Merge "Change Special:Mute link on email footer to be in plain text"
[lhc/web/wiklou.git] / includes / Revision / 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\Revision;
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\Storage\BlobAccessException;
42 use MediaWiki\Storage\BlobStore;
43 use MediaWiki\Storage\NameTableAccessException;
44 use MediaWiki\Storage\NameTableStore;
45 use MediaWiki\Storage\SqlBlobStore;
46 use MediaWiki\User\UserIdentity;
47 use MediaWiki\User\UserIdentityValue;
48 use Message;
49 use MWException;
50 use MWUnknownContentModelException;
51 use Psr\Log\LoggerAwareInterface;
52 use Psr\Log\LoggerInterface;
53 use Psr\Log\NullLogger;
54 use RecentChange;
55 use Revision;
56 use RuntimeException;
57 use stdClass;
58 use Title;
59 use User;
60 use WANObjectCache;
61 use Wikimedia\Assert\Assert;
62 use Wikimedia\Rdbms\Database;
63 use Wikimedia\Rdbms\DBConnRef;
64 use Wikimedia\Rdbms\IDatabase;
65 use Wikimedia\Rdbms\ILoadBalancer;
66 use Wikimedia\Rdbms\ResultWrapper;
67
68 /**
69 * Service for looking up page revisions.
70 *
71 * @since 1.31
72 * @since 1.32 Renamed from MediaWiki\Storage\RevisionStore
73 *
74 * @note This was written to act as a drop-in replacement for the corresponding
75 * static methods in Revision.
76 */
77 class RevisionStore
78 implements IDBAccessObject, RevisionFactory, RevisionLookup, LoggerAwareInterface {
79
80 const ROW_CACHE_KEY = 'revision-row-1.29';
81
82 /**
83 * @var SqlBlobStore
84 */
85 private $blobStore;
86
87 /**
88 * @var bool|string
89 */
90 private $wikiId;
91
92 /**
93 * @var boolean
94 * @see $wgContentHandlerUseDB
95 */
96 private $contentHandlerUseDB = true;
97
98 /**
99 * @var ILoadBalancer
100 */
101 private $loadBalancer;
102
103 /**
104 * @var WANObjectCache
105 */
106 private $cache;
107
108 /**
109 * @var CommentStore
110 */
111 private $commentStore;
112
113 /**
114 * @var ActorMigration
115 */
116 private $actorMigration;
117
118 /**
119 * @var LoggerInterface
120 */
121 private $logger;
122
123 /**
124 * @var NameTableStore
125 */
126 private $contentModelStore;
127
128 /**
129 * @var NameTableStore
130 */
131 private $slotRoleStore;
132
133 /** @var int An appropriate combination of SCHEMA_COMPAT_XXX flags. */
134 private $mcrMigrationStage;
135
136 /** @var SlotRoleRegistry */
137 private $slotRoleRegistry;
138
139 /**
140 * @todo $blobStore should be allowed to be any BlobStore!
141 *
142 * @param ILoadBalancer $loadBalancer
143 * @param SqlBlobStore $blobStore
144 * @param WANObjectCache $cache A cache for caching revision rows. This can be the local
145 * wiki's default instance even if $wikiId refers to a different wiki, since
146 * makeGlobalKey() is used to constructed a key that allows cached revision rows from
147 * the same database to be re-used between wikis. For example, enwiki and frwiki will
148 * use the same cache keys for revision rows from the wikidatawiki database, regardless
149 * of the cache's default key space.
150 * @param CommentStore $commentStore
151 * @param NameTableStore $contentModelStore
152 * @param NameTableStore $slotRoleStore
153 * @param SlotRoleRegistry $slotRoleRegistry
154 * @param int $mcrMigrationStage An appropriate combination of SCHEMA_COMPAT_XXX flags
155 * @param ActorMigration $actorMigration
156 * @param bool|string $wikiId
157 *
158 */
159 public function __construct(
160 ILoadBalancer $loadBalancer,
161 SqlBlobStore $blobStore,
162 WANObjectCache $cache,
163 CommentStore $commentStore,
164 NameTableStore $contentModelStore,
165 NameTableStore $slotRoleStore,
166 SlotRoleRegistry $slotRoleRegistry,
167 $mcrMigrationStage,
168 ActorMigration $actorMigration,
169 $wikiId = false
170 ) {
171 Assert::parameterType( 'string|boolean', $wikiId, '$wikiId' );
172 Assert::parameterType( 'integer', $mcrMigrationStage, '$mcrMigrationStage' );
173 Assert::parameter(
174 ( $mcrMigrationStage & SCHEMA_COMPAT_READ_BOTH ) !== SCHEMA_COMPAT_READ_BOTH,
175 '$mcrMigrationStage',
176 'Reading from the old and the new schema at the same time is not supported.'
177 );
178 Assert::parameter(
179 ( $mcrMigrationStage & SCHEMA_COMPAT_READ_BOTH ) !== 0,
180 '$mcrMigrationStage',
181 'Reading needs to be enabled for the old or the new schema.'
182 );
183 Assert::parameter(
184 ( $mcrMigrationStage & SCHEMA_COMPAT_WRITE_BOTH ) !== 0,
185 '$mcrMigrationStage',
186 'Writing needs to be enabled for the old or the new schema.'
187 );
188 Assert::parameter(
189 ( $mcrMigrationStage & SCHEMA_COMPAT_READ_OLD ) === 0
190 || ( $mcrMigrationStage & SCHEMA_COMPAT_WRITE_OLD ) !== 0,
191 '$mcrMigrationStage',
192 'Cannot read the old schema when not also writing it.'
193 );
194 Assert::parameter(
195 ( $mcrMigrationStage & SCHEMA_COMPAT_READ_NEW ) === 0
196 || ( $mcrMigrationStage & SCHEMA_COMPAT_WRITE_NEW ) !== 0,
197 '$mcrMigrationStage',
198 'Cannot read the new schema when not also writing it.'
199 );
200
201 $this->loadBalancer = $loadBalancer;
202 $this->blobStore = $blobStore;
203 $this->cache = $cache;
204 $this->commentStore = $commentStore;
205 $this->contentModelStore = $contentModelStore;
206 $this->slotRoleStore = $slotRoleStore;
207 $this->slotRoleRegistry = $slotRoleRegistry;
208 $this->mcrMigrationStage = $mcrMigrationStage;
209 $this->actorMigration = $actorMigration;
210 $this->wikiId = $wikiId;
211 $this->logger = new NullLogger();
212 }
213
214 /**
215 * @param int $flags A combination of SCHEMA_COMPAT_XXX flags.
216 * @return bool True if all the given flags were set in the $mcrMigrationStage
217 * parameter passed to the constructor.
218 */
219 private function hasMcrSchemaFlags( $flags ) {
220 return ( $this->mcrMigrationStage & $flags ) === $flags;
221 }
222
223 /**
224 * Throws a RevisionAccessException if this RevisionStore is configured for cross-wiki loading
225 * and still reading from the old DB schema.
226 *
227 * @throws RevisionAccessException
228 */
229 private function assertCrossWikiContentLoadingIsSafe() {
230 if ( $this->wikiId !== false && $this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_OLD ) ) {
231 throw new RevisionAccessException(
232 "Cross-wiki content loading is not supported by the pre-MCR schema"
233 );
234 }
235 }
236
237 public function setLogger( LoggerInterface $logger ) {
238 $this->logger = $logger;
239 }
240
241 /**
242 * @return bool Whether the store is read-only
243 */
244 public function isReadOnly() {
245 return $this->blobStore->isReadOnly();
246 }
247
248 /**
249 * @return bool
250 */
251 public function getContentHandlerUseDB() {
252 return $this->contentHandlerUseDB;
253 }
254
255 /**
256 * @see $wgContentHandlerUseDB
257 * @param bool $contentHandlerUseDB
258 * @throws MWException
259 */
260 public function setContentHandlerUseDB( $contentHandlerUseDB ) {
261 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW )
262 || $this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_NEW )
263 ) {
264 if ( !$contentHandlerUseDB ) {
265 throw new MWException(
266 'Content model must be stored in the database for multi content revision migration.'
267 );
268 }
269 }
270 $this->contentHandlerUseDB = $contentHandlerUseDB;
271 }
272
273 /**
274 * @return ILoadBalancer
275 */
276 private function getDBLoadBalancer() {
277 return $this->loadBalancer;
278 }
279
280 /**
281 * @param int $mode DB_MASTER or DB_REPLICA
282 * @param array $groups
283 *
284 * @return IDatabase
285 */
286 private function getDBConnection( $mode, $groups = [] ) {
287 $lb = $this->getDBLoadBalancer();
288 return $lb->getConnection( $mode, $groups, $this->wikiId );
289 }
290
291 /**
292 * @param int $queryFlags a bit field composed of READ_XXX flags
293 *
294 * @return DBConnRef
295 */
296 private function getDBConnectionRefForQueryFlags( $queryFlags ) {
297 list( $mode, ) = DBAccessObjectUtils::getDBOptions( $queryFlags );
298 return $this->getDBConnectionRef( $mode );
299 }
300
301 /**
302 * @param IDatabase $connection
303 */
304 private function releaseDBConnection( IDatabase $connection ) {
305 $lb = $this->getDBLoadBalancer();
306 $lb->reuseConnection( $connection );
307 }
308
309 /**
310 * @param int $mode DB_MASTER or DB_REPLICA
311 *
312 * @return DBConnRef
313 */
314 private function getDBConnectionRef( $mode ) {
315 $lb = $this->getDBLoadBalancer();
316 return $lb->getConnectionRef( $mode, [], $this->wikiId );
317 }
318
319 /**
320 * Determines the page Title based on the available information.
321 *
322 * MCR migration note: this corresponds to Revision::getTitle
323 *
324 * @note this method should be private, external use should be avoided!
325 *
326 * @param int|null $pageId
327 * @param int|null $revId
328 * @param int $queryFlags
329 *
330 * @return Title
331 * @throws RevisionAccessException
332 */
333 public function getTitle( $pageId, $revId, $queryFlags = self::READ_NORMAL ) {
334 if ( !$pageId && !$revId ) {
335 throw new InvalidArgumentException( '$pageId and $revId cannot both be 0 or null' );
336 }
337
338 // This method recalls itself with READ_LATEST if READ_NORMAL doesn't get us a Title
339 // So ignore READ_LATEST_IMMUTABLE flags and handle the fallback logic in this method
340 if ( DBAccessObjectUtils::hasFlags( $queryFlags, self::READ_LATEST_IMMUTABLE ) ) {
341 $queryFlags = self::READ_NORMAL;
342 }
343
344 $canUseTitleNewFromId = ( $pageId !== null && $pageId > 0 && $this->wikiId === false );
345 list( $dbMode, $dbOptions ) = DBAccessObjectUtils::getDBOptions( $queryFlags );
346 $titleFlags = ( $dbMode == DB_MASTER ? Title::GAID_FOR_UPDATE : 0 );
347
348 // Loading by ID is best, but Title::newFromID does not support that for foreign IDs.
349 if ( $canUseTitleNewFromId ) {
350 // TODO: better foreign title handling (introduce TitleFactory)
351 $title = Title::newFromID( $pageId, $titleFlags );
352 if ( $title ) {
353 return $title;
354 }
355 }
356
357 // rev_id is defined as NOT NULL, but this revision may not yet have been inserted.
358 $canUseRevId = ( $revId !== null && $revId > 0 );
359
360 if ( $canUseRevId ) {
361 $dbr = $this->getDBConnectionRef( $dbMode );
362 // @todo: Title::getSelectFields(), or Title::getQueryInfo(), or something like that
363 $row = $dbr->selectRow(
364 [ 'revision', 'page' ],
365 [
366 'page_namespace',
367 'page_title',
368 'page_id',
369 'page_latest',
370 'page_is_redirect',
371 'page_len',
372 ],
373 [ 'rev_id' => $revId ],
374 __METHOD__,
375 $dbOptions,
376 [ 'page' => [ 'JOIN', 'page_id=rev_page' ] ]
377 );
378 if ( $row ) {
379 // TODO: better foreign title handling (introduce TitleFactory)
380 return Title::newFromRow( $row );
381 }
382 }
383
384 // If we still don't have a title, fallback to master if that wasn't already happening.
385 if ( $dbMode !== DB_MASTER ) {
386 $title = $this->getTitle( $pageId, $revId, self::READ_LATEST );
387 if ( $title ) {
388 $this->logger->info(
389 __METHOD__ . ' fell back to READ_LATEST and got a Title.',
390 [ 'trace' => wfBacktrace() ]
391 );
392 return $title;
393 }
394 }
395
396 throw new RevisionAccessException(
397 "Could not determine title for page ID $pageId and revision ID $revId"
398 );
399 }
400
401 /**
402 * @param mixed $value
403 * @param string $name
404 *
405 * @throws IncompleteRevisionException if $value is null
406 * @return mixed $value, if $value is not null
407 */
408 private function failOnNull( $value, $name ) {
409 if ( $value === null ) {
410 throw new IncompleteRevisionException(
411 "$name must not be " . var_export( $value, true ) . "!"
412 );
413 }
414
415 return $value;
416 }
417
418 /**
419 * @param mixed $value
420 * @param string $name
421 *
422 * @throws IncompleteRevisionException if $value is empty
423 * @return mixed $value, if $value is not null
424 */
425 private function failOnEmpty( $value, $name ) {
426 if ( $value === null || $value === 0 || $value === '' ) {
427 throw new IncompleteRevisionException(
428 "$name must not be " . var_export( $value, true ) . "!"
429 );
430 }
431
432 return $value;
433 }
434
435 /**
436 * Insert a new revision into the database, returning the new revision record
437 * on success and dies horribly on failure.
