamp
1052 * @return RevisionRecord|null
1053 */
1054 public function getRevisionByTimestamp( $title, $timestamp ) {
1055 $db = $this->getDBConnection( DB_REPLICA );
1056 return $this->newRevisionFromConds(
1057 [
1058 'rev_timestamp' => $db->timestamp( $timestamp ),
1059 'page_namespace' => $title->getNamespace(),
1060 'page_title' => $title->getDBkey()
1061 ],
1062 0,
1063 $title
1064 );
1065 }
1066
1067 /**
1068 * Make a fake revision object from an archive table row. This is queried
1069 * for permissions or even inserted (as in Special:Undelete)
1070 *
1071 * MCR migration note: this replaces Revision::newFromArchiveRow
1072 *
1073 * @param object $row
1074 * @param int $queryFlags
1075 * @param Title|null $title
1076 * @param array $overrides associative array with fields of $row to override. This may be
1077 * used e.g. to force the parent revision ID or page ID. Keys in the array are fields
1078 * names from the archive table without the 'ar_' prefix, i.e. use 'parent_id' to
1079 * override ar_parent_id.
1080 *
1081 * @return RevisionRecord
1082 * @throws MWException
1083 */
1084 public function newRevisionFromArchiveRow(
1085 $row,
1086 $queryFlags = 0,
1087 Title $title = null,
1088 array $overrides = []
1089 ) {
1090 Assert::parameterType( 'object', $row, '$row' );
1091
1092 // check second argument, since Revision::newFromArchiveRow had $overrides in that spot.
1093 Assert::parameterType( 'integer', $queryFlags, '$queryFlags' );
1094
1095 if ( !$title && isset( $overrides['title'] ) ) {
1096 if ( !( $overrides['title'] instanceof Title ) ) {
1097 throw new MWException( 'title field override must contain a Title object.' );
1098 }
1099
1100 $title = $overrides['title'];
1101 }
1102
1103 if ( !isset( $title ) ) {
1104 if ( isset( $row->ar_namespace ) && isset( $row->ar_title ) ) {
1105 $title = Title::makeTitle( $row->ar_namespace, $row->ar_title );
1106 } else {
1107 throw new InvalidArgumentException(
1108 'A Title or ar_namespace and ar_title must be given'
1109 );
1110 }
1111 }
1112
1113 foreach ( $overrides as $key => $value ) {
1114 $field = "ar_$key";
1115 $row->$field = $value;
1116 }
1117
1118 try {
1119 $user = User::newFromAnyId(
1120 $row->ar_user ?? null,
1121 $row->ar_user_text ?? null,
1122 $row->ar_actor ?? null
1123 );
1124 } catch ( InvalidArgumentException $ex ) {
1125 wfWarn( __METHOD__ . ': ' . $ex->getMessage() );
1126 $user = new UserIdentityValue( 0, '', 0 );
1127 }
1128
1129 $comment = $this->commentStore
1130 // Legacy because $row may have come from self::selectFields()
1131 ->getCommentLegacy( $this->getDBConnection( DB_REPLICA ), 'ar_comment', $row, true );
1132
1133 $mainSlot = $this->emulateMainSlot_1_29( $row, $queryFlags, $title );
1134 $slots = new RevisionSlots( [ 'main' => $mainSlot ] );
1135
1136 return new RevisionArchiveRecord( $title, $user, $comment, $row, $slots, $this->wikiId );
1137 }
1138
1139 /**
1140 * @see RevisionFactory::newRevisionFromRow_1_29
1141 *
1142 * MCR migration note: this replaces Revision::newFromRow
1143 *
1144 * @param object $row
1145 * @param int $queryFlags
1146 * @param Title|null $title
1147 *
1148 * @return RevisionRecord
1149 * @throws MWException
1150 * @throws RevisionAccessException
1151 */
1152 private function newRevisionFromRow_1_29( $row, $queryFlags = 0, Title $title = null ) {
1153 Assert::parameterType( 'object', $row, '$row' );
1154
1155 if ( !$title ) {
1156 $pageId = $row->rev_page ?? 0; // XXX: also check page_id?
