(bug 17947) Use current date/time on Special:Preferences rather than Wikipedia day...
[lhc/web/wiklou.git] / includes / LinksUpdate.php
1 <?php
2 /**
3 * See docs/deferred.txt
4 *
5 * @todo document (e.g. one-sentence top-level class description).
6 */
7 class LinksUpdate {
8
9 /**@{{
10 * @private
11 */
12 var $mId, //!< Page ID of the article linked from
13 $mTitle, //!< Title object of the article linked from
14 $mLinks, //!< Map of title strings to IDs for the links in the document
15 $mImages, //!< DB keys of the images used, in the array key only
16 $mTemplates, //!< Map of title strings to IDs for the template references, including broken ones
17 $mExternals, //!< URLs of external links, array key only
18 $mCategories, //!< Map of category names to sort keys
19 $mInterlangs, //!< Map of language codes to titles
20 $mProperties, //!< Map of arbitrary name to value
21 $mDb, //!< Database connection reference
22 $mOptions, //!< SELECT options to be used (array)
23 $mRecursive; //!< Whether to queue jobs for recursive updates
24 /**@}}*/
25
26 /**
27 * Constructor
28 *
29 * @param Title $title Title of the page we're updating
30 * @param ParserOutput $parserOutput Output from a full parse of this page
31 * @param bool $recursive Queue jobs for recursive updates?
32 */
33 function LinksUpdate( $title, $parserOutput, $recursive = true ) {
34 global $wgAntiLockFlags;
35
36 if ( $wgAntiLockFlags & ALF_NO_LINK_LOCK ) {
37 $this->mOptions = array();
38 } else {
39 $this->mOptions = array( 'FOR UPDATE' );
40 }
41 $this->mDb = wfGetDB( DB_MASTER );
42
43 if ( !is_object( $title ) ) {
44 throw new MWException( "The calling convention to LinksUpdate::LinksUpdate() has changed. " .
45 "Please see Article::editUpdates() for an invocation example.\n" );
46 }
47 $this->mTitle = $title;
48 $this->mId = $title->getArticleID();
49
50 $this->mParserOutput = $parserOutput;
51 $this->mLinks = $parserOutput->getLinks();
52 $this->mImages = $parserOutput->getImages();
53 $this->mTemplates = $parserOutput->getTemplates();
54 $this->mExternals = $parserOutput->getExternalLinks();
55 $this->mCategories = $parserOutput->getCategories();
56 $this->mProperties = $parserOutput->getProperties();
57
58 # Convert the format of the interlanguage links
59 # I didn't want to change it in the ParserOutput, because that array is passed all
60 # the way back to the skin, so either a skin API break would be required, or an
61 # inefficient back-conversion.
62 $ill = $parserOutput->getLanguageLinks();
63 $this->mInterlangs = array();
64 foreach ( $ill as $link ) {
65 list( $key, $title ) = explode( ':', $link, 2 );
66 $this->mInterlangs[$key] = $title;
67 }
68
69 $this->mRecursive = $recursive;
70 $this->mTouchTmplLinks = false;
71
72 wfRunHooks( 'LinksUpdateConstructed', array( &$this ) );
73 }
74
75 /**
76 * Update link tables with outgoing links from an updated article
77 */
78 public function doUpdate() {
79 global $wgUseDumbLinkUpdate;
80
81 wfRunHooks( 'LinksUpdate', array( &$this ) );
82 if ( $wgUseDumbLinkUpdate ) {
83 $this->doDumbUpdate();
84 } else {
85 $this->doIncrementalUpdate();
86 }
87 wfRunHooks( 'LinksUpdateComplete', array( &$this ) );
88 }
89
90 protected function doIncrementalUpdate() {
91 wfProfileIn( __METHOD__ );
92
93 # Page links
94 $existing = $this->getExistingLinks();
95 $this->incrTableUpdate( 'pagelinks', 'pl', $this->getLinkDeletions( $existing ),
96 $this->getLinkInsertions( $existing ) );
97
98 # Image links
99 $existing = $this->getExistingImages();
100
101 $imageDeletes = $this->getImageDeletions( $existing );