438 *
439 * MCR migration note: this replaces Revision::insertOn
440 *
441 * @param RevisionRecord $rev
442 * @param IDatabase $dbw (master connection)
443 *
444 * @throws InvalidArgumentException
445 * @return RevisionRecord the new revision record.
446 */
447 public function insertRevisionOn( RevisionRecord $rev, IDatabase $dbw ) {
448 // TODO: pass in a DBTransactionContext instead of a database connection.
449 $this->checkDatabaseWikiId( $dbw );
450
451 $slotRoles = $rev->getSlotRoles();
452
453 // Make sure the main slot is always provided throughout migration
454 if ( !in_array( SlotRecord::MAIN, $slotRoles ) ) {
455 throw new InvalidArgumentException(
456 'main slot must be provided'
457 );
458 }
459
460 // If we are not writing into the new schema, we can't support extra slots.
461 if ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW )
462 && $slotRoles !== [ SlotRecord::MAIN ]
463 ) {
464 throw new InvalidArgumentException(
465 'Only the main slot is supported when not writing to the MCR enabled schema!'
466 );
467 }
468
469 // As long as we are not reading from the new schema, we don't want to write extra slots.
470 if ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_NEW )
471 && $slotRoles !== [ SlotRecord::MAIN ]
472 ) {
473 throw new InvalidArgumentException(
474 'Only the main slot is supported when not reading from the MCR enabled schema!'
475 );
476 }
477
478 // Checks
479 $this->failOnNull( $rev->getSize(), 'size field' );
480 $this->failOnEmpty( $rev->getSha1(), 'sha1 field' );
481 $this->failOnEmpty( $rev->getTimestamp(), 'timestamp field' );
482 $comment = $this->failOnNull( $rev->getComment( RevisionRecord::RAW ), 'comment' );
483 $user = $this->failOnNull( $rev->getUser( RevisionRecord::RAW ), 'user' );
484 $this->failOnNull( $user->getId(), 'user field' );
485 $this->failOnEmpty( $user->getName(), 'user_text field' );
486
487 if ( !$rev->isReadyForInsertion() ) {
488 // This is here for future-proofing. At the time this check being added, it
489 // was redundant to the individual checks above.
490 throw new IncompleteRevisionException( 'Revision is incomplete' );
491 }
492
493 // TODO: we shouldn't need an actual Title here.
494 $title = Title::newFromLinkTarget( $rev->getPageAsLinkTarget() );
495 $pageId = $this->failOnEmpty( $rev->getPageId(), 'rev_page field' ); // check this early
496
497 $parentId = $rev->getParentId() === null
498 ? $this->getPreviousRevisionId( $dbw, $rev )
499 : $rev->getParentId();
500
501 /** @var RevisionRecord $rev */
502 $rev = $dbw->doAtomicSection(
503 __METHOD__,
504 function ( IDatabase $dbw, $fname ) use (
505 $rev,
506 $user,
507 $comment,
508 $title,
509 $pageId,
510 $parentId
511 ) {
512 return $this->insertRevisionInternal(
513 $rev,
514 $dbw,
515 $user,
516 $comment,
517 $title,
518 $pageId,
519 $parentId
520 );
521 }
522 );
523
524 // sanity checks
525 Assert::postcondition( $rev->getId() > 0, 'revision must have an ID' );
526 Assert::postcondition( $rev->getPageId() > 0, 'revision must have a page ID' );
527 Assert::postcondition(
528 $rev->getComment( RevisionRecord::RAW ) !== null,
529 'revision must have a comment'
530 );
531 Assert::postcondition(
532 $rev->getUser( RevisionRecord::RAW ) !== null,
533 'revision must have a user'
534 );
535
536 // Trigger exception if the main slot is missing.
537 // Technically, this could go away after MCR migration: while
538 // calling code may require a main slot to exist, RevisionStore
539 // really should not know or care about that requirement.
540 $rev->getSlot( SlotRecord::MAIN, RevisionRecord::RAW );
541
542 foreach ( $slotRoles as $role ) {
543 $slot = $rev->getSlot( $role, RevisionRecord::RAW );
544 Assert::postcondition(
545 $slot->getContent() !== null,
546 $role . ' slot must have content'
547 );
548 Assert::postcondition(
549 $slot->hasRevision(),
550 $role . ' slot must have a revision associated'
551 );
552 }
553
554 Hooks::run( 'RevisionRecordInserted', [ $rev ] );
555
556 // TODO: deprecate in 1.32!
557 $legacyRevision = new Revision( $rev );
558 Hooks::run( 'RevisionInsertComplete', [ &$legacyRevision, null, null ] );
559
560 return $rev;
561 }
562
563 private function insertRevisionInternal(
564 RevisionRecord $rev,
565 IDatabase $dbw,
566 User $user,
567 CommentStoreComment $comment,
568 Title $title,
569 $pageId,
570 $parentId
571 ) {
572 $slotRoles = $rev->getSlotRoles();
573
574 $revisionRow = $this->insertRevisionRowOn(
575 $dbw,
576 $rev,
577 $title,
578 $parentId
579 );
580
581 $revisionId = $revisionRow['rev_id'];
582
583 $blobHints = [
584 BlobStore::PAGE_HINT => $pageId,
585 BlobStore::REVISION_HINT => $revisionId,
586 BlobStore::PARENT_HINT => $parentId,
587 ];
588
589 $newSlots = [];
590 foreach ( $slotRoles as $role ) {
591 $slot = $rev->getSlot( $role, RevisionRecord::RAW );
592
593 // If the SlotRecord already has a revision ID set, this means it already exists
594 // in the database, and should already belong to the current revision.
595 // However, a slot may already have a revision, but no content ID, if the slot
596 // is emulated based on the archive table, because we are in SCHEMA_COMPAT_READ_OLD
597 // mode, and the respective archive row was not yet migrated to the new schema.
598 // In that case, a new slot row (and content row) must be inserted even during
599 // undeletion.
600 if ( $slot->hasRevision() && $slot->hasContentId() ) {
601 // TODO: properly abort transaction if the assertion fails!
602 Assert::parameter(
603 $slot->getRevision() === $revisionId,
604 'slot role ' . $slot->getRole(),
605 'Existing slot should belong to revision '
606 . $revisionId . ', but belongs to revision ' . $slot->getRevision() . '!'
607 );
608
609 // Slot exists, nothing to do, move along.
610 // This happens when restoring archived revisions.
611
612 $newSlots[$role] = $slot;
613
614 // Write the main slot's text ID to the revision table for backwards compatibility
615 if ( $slot->getRole() === SlotRecord::MAIN
616 && $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_OLD )
617 ) {
618 $blobAddress = $slot->getAddress();
619 $this->updateRevisionTextId( $dbw, $revisionId, $blobAddress );
620 }
621 } else {
622 $newSlots[$role] = $this->insertSlotOn( $dbw, $revisionId, $slot, $title, $blobHints );
623 }
624 }
625
626 $this->insertIpChangesRow( $dbw, $user, $rev, $revisionId );
627
628 $rev = new RevisionStoreRecord(
629 $title,
630 $user,
631 $comment,
632 (object)$revisionRow,
633 new RevisionSlots( $newSlots ),
634 $this->wikiId
635 );
636
637 return $rev;
638 }
639
640 /**
641 * @param IDatabase $dbw
642 * @param int $revisionId
643 * @param string &$blobAddress (may change!)
644 *
645 * @return int the text row id
646 */
647 private function updateRevisionTextId( IDatabase $dbw, $revisionId, &$blobAddress ) {
648 $textId = $this->blobStore->getTextIdFromAddress( $blobAddress );
649 if ( !$textId ) {
650 throw new LogicException(
651 'Blob address not supported in 1.29 database schema: ' . $blobAddress
652 );
653 }
654
655 // getTextIdFromAddress() is free to insert something into the text table, so $textId
656 // may be a new value, not anything already contained in $blobAddress.
657 $blobAddress = SqlBlobStore::makeAddressFromTextId( $textId );
658
659 $dbw->update(
660 'revision',
661 [ 'rev_text_id' => $textId ],
662 [ 'rev_id' => $revisionId ],
663 __METHOD__
664 );
665
666 return $textId;
667 }
668
669 /**
670 * @param IDatabase $dbw
671 * @param int $revisionId
672 * @param SlotRecord $protoSlot
673 * @param Title $title
674 * @param array $blobHints See the BlobStore::XXX_HINT constants
675 * @return SlotRecord
676 */
677 private function insertSlotOn(
678 IDatabase $dbw,
679 $revisionId,
680 SlotRecord $protoSlot,
681 Title $title,
682 array $blobHints = []
683 ) {
684 if ( $protoSlot->hasAddress() ) {
685 $blobAddress = $protoSlot->getAddress();
686 } else {
687 $blobAddress = $this->storeContentBlob( $protoSlot, $title, $blobHints );
688 }
689
690 $contentId = null;
691
692 // Write the main slot's text ID to the revision table for backwards compatibility
693 if ( $protoSlot->getRole() === SlotRecord::MAIN
694 && $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_OLD )
695 ) {
696 // If SCHEMA_COMPAT_WRITE_NEW is also set, the fake content ID is overwritten
697 // with the real content ID below.
698 $textId = $this->updateRevisionTextId( $dbw, $revisionId, $blobAddress );
699 $contentId = $this->emulateContentId( $textId );
700 }
701
702 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW ) ) {
703 if ( $protoSlot->hasContentId() ) {
704 $contentId = $protoSlot->getContentId();
705 } else {
706 $contentId = $this->insertContentRowOn( $protoSlot, $dbw, $blobAddress );
707 }
708
709 $this->insertSlotRowOn( $protoSlot, $dbw, $revisionId, $contentId );
710 }
711
712 $savedSlot = SlotRecord::newSaved(
713 $revisionId,
714 $contentId,
715 $blobAddress,
716 $protoSlot
717 );
718
719 return $savedSlot;
720 }
721
722 /**
723 * Insert IP revision into ip_changes for use when querying for a range.
724 * @param IDatabase $dbw
725 * @param User $user
726 * @param RevisionRecord $rev
727 * @param int $revisionId
728 */
729 private function insertIpChangesRow(
730 IDatabase $dbw,
731 User $user,
732 RevisionRecord $rev,
733 $revisionId
734 ) {
735 if ( $user->getId() === 0 && IP::isValid( $user->getName() ) ) {
736 $ipcRow = [
737 'ipc_rev_id' => $revisionId,
738 'ipc_rev_timestamp' => $dbw->timestamp( $rev->getTimestamp() ),
739 'ipc_hex' => IP::toHex( $user->getName() ),
740 ];
741 $dbw->insert( 'ip_changes', $ipcRow, __METHOD__ );
742 }
743 }
744
745 /**
746 * @param IDatabase $dbw
747 * @param RevisionRecord $rev
748 * @param Title $title
749 * @param int $parentId
750 *
751 * @return array a revision table row
752 *
753 * @throws MWException
754 * @throws MWUnknownContentModelException
755 */
756 private function insertRevisionRowOn(
757 IDatabase $dbw,
758 RevisionRecord $rev,
759 Title $title,
760 $parentId
761 ) {
762 $revisionRow = $this->getBaseRevisionRow( $dbw, $rev, $title, $parentId );
763
764 list( $commentFields, $commentCallback ) =
765 $this->commentStore->insertWithTempTable(
766 $dbw,
767 'rev_comment',
768 $rev->getComment( RevisionRecord::RAW )
769 );
770 $revisionRow += $commentFields;
771
772 list( $actorFields, $actorCallback ) =
773 $this->actorMigration->getInsertValuesWithTempTable(
774 $dbw,
775 'rev_user',
776 $rev->getUser( RevisionRecord::RAW )
777 );
778 $revisionRow += $actorFields;
779
780 $dbw->insert( 'revision', $revisionRow, __METHOD__ );
781
782 if ( !isset( $revisionRow['rev_id'] ) ) {
783 // only if auto-increment was used
784 $revisionRow['rev_id'] = intval( $dbw->insertId() );
785
786 if ( $dbw->getType() === 'mysql' ) {
787 // (T202032) MySQL until 8.0 and MariaDB until some version after 10.1.34 don't save the
788 // auto-increment value to disk, so on server restart it might reuse IDs from deleted
789 // revisions. We can fix that with an insert with an explicit rev_id value, if necessary.
790
791 $maxRevId = intval( $dbw->selectField( 'archive', 'MAX(ar_rev_id)', '', __METHOD__ ) );
792 $table = 'archive';
793 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW ) ) {
794 $maxRevId2 = intval( $dbw->selectField( 'slots', 'MAX(slot_revision_id)', '', __METHOD__ ) );
795 if ( $maxRevId2 >= $maxRevId ) {
796 $maxRevId = $maxRevId2;
797 $table = 'slots';
798 }
799 }
800
801 if ( $maxRevId >= $revisionRow['rev_id'] ) {
802 $this->logger->debug(
803 '__METHOD__: Inserted revision {revid} but {table} has revisions up to {maxrevid}.'
804 . ' Trying to fix it.',
805 [
806 'revid' => $revisionRow['rev_id'],
807 'table' => $table,
808 'maxrevid' => $maxRevId,
809 ]
810 );
811
812 if ( !$dbw->lock( 'fix-for-T202032', __METHOD__ ) ) {
813 throw new MWException( 'Failed to get database lock for T202032' );
814 }
815 $fname = __METHOD__;
816 $dbw->onTransactionResolution( function ( $trigger, $dbw ) use ( $fname ) {
817 $dbw->unlock( 'fix-for-T202032', $fname );
818 } );
819
820 $dbw->delete( 'revision', [ 'rev_id' => $revisionRow['rev_id'] ], __METHOD__ );
821
822 // The locking here is mostly to make MySQL bypass the REPEATABLE-READ transaction
823 // isolation (weird MySQL "feature"). It does seem to block concurrent auto-incrementing
824 // inserts too, though, at least on MariaDB 10.1.29.