1157 $revId = $row->rev_id ?? 0;
1158
1159 $title = $this->getTitle( $pageId, $revId, $queryFlags );
1160 }
1161
1162 if ( !isset( $row->page_latest ) ) {
1163 $row->page_latest = $title->getLatestRevID();
1164 if ( $row->page_latest === 0 && $title->exists() ) {
1165 wfWarn( 'Encountered title object in limbo: ID ' . $title->getArticleID() );
1166 }
1167 }
1168
1169 try {
1170 $user = User::newFromAnyId(
1171 $row->rev_user ?? null,
1172 $row->rev_user_text ?? null,
1173 $row->rev_actor ?? null
1174 );
1175 } catch ( InvalidArgumentException $ex ) {
1176 wfWarn( __METHOD__ . ': ' . $ex->getMessage() );
1177 $user = new UserIdentityValue( 0, '', 0 );
1178 }
1179
1180 $comment = $this->commentStore
1181 // Legacy because $row may have come from self::selectFields()
1182 ->getCommentLegacy( $this->getDBConnection( DB_REPLICA ), 'rev_comment', $row, true );
1183
1184 $mainSlot = $this->emulateMainSlot_1_29( $row, $queryFlags, $title );
1185 $slots = new RevisionSlots( [ 'main' => $mainSlot ] );
1186
1187 return new RevisionStoreRecord( $title, $user, $comment, $row, $slots, $this->wikiId );
1188 }
1189
1190 /**
1191 * @see RevisionFactory::newRevisionFromRow
1192 *
1193 * MCR migration note: this replaces Revision::newFromRow
1194 *
1195 * @param object $row
1196 * @param int $queryFlags
1197 * @param Title|null $title
1198 *
1199 * @return RevisionRecord
1200 */
1201 public function newRevisionFromRow( $row, $queryFlags = 0, Title $title = null ) {
1202 return $this->newRevisionFromRow_1_29( $row, $queryFlags, $title );
1203 }
1204
1205 /**
1206 * Constructs a new MutableRevisionRecord based on the given associative array following
1207 * the MW1.29 convention for the Revision constructor.
1208 *
1209 * MCR migration note: this replaces Revision::newFromRow
1210 *
1211 * @param array $fields
1212 * @param int $queryFlags
1213 * @param Title|null $title
1214 *
1215 * @return MutableRevisionRecord
1216 * @throws MWException
1217 * @throws RevisionAccessException
1218 */
1219 public function newMutableRevisionFromArray(
1220 array $fields,
1221 $queryFlags = 0,
1222 Title $title = null
1223 ) {
1224 if ( !$title && isset( $fields['title'] ) ) {
1225 if ( !( $fields['title'] instanceof Title ) ) {
1226 throw new MWException( 'title field must contain a Title object.' );
1227 }
1228
1229 $title = $fields['title'];
1230 }
1231
1232 if ( !$title ) {
1233 $pageId = $fields['page'] ?? 0;
1234 $revId = $fields['id'] ?? 0;
1235
1236 $title = $this->getTitle( $pageId, $revId, $queryFlags );
1237 }
1238
1239 if ( !isset( $fields['page'] ) ) {
1240 $fields['page'] = $title->getArticleID( $queryFlags );
1241 }
1242
1243 // if we have a content object, use it to set the model and type
1244 if ( !empty( $fields['content'] ) ) {
1245 if ( !( $fields['content'] instanceof Content ) ) {
1246 throw new MWException( 'content field must contain a Content object.' );
1247 }
1248
1249 if ( !empty( $fields['text_id'] ) ) {
1250 throw new MWException(
1251 "Text already stored in external store (id {$fields['text_id']}), " .