102 $this->incrTableUpdate( 'imagelinks', 'il', $imageDeletes, $this->getImageInsertions( $existing ) );
103
104 # Invalidate all image description pages which had links added or removed
105 $imageUpdates = $imageDeletes + array_diff_key( $this->mImages, $existing );
106 $this->invalidateImageDescriptions( $imageUpdates );
107
108 # External links
109 $existing = $this->getExistingExternals();
110 $this->incrTableUpdate( 'externallinks', 'el', $this->getExternalDeletions( $existing ),
111 $this->getExternalInsertions( $existing ) );
112
113 # Language links
114 $existing = $this->getExistingInterlangs();
115 $this->incrTableUpdate( 'langlinks', 'll', $this->getInterlangDeletions( $existing ),
116 $this->getInterlangInsertions( $existing ) );
117
118 # Template links
119 $existing = $this->getExistingTemplates();
120 $this->incrTableUpdate( 'templatelinks', 'tl', $this->getTemplateDeletions( $existing ),
121 $this->getTemplateInsertions( $existing ) );
122
123 # Category links
124 $existing = $this->getExistingCategories();
125
126 $categoryDeletes = $this->getCategoryDeletions( $existing );
127
128 $this->incrTableUpdate( 'categorylinks', 'cl', $categoryDeletes, $this->getCategoryInsertions( $existing ) );
129
130 # Invalidate all categories which were added, deleted or changed (set symmetric difference)
131 $categoryInserts = array_diff_assoc( $this->mCategories, $existing );
132 $categoryUpdates = $categoryInserts + $categoryDeletes;
133 $this->invalidateCategories( $categoryUpdates );
134 $this->updateCategoryCounts( $categoryInserts, $categoryDeletes );
135
136 # Page properties
137 $existing = $this->getExistingProperties();
138
139 $propertiesDeletes = $this->getPropertyDeletions( $existing );
140
141 $this->incrTableUpdate( 'page_props', 'pp', $propertiesDeletes, $this->getPropertyInsertions( $existing ) );
142
143 # Invalidate the necessary pages
144 $changed = $propertiesDeletes + array_diff_assoc( $this->mProperties, $existing );
145 $this->invalidateProperties( $changed );
146
147 # Refresh links of all pages including this page
148 # This will be in a separate transaction
149 if ( $this->mRecursive ) {
150 $this->queueRecursiveJobs();
151 }
152
153 wfProfileOut( __METHOD__ );
154 }
155
156 /**
157 * Link update which clears the previous entries and inserts new ones
158 * May be slower or faster depending on level of lock contention and write speed of DB
159 * Also useful where link table corruption needs to be repaired, e.g. in refreshLinks.php
160 */
161 protected function doDumbUpdate() {
162 wfProfileIn( __METHOD__ );
163
164 # Refresh category pages and image description pages
165 $existing = $this->getExistingCategories();
166 $categoryInserts = array_diff_assoc( $this->mCategories, $existing );
167 $categoryDeletes = array_diff_assoc( $existing, $this->mCategories );
168 $categoryUpdates = $categoryInserts + $categoryDeletes;
169 $existing = $this->getExistingImages();
170 $imageUpdates = array_diff_key( $existing, $this->mImages ) + array_diff_key( $this->mImages, $existing );
171
172 $this->dumbTableUpdate( 'pagelinks', $this->getLinkInsertions(), 'pl_from' );
173 $this->dumbTableUpdate( 'imagelinks', $this->getImageInsertions(), 'il_from' );
174 $this->dumbTableUpdate( 'categorylinks', $this->getCategoryInsertions(), 'cl_from' );
175 $this->dumbTableUpdate( 'templatelinks', $this->getTemplateInsertions(), 'tl_from' );
176 $this->dumbTableUpdate( 'externallinks', $this->getExternalInsertions(), 'el_from' );
177 $this->dumbTableUpdate( 'langlinks', $this->getInterlangInsertions(),'ll_from' );