825 //
826 // Don't try to lock `revision` in this way, it'll deadlock if there are concurrent
827 // transactions in this code path thanks to the row lock from the original ->insert() above.
828 //
829 // And we have to use raw SQL to bypass the "aggregation used with a locking SELECT" warning
830 // that's for non-MySQL DBs.
831 $row1 = $dbw->query(
832 $dbw->selectSQLText( 'archive', [ 'v' => "MAX(ar_rev_id)" ], '', __METHOD__ ) . ' FOR UPDATE'
833 )->fetchObject();
834 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW ) ) {
835 $row2 = $dbw->query(
836 $dbw->selectSQLText( 'slots', [ 'v' => "MAX(slot_revision_id)" ], '', __METHOD__ )
837 . ' FOR UPDATE'
838 )->fetchObject();
839 } else {
840 $row2 = null;
841 }
842 $maxRevId = max(
843 $maxRevId,
844 $row1 ? intval( $row1->v ) : 0,
845 $row2 ? intval( $row2->v ) : 0
846 );
847
848 // If we don't have SCHEMA_COMPAT_WRITE_NEW, all except the first of any concurrent
849 // transactions will throw a duplicate key error here. It doesn't seem worth trying
850 // to avoid that.
851 $revisionRow['rev_id'] = $maxRevId + 1;
852 $dbw->insert( 'revision', $revisionRow, __METHOD__ );
853 }
854 }
855 }
856
857 $commentCallback( $revisionRow['rev_id'] );
858 $actorCallback( $revisionRow['rev_id'], $revisionRow );
859
860 return $revisionRow;
861 }
862
863 /**
864 * @param IDatabase $dbw
865 * @param RevisionRecord $rev
866 * @param Title $title
867 * @param int $parentId
868 *
869 * @return array [ 0 => array $revisionRow, 1 => callable ]
870 * @throws MWException
871 * @throws MWUnknownContentModelException
872 */
873 private function getBaseRevisionRow(
874 IDatabase $dbw,
875 RevisionRecord $rev,
876 Title $title,
877 $parentId
878 ) {
879 // Record the edit in revisions
880 $revisionRow = [
881 'rev_page' => $rev->getPageId(),
882 'rev_parent_id' => $parentId,
883 'rev_minor_edit' => $rev->isMinor() ? 1 : 0,
884 'rev_timestamp' => $dbw->timestamp( $rev->getTimestamp() ),
885 'rev_deleted' => $rev->getVisibility(),
886 'rev_len' => $rev->getSize(),
887 'rev_sha1' => $rev->getSha1(),
888 ];
889
890 if ( $rev->getId() !== null ) {
891 // Needed to restore revisions with their original ID
892 $revisionRow['rev_id'] = $rev->getId();
893 }
894
895 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_OLD ) ) {
896 // In non MCR mode this IF section will relate to the main slot
897 $mainSlot = $rev->getSlot( SlotRecord::MAIN );
898 $model = $mainSlot->getModel();
899 $format = $mainSlot->getFormat();
900
901 // MCR migration note: rev_content_model and rev_content_format will go away
902 if ( $this->contentHandlerUseDB ) {
903 $this->assertCrossWikiContentLoadingIsSafe();
904
905 $defaultModel = ContentHandler::getDefaultModelFor( $title );
906 $defaultFormat = ContentHandler::getForModelID( $defaultModel )->getDefaultFormat();
907
908 $revisionRow['rev_content_model'] = ( $model === $defaultModel ) ? null : $model;
909 $revisionRow['rev_content_format'] = ( $format === $defaultFormat ) ? null : $format;
910 }
911 }
912
913 return $revisionRow;
914 }
915
916 /**
917 * @param SlotRecord $slot
918 * @param Title $title
919 * @param array $blobHints See the BlobStore::XXX_HINT constants
920 *
921 * @throws MWException
922 * @return string the blob address
923 */
924 private function storeContentBlob(
925 SlotRecord $slot,
926 Title $title,
927 array $blobHints = []
928 ) {
929 $content = $slot->getContent();
930 $format = $content->getDefaultFormat();
931 $model = $content->getModel();
932
933 $this->checkContent( $content, $title, $slot->getRole() );
934
935 return $this->blobStore->storeBlob(
936 $content->serialize( $format ),
937 // These hints "leak" some information from the higher abstraction layer to
938 // low level storage to allow for optimization.
939 array_merge(
940 $blobHints,
941 [
942 BlobStore::DESIGNATION_HINT => 'page-content',
943 BlobStore::ROLE_HINT => $slot->getRole(),
944 BlobStore::SHA1_HINT => $slot->getSha1(),
945 BlobStore::MODEL_HINT => $model,
946 BlobStore::FORMAT_HINT => $format,
947 ]
948 )
949 );
950 }
951
952 /**
953 * @param SlotRecord $slot
954 * @param IDatabase $dbw
955 * @param int $revisionId
956 * @param int $contentId
957 */
958 private function insertSlotRowOn( SlotRecord $slot, IDatabase $dbw, $revisionId, $contentId ) {
959 $slotRow = [
960 'slot_revision_id' => $revisionId,
961 'slot_role_id' => $this->slotRoleStore->acquireId( $slot->getRole() ),
962 'slot_content_id' => $contentId,
963 // If the slot has a specific origin use that ID, otherwise use the ID of the revision
964 // that we just inserted.
965 'slot_origin' => $slot->hasOrigin() ? $slot->getOrigin() : $revisionId,
966 ];
967 $dbw->insert( 'slots', $slotRow, __METHOD__ );
968 }
969
970 /**
971 * @param SlotRecord $slot
972 * @param IDatabase $dbw
973 * @param string $blobAddress
974 * @return int content row ID
975 */
976 private function insertContentRowOn( SlotRecord $slot, IDatabase $dbw, $blobAddress ) {
977 $contentRow = [
978 'content_size' => $slot->getSize(),
979 'content_sha1' => $slot->getSha1(),
980 'content_model' => $this->contentModelStore->acquireId( $slot->getModel() ),
981 'content_address' => $blobAddress,
982 ];
983 $dbw->insert( 'content', $contentRow, __METHOD__ );
984 return intval( $dbw->insertId() );
985 }
986
987 /**
988 * MCR migration note: this corresponds to Revision::checkContentModel
989 *
990 * @param Content $content
991 * @param Title $title
992 * @param string $role
993 *
994 * @throws MWException
995 * @throws MWUnknownContentModelException
996 */
997 private function checkContent( Content $content, Title $title, $role ) {
998 // Note: may return null for revisions that have not yet been inserted
999
1000 $model = $content->getModel();
1001 $format = $content->getDefaultFormat();
1002 $handler = $content->getContentHandler();
1003
1004 $name = "$title";
1005
1006 if ( !$handler->isSupportedFormat( $format ) ) {
1007 throw new MWException( "Can't use format $format with content model $model on $name" );
1008 }
1009
1010 if ( !$this->contentHandlerUseDB ) {
1011 // if $wgContentHandlerUseDB is not set,
1012 // all revisions must use the default content model and format.
1013
1014 $this->assertCrossWikiContentLoadingIsSafe();
1015
1016 $roleHandler = $this->slotRoleRegistry->getRoleHandler( $role );
1017 $defaultModel = $roleHandler->getDefaultModel( $title );
1018 $defaultHandler = ContentHandler::getForModelID( $defaultModel );
1019 $defaultFormat = $defaultHandler->getDefaultFormat();
1020
1021 if ( $model != $defaultModel ) {
1022 throw new MWException( "Can't save non-default content model with "
1023 . "\$wgContentHandlerUseDB disabled: model is $model, "
1024 . "default for $name is $defaultModel"
1025 );
1026 }
1027
1028 if ( $format != $defaultFormat ) {
1029 throw new MWException( "Can't use non-default content format with "
1030 . "\$wgContentHandlerUseDB disabled: format is $format, "
1031 . "default for $name is $defaultFormat"
1032 );
1033 }
1034 }
1035
1036 if ( !$content->isValid() ) {
1037 throw new MWException(
1038 "New content for $name is not valid! Content model is $model"
1039 );
1040 }
1041 }
1042
1043 /**
1044 * Create a new null-revision for insertion into a page's
1045 * history. This will not re-save the text, but simply refer
1046 * to the text from the previous version.
1047 *
1048 * Such revisions can for instance identify page rename
1049 * operations and other such meta-modifications.
1050 *
1051 * @note This method grabs a FOR UPDATE lock on the relevant row of the page table,
1052 * to prevent a new revision from being inserted before the null revision has been written
1053 * to the database.
1054 *
1055 * MCR migration note: this replaces Revision::newNullRevision
1056 *
1057 * @todo Introduce newFromParentRevision(). newNullRevision can then be based on that
1058 * (or go away).
1059 *
1060 * @param IDatabase $dbw used for obtaining the lock on the page table row
1061 * @param Title $title Title of the page to read from
1062 * @param CommentStoreComment $comment RevisionRecord's summary
1063 * @param bool $minor Whether the revision should be considered as minor
1064 * @param User $user The user to attribute the revision to
1065 *
1066 * @return RevisionRecord|null RevisionRecord or null on error
1067 */
1068 public function newNullRevision(
1069 IDatabase $dbw,
1070 Title $title,
1071 CommentStoreComment $comment,
1072 $minor,
1073 User $user
1074 ) {
1075 $this->checkDatabaseWikiId( $dbw );
1076
1077 $pageId = $title->getArticleID();
1078
1079 // T51581: Lock the page table row to ensure no other process
1080 // is adding a revision to the page at the same time.
1081 // Avoid locking extra tables, compare T191892.
1082 $pageLatest = $dbw->selectField(
1083 'page',
1084 'page_latest',
1085 [ 'page_id' => $pageId ],
1086 __METHOD__,
1087 [ 'FOR UPDATE' ]
1088 );
1089
1090 if ( !$pageLatest ) {
1091 return null;
1092 }
1093
1094 // Fetch the actual revision row from master, without locking all extra tables.
1095 $oldRevision = $this->loadRevisionFromConds(
1096 $dbw,
1097 [ 'rev_id' => intval( $pageLatest ) ],
1098 self::READ_LATEST,
1099 $title
1100 );
1101
1102 if ( !$oldRevision ) {
1103 $msg = "Failed to load latest revision ID $pageLatest of page ID $pageId.";
1104 $this->logger->error(
1105 $msg,
1106 [ 'exception' => new RuntimeException( $msg ) ]
1107 );
1108 return null;
1109 }
1110
1111 // Construct the new revision
1112 $timestamp = wfTimestampNow(); // TODO: use a callback, so we can override it for testing.
1113 $newRevision = MutableRevisionRecord::newFromParentRevision( $oldRevision );
1114
1115 $newRevision->setComment( $comment );
1116 $newRevision->setUser( $user );
1117 $newRevision->setTimestamp( $timestamp );
1118 $newRevision->setMinorEdit( $minor );
1119
1120 return $newRevision;
1121 }
1122
1123 /**
1124 * MCR migration note: this replaces Revision::isUnpatrolled
1125 *
1126 * @todo This is overly specific, so move or kill this method.
1127 *
1128 * @param RevisionRecord $rev
1129 *
1130 * @return int Rcid of the unpatrolled row, zero if there isn't one
1131 */
1132 public function getRcIdIfUnpatrolled( RevisionRecord $rev ) {
1133 $rc = $this->getRecentChange( $rev );
1134 if ( $rc && $rc->getAttribute( 'rc_patrolled' ) == RecentChange::PRC_UNPATROLLED ) {
1135 return $rc->getAttribute( 'rc_id' );
1136 } else {
1137 return 0;
1138 }
1139 }
1140
1141 /**
1142 * Get the RC object belonging to the current revision, if there's one
1143 *
1144 * MCR migration note: this replaces Revision::getRecentChange
1145 *
1146 * @todo move this somewhere else?
1147 *
1148 * @param RevisionRecord $rev
1149 * @param int $flags (optional) $flags include:
1150 * IDBAccessObject::READ_LATEST: Select the data from the master
1151 *
1152 * @return null|RecentChange
1153 */
1154 public function getRecentChange( RevisionRecord $rev, $flags = 0 ) {
1155 list( $dbType, ) = DBAccessObjectUtils::getDBOptions( $flags );
1156 $db = $this->getDBConnection( $dbType );
1157
1158 $userIdentity = $rev->getUser( RevisionRecord::RAW );
1159
1160 if ( !$userIdentity ) {
1161 // If the revision has no user identity, chances are it never went
1162 // into the database, and doesn't have an RC entry.
1163 return null;
1164 }
1165
1166 // TODO: Select by rc_this_oldid alone - but as of Nov 2017, there is no index on that!
1167 $actorWhere = $this->actorMigration->getWhere( $db, 'rc_user', $rev->getUser(), false );
1168 $rc = RecentChange::newFromConds(
1169 [
1170 $actorWhere['conds'],
1171 'rc_timestamp' => $db->timestamp( $rev->getTimestamp() ),
1172 'rc_this_oldid' => $rev->getId()
1173 ],
1174 __METHOD__,
1175 $dbType
1176 );
1177
1178 $this->releaseDBConnection( $db );
1179
1180 // XXX: cache this locally? Glue it to the RevisionRecord?
1181 return $rc;
1182 }
1183
1184 /**
1185 * Maps fields of the archive row to corresponding revision rows.
1186 *
1187 * @param object $archiveRow
1188 *
1189 * @return object a revision row object, corresponding to $archiveRow.