1252 "can't serialize content object"
1253 );
1254 }
1255 }
1256
1257 if (
1258 isset( $fields['comment'] )
1259 && !( $fields['comment'] instanceof CommentStoreComment )
1260 ) {
1261 $commentData = $fields['comment_data'] ?? null;
1262
1263 if ( $fields['comment'] instanceof Message ) {
1264 $fields['comment'] = CommentStoreComment::newUnsavedComment(
1265 $fields['comment'],
1266 $commentData
1267 );
1268 } else {
1269 $commentText = trim( strval( $fields['comment'] ) );
1270 $fields['comment'] = CommentStoreComment::newUnsavedComment(
1271 $commentText,
1272 $commentData
1273 );
1274 }
1275 }
1276
1277 $mainSlot = $this->emulateMainSlot_1_29( $fields, $queryFlags, $title );
1278
1279 $revision = new MutableRevisionRecord( $title, $this->wikiId );
1280 $this->initializeMutableRevisionFromArray( $revision, $fields );
1281 $revision->setSlot( $mainSlot );
1282
1283 return $revision;
1284 }
1285
1286 /**
1287 * @param MutableRevisionRecord $record
1288 * @param array $fields
1289 */
1290 private function initializeMutableRevisionFromArray(
1291 MutableRevisionRecord $record,
1292 array $fields
1293 ) {
1294 /** @var UserIdentity $user */
1295 $user = null;
1296
1297 if ( isset( $fields['user'] ) && ( $fields['user'] instanceof UserIdentity ) ) {
1298 $user = $fields['user'];
1299 } else {
1300 try {
1301 $user = User::newFromAnyId(
1302 $fields['user'] ?? null,
1303 $fields['user_text'] ?? null,
1304 $fields['actor'] ?? null
1305 );
1306 } catch ( InvalidArgumentException $ex ) {
1307 $user = null;
1308 }
1309 }
1310
1311 if ( $user ) {
1312 $record->setUser( $user );
1313 }
1314
1315 $timestamp = isset( $fields['timestamp'] )
1316 ? strval( $fields['timestamp'] )
1317 : wfTimestampNow(); // TODO: use a callback, so we can override it for testing.
1318
1319 $record->setTimestamp( $timestamp );
1320
1321 if ( isset( $fields['page'] ) ) {
1322 $record->setPageId( intval( $fields['page'] ) );
1323 }
1324
1325 if ( isset( $fields['id'] ) ) {
1326 $record->setId( intval( $fields['id'] ) );
1327 }
1328 if ( isset( $fields['parent_id'] ) ) {
1329 $record->setParentId( intval( $fields['parent_id'] ) );
1330 }
1331
1332 if ( isset( $fields['sha1'] ) ) {
1333 $record->setSha1( $fields['sha1'] );
1334 }
1335 if ( isset( $fields['size'] ) ) {
1336 $record->setSize( intval( $fields['size'] ) );
1337 }
1338
1339 if ( isset( $fields['minor_edit'] ) ) {
1340 $record->setMinorEdit( intval( $fields['minor_edit'] ) !== 0 );
1341 }
1342 if ( isset( $fields['deleted'] ) ) {
1343 $record->setVisibility( intval( $fields['deleted'] ) );
1344 }
1345
1346 if ( isset( $fields['comment'] ) ) {
1347 Assert::parameterType(
1348 CommentStoreComment::class,
1349 $fields['comment'],
1350 '$row[\'comment\']'
1351 );
1352 $record->setComment( $fields['comment'] );
1353 }
1354 }
1355
1356 /**
1357 * Load a page revision from a given revision ID number.
1358 * Returns null if no such revision can be found.
1359 *
1360 * MCR migration note: this corresponds to Revision::loadFromId
1361 *
1362 * @note direct use is deprecated!
1363 * @todo remove when unused! there seem to be no callers of Revision::loadFromId
1364 *
1365 * @param IDatabase $db
1366 * @param int $id
1367 *
1368 * @return RevisionRecord|null
1369 */
1370 public function loadRevisionFromId( IDatabase $db, $id ) {
1371 return $this->loadRevisionFromConds( $db, [ 'rev_id' => intval( $id ) ] );
1372 }
1373
1374 /**
1375 * Load either the current, or a specified, revision
1376 * that's attached to a given page. If not attached
1377 * to that page, will return null.
1378 *
1379 * MCR migration note: this replaces Revision::loadFromPageId
1380 *
1381 * @note direct use is deprecated!
1382 * @todo remove when unused!
1383 *
1384 * @param IDatabase $db
1385 * @param int $pageid
1386 * @param int $id
1387 * @return RevisionRecord|null
1388 */
1389 public function loadRevisionFromPageId( IDatabase $db, $pageid, $id = 0 ) {
1390 $conds = [ 'rev_page' => intval( $pageid ), 'page_id' => intval( $pageid ) ];
1391 if ( $id ) {
1392 $conds['rev_id'] = intval( $id );
1393 } else {
1394 $conds[] = 'rev_id=page_latest';
1395 }
1396 return $this->loadRevisionFromConds( $db, $conds );
1397 }
1398
1399 /**
1400 * Load either the current, or a specified, revision
1401 * that's attached to a given page. If not attached
1402 * to that page, will return null.
1403 *
1404 * MCR migration note: this replaces Revision::loadFromTitle
1405 *
1406 * @note direct use is deprecated!
