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