1190 */
1191 private static function mapArchiveFields( $archiveRow ) {
1192 $fieldMap = [
1193 // keep with ar prefix:
1194 'ar_id' => 'ar_id',
1195
1196 // not the same suffix:
1197 'ar_page_id' => 'rev_page',
1198 'ar_rev_id' => 'rev_id',
1199
1200 // same suffix:
1201 'ar_text_id' => 'rev_text_id',
1202 'ar_timestamp' => 'rev_timestamp',
1203 'ar_user_text' => 'rev_user_text',
1204 'ar_user' => 'rev_user',
1205 'ar_actor' => 'rev_actor',
1206 'ar_minor_edit' => 'rev_minor_edit',
1207 'ar_deleted' => 'rev_deleted',
1208 'ar_len' => 'rev_len',
1209 'ar_parent_id' => 'rev_parent_id',
1210 'ar_sha1' => 'rev_sha1',
1211 'ar_comment' => 'rev_comment',
1212 'ar_comment_cid' => 'rev_comment_cid',
1213 'ar_comment_id' => 'rev_comment_id',
1214 'ar_comment_text' => 'rev_comment_text',
1215 'ar_comment_data' => 'rev_comment_data',
1216 'ar_comment_old' => 'rev_comment_old',
1217 'ar_content_format' => 'rev_content_format',
1218 'ar_content_model' => 'rev_content_model',
1219 ];
1220
1221 $revRow = new stdClass();
1222 foreach ( $fieldMap as $arKey => $revKey ) {
1223 if ( property_exists( $archiveRow, $arKey ) ) {
1224 $revRow->$revKey = $archiveRow->$arKey;
1225 }
1226 }
1227
1228 return $revRow;
1229 }
1230
1231 /**
1232 * Constructs a RevisionRecord for the revisions main slot, based on the MW1.29 schema.
1233 *
1234 * @param object|array $row Either a database row or an array
1235 * @param int $queryFlags for callbacks
1236 * @param Title $title
1237 *
1238 * @return SlotRecord The main slot, extracted from the MW 1.29 style row.
1239 * @throws MWException
1240 */
1241 private function emulateMainSlot_1_29( $row, $queryFlags, Title $title ) {
1242 $mainSlotRow = new stdClass();
1243 $mainSlotRow->role_name = SlotRecord::MAIN;
1244 $mainSlotRow->model_name = null;
1245 $mainSlotRow->slot_revision_id = null;
1246 $mainSlotRow->slot_content_id = null;
1247 $mainSlotRow->content_address = null;
1248
1249 $content = null;
1250 $blobData = null;
1251 $blobFlags = null;
1252
1253 if ( is_object( $row ) ) {
1254 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_NEW ) ) {
1255 // Don't emulate from a row when using the new schema.
1256 // Emulating from an array is still OK.
1257 throw new LogicException( 'Can\'t emulate the main slot when using MCR schema.' );
1258 }
1259
1260 // archive row
1261 if ( !isset( $row->rev_id ) && ( isset( $row->ar_user ) || isset( $row->ar_actor ) ) ) {
1262 $row = $this->mapArchiveFields( $row );
1263 }
1264
1265 if ( isset( $row->rev_text_id ) && $row->rev_text_id > 0 ) {
1266 $mainSlotRow->content_address = SqlBlobStore::makeAddressFromTextId(
1267 $row->rev_text_id
1268 );
1269 }
1270
1271 // This is used by null-revisions
1272 $mainSlotRow->slot_origin = isset( $row->slot_origin )
1273 ? intval( $row->slot_origin )
1274 : null;
1275
1276 if ( isset( $row->old_text ) ) {
1277 // this happens when the text-table gets joined directly, in the pre-1.30 schema
1278 $blobData = isset( $row->old_text ) ? strval( $row->old_text ) : null;
1279 // Check against selects that might have not included old_flags
1280 if ( !property_exists( $row, 'old_flags' ) ) {
1281 throw new InvalidArgumentException( 'old_flags was not set in $row' );
1282 }
1283 $blobFlags = $row->old_flags ?? '';
1284 }
1285
1286 $mainSlotRow->slot_revision_id = intval( $row->rev_id );
1287
1288 $mainSlotRow->content_size = isset( $row->rev_len ) ? intval( $row->rev_len ) : null;
1289 $mainSlotRow->content_sha1 = isset( $row->rev_sha1 ) ? strval( $row->rev_sha1 ) : null;
1290 $mainSlotRow->model_name = isset( $row->rev_content_model )
1291 ? strval( $row->rev_content_model )
1292 : null;
1293 // XXX: in the future, we'll probably always use the default format, and drop content_format
1294 $mainSlotRow->format_name = isset( $row->rev_content_format )
1295 ? strval( $row->rev_content_format )
1296 : null;
1297
1298 if ( isset( $row->rev_text_id ) && intval( $row->rev_text_id ) > 0 ) {
1299 // Overwritten below for SCHEMA_COMPAT_WRITE_NEW
1300 $mainSlotRow->slot_content_id
1301 = $this->emulateContentId( intval( $row->rev_text_id ) );
1302 }
1303 } elseif ( is_array( $row ) ) {
1304 $mainSlotRow->slot_revision_id = isset( $row['id'] ) ? intval( $row['id'] ) : null;
1305
1306 $mainSlotRow->slot_origin = isset( $row['slot_origin'] )
1307 ? intval( $row['slot_origin'] )
1308 : null;
1309 $mainSlotRow->content_address = isset( $row['text_id'] )
1310 ? SqlBlobStore::makeAddressFromTextId( intval( $row['text_id'] ) )
1311 : null;
1312 $mainSlotRow->content_size = isset( $row['len'] ) ? intval( $row['len'] ) : null;
1313 $mainSlotRow->content_sha1 = isset( $row['sha1'] ) ? strval( $row['sha1'] ) : null;
1314
1315 $mainSlotRow->model_name = isset( $row['content_model'] )
1316 ? strval( $row['content_model'] ) : null; // XXX: must be a string!
1317 // XXX: in the future, we'll probably always use the default format, and drop content_format
1318 $mainSlotRow->format_name = isset( $row['content_format'] )
1319 ? strval( $row['content_format'] ) : null;
1320 $blobData = isset( $row['text'] ) ? rtrim( strval( $row['text'] ) ) : null;
1321 // XXX: If the flags field is not set then $blobFlags should be null so that no
1322 // decoding will happen. An empty string will result in default decodings.
1323 $blobFlags = isset( $row['flags'] ) ? trim( strval( $row['flags'] ) ) : null;
1324
1325 // if we have a Content object, override mText and mContentModel
1326 if ( !empty( $row['content'] ) ) {
1327 if ( !( $row['content'] instanceof Content ) ) {
1328 throw new MWException( 'content field must contain a Content object.' );
1329 }
1330
1331 /** @var Content $content */
1332 $content = $row['content'];
1333 $handler = $content->getContentHandler();
1334
1335 $mainSlotRow->model_name = $content->getModel();
1336
1337 // XXX: in the future, we'll probably always use the default format.
1338 if ( $mainSlotRow->format_name === null ) {
1339 $mainSlotRow->format_name = $handler->getDefaultFormat();
1340 }
1341 }
1342
1343 if ( isset( $row['text_id'] ) && intval( $row['text_id'] ) > 0 ) {
1344 // Overwritten below for SCHEMA_COMPAT_WRITE_NEW
1345 $mainSlotRow->slot_content_id
1346 = $this->emulateContentId( intval( $row['text_id'] ) );
1347 }
1348 } else {
1349 throw new MWException( 'Revision constructor passed invalid row format.' );
1350 }
1351
1352 // With the old schema, the content changes with every revision,
1353 // except for null-revisions.
1354 if ( !isset( $mainSlotRow->slot_origin ) ) {
1355 $mainSlotRow->slot_origin = $mainSlotRow->slot_revision_id;
1356 }
1357
1358 if ( $mainSlotRow->model_name === null ) {
1359 $mainSlotRow->model_name = function ( SlotRecord $slot ) use ( $title ) {
1360 $this->assertCrossWikiContentLoadingIsSafe();
1361
1362 return $this->slotRoleRegistry->getRoleHandler( $slot->getRole() )
1363 ->getDefaultModel( $title );
1364 };
1365 }
1366
1367 if ( !$content ) {
1368 // XXX: We should perhaps fail if $blobData is null and $mainSlotRow->content_address
1369 // is missing, but "empty revisions" with no content are used in some edge cases.
1370
1371 $content = function ( SlotRecord $slot )
1372 use ( $blobData, $blobFlags, $queryFlags, $mainSlotRow )
1373 {
1374 return $this->loadSlotContent(
1375 $slot,
1376 $blobData,
1377 $blobFlags,
1378 $mainSlotRow->format_name,
1379 $queryFlags
1380 );
1381 };
1382 }
1383
1384 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW ) ) {
1385 // NOTE: this callback will be looped through RevisionSlot::newInherited(), allowing
1386 // the inherited slot to have the same content_id as the original slot. In that case,
1387 // $slot will be the inherited slot, while $mainSlotRow still refers to the original slot.
1388 $mainSlotRow->slot_content_id =
1389 function ( SlotRecord $slot ) use ( $queryFlags, $mainSlotRow ) {
1390 $db = $this->getDBConnectionRefForQueryFlags( $queryFlags );
1391 return $this->findSlotContentId( $db, $mainSlotRow->slot_revision_id, SlotRecord::MAIN );
1392 };
1393 }
1394
1395 return new SlotRecord( $mainSlotRow, $content );
1396 }
1397
1398 /**
1399 * Provides a content ID to use with emulated SlotRecords in SCHEMA_COMPAT_OLD mode,
1400 * based on the revision's text ID (rev_text_id or ar_text_id, respectively).
1401 * Note that in SCHEMA_COMPAT_WRITE_BOTH, a callback to findSlotContentId() should be used
1402 * instead, since in that mode, some revision rows may already have a real content ID,
1403 * while other's don't - and for the ones that don't, we should indicate that it
1404 * is missing and cause SlotRecords::hasContentId() to return false.
1405 *
1406 * @param int $textId
1407 * @return int The emulated content ID
1408 */
1409 private function emulateContentId( $textId ) {
1410 // Return a negative number to ensure the ID is distinct from any real content IDs
1411 // that will be assigned in SCHEMA_COMPAT_WRITE_NEW mode and read in SCHEMA_COMPAT_READ_NEW
1412 // mode.
1413 return -$textId;
1414 }
1415
1416 /**
1417 * Loads a Content object based on a slot row.
1418 *
1419 * This method does not call $slot->getContent(), and may be used as a callback
1420 * called by $slot->getContent().
1421 *
1422 * MCR migration note: this roughly corresponds to Revision::getContentInternal
1423 *
1424 * @param SlotRecord $slot The SlotRecord to load content for
1425 * @param string|null $blobData The content blob, in the form indicated by $blobFlags
1426 * @param string|null $blobFlags Flags indicating how $blobData needs to be processed.
1427 * Use null if no processing should happen. That is in constrast to the empty string,
1428 * which causes the blob to be decoded according to the configured legacy encoding.
1429 * @param string|null $blobFormat MIME type indicating how $dataBlob is encoded
1430 * @param int $queryFlags
1431 *
1432 * @throws RevisionAccessException
1433 * @return Content
1434 */
1435 private function loadSlotContent(
1436 SlotRecord $slot,
1437 $blobData = null,
1438 $blobFlags = null,
1439 $blobFormat = null,
1440 $queryFlags = 0
1441 ) {
1442 if ( $blobData !== null ) {
1443 Assert::parameterType( 'string', $blobData, '$blobData' );
1444 Assert::parameterType( 'string|null', $blobFlags, '$blobFlags' );
1445
1446 $cacheKey = $slot->hasAddress() ? $slot->getAddress() : null;
1447
1448 if ( $blobFlags === null ) {
1449 // No blob flags, so use the blob verbatim.
1450 $data = $blobData;
1451 } else {
1452 $data = $this->blobStore->expandBlob( $blobData, $blobFlags, $cacheKey );
1453 if ( $data === false ) {
1454 throw new RevisionAccessException(
1455 "Failed to expand blob data using flags $blobFlags (key: $cacheKey)"
1456 );
1457 }
1458 }
1459
1460 } else {
1461 $address = $slot->getAddress();
1462 try {
1463 $data = $this->blobStore->getBlob( $address, $queryFlags );
1464 } catch ( BlobAccessException $e ) {
1465 throw new RevisionAccessException(
1466 "Failed to load data blob from $address: " . $e->getMessage(), 0, $e
1467 );
1468 }
1469 }
1470
1471 // Unserialize content
1472 $handler = ContentHandler::getForModelID( $slot->getModel() );
1473
1474 $content = $handler->unserializeContent( $data, $blobFormat );
1475 return $content;
1476 }
1477
1478 /**
1479 * Load a page revision from a given revision ID number.
1480 * Returns null if no such revision can be found.
1481 *
1482 * MCR migration note: this replaces Revision::newFromId
1483 *
1484 * $flags include:
1485 * IDBAccessObject::READ_LATEST: Select the data from the master
1486 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
1487 *
1488 * @param int $id
1489 * @param int $flags (optional)
1490 * @return RevisionRecord|null
1491 */
1492 public function getRevisionById( $id, $flags = 0 ) {
1493 return $this->newRevisionFromConds( [ 'rev_id' => intval( $id ) ], $flags );
1494 }
1495
1496 /**
1497 * Load either the current, or a specified, revision
1498 * that's attached to a given link target. If not attached
1499 * to that link target, will return null.