1407 * @todo remove when unused!
1408 *
1409 * @param IDatabase $db
1410 * @param Title $title
1411 * @param int $id
1412 *
1413 * @return RevisionRecord|null
1414 */
1415 public function loadRevisionFromTitle( IDatabase $db, $title, $id = 0 ) {
1416 if ( $id ) {
1417 $matchId = intval( $id );
1418 } else {
1419 $matchId = 'page_latest';
1420 }
1421
1422 return $this->loadRevisionFromConds(
1423 $db,
1424 [
1425 "rev_id=$matchId",
1426 'page_namespace' => $title->getNamespace(),
1427 'page_title' => $title->getDBkey()
1428 ],
1429 0,
1430 $title
1431 );
1432 }
1433
1434 /**
1435 * Load the revision for the given title with the given timestamp.
1436 * WARNING: Timestamps may in some circumstances not be unique,
1437 * so this isn't the best key to use.
1438 *
1439 * MCR migration note: this replaces Revision::loadFromTimestamp
1440 *
1441 * @note direct use is deprecated! Use getRevisionFromTimestamp instead!
1442 * @todo remove when unused!
1443 *
1444 * @param IDatabase $db
1445 * @param Title $title
1446 * @param string $timestamp
1447 * @return RevisionRecord|null
1448 */
1449 public function loadRevisionFromTimestamp( IDatabase $db, $title, $timestamp ) {
1450 return $this->loadRevisionFromConds( $db,
1451 [
1452 'rev_timestamp' => $db->timestamp( $timestamp ),
1453 'page_namespace' => $title->getNamespace(),
1454 'page_title' => $title->getDBkey()
1455 ],
1456 0,
1457 $title
1458 );
1459 }
1460
1461 /**
1462 * Given a set of conditions, fetch a revision
1463 *
1464 * This method should be used if we are pretty sure the revision exists.
1465 * Unless $flags has READ_LATEST set, this method will first try to find the revision
1466 * on a replica before hitting the master database.
1467 *
1468 * MCR migration note: this corresponds to Revision::newFromConds
1469 *
1470 * @param array $conditions
1471 * @param int $flags (optional)
1472 * @param Title $title
1473 *
1474 * @return RevisionRecord|null
1475 */
1476 private function newRevisionFromConds( $conditions, $flags = 0, Title $title = null ) {
1477 $db = $this->getDBConnection( ( $flags & self::READ_LATEST ) ? DB_MASTER : DB_REPLICA );
1478 $rev = $this->loadRevisionFromConds( $db, $conditions, $flags, $title );
1479 $this->releaseDBConnection( $db );
1480
1481 $lb = $this->getDBLoadBalancer();
1482
1483 // Make sure new pending/committed revision are visibile later on
1484 // within web requests to certain avoid bugs like T93866 and T94407.
1485 if ( !$rev
1486 && !( $flags & self::READ_LATEST )
1487 && $lb->getServerCount() > 1
1488 && $lb->hasOrMadeRecentMasterChanges()
1489 ) {
1490 $flags = self::READ_LATEST;
1491 $db = $this->getDBConnection( DB_MASTER );
1492 $rev = $this->loadRevisionFromConds( $db, $conditions, $flags, $title );
1493 $this->releaseDBConnection( $db );
1494 }
1495
1496 return $rev;
1497 }
1498
1499 /**
1500 * Given a set of conditions, fetch a revision from
1501 * the given database connection.
1502 *
1503 * MCR migration note: this corresponds to Revision::loadFromConds
1504 *
1505 * @param IDatabase $db
1506 * @param array $conditions
1507 * @param int $flags (optional)
1508 * @param Title $title
1509 *
1510 * @return RevisionRecord|null
1511 */
1512 private function loadRevisionFromConds(
1513 IDatabase $db,
1514 $conditions,
1515 $flags = 0,
1516 Title $title = null
1517 ) {
1518 $row = $this->fetchRevisionRowFromConds( $db, $conditions, $flags );
1519 if ( $row ) {
1520 $rev = $this->newRevisionFromRow( $row, $flags, $title );
1521
1522 return $rev;
1523 }
1524
1525 return null;
1526 }
1527
1528 /**
1529 * Throws an exception if the given database connection does not belong to the wiki this
1530 * RevisionStore is bound to.