178 $this->dumbTableUpdate( 'page_props', $this->getPropertyInsertions(), 'pp_page' );
179
180 # Update the cache of all the category pages and image description
181 # pages which were changed, and fix the category table count
182 $this->invalidateCategories( $categoryUpdates );
183 $this->updateCategoryCounts( $categoryInserts, $categoryDeletes );
184 $this->invalidateImageDescriptions( $imageUpdates );
185
186 # Refresh links of all pages including this page
187 # This will be in a separate transaction
188 if ( $this->mRecursive ) {
189 $this->queueRecursiveJobs();
190 }
191
192 wfProfileOut( __METHOD__ );
193 }
194
195 function queueRecursiveJobs() {
196 global $wgUpdateRowsPerJob;
197 wfProfileIn( __METHOD__ );
198
199 $cache = $this->mTitle->getBacklinkCache();
200 $batches = $cache->partition( 'templatelinks', $wgUpdateRowsPerJob );
201 if ( !$batches ) {
202 wfProfileOut( __METHOD__ );
203 return;
204 }
205 $jobs = array();
206 foreach ( $batches as $batch ) {
207 list( $start, $end ) = $batch;
208 $params = array(
209 'start' => $start,
210 'end' => $end,
211 );
212 $jobs[] = new RefreshLinksJob2( $this->mTitle, $params );
213 }
214 Job::batchInsert( $jobs );
215
216 wfProfileOut( __METHOD__ );
217 }
218
219 /**
220 * Invalidate the cache of a list of pages from a single namespace
221 *
222 * @param integer $namespace
223 * @param array $dbkeys
224 */
225 function invalidatePages( $namespace, $dbkeys ) {
226 if ( !count( $dbkeys ) ) {
227 return;
228 }
229
230 /**
231 * Determine which pages need to be updated
232 * This is necessary to prevent the job queue from smashing the DB with
233 * large numbers of concurrent invalidations of the same page
234 */
235 $now = $this->mDb->timestamp();
236 $ids = array();
237 $res = $this->mDb->select( 'page', array( 'page_id' ),
238 array(
239 'page_namespace' => $namespace,
240 'page_title IN (' . $this->mDb->makeList( $dbkeys ) . ')',
241 'page_touched < ' . $this->mDb->addQuotes( $now )
242 ), __METHOD__
243 );
244 while ( $row = $this->mDb->fetchObject( $res ) ) {
245 $ids[] = $row->page_id;
246 }
247 if ( !count( $ids ) ) {
248 return;
249 }
250
251 /**
252 * Do the update
253 * We still need the page_touched condition, in case the row has changed since
254 * the non-locking select above.
255 */
256 $this->mDb->update( 'page', array( 'page_touched' => $now ),
257 array(
258 'page_id IN (' . $this->mDb->makeList( $ids ) . ')',
259 'page_touched < ' . $this->mDb->addQuotes( $now )
260 ), __METHOD__
261 );
262 }
263
264 function invalidateCategories( $cats ) {
265 $this->invalidatePages( NS_CATEGORY, array_keys( $cats ) );
266 }
267
268 /**
269 * Update all the appropriate counts in the category table.
270 * @param $added associative array of category name => sort key
271 * @param $deleted associative array of category name => sort key
272 */
273 function updateCategoryCounts( $added, $deleted ) {
274 $a = new Article($this->mTitle);
275 $a->updateCategoryCounts(
276 array_keys( $added ), array_keys( $deleted )
277 );
278 }
279
280 function invalidateImageDescriptions( $images ) {
281 $this->invalidatePages( NS_FILE, array_keys( $images ) );
282 }
283
284 function dumbTableUpdate( $table, $insertions, $fromField ) {
285 $this->mDb->delete( $table, array( $fromField => $this->mId ), __METHOD__ );
286 if ( count( $insertions ) ) {
287 # The link array was constructed without FOR UPDATE, so there may
288 # be collisions. This may cause minor link table inconsistencies,
289 # which is better than crippling the site with lock contention.