1500 *
1501 * MCR migration note: this replaces Revision::newFromTitle
1502 *
1503 * $flags include:
1504 * IDBAccessObject::READ_LATEST: Select the data from the master
1505 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
1506 *
1507 * @param LinkTarget $linkTarget
1508 * @param int $revId (optional)
1509 * @param int $flags Bitfield (optional)
1510 * @return RevisionRecord|null
1511 */
1512 public function getRevisionByTitle( LinkTarget $linkTarget, $revId = 0, $flags = 0 ) {
1513 // TODO should not require Title in future (T206498)
1514 $title = Title::newFromLinkTarget( $linkTarget );
1515 $conds = [
1516 'page_namespace' => $title->getNamespace(),
1517 'page_title' => $title->getDBkey()
1518 ];
1519 if ( $revId ) {
1520 // Use the specified revision ID.
1521 // Note that we use newRevisionFromConds here because we want to retry
1522 // and fall back to master if the page is not found on a replica.
1523 // Since the caller supplied a revision ID, we are pretty sure the revision is
1524 // supposed to exist, so we should try hard to find it.
1525 $conds['rev_id'] = $revId;
1526 return $this->newRevisionFromConds( $conds, $flags, $title );
1527 } else {
1528 // Use a join to get the latest revision.
1529 // Note that we don't use newRevisionFromConds here because we don't want to retry
1530 // and fall back to master. The assumption is that we only want to force the fallback
1531 // if we are quite sure the revision exists because the caller supplied a revision ID.
1532 // If the page isn't found at all on a replica, it probably simply does not exist.
1533 $db = $this->getDBConnectionRefForQueryFlags( $flags );
1534
1535 $conds[] = 'rev_id=page_latest';
1536 $rev = $this->loadRevisionFromConds( $db, $conds, $flags, $title );
1537
1538 return $rev;
1539 }
1540 }
1541
1542 /**
1543 * Load either the current, or a specified, revision
1544 * that's attached to a given page ID.
1545 * Returns null if no such revision can be found.
1546 *
1547 * MCR migration note: this replaces Revision::newFromPageId
1548 *
1549 * $flags include:
1550 * IDBAccessObject::READ_LATEST: Select the data from the master (since 1.20)
1551 * IDBAccessObject::READ_LOCKING : Select & lock the data from the master
1552 *
1553 * @param int $pageId
1554 * @param int $revId (optional)
1555 * @param int $flags Bitfield (optional)
1556 * @return RevisionRecord|null
1557 */
1558 public function getRevisionByPageId( $pageId, $revId = 0, $flags = 0 ) {
1559 $conds = [ 'page_id' => $pageId ];
1560 if ( $revId ) {
1561 // Use the specified revision ID.
1562 // Note that we use newRevisionFromConds here because we want to retry
1563 // and fall back to master if the page is not found on a replica.
1564 // Since the caller supplied a revision ID, we are pretty sure the revision is
1565 // supposed to exist, so we should try hard to find it.
1566 $conds['rev_id'] = $revId;
1567 return $this->newRevisionFromConds( $conds, $flags );
1568 } else {
1569 // Use a join to get the latest revision.
1570 // Note that we don't use newRevisionFromConds here because we don't want to retry
1571 // and fall back to master. The assumption is that we only want to force the fallback
1572 // if we are quite sure the revision exists because the caller supplied a revision ID.
1573 // If the page isn't found at all on a replica, it probably simply does not exist.
1574 $db = $this->getDBConnectionRefForQueryFlags( $flags );
1575
1576 $conds[] = 'rev_id=page_latest';
1577 $rev = $this->loadRevisionFromConds( $db, $conds, $flags );
1578
1579 return $rev;
1580 }
1581 }
1582
1583 /**
1584 * Load the revision for the given title with the given timestamp.
1585 * WARNING: Timestamps may in some circumstances not be unique,
1586 * so this isn't the best key to use.
1587 *
1588 * MCR migration note: this replaces Revision::loadFromTimestamp
1589 *
1590 * @param Title $title
1591 * @param string $timestamp
1592 * @return RevisionRecord|null
1593 */
1594 public function getRevisionByTimestamp( $title, $timestamp ) {
1595 $db = $this->getDBConnection( DB_REPLICA );
1596 return $this->newRevisionFromConds(
1597 [
1598 'rev_timestamp' => $db->timestamp( $timestamp ),
1599 'page_namespace' => $title->getNamespace(),
1600 'page_title' => $title->getDBkey()
1601 ],
1602 0,
1603 $title
1604 );
1605 }
1606
1607 /**
1608 * @param int $revId The revision to load slots for.
1609 * @param int $queryFlags
1610 * @param Title $title
1611 *
1612 * @return SlotRecord[]
1613 */
1614 private function loadSlotRecords( $revId, $queryFlags, Title $title ) {
1615 $revQuery = self::getSlotsQueryInfo( [ 'content' ] );
1616
1617 list( $dbMode, $dbOptions ) = DBAccessObjectUtils::getDBOptions( $queryFlags );
1618 $db = $this->getDBConnectionRef( $dbMode );
1619
1620 $res = $db->select(
1621 $revQuery['tables'],
1622 $revQuery['fields'],
1623 [
1624 'slot_revision_id' => $revId,
1625 ],
1626 __METHOD__,
1627 $dbOptions,
1628 $revQuery['joins']
1629 );
1630
1631 $slots = $this->constructSlotRecords( $revId, $res, $queryFlags, $title );
1632
1633 return $slots;
1634 }
1635
1636 /**
1637 * Factory method for SlotRecords based on known slot rows.
1638 *
1639 * @param int $revId The revision to load slots for.
1640 * @param object[]|ResultWrapper $slotRows
1641 * @param int $queryFlags
1642 * @param Title $title
1643 *
1644 * @return SlotRecord[]
1645 */
1646 private function constructSlotRecords( $revId, $slotRows, $queryFlags, Title $title ) {
1647 $slots = [];
1648
1649 foreach ( $slotRows as $row ) {
1650 // Resolve role names and model names from in-memory cache, if they were not joined in.
1651 if ( !isset( $row->role_name ) ) {
1652 $row->role_name = $this->slotRoleStore->getName( (int)$row->slot_role_id );
1653 }
1654
1655 if ( !isset( $row->model_name ) ) {
1656 if ( isset( $row->content_model ) ) {
1657 $row->model_name = $this->contentModelStore->getName( (int)$row->content_model );
1658 } else {
1659 // We may get here if $row->model_name is set but null, perhaps because it
1660 // came from rev_content_model, which is NULL for the default model.
1661 $slotRoleHandler = $this->slotRoleRegistry->getRoleHandler( $row->role_name );
1662 $row->model_name = $slotRoleHandler->getDefaultModel( $title );
1663 }
1664 }
1665
1666 if ( !isset( $row->content_id ) && isset( $row->rev_text_id ) ) {
1667 $row->slot_content_id
1668 = $this->emulateContentId( intval( $row->rev_text_id ) );
1669 }
1670
1671 $contentCallback = function ( SlotRecord $slot ) use ( $queryFlags ) {
1672 return $this->loadSlotContent( $slot, null, null, null, $queryFlags );
1673 };
1674
1675 $slots[$row->role_name] = new SlotRecord( $row, $contentCallback );
1676 }
1677
1678 if ( !isset( $slots[SlotRecord::MAIN] ) ) {
1679 throw new RevisionAccessException(
1680 'Main slot of revision ' . $revId . ' not found in database!'
1681 );
1682 }
1683
1684 return $slots;
1685 }
1686
1687 /**
1688 * Factory method for RevisionSlots based on a revision ID.
1689 *
1690 * @note If other code has a need to construct RevisionSlots objects, this should be made
1691 * public, since RevisionSlots instances should not be constructed directly.
1692 *
1693 * @param int $revId
1694 * @param object $revisionRow
1695 * @param object[]|null $slotRows
1696 * @param int $queryFlags
1697 * @param Title $title
1698 *
1699 * @return RevisionSlots
1700 * @throws MWException
1701 */
1702 private function newRevisionSlots(
1703 $revId,
1704 $revisionRow,
1705 $slotRows,
1706 $queryFlags,
1707 Title $title
1708 ) {
1709 if ( $slotRows ) {
1710 $slots = new RevisionSlots(
1711 $this->constructSlotRecords( $revId, $slotRows, $queryFlags, $title )
1712 );
1713 } elseif ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_NEW ) ) {
1714 $mainSlot = $this->emulateMainSlot_1_29( $revisionRow, $queryFlags, $title );
1715 // @phan-suppress-next-line PhanTypeInvalidCallableArraySize false positive
1716 $slots = new RevisionSlots( [ SlotRecord::MAIN => $mainSlot ] );
1717 } else {
1718 // XXX: do we need the same kind of caching here
1719 // that getKnownCurrentRevision uses (if $revId == page_latest?)
1720
1721 $slots = new RevisionSlots( function () use( $revId, $queryFlags, $title ) {
1722 return $this->loadSlotRecords( $revId, $queryFlags, $title );
1723 } );
1724 }
1725
1726 return $slots;
1727 }
1728
1729 /**
1730 * Make a fake revision object from an archive table row. This is queried
1731 * for permissions or even inserted (as in Special:Undelete)
1732 *
1733 * MCR migration note: this replaces Revision::newFromArchiveRow
1734 *
1735 * @param object $row
1736 * @param int $queryFlags
1737 * @param Title|null $title
1738 * @param array $overrides associative array with fields of $row to override. This may be
1739 * used e.g. to force the parent revision ID or page ID. Keys in the array are fields
1740 * names from the archive table without the 'ar_' prefix, i.e. use 'parent_id' to
1741 * override ar_parent_id.
1742 *
1743 * @return RevisionRecord
1744 * @throws MWException
1745 */
1746 public function newRevisionFromArchiveRow(
1747 $row,
1748 $queryFlags = 0,
1749 Title $title = null,
1750 array $overrides = []
1751 ) {
1752 Assert::parameterType( 'object', $row, '$row' );
1753
1754 // check second argument, since Revision::newFromArchiveRow had $overrides in that spot.
1755 Assert::parameterType( 'integer', $queryFlags, '$queryFlags' );
1756
1757 if ( !$title && isset( $overrides['title'] ) ) {
1758 if ( !( $overrides['title'] instanceof Title ) ) {
1759 throw new MWException( 'title field override must contain a Title object.' );
1760 }
1761
1762 $title = $overrides['title'];
1763 }
1764
1765 if ( !isset( $title ) ) {
1766 if ( isset( $row->ar_namespace ) && isset( $row->ar_title ) ) {
1767 $title = Title::makeTitle( $row->ar_namespace, $row->ar_title );
1768 } else {
1769 throw new InvalidArgumentException(
1770 'A Title or ar_namespace and ar_title must be given'
1771 );
1772 }
1773 }
1774
1775 foreach ( $overrides as $key => $value ) {
1776 $field = "ar_$key";
1777 $row->$field = $value;
1778 }
1779
1780 try {
1781 $user = User::newFromAnyId(
1782 $row->ar_user ?? null,
1783 $row->ar_user_text ?? null,
1784 $row->ar_actor ?? null,
1785 $this->wikiId
1786 );
1787 } catch ( InvalidArgumentException $ex ) {
1788 wfWarn( __METHOD__ . ': ' . $title->getPrefixedDBkey() . ': ' . $ex->getMessage() );
1789 $user = new UserIdentityValue( 0, 'Unknown user', 0 );
1790 }
1791
1792 $db = $this->getDBConnectionRefForQueryFlags( $queryFlags );
1793 // Legacy because $row may have come from self::selectFields()
1794 $comment = $this->commentStore->getCommentLegacy( $db, 'ar_comment', $row, true );
1795
1796 $slots = $this->newRevisionSlots( $row->ar_rev_id, $row, null, $queryFlags, $title );
1797
1798 return new RevisionArchiveRecord( $title, $user, $comment, $row, $slots, $this->wikiId );
1799 }
1800
1801 /**
1802 * @see RevisionFactory::newRevisionFromRow
1803 *
1804 * MCR migration note: this replaces Revision::newFromRow
1805 *
1806 * @param object $row A database row generated from a query based on getQueryInfo()
1807 * @param int $queryFlags
1808 * @param Title|null $title
1809 * @param bool $fromCache if true, the returned RevisionRecord will ensure that no stale
1810 * data is returned from getters, by querying the database as needed
1811 * @return RevisionRecord
1812 */
1813 public function newRevisionFromRow(
1814 $row,
1815 $queryFlags = 0,
1816 Title $title = null,
1817 $fromCache = false
1818 ) {
1819 return $this->newRevisionFromRowAndSlots( $row, null, $queryFlags, $title, $fromCache );
1820 }
1821
1822 /**
1823 * @param object $row A database row generated from a query based on getQueryInfo()
1824 * @param null|object[] $slotRows Database rows generated from a query based on
1825 * getSlotsQueryInfo with the 'content' flag set.
1826 * @param int $queryFlags
1827 * @param Title|null $title
1828 * @param bool $fromCache if true, the returned RevisionRecord will ensure that no stale
1829 * data is returned from getters, by querying the database as needed
1830 *
1831 * @return RevisionRecord
1832 * @throws MWException
1833 * @see RevisionFactory::newRevisionFromRow
1834 *
1835 * MCR migration note: this replaces Revision::newFromRow
1836 *
1837 */
1838 public function newRevisionFromRowAndSlots(
1839 $row,
1840 $slotRows,
1841 $queryFlags = 0,
1842 Title $title = null,
1843 $fromCache = false
1844 ) {
1845 Assert::parameterType( 'object', $row, '$row' );
1846
1847 if ( !$title ) {
1848 $pageId = $row->rev_page ?? 0; // XXX: also check page_id?