1531 *
1532 * @param IDatabase $db
1533 * @throws MWException
1534 */
1535 private function checkDatabaseWikiId( IDatabase $db ) {
1536 $storeWiki = $this->wikiId;
1537 $dbWiki = $db->getDomainID();
1538
1539 if ( $dbWiki === $storeWiki ) {
1540 return;
1541 }
1542
1543 // XXX: we really want the default database ID...
1544 $storeWiki = $storeWiki ?: wfWikiID();
1545 $dbWiki = $dbWiki ?: wfWikiID();
1546
1547 if ( $dbWiki === $storeWiki ) {
1548 return;
1549 }
1550
1551 // HACK: counteract encoding imposed by DatabaseDomain
1552 $storeWiki = str_replace( '?h', '-', $storeWiki );
1553 $dbWiki = str_replace( '?h', '-', $dbWiki );
1554
1555 if ( $dbWiki === $storeWiki ) {
1556 return;
1557 }
1558
1559 throw new MWException( "RevisionStore for $storeWiki "
1560 . "cannot be used with a DB connection for $dbWiki" );
1561 }
1562
1563 /**
1564 * Given a set of conditions, return a row with the
1565 * fields necessary to build RevisionRecord objects.
1566 *
1567 * MCR migration note: this corresponds to Revision::fetchFromConds
1568 *
1569 * @param IDatabase $db
1570 * @param array $conditions
1571 * @param int $flags (optional)
1572 *
1573 * @return object|false data row as a raw object
1574 */
1575 private function fetchRevisionRowFromConds( IDatabase $db, $conditions, $flags = 0 ) {
1576 $this->checkDatabaseWikiId( $db );
1577
1578 $revQuery = self::getQueryInfo( [ 'page', 'user' ] );
1579 $options = [];
1580 if ( ( $flags & self::READ_LOCKING ) == self::READ_LOCKING ) {
1581 $options[] = 'FOR UPDATE';
1582 }
1583 return $db->selectRow(
1584 $revQuery['tables'],
1585 $revQuery['fields'],
1586 $conditions,
1587 __METHOD__,
1588 $options,
1589 $revQuery['joins']
1590 );
1591 }
1592
1593 /**
1594 * Return the tables, fields, and join conditions to be selected to create
1595 * a new revision object.
1596 *
1597 * MCR migration note: this replaces Revision::getQueryInfo
1598 *
1599 * @since 1.31
1600 *
1601 * @param array $options Any combination of the following strings
1602 * - 'page': Join with the page table, and select fields to identify the page
1603 * - 'user': Join with the user table, and select the user name
1604 * - 'text': Join with the text table, and select fields to load page text
1605 *
1606 * @return array With three keys:
1607 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
1608 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
1609 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
1610 */
1611 public function getQueryInfo( $options = [] ) {
1612 $ret = [
1613 'tables' => [],
1614 'fields' => [],
1615 'joins' => [],
1616 ];
1617
1618 $ret['tables'][] = 'revision';
1619 $ret['fields'] = array_merge( $ret['fields'], [
1620 'rev_id',
1621 'rev_page',
1622 'rev_text_id',
1623 'rev_timestamp',
1624 'rev_minor_edit',
1625 'rev_deleted',
1626 'rev_len',
1627 'rev_parent_id',
1628 'rev_sha1',
1629 ] );
1630
1631 $commentQuery = $this->commentStore->getJoin( 'rev_comment' );
1632 $ret['tables'] = array_merge( $ret['tables'], $commentQuery['tables'] );
1633 $ret['fields'] = array_merge( $ret['fields'], $commentQuery['fields'] );
1634 $ret['joins'] = array_merge( $ret['joins'], $commentQuery['joins'] );
1635
1636 $actorQuery = $this->actorMigration->getJoin( 'rev_user' );
1637 $ret['tables'] = array_merge( $ret['tables'], $actorQuery['tables'] );
1638 $ret['fields'] = array_merge( $ret['fields'], $actorQuery['fields'] );
1639 $ret['joins'] = array_merge( $ret['joins'], $actorQuery['joins'] );
1640
1641 if ( $this->contentHandlerUseDB ) {
1642 $ret['fields'][] = 'rev_content_format';
1643 $ret['fields'][] = 'rev_content_model';
1644 }
1645
1646 if ( in_array( 'page', $options, true ) ) {
1647 $ret['tables'][] = 'page';
1648 $ret['fields'] = array_merge( $ret['fields'], [
1649 'page_namespace',
1650 'page_title',
1651 'page_id',
1652 'page_latest',
1653 'page_is_redirect',
1654 'page_len',
1655 ] );
1656 $ret['joins']['page'] = [ 'INNER JOIN', [ 'page_id = rev_page' ] ];
1657 }
1658
1659 if ( in_array( 'user', $options, true ) ) {
1660 $ret['tables'][] = 'user';
1661 $ret['fields'] = array_merge( $ret['fields'], [
1662 'user_name',
1663 ] );
1664 $u = $actorQuery['fields']['rev_user'];
1665 $ret['joins']['user'] = [ 'LEFT JOIN', [ "$u != 0", "user_id = $u" ] ];
1666 }
1667
1668 if ( in_array( 'text', $options, true ) ) {
1669 $ret['tables'][] = 'text';
1670 $ret['fields'] = array_merge( $ret['fields'], [
1671 'old_text',
1672 'old_flags'
1673 ] );
1674 $ret['joins']['text'] = [ 'INNER JOIN', [ 'rev_text_id=old_id' ] ];
1675 }
1676
1677 return $ret;
1678 }
1679
1680 /**
1681 * Return the tables, fields, and join conditions to be selected to create
1682 * a new archived revision object.