290 $this->mDb->insert( $table, $insertions, __METHOD__, array( 'IGNORE' ) );
291 }
292 }
293
294 /**
295 * Make a WHERE clause from a 2-d NS/dbkey array
296 *
297 * @param array $arr 2-d array indexed by namespace and DB key
298 * @param string $prefix Field name prefix, without the underscore
299 */
300 function makeWhereFrom2d( &$arr, $prefix ) {
301 $lb = new LinkBatch;
302 $lb->setArray( $arr );
303 return $lb->constructSet( $prefix, $this->mDb );
304 }
305
306 /**
307 * Update a table by doing a delete query then an insert query
308 * @private
309 */
310 function incrTableUpdate( $table, $prefix, $deletions, $insertions ) {
311 if ( $table == 'page_props' ) {
312 $fromField = 'pp_page';
313 } else {
314 $fromField = "{$prefix}_from";
315 }
316 $where = array( $fromField => $this->mId );
317 if ( $table == 'pagelinks' || $table == 'templatelinks' ) {
318 $clause = $this->makeWhereFrom2d( $deletions, $prefix );
319 if ( $clause ) {
320 $where[] = $clause;
321 } else {
322 $where = false;
323 }
324 } else {
325 if ( $table == 'langlinks' ) {
326 $toField = 'll_lang';
327 } elseif ( $table == 'page_props' ) {
328 $toField = 'pp_propname';
329 } else {
330 $toField = $prefix . '_to';
331 }
332 if ( count( $deletions ) ) {
333 $where[] = "$toField IN (" . $this->mDb->makeList( array_keys( $deletions ) ) . ')';
334 } else {
335 $where = false;
336 }
337 }
338 if ( $where ) {
339 $this->mDb->delete( $table, $where, __METHOD__ );
340 }
341 if ( count( $insertions ) ) {
342 $this->mDb->insert( $table, $insertions, __METHOD__, 'IGNORE' );
343 }
344 }
345
346
347 /**
348 * Get an array of pagelinks insertions for passing to the DB
349 * Skips the titles specified by the 2-D array $existing
350 * @private
351 */
352 function getLinkInsertions( $existing = array() ) {
353 $arr = array();
354 foreach( $this->mLinks as $ns => $dbkeys ) {
355 # array_diff_key() was introduced in PHP 5.1, there is a compatibility function
356 # in GlobalFunctions.php
357 $diffs = isset( $existing[$ns] ) ? array_diff_key( $dbkeys, $existing[$ns] ) : $dbkeys;
358 foreach ( $diffs as $dbk => $id ) {
359 $arr[] = array(
360 'pl_from' => $this->mId,
361 'pl_namespace' => $ns,
362 'pl_title' => $dbk
363 );
364 }
365 }
366 return $arr;
367 }
368
369 /**
370 * Get an array of template insertions. Like getLinkInsertions()
371 * @private
372 */
373 function getTemplateInsertions( $existing = array() ) {
374 $arr = array();
375 foreach( $this->mTemplates as $ns => $dbkeys ) {
376 $diffs = isset( $existing[$ns] ) ? array_diff_key( $dbkeys, $existing[$ns] ) : $dbkeys;
377 foreach ( $diffs as $dbk => $id ) {
378 $arr[] = array(
379 'tl_from' => $this->mId,
380 'tl_namespace' => $ns,
381 'tl_title' => $dbk
382 );
383 }
384 }
385 return $arr;
386 }
387
388 /**
389 * Get an array of image insertions
390 * Skips the names specified in $existing
391 * @private
392 */
393 function getImageInsertions( $existing = array() ) {
394 $arr = array();
395 $diffs = array_diff_key( $this->mImages, $existing );
396 foreach( $diffs as $iname => $dummy ) {
397 $arr[] = array(
398 'il_from' => $this->mId,
399 'il_to' => $iname
400 );
401 }
402 return $arr;
403 }
404
405 /**
406 * Get an array of externallinks insertions. Skips the names specified in $existing
407 * @private
408 */
409 function getExternalInsertions( $existing = array() ) {
410 $arr = array();
411 $diffs = array_diff_key( $this->mExternals, $existing );
412 foreach( $diffs as $url => $dummy ) {
413 $arr[] = array(
414 'el_from' => $this->mId,
415 'el_to' => $url,
416 'el_index' => wfMakeUrlIndex( $url ),
417 );
418 }
419 return $arr;
420 }
421
422 /**
423 * Get an array of category insertions
424 * @param array $existing Array mapping existing category names to sort keys. If both
425 * match a link in $this, the link will be omitted from the output
426 * @private
427 */
428 function getCategoryInsertions( $existing = array() ) {
429 global $wgContLang;
430 $diffs = array_diff_assoc( $this->mCategories, $existing );
431 $arr = array();
432 foreach ( $diffs as $name => $sortkey ) {
433 $nt = Title::makeTitleSafe( NS_CATEGORY, $name );
434 $wgContLang->findVariantLink( $name, $nt, true );
435 $arr[] = array(
436 'cl_from' => $this->mId,
437 'cl_to' => $name,
438 'cl_sortkey' => $sortkey,
439 'cl_timestamp' => $this->mDb->timestamp()
440 );
441 }
442 return $arr;
443 }
444
445 /**
446 * Get an array of interlanguage link insertions
447 * @param array $existing Array mapping existing language codes to titles
448 * @private
449 */
450 function getInterlangInsertions( $existing = array() ) {
451 $diffs = array_diff_assoc( $this->mInterlangs, $existing );