1849 $revId = $row->rev_id ?? 0;
1850
1851 $title = $this->getTitle( $pageId, $revId, $queryFlags );
1852 }
1853
1854 if ( !isset( $row->page_latest ) ) {
1855 $row->page_latest = $title->getLatestRevID();
1856 if ( $row->page_latest === 0 && $title->exists() ) {
1857 wfWarn( 'Encountered title object in limbo: ID ' . $title->getArticleID() );
1858 }
1859 }
1860
1861 try {
1862 $user = User::newFromAnyId(
1863 $row->rev_user ?? null,
1864 $row->rev_user_text ?? null,
1865 $row->rev_actor ?? null,
1866 $this->wikiId
1867 );
1868 } catch ( InvalidArgumentException $ex ) {
1869 wfWarn( __METHOD__ . ': ' . $title->getPrefixedDBkey() . ': ' . $ex->getMessage() );
1870 $user = new UserIdentityValue( 0, 'Unknown user', 0 );
1871 }
1872
1873 $db = $this->getDBConnectionRefForQueryFlags( $queryFlags );
1874 // Legacy because $row may have come from self::selectFields()
1875 $comment = $this->commentStore->getCommentLegacy( $db, 'rev_comment', $row, true );
1876
1877 $slots = $this->newRevisionSlots( $row->rev_id, $row, $slotRows, $queryFlags, $title );
1878
1879 // If this is a cached row, instantiate a cache-aware revision class to avoid stale data.
1880 if ( $fromCache ) {
1881 $rev = new RevisionStoreCacheRecord(
1882 function ( $revId ) use ( $queryFlags ) {
1883 $db = $this->getDBConnectionRefForQueryFlags( $queryFlags );
1884 return $this->fetchRevisionRowFromConds(
1885 $db,
1886 [ 'rev_id' => intval( $revId ) ]
1887 );
1888 },
1889 $title, $user, $comment, $row, $slots, $this->wikiId
1890 );
1891 } else {
1892 $rev = new RevisionStoreRecord(
1893 $title, $user, $comment, $row, $slots, $this->wikiId );
1894 }
1895 return $rev;
1896 }
1897
1898 /**
1899 * Constructs a new MutableRevisionRecord based on the given associative array following
1900 * the MW1.29 convention for the Revision constructor.
1901 *
1902 * MCR migration note: this replaces Revision::newFromRow
1903 *
1904 * @param array $fields
1905 * @param int $queryFlags
1906 * @param Title|null $title
1907 *
1908 * @return MutableRevisionRecord
1909 * @throws MWException
1910 * @throws RevisionAccessException
1911 */
1912 public function newMutableRevisionFromArray(
1913 array $fields,
1914 $queryFlags = 0,
1915 Title $title = null
1916 ) {
1917 if ( !$title && isset( $fields['title'] ) ) {
1918 if ( !( $fields['title'] instanceof Title ) ) {
1919 throw new MWException( 'title field must contain a Title object.' );
1920 }
1921
1922 $title = $fields['title'];
1923 }
1924
1925 if ( !$title ) {
1926 $pageId = $fields['page'] ?? 0;
1927 $revId = $fields['id'] ?? 0;
1928
1929 $title = $this->getTitle( $pageId, $revId, $queryFlags );
1930 }
1931
1932 if ( !isset( $fields['page'] ) ) {
1933 $fields['page'] = $title->getArticleID( $queryFlags );
1934 }
1935
1936 // if we have a content object, use it to set the model and type
1937 if ( !empty( $fields['content'] ) && !( $fields['content'] instanceof Content )
1938 && !is_array( $fields['content'] )
1939 ) {
1940 throw new MWException(
1941 'content field must contain a Content object or an array of Content objects.'
1942 );
1943 }
1944
1945 if ( !empty( $fields['text_id'] ) ) {
1946 if ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_OLD ) ) {
1947 throw new MWException( "The text_id field is only available in the pre-MCR schema" );
1948 }
1949
1950 if ( !empty( $fields['content'] ) ) {
1951 throw new MWException(
1952 "Text already stored in external store (id {$fields['text_id']}), " .
1953 "can't specify content object"
1954 );
1955 }
1956 }
1957
1958 if (
1959 isset( $fields['comment'] )
1960 && !( $fields['comment'] instanceof CommentStoreComment )
1961 ) {
1962 $commentData = $fields['comment_data'] ?? null;
1963
1964 if ( $fields['comment'] instanceof Message ) {
1965 $fields['comment'] = CommentStoreComment::newUnsavedComment(
1966 $fields['comment'],
1967 $commentData
1968 );
1969 } else {
1970 $commentText = trim( strval( $fields['comment'] ) );
1971 $fields['comment'] = CommentStoreComment::newUnsavedComment(
1972 $commentText,
1973 $commentData
1974 );
1975 }
1976 }
1977
1978 $revision = new MutableRevisionRecord( $title, $this->wikiId );
1979 $this->initializeMutableRevisionFromArray( $revision, $fields );
1980
1981 if ( isset( $fields['content'] ) && is_array( $fields['content'] ) ) {
1982 // @phan-suppress-next-line PhanTypeNoPropertiesForeach
1983 foreach ( $fields['content'] as $role => $content ) {
1984 $revision->setContent( $role, $content );
1985 }
1986 } else {
1987 $mainSlot = $this->emulateMainSlot_1_29( $fields, $queryFlags, $title );
1988 $revision->setSlot( $mainSlot );
1989 }
1990
1991 return $revision;
1992 }
1993
1994 /**
1995 * @param MutableRevisionRecord $record
1996 * @param array $fields
1997 */
1998 private function initializeMutableRevisionFromArray(
1999 MutableRevisionRecord $record,
2000 array $fields
2001 ) {
2002 /** @var UserIdentity $user */
2003 $user = null;
2004
2005 // If a user is passed in, use it if possible. We cannot use a user from a
2006 // remote wiki with unsuppressed ids, due to issues described in T222212.
2007 if ( isset( $fields['user'] ) &&
2008 ( $fields['user'] instanceof UserIdentity ) &&
2009 ( $this->wikiId === false ||
2010 ( !$fields['user']->getId() && !$fields['user']->getActorId() ) )
2011 ) {
2012 $user = $fields['user'];
2013 } else {
2014 try {
2015 $user = User::newFromAnyId(
2016 $fields['user'] ?? null,
2017 $fields['user_text'] ?? null,
2018 $fields['actor'] ?? null,
2019 $this->wikiId
2020 );
2021 } catch ( InvalidArgumentException $ex ) {
2022 $user = null;
2023 }
2024 }
2025
2026 if ( $user ) {
2027 $record->setUser( $user );
2028 }
2029
2030 $timestamp = isset( $fields['timestamp'] )
2031 ? strval( $fields['timestamp'] )
2032 : wfTimestampNow(); // TODO: use a callback, so we can override it for testing.
2033
2034 $record->setTimestamp( $timestamp );
2035
2036 if ( isset( $fields['page'] ) ) {
2037 $record->setPageId( intval( $fields['page'] ) );
2038 }
2039
2040 if ( isset( $fields['id'] ) ) {
2041 $record->setId( intval( $fields['id'] ) );
2042 }
2043 if ( isset( $fields['parent_id'] ) ) {
2044 $record->setParentId( intval( $fields['parent_id'] ) );
2045 }
2046
2047 if ( isset( $fields['sha1'] ) ) {
2048 $record->setSha1( $fields['sha1'] );
2049 }
2050 if ( isset( $fields['size'] ) ) {
2051 $record->setSize( intval( $fields['size'] ) );
2052 }
2053
2054 if ( isset( $fields['minor_edit'] ) ) {
2055 $record->setMinorEdit( intval( $fields['minor_edit'] ) !== 0 );
2056 }
2057 if ( isset( $fields['deleted'] ) ) {
2058 $record->setVisibility( intval( $fields['deleted'] ) );
2059 }
2060
2061 if ( isset( $fields['comment'] ) ) {
2062 Assert::parameterType(
2063 CommentStoreComment::class,
2064 $fields['comment'],
2065 '$row[\'comment\']'
2066 );
2067 $record->setComment( $fields['comment'] );
2068 }
2069 }
2070
2071 /**
2072 * Load a page revision from a given revision ID number.
2073 * Returns null if no such revision can be found.
2074 *
2075 * MCR migration note: this corresponds to Revision::loadFromId
2076 *
2077 * @note direct use is deprecated!
2078 * @todo remove when unused! there seem to be no callers of Revision::loadFromId
2079 *
2080 * @param IDatabase $db
2081 * @param int $id
2082 *
2083 * @return RevisionRecord|null
2084 */
2085 public function loadRevisionFromId( IDatabase $db, $id ) {
2086 return $this->loadRevisionFromConds( $db, [ 'rev_id' => intval( $id ) ] );
2087 }
2088
2089 /**
2090 * Load either the current, or a specified, revision
2091 * that's attached to a given page. If not attached
2092 * to that page, will return null.
2093 *
2094 * MCR migration note: this replaces Revision::loadFromPageId
2095 *
2096 * @note direct use is deprecated!
2097 * @todo remove when unused!
2098 *
2099 * @param IDatabase $db
2100 * @param int $pageid
2101 * @param int $id
2102 * @return RevisionRecord|null
2103 */
2104 public function loadRevisionFromPageId( IDatabase $db, $pageid, $id = 0 ) {
2105 $conds = [ 'rev_page' => intval( $pageid ), 'page_id' => intval( $pageid ) ];
2106 if ( $id ) {
2107 $conds['rev_id'] = intval( $id );
2108 } else {
2109 $conds[] = 'rev_id=page_latest';
2110 }
2111 return $this->loadRevisionFromConds( $db, $conds );
2112 }
2113
2114 /**
2115 * Load either the current, or a specified, revision
2116 * that's attached to a given page. If not attached
2117 * to that page, will return null.
2118 *
2119 * MCR migration note: this replaces Revision::loadFromTitle
2120 *
2121 * @note direct use is deprecated!
2122 * @todo remove when unused!
2123 *
2124 * @param IDatabase $db
2125 * @param Title $title
2126 * @param int $id
2127 *
2128 * @return RevisionRecord|null
2129 */
2130 public function loadRevisionFromTitle( IDatabase $db, $title, $id = 0 ) {
2131 if ( $id ) {
2132 $matchId = intval( $id );
2133 } else {
2134 $matchId = 'page_latest';
2135 }
2136
2137 return $this->loadRevisionFromConds(
2138 $db,
2139 [
2140 "rev_id=$matchId",
2141 'page_namespace' => $title->getNamespace(),
2142 'page_title' => $title->getDBkey()
2143 ],
2144 0,
2145 $title
2146 );
2147 }
2148
2149 /**
2150 * Load the revision for the given title with the given timestamp.
2151 * WARNING: Timestamps may in some circumstances not be unique,
2152 * so this isn't the best key to use.
2153 *
2154 * MCR migration note: this replaces Revision::loadFromTimestamp
2155 *
2156 * @note direct use is deprecated! Use getRevisionFromTimestamp instead!
2157 * @todo remove when unused!
2158 *
2159 * @param IDatabase $db
2160 * @param Title $title
2161 * @param string $timestamp
2162 * @return RevisionRecord|null
2163 */
2164 public function loadRevisionFromTimestamp( IDatabase $db, $title, $timestamp ) {
2165 return $this->loadRevisionFromConds( $db,
2166 [
2167 'rev_timestamp' => $db->timestamp( $timestamp ),
2168 'page_namespace' => $title->getNamespace(),
2169 'page_title' => $title->getDBkey()
2170 ],
2171 0,
2172 $title
2173 );
2174 }
2175
2176 /**
2177 * Given a set of conditions, fetch a revision
2178 *
2179 * This method should be used if we are pretty sure the revision exists.
2180 * Unless $flags has READ_LATEST set, this method will first try to find the revision
2181 * on a replica before hitting the master database.
2182 *
2183 * MCR migration note: this corresponds to Revision::newFromConds
2184 *
2185 * @param array $conditions
2186 * @param int $flags (optional)
2187 * @param Title|null $title
2188 *
2189 * @return RevisionRecord|null
2190 */
2191 private function newRevisionFromConds( $conditions, $flags = 0, Title $title = null ) {
2192 $db = $this->getDBConnectionRefForQueryFlags( $flags );
2193 $rev = $this->loadRevisionFromConds( $db, $conditions, $flags, $title );
2194
2195 $lb = $this->getDBLoadBalancer();
2196
2197 // Make sure new pending/committed revision are visibile later on
2198 // within web requests to certain avoid bugs like T93866 and T94407.
2199 if ( !$rev
2200 && !( $flags & self::READ_LATEST )
2201 && $lb->hasStreamingReplicaServers()
2202 && $lb->hasOrMadeRecentMasterChanges()
2203 ) {
2204 $flags = self::READ_LATEST;
2205 $dbw = $this->getDBConnection( DB_MASTER );
2206 $rev = $this->loadRevisionFromConds( $dbw, $conditions, $flags, $title );
2207 $this->releaseDBConnection( $dbw );
2208 }
2209
2210 return $rev;
2211 }
2212
2213 /**
2214 * Given a set of conditions, fetch a revision from
2215 * the given database connection.
2216 *
2217 * MCR migration note: this corresponds to Revision::loadFromConds
2218 *
2219 * @param IDatabase $db
2220 * @param array $conditions
2221 * @param int $flags (optional)
2222 * @param Title|null $title
2223 *
2224 * @return RevisionRecord|null
2225 */
2226 private function loadRevisionFromConds(
2227 IDatabase $db,
2228 $conditions,
2229 $flags = 0,
2230 Title $title = null
2231 ) {
2232 $row = $this->fetchRevisionRowFromConds( $db, $conditions, $flags );
2233 if ( $row ) {
2234 $rev = $this->newRevisionFromRow( $row, $flags, $title );
2235
2236 return $rev;
2237 }
2238
2239 return null;
2240 }
2241
2242 /**
2243 * Throws an exception if the given database connection does not belong to the wiki this
2244 * RevisionStore is bound to.