1683 *
1684 * MCR migration note: this replaces Revision::getArchiveQueryInfo
1685 *
1686 * @since 1.31
1687 *
1688 * @return array With three keys:
1689 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
1690 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
1691 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
1692 */
1693 public function getArchiveQueryInfo() {
1694 $commentQuery = $this->commentStore->getJoin( 'ar_comment' );
1695 $actorQuery = $this->actorMigration->getJoin( 'ar_user' );
1696 $ret = [
1697 'tables' => [ 'archive' ] + $commentQuery['tables'] + $actorQuery['tables'],
1698 'fields' => [
1699 'ar_id',
1700 'ar_page_id',
1701 'ar_namespace',
1702 'ar_title',
1703 'ar_rev_id',
1704 'ar_text_id',
1705 'ar_timestamp',
1706 'ar_minor_edit',
1707 'ar_deleted',
1708 'ar_len',
1709 'ar_parent_id',
1710 'ar_sha1',
1711 ] + $commentQuery['fields'] + $actorQuery['fields'],
1712 'joins' => $commentQuery['joins'] + $actorQuery['joins'],
1713 ];
1714
1715 if ( $this->contentHandlerUseDB ) {
1716 $ret['fields'][] = 'ar_content_format';
1717 $ret['fields'][] = 'ar_content_model';
1718 }
1719
1720 return $ret;
1721 }
1722
1723 /**
1724 * Do a batched query for the sizes of a set of revisions.
1725 *
1726 * MCR migration note: this replaces Revision::getParentLengths
1727 *
1728 * @param int[] $revIds
1729 * @return int[] associative array mapping revision IDs from $revIds to the nominal size
1730 * of the corresponding revision.
1731 */
1732 public function getRevisionSizes( array $revIds ) {
1733 return $this->listRevisionSizes( $this->getDBConnection( DB_REPLICA ), $revIds );
1734 }
1735
1736 /**
1737 * Do a batched query for the sizes of a set of revisions.
1738 *
1739 * MCR migration note: this replaces Revision::getParentLengths
1740 *
1741 * @deprecated use RevisionStore::getRevisionSizes instead.
1742 *
1743 * @param IDatabase $db
1744 * @param int[] $revIds
1745 * @return int[] associative array mapping revision IDs from $revIds to the nominal size
1746 * of the corresponding revision.