2245 *
2246 * @param IDatabase $db
2247 * @throws MWException
2248 */
2249 private function checkDatabaseWikiId( IDatabase $db ) {
2250 $storeWiki = $this->wikiId;
2251 $dbWiki = $db->getDomainID();
2252
2253 if ( $dbWiki === $storeWiki ) {
2254 return;
2255 }
2256
2257 $storeWiki = $storeWiki ?: $this->loadBalancer->getLocalDomainID();
2258 // @FIXME: when would getDomainID() be false here?
2259 $dbWiki = $dbWiki ?: wfWikiID();
2260
2261 if ( $dbWiki === $storeWiki ) {
2262 return;
2263 }
2264
2265 // HACK: counteract encoding imposed by DatabaseDomain
2266 $storeWiki = str_replace( '?h', '-', $storeWiki );
2267 $dbWiki = str_replace( '?h', '-', $dbWiki );
2268
2269 if ( $dbWiki === $storeWiki ) {
2270 return;
2271 }
2272
2273 throw new MWException( "RevisionStore for $storeWiki "
2274 . "cannot be used with a DB connection for $dbWiki" );
2275 }
2276
2277 /**
2278 * Given a set of conditions, return a row with the
2279 * fields necessary to build RevisionRecord objects.
2280 *
2281 * MCR migration note: this corresponds to Revision::fetchFromConds
2282 *
2283 * @param IDatabase $db
2284 * @param array $conditions
2285 * @param int $flags (optional)
2286 *
2287 * @return object|false data row as a raw object
2288 */
2289 private function fetchRevisionRowFromConds( IDatabase $db, $conditions, $flags = 0 ) {
2290 $this->checkDatabaseWikiId( $db );
2291
2292 $revQuery = $this->getQueryInfo( [ 'page', 'user' ] );
2293 $options = [];
2294 if ( ( $flags & self::READ_LOCKING ) == self::READ_LOCKING ) {
2295 $options[] = 'FOR UPDATE';
2296 }
2297 return $db->selectRow(
2298 $revQuery['tables'],
2299 $revQuery['fields'],
2300 $conditions,
2301 __METHOD__,
2302 $options,
2303 $revQuery['joins']
2304 );
2305 }
2306
2307 /**
2308 * Finds the ID of a content row for a given revision and slot role.
2309 * This can be used to re-use content rows even while the content ID
2310 * is still missing from SlotRecords, when writing to both the old and
2311 * the new schema during MCR schema migration.
2312 *
2313 * @todo remove after MCR schema migration is complete.
2314 *
2315 * @param IDatabase $db
2316 * @param int $revId
2317 * @param string $role
2318 *
2319 * @return int|null
2320 */
2321 private function findSlotContentId( IDatabase $db, $revId, $role ) {
2322 if ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_NEW ) ) {
2323 return null;
2324 }
2325
2326 try {
2327 $roleId = $this->slotRoleStore->getId( $role );
2328 $conditions = [
2329 'slot_revision_id' => $revId,
2330 'slot_role_id' => $roleId,
2331 ];
2332
2333 $contentId = $db->selectField( 'slots', 'slot_content_id', $conditions, __METHOD__ );
2334
2335 return $contentId ?: null;
2336 } catch ( NameTableAccessException $ex ) {
2337 // If the role is missing from the slot_roles table,
2338 // the corresponding row in slots cannot exist.
2339 return null;
2340 }
2341 }
2342
2343 /**
2344 * Return the tables, fields, and join conditions to be selected to create
2345 * a new RevisionStoreRecord object.
2346 *
2347 * MCR migration note: this replaces Revision::getQueryInfo
2348 *
2349 * If the format of fields returned changes in any way then the cache key provided by
2350 * self::getRevisionRowCacheKey should be updated.
2351 *
2352 * @since 1.31
2353 *
2354 * @param array $options Any combination of the following strings
2355 * - 'page': Join with the page table, and select fields to identify the page
2356 * - 'user': Join with the user table, and select the user name
2357 * - 'text': Join with the text table, and select fields to load page text. This
2358 * option is deprecated in MW 1.32 when the MCR migration flag SCHEMA_COMPAT_WRITE_NEW
2359 * is set, and disallowed when SCHEMA_COMPAT_READ_OLD is not set.
2360 *
2361 * @return array With three keys:
2362 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
2363 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
2364 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
2365 */
2366 public function getQueryInfo( $options = [] ) {
2367 $ret = [
2368 'tables' => [],
2369 'fields' => [],
2370 'joins' => [],
2371 ];
2372
2373 $ret['tables'][] = 'revision';
2374 $ret['fields'] = array_merge( $ret['fields'], [
2375 'rev_id',
2376 'rev_page',
2377 'rev_timestamp',
2378 'rev_minor_edit',
2379 'rev_deleted',
2380 'rev_len',
2381 'rev_parent_id',
2382 'rev_sha1',
2383 ] );
2384
2385 $commentQuery = $this->commentStore->getJoin( 'rev_comment' );
2386 $ret['tables'] = array_merge( $ret['tables'], $commentQuery['tables'] );
2387 $ret['fields'] = array_merge( $ret['fields'], $commentQuery['fields'] );
2388 $ret['joins'] = array_merge( $ret['joins'], $commentQuery['joins'] );
2389
2390 $actorQuery = $this->actorMigration->getJoin( 'rev_user' );
2391 $ret['tables'] = array_merge( $ret['tables'], $actorQuery['tables'] );
2392 $ret['fields'] = array_merge( $ret['fields'], $actorQuery['fields'] );
2393 $ret['joins'] = array_merge( $ret['joins'], $actorQuery['joins'] );
2394
2395 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_OLD ) ) {
2396 $ret['fields'][] = 'rev_text_id';
2397
2398 if ( $this->contentHandlerUseDB ) {
2399 $ret['fields'][] = 'rev_content_format';
2400 $ret['fields'][] = 'rev_content_model';
2401 }
2402 }
2403
2404 if ( in_array( 'page', $options, true ) ) {
2405 $ret['tables'][] = 'page';
2406 $ret['fields'] = array_merge( $ret['fields'], [
2407 'page_namespace',
2408 'page_title',
2409 'page_id',
2410 'page_latest',
2411 'page_is_redirect',
2412 'page_len',
2413 ] );
2414 $ret['joins']['page'] = [ 'JOIN', [ 'page_id = rev_page' ] ];
2415 }
2416
2417 if ( in_array( 'user', $options, true ) ) {
2418 $ret['tables'][] = 'user';
2419 $ret['fields'] = array_merge( $ret['fields'], [
2420 'user_name',
2421 ] );
2422 $u = $actorQuery['fields']['rev_user'];
2423 $ret['joins']['user'] = [ 'LEFT JOIN', [ "$u != 0", "user_id = $u" ] ];
2424 }
2425
2426 if ( in_array( 'text', $options, true ) ) {
2427 if ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_WRITE_OLD ) ) {
2428 throw new InvalidArgumentException( 'text table can no longer be joined directly' );
2429 } elseif ( !$this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_OLD ) ) {
2430 // NOTE: even when this class is set to not read from the old schema, callers
2431 // should still be able to join against the text table, as long as we are still
2432 // writing the old schema for compatibility.
2433 // TODO: This should trigger a deprecation warning eventually (T200918), but not
2434 // before all known usages are removed (see T198341 and T201164).
2435 // wfDeprecated( __METHOD__ . ' with `text` option', '1.32' );
2436 }
2437
2438 $ret['tables'][] = 'text';
2439 $ret['fields'] = array_merge( $ret['fields'], [
2440 'old_text',
2441 'old_flags'
2442 ] );
2443 $ret['joins']['text'] = [ 'JOIN', [ 'rev_text_id=old_id' ] ];
2444 }
2445
2446 return $ret;
2447 }
2448
2449 /**
2450 * Return the tables, fields, and join conditions to be selected to create
2451 * a new SlotRecord.
2452 *
2453 * @since 1.32
2454 *
2455 * @param array $options Any combination of the following strings
2456 * - 'content': Join with the content table, and select content meta-data fields
2457 * - 'model': Join with the content_models table, and select the model_name field.
2458 * Only applicable if 'content' is also set.
2459 * - 'role': Join with the slot_roles table, and select the role_name field
2460 *
2461 * @return array With three keys:
2462 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
2463 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
2464 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
2465 */
2466 public function getSlotsQueryInfo( $options = [] ) {
2467 $ret = [
2468 'tables' => [],
2469 'fields' => [],
2470 'joins' => [],
2471 ];
2472
2473 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_OLD ) ) {
2474 $db = $this->getDBConnectionRef( DB_REPLICA );
2475 $ret['tables'][] = 'revision';
2476
2477 $ret['fields']['slot_revision_id'] = 'rev_id';
2478 $ret['fields']['slot_content_id'] = 'NULL';
2479 $ret['fields']['slot_origin'] = 'rev_id';
2480 $ret['fields']['role_name'] = $db->addQuotes( SlotRecord::MAIN );
2481
2482 if ( in_array( 'content', $options, true ) ) {
2483 $ret['fields']['content_size'] = 'rev_len';
2484 $ret['fields']['content_sha1'] = 'rev_sha1';
2485 $ret['fields']['content_address']
2486 = $db->buildConcat( [ $db->addQuotes( 'tt:' ), 'rev_text_id' ] );
2487
2488 // Allow the content_id field to be emulated later
2489 $ret['fields']['rev_text_id'] = 'rev_text_id';
2490
2491 if ( $this->contentHandlerUseDB ) {
2492 $ret['fields']['model_name'] = 'rev_content_model';
2493 } else {
2494 $ret['fields']['model_name'] = 'NULL';
2495 }
2496 }
2497 } else {
2498 $ret['tables'][] = 'slots';
2499 $ret['fields'] = array_merge( $ret['fields'], [
2500 'slot_revision_id',
2501 'slot_content_id',
2502 'slot_origin',
2503 'slot_role_id',
2504 ] );
2505
2506 if ( in_array( 'role', $options, true ) ) {
2507 // Use left join to attach role name, so we still find the revision row even
2508 // if the role name is missing. This triggers a more obvious failure mode.
2509 $ret['tables'][] = 'slot_roles';
2510 $ret['joins']['slot_roles'] = [ 'LEFT JOIN', [ 'slot_role_id = role_id' ] ];
2511 $ret['fields'][] = 'role_name';
2512 }
2513
2514 if ( in_array( 'content', $options, true ) ) {
2515 $ret['tables'][] = 'content';
2516 $ret['fields'] = array_merge( $ret['fields'], [
2517 'content_size',
2518 'content_sha1',
2519 'content_address',
2520 'content_model',
2521 ] );
2522 $ret['joins']['content'] = [ 'JOIN', [ 'slot_content_id = content_id' ] ];
2523
2524 if ( in_array( 'model', $options, true ) ) {
2525 // Use left join to attach model name, so we still find the revision row even
2526 // if the model name is missing. This triggers a more obvious failure mode.
2527 $ret['tables'][] = 'content_models';
2528 $ret['joins']['content_models'] = [ 'LEFT JOIN', [ 'content_model = model_id' ] ];
2529 $ret['fields'][] = 'model_name';
2530 }
2531
2532 }
2533 }
2534
2535 return $ret;
2536 }
2537
2538 /**
2539 * Return the tables, fields, and join conditions to be selected to create
2540 * a new RevisionArchiveRecord object.
2541 *
2542 * MCR migration note: this replaces Revision::getArchiveQueryInfo
2543 *
2544 * @since 1.31
2545 *
2546 * @return array With three keys:
2547 * - tables: (string[]) to include in the `$table` to `IDatabase->select()`
2548 * - fields: (string[]) to include in the `$vars` to `IDatabase->select()`
2549 * - joins: (array) to include in the `$join_conds` to `IDatabase->select()`
2550 */
2551 public function getArchiveQueryInfo() {
2552 $commentQuery = $this->commentStore->getJoin( 'ar_comment' );
2553 $actorQuery = $this->actorMigration->getJoin( 'ar_user' );
2554 $ret = [
2555 'tables' => [ 'archive' ] + $commentQuery['tables'] + $actorQuery['tables'],
2556 'fields' => [
2557 'ar_id',
2558 'ar_page_id',
2559 'ar_namespace',
2560 'ar_title',
2561 'ar_rev_id',
2562 'ar_timestamp',
2563 'ar_minor_edit',
2564 'ar_deleted',
2565 'ar_len',
2566 'ar_parent_id',
2567 'ar_sha1',
2568 ] + $commentQuery['fields'] + $actorQuery['fields'],
2569 'joins' => $commentQuery['joins'] + $actorQuery['joins'],
2570 ];
2571
2572 if ( $this->hasMcrSchemaFlags( SCHEMA_COMPAT_READ_OLD ) ) {
2573 $ret['fields'][] = 'ar_text_id';
2574
2575 if ( $this->contentHandlerUseDB ) {
2576 $ret['fields'][] = 'ar_content_format';
2577 $ret['fields'][] = 'ar_content_model';
2578 }
2579 }
2580
2581 return $ret;
2582 }
2583
2584 /**
2585 * Do a batched query for the sizes of a set of revisions.
2586 *
2587 * MCR migration note: this replaces Revision::getParentLengths
2588 *
2589 * @param int[] $revIds
2590 * @return int[] associative array mapping revision IDs from $revIds to the nominal size
2591 * of the corresponding revision.
2592 */
2593 public function getRevisionSizes( array $revIds ) {
2594 return $this->listRevisionSizes( $this->getDBConnection( DB_REPLICA ), $revIds );
2595 }
2596
2597 /**
2598 * Do a batched query for the sizes of a set of revisions.
2599 *
2600 * MCR migration note: this replaces Revision::getParentLengths
2601 *
2602 * @deprecated use RevisionStore::getRevisionSizes instead.
2603 *
2604 * @param IDatabase $db
2605 * @param int[] $revIds
2606 * @return int[] associative array mapping revision IDs from $revIds to the nominal size
2607 * of the corresponding revision.
2608 */
2609 public function listRevisionSizes( IDatabase $db, array $revIds ) {
2610 $this->checkDatabaseWikiId( $db );
2611
2612 $revLens = [];
2613 if ( !$revIds ) {
2614 return $revLens; // empty
2615 }
2616
2617 $res = $db->select(
2618 'revision',
2619 [ 'rev_id', 'rev_len' ],
2620 [ 'rev_id' => $revIds ],
2621 __METHOD__
2622 );
2623
2624 foreach ( $res as $row ) {
2625 $revLens[$row->rev_id] = intval( $row->rev_len );
2626 }
2627
2628 return $revLens;
2629 }
2630
2631 /**
2632 * Implementation of getPreviousRevision and getNextRevision.
2633 *
2634 * @param RevisionRecord $rev
2635 * @param int $flags
2636 * @param string $dir 'next' or 'prev'
2637 * @return RevisionRecord|null
2638 */
2639 private function getRelativeRevision( RevisionRecord $rev, $flags, $dir ) {
2640 $op = $dir === 'next' ? '>' : '<';
2641 $sort = $dir === 'next' ? 'ASC' : 'DESC';
2642
2643 if ( !$rev->getId() || !$rev->getPageId() ) {
2644 // revision is unsaved or otherwise incomplete
2645 return null;
2646 }
2647
2648 if ( $rev instanceof RevisionArchiveRecord ) {
2649 // revision is deleted, so it's not part of the page history
2650 return null;
2651 }
2652
2653 list( $dbType, ) = DBAccessObjectUtils::getDBOptions( $flags );
2654 $db = $this->getDBConnection( $dbType, [ 'contributions' ] );
2655
2656 $ts = $this->getTimestampFromId( $rev->getId(), $flags );
2657 if ( $ts === false ) {
2658 // XXX Should this be moved into getTimestampFromId?
2659 $ts = $db->selectField( 'archive', 'ar_timestamp',
2660 [ 'ar_rev_id' => $rev->getId() ], __METHOD__ );
2661 if ( $ts === false ) {
2662 // XXX Is this reachable? How can we have a page id but no timestamp?
2663 return null;
2664 }
2665 }
2666 $ts = $db->addQuotes( $db->timestamp( $ts ) );
2667
2668 $revId = $db->selectField( 'revision', 'rev_id',
2669 [
2670 'rev_page' => $rev->getPageId(),
2671 "rev_timestamp $op $ts OR (rev_timestamp = $ts AND rev_id $op {$rev->getId()})"
2672 ],
2673 __METHOD__,
2674 [
2675 'ORDER BY' => "rev_timestamp $sort, rev_id $sort",
2676 'IGNORE INDEX' => 'rev_timestamp', // Probably needed for T159319
2677 ]
2678 );
2679
2680 if ( $revId === false ) {
2681 return null;
2682 }
2683
2684 return $this->getRevisionById( intval( $revId ) );
2685 }
2686
2687 /**
2688 * Get the revision before $rev in the page's history, if any.
2689 * Will return null for the first revision but also for deleted or unsaved revisions.
2690 *
2691 * MCR migration note: this replaces Revision::getPrevious
2692 *
2693 * @see Title::getPreviousRevisionID
2694 * @see PageArchive::getPreviousRevision
2695 *
2696 * @param RevisionRecord $rev
2697 * @param int $flags (optional) $flags include:
2698 * IDBAccessObject::READ_LATEST: Select the data from the master
2699 *
2700 * @return RevisionRecord|null
2701 */
2702 public function getPreviousRevision( RevisionRecord $rev, $flags = 0 ) {
2703 if ( $flags instanceof Title ) {
2704 // Old calling convention, we don't use Title here anymore
2705 wfDeprecated( __METHOD__ . ' with Title', '1.34' );
2706 $flags = 0;
2707 }
2708
2709 return $this->getRelativeRevision( $rev, $flags, 'prev' );
2710 }
2711
2712 /**
2713 * Get the revision after $rev in the page's history, if any.
2714 * Will return null for the latest revision but also for deleted or unsaved revisions.
2715 *
2716 * MCR migration note: this replaces Revision::getNext
2717 *
2718 * @see Title::getNextRevisionID
2719 *
2720 * @param RevisionRecord $rev
2721 * @param int $flags (optional) $flags include:
2722 * IDBAccessObject::READ_LATEST: Select the data from the master
2723 * @return RevisionRecord|null
2724 */
2725 public function getNextRevision( RevisionRecord $rev, $flags = 0 ) {
2726 if ( $flags instanceof Title ) {
2727 // Old calling convention, we don't use Title here anymore
2728 wfDeprecated( __METHOD__ . ' with Title', '1.34' );
2729 $flags = 0;
2730 }
2731
2732 return $this->getRelativeRevision( $rev, $flags, 'next' );
2733 }
2734
2735 /**
2736 * Get previous revision Id for this page_id
2737 * This is used to populate rev_parent_id on save
2738 *
2739 * MCR migration note: this corresponds to Revision::getPreviousRevisionId
2740 *
2741 * @param IDatabase $db
2742 * @param RevisionRecord $rev
2743 *
2744 * @return int
2745 */
2746 private function getPreviousRevisionId( IDatabase $db, RevisionRecord $rev ) {
2747 $this->checkDatabaseWikiId( $db );
2748
2749 if ( $rev->getPageId() === null ) {
2750 return 0;
2751 }
2752 # Use page_latest if ID is not given
2753 if ( !$rev->getId() ) {
2754 $prevId = $db->selectField(
2755 'page', 'page_latest',
2756 [ 'page_id' => $rev->getPageId() ],
2757 __METHOD__
2758 );
2759 } else {
2760 $prevId = $db->selectField(
2761 'revision', 'rev_id',
2762 [ 'rev_page' => $rev->getPageId(), 'rev_id < ' . $rev->getId() ],
2763 __METHOD__,
2764 [ 'ORDER BY' => 'rev_id DESC' ]
2765 );
2766 }
2767 return intval( $prevId );
2768 }
2769
2770 /**
2771 * Get rev_timestamp from rev_id, without loading the rest of the row.
2772 *
2773 * Historically, there was an extra Title parameter that was passed before $id. This is no
2774 * longer needed and is deprecated in 1.34.
2775 *
2776 * MCR migration note: this replaces Revision::getTimestampFromId
2777 *
2778 * @param int $id
2779 * @param int $flags
2780 * @return string|bool False if not found
2781 */
2782 public function getTimestampFromId( $id, $flags = 0 ) {
2783 if ( $id instanceof Title ) {
2784 // Old deprecated calling convention supported for backwards compatibility
2785 $id = $flags;
2786 $flags = func_num_args() > 2 ? func_get_arg( 2 ) : 0;
2787 }
2788 $db = $this->getDBConnectionRefForQueryFlags( $flags );
2789
2790 $timestamp =
2791 $db->selectField( 'revision', 'rev_timestamp', [ 'rev_id' => $id ], __METHOD__ );
2792
2793 return ( $timestamp !== false ) ? wfTimestamp( TS_MW, $timestamp ) : false;
2794 }
2795
2796 /**
2797 * Get count of revisions per page...not very efficient
2798 *
2799 * MCR migration note: this replaces Revision::countByPageId
2800 *
2801 * @param IDatabase $db
2802 * @param int $id Page id
2803 * @return int
2804 */
2805 public function countRevisionsByPageId( IDatabase $db, $id ) {
2806 $this->checkDatabaseWikiId( $db );
2807
2808 $row = $db->selectRow( 'revision',
2809 [ 'revCount' => 'COUNT(*)' ],
2810 [ 'rev_page' => $id ],
2811 __METHOD__
2812 );
2813 if ( $row ) {
2814 return intval( $row->revCount );
2815 }
2816 return 0;
2817 }
2818
2819 /**
2820 * Get count of revisions per page...not very efficient
2821 *
2822 * MCR migration note: this replaces Revision::countByTitle
2823 *
2824 * @param IDatabase $db
2825 * @param Title $title
2826 * @return int
2827 */
2828 public function countRevisionsByTitle( IDatabase $db, $title ) {
2829 $id = $title->getArticleID();
2830 if ( $id ) {
2831 return $this->countRevisionsByPageId( $db, $id );
2832 }
2833 return 0;
2834 }
2835
2836 /**
2837 * Check if no edits were made by other users since
2838 * the time a user started editing the page. Limit to
2839 * 50 revisions for the sake of performance.
2840 *
2841 * MCR migration note: this replaces Revision::userWasLastToEdit
2842 *
2843 * @deprecated since 1.31; Can possibly be removed, since the self-conflict suppression
2844 * logic in EditPage that uses this seems conceptually dubious. Revision::userWasLastToEdit
2845 * has been deprecated since 1.24.
2846 *
2847 * @param IDatabase $db The Database to perform the check on.
2848 * @param int $pageId The ID of the page in question
2849 * @param int $userId The ID of the user in question
2850 * @param string $since Look at edits since this time
2851 *
2852 * @return bool True if the given user was the only one to edit since the given timestamp
2853 */
2854 public function userWasLastToEdit( IDatabase $db, $pageId, $userId, $since ) {
2855 $this->checkDatabaseWikiId( $db );
2856
2857 if ( !$userId ) {
2858 return false;
2859 }
2860
2861 $revQuery = $this->getQueryInfo();
2862 $res = $db->select(
2863 $revQuery['tables'],
2864 [
2865 'rev_user' => $revQuery['fields']['rev_user'],
2866 ],
2867 [
2868 'rev_page' => $pageId,
2869 'rev_timestamp > ' . $db->addQuotes( $db->timestamp( $since ) )
2870 ],
2871 __METHOD__,
2872 [ 'ORDER BY' => 'rev_timestamp ASC', 'LIMIT' => 50 ],
2873 $revQuery['joins']
2874 );
2875 foreach ( $res as $row ) {
2876 if ( $row->rev_user != $userId ) {
2877 return false;
2878 }
2879 }
2880 return true;
2881 }
2882
2883 /**
2884 * Load a revision based on a known page ID and current revision ID from the DB
2885 *
2886 * This method allows for the use of caching, though accessing anything that normally
2887 * requires permission checks (aside from the text) will trigger a small DB lookup.
2888 *
2889 * MCR migration note: this replaces Revision::newKnownCurrent
2890 *
2891 * @param Title $title the associated page title
2892 * @param int $revId current revision of this page. Defaults to $title->getLatestRevID().
2893 *
2894 * @return RevisionRecord|bool Returns false if missing
2895 */
2896 public function getKnownCurrentRevision( Title $title, $revId ) {
2897 $db = $this->getDBConnectionRef( DB_REPLICA );
2898
2899 $pageId = $title->getArticleID();
2900
2901 if ( !$pageId ) {
2902 return false;
2903 }
2904
2905 if ( !$revId ) {
2906 $revId = $title->getLatestRevID();
2907 }
2908
2909 if ( !$revId ) {
2910 wfWarn(
2911 'No latest revision known for page ' . $title->getPrefixedDBkey()
2912 . ' even though it exists with page ID ' . $pageId
2913 );
2914 return false;
2915 }
2916
2917 // Load the row from cache if possible. If not possible, populate the cache.
2918 // As a minor optimization, remember if this was a cache hit or miss.
2919 // We can sometimes avoid a database query later if this is a cache miss.
2920 $fromCache = true;
2921 $row = $this->cache->getWithSetCallback(
2922 // Page/rev IDs passed in from DB to reflect history merges
2923 $this->getRevisionRowCacheKey( $db, $pageId, $revId ),
2924 WANObjectCache::TTL_WEEK,
2925 function ( $curValue, &$ttl, array &$setOpts ) use (
2926 $db, $pageId, $revId, &$fromCache
2927 ) {
2928 $setOpts += Database::getCacheSetOptions( $db );
2929 $row = $this->fetchRevisionRowFromConds( $db, [ 'rev_id' => intval( $revId ) ] );
2930 if ( $row ) {
2931 $fromCache = false;
2932 }
2933 return $row; // don't cache negatives
2934 }
2935 );
2936
2937 // Reflect revision deletion and user renames.
2938 if ( $row ) {
2939 return $this->newRevisionFromRow( $row, 0, $title, $fromCache );
2940 } else {
2941 return false;
2942 }
2943 }
2944
2945 /**
2946 * Get a cache key for use with a row as selected with getQueryInfo( [ 'page', 'user' ] )
2947 * Caching rows without 'page' or 'user' could lead to issues.
2948 * If the format of the rows returned by the query provided by getQueryInfo changes the
2949 * cache key should be updated to avoid conflicts.
2950 *
2951 * @param IDatabase $db
2952 * @param int $pageId
2953 * @param int $revId
2954 * @return string
2955 */
2956 private function getRevisionRowCacheKey( IDatabase $db, $pageId, $revId ) {
2957 return $this->cache->makeGlobalKey(
2958 self::ROW_CACHE_KEY,
2959 $db->getDomainID(),
2960 $pageId,
2961 $revId
2962 );
2963 }
2964
2965 // TODO: move relevant methods from Title here, e.g. getFirstRevision, isBigDeletion, etc.
2966
2967 }
2968
2969 /**
2970 * Retain the old class name for backwards compatibility.
2971 * @deprecated since 1.32
2972 */
2973 class_alias( RevisionStore::class, 'MediaWiki\Storage\RevisionStore' );