1747 */
1748 public function listRevisionSizes( IDatabase $db, array $revIds ) {
1749 $this->checkDatabaseWikiId( $db );
1750
1751 $revLens = [];
1752 if ( !$revIds ) {
1753 return $revLens; // empty
1754 }
1755
1756 $res = $db->select(
1757 'revision',
1758 [ 'rev_id', 'rev_len' ],
1759 [ 'rev_id' => $revIds ],
1760 __METHOD__
1761 );
1762
1763 foreach ( $res as $row ) {
1764 $revLens[$row->rev_id] = intval( $row->rev_len );
1765 }
1766
1767 return $revLens;
1768 }
1769
1770 /**
1771 * Get previous revision for this title
1772 *
1773 * MCR migration note: this replaces Revision::getPrevious
1774 *
1775 * @param RevisionRecord $rev
1776 * @param Title $title if known (optional)
1777 *
1778 * @return RevisionRecord|null
1779 */
1780 public function getPreviousRevision( RevisionRecord $rev, Title $title = null ) {
1781 if ( $title === null ) {
1782 $title = $this->getTitle( $rev->getPageId(), $rev->getId() );
1783 }
1784 $prev = $title->getPreviousRevisionID( $rev->getId() );
1785 if ( $prev ) {
1786 return $this->getRevisionByTitle( $title, $prev );
1787 }
1788 return null;
1789 }
1790
1791 /**
1792 * Get next revision for this title
1793 *
1794 * MCR migration note: this replaces Revision::getNext
1795 *
1796 * @param RevisionRecord $rev
1797 * @param Title $title if known (optional)
1798 *
1799 * @return RevisionRecord|null
1800 */
1801 public function getNextRevision( RevisionRecord $rev, Title $title = null ) {
1802 if ( $title === null ) {
1803 $title = $this->getTitle( $rev->getPageId(), $rev->getId() );
1804 }
1805 $next = $title->getNextRevisionID( $rev->getId() );
1806 if ( $next ) {
1807 return $this->getRevisionByTitle( $title, $next );
1808 }
1809 return null;
1810 }
1811
1812 /**
1813 * Get previous revision Id for this page_id
1814 * This is used to populate rev_parent_id on save
1815 *
1816 * MCR migration note: this corresponds to Revision::getPreviousRevisionId
1817 *
1818 * @param IDatabase $db
1819 * @param RevisionRecord $rev
1820 *
1821 * @return int
1822 */
1823 private function getPreviousRevisionId( IDatabase $db, RevisionRecord $rev ) {
1824 $this->checkDatabaseWikiId( $db );
1825
1826 if ( $rev->getPageId() === null ) {
1827 return 0;
1828 }
1829 # Use page_latest if ID is not given
1830 if ( !$rev->getId() ) {
1831 $prevId = $db->selectField(
1832 'page', 'page_latest',
1833 [ 'page_id' => $rev->getPageId() ],
1834 __METHOD__
1835 );
1836 } else {
1837 $prevId = $db->selectField(
1838 'revision', 'rev_id',
1839 [ 'rev_page' => $rev->getPageId(), 'rev_id < ' . $rev->getId() ],
1840 __METHOD__,
1841 [ 'ORDER BY' => 'rev_id DESC' ]
1842 );
1843 }
1844 return intval( $prevId );
1845 }
1846
1847 /**
1848 * Get rev_timestamp from rev_id, without loading the rest of the row
1849 *
1850 * MCR migration note: this replaces Revision::getTimestampFromId
1851 *
1852 * @param Title $title
1853 * @param int $id
1854 * @param int $flags
1855 * @return string|bool False if not found
1856 */
1857 public function getTimestampFromId( $title, $id, $flags = 0 ) {
1858 $db = $this->getDBConnection(
1859 ( $flags & IDBAccessObject::READ_LATEST ) ? DB_MASTER : DB_REPLICA
1860 );
1861
1862 $conds = [ 'rev_id' => $id ];
1863 $conds['rev_page'] = $title->getArticleID();
1864 $timestamp = $db->selectField( 'revision', 'rev_timestamp', $conds, __METHOD__ );
1865
1866 $this->releaseDBConnection( $db );
1867 return ( $timestamp !== false ) ? wfTimestamp( TS_MW, $timestamp ) : false;
1868 }
1869
1870 /**
1871 * Get count of revisions per page...not very efficient
1872 *
1873 * MCR migration note: this replaces Revision::countByPageId
1874 *
1875 * @param IDatabase $db
1876 * @param int $id Page id
1877 * @return int
1878 */
1879 public function countRevisionsByPageId( IDatabase $db, $id ) {
1880 $this->checkDatabaseWikiId( $db );
1881
1882 $row = $db->selectRow( 'revision',
1883 [ 'revCount' => 'COUNT(*)' ],
1884 [ 'rev_page' => $id ],
1885 __METHOD__
1886 );
1887 if ( $row ) {
1888 return intval( $row->revCount );
1889 }
1890 return 0;
1891 }
1892
1893 /**
1894 * Get count of revisions per page...not very efficient
1895 *
1896 * MCR migration note: this replaces Revision::countByTitle
1897 *
1898 * @param IDatabase $db
1899 * @param Title $title
1900 * @return int
1901 */
1902 public function countRevisionsByTitle( IDatabase $db, $title ) {
1903 $id = $title->getArticleID();
1904 if ( $id ) {
1905 return $this->countRevisionsByPageId( $db, $id );
1906 }
1907 return 0;
1908 }
1909
1910 /**
1911 * Check if no edits were made by other users since
1912 * the time a user started editing the page. Limit to
1913 * 50 revisions for the sake of performance.
1914 *
1915 * MCR migration note: this replaces Revision::userWasLastToEdit
1916 *
1917 * @deprecated since 1.31; Can possibly be removed, since the self-conflict suppression
1918 * logic in EditPage that uses this seems conceptually dubious. Revision::userWasLastToEdit
1919 * has been deprecated since 1.24.
1920 *
1921 * @param IDatabase $db The Database to perform the check on.
1922 * @param int $pageId The ID of the page in question
1923 * @param int $userId The ID of the user in question
1924 * @param string $since Look at edits since this time
1925 *
1926 * @return bool True if the given user was the only one to edit since the given timestamp
1927 */
1928 public function userWasLastToEdit( IDatabase $db, $pageId, $userId, $since ) {
1929 $this->checkDatabaseWikiId( $db );
1930
1931 if ( !$userId ) {
1932 return false;
1933 }
1934
1935 $revQuery = self::getQueryInfo();
1936 $res = $db->select(
1937 $revQuery['tables'],
1938 [
1939 'rev_user' => $revQuery['fields']['rev_user'],
1940 ],
1941 [
1942 'rev_page' => $pageId,
1943 'rev_timestamp > ' . $db->addQuotes( $db->timestamp( $since ) )
1944 ],
1945 __METHOD__,
1946 [ 'ORDER BY' => 'rev_timestamp ASC', 'LIMIT' => 50 ],
1947 $revQuery['joins']
1948 );
1949 foreach ( $res as $row ) {
1950 if ( $row->rev_user != $userId ) {
1951 return false;
1952 }
1953 }
1954 return true;
1955 }
1956
1957 /**
1958 * Load a revision based on a known page ID and current revision ID from the DB
1959 *
1960 * This method allows for the use of caching, though accessing anything that normally
1961 * requires permission checks (aside from the text) will trigger a small DB lookup.
1962 *
1963 * MCR migration note: this replaces Revision::newKnownCurrent
1964 *
1965 * @param Title $title the associated page title
1966 * @param int $revId current revision of this page. Defaults to $title->getLatestRevID().
1967 *
1968 * @return RevisionRecord|bool Returns false if missing
1969 */
1970 public function getKnownCurrentRevision( Title $title, $revId ) {
1971 $db = $this->getDBConnectionRef( DB_REPLICA );
1972
1973 $pageId = $title->getArticleID();
1974
1975 if ( !$pageId ) {
1976 return false;
1977 }
1978
1979 if ( !$revId ) {
1980 $revId = $title->getLatestRevID();
1981 }
1982
1983 if ( !$revId ) {
1984 wfWarn(
1985 'No latest revision known for page ' . $title->getPrefixedDBkey()
1986 . ' even though it exists with page ID ' . $pageId
1987 );
1988 return false;
1989 }
1990
1991 $row = $this->cache->getWithSetCallback(
1992 // Page/rev IDs passed in from DB to reflect history merges
1993 $this->cache->makeGlobalKey( 'revision-row-1.29', $db->getDomainID(), $pageId, $revId ),
1994 WANObjectCache::TTL_WEEK,
1995 function ( $curValue, &$ttl, array &$setOpts ) use ( $db, $pageId, $revId ) {
1996 $setOpts += Database::getCacheSetOptions( $db );
1997
1998 $conds = [
1999 'rev_page' => intval( $pageId ),
2000 'page_id' => intval( $pageId ),
2001 'rev_id' => intval( $revId ),
2002 ];
2003
2004 $row = $this->fetchRevisionRowFromConds( $db, $conds );
2005 return $row ?: false; // don't cache negatives
2006 }
2007 );
2008
2009 // Reflect revision deletion and user renames
2010 if ( $row ) {
2011 return $this->newRevisionFromRow( $row, 0, $title );
2012 } else {
2013 return false;
2014 }
2015 }
2016
2017 // TODO: move relevant methods from Title here, e.g. getFirstRevision, isBigDeletion, etc.
2018
2019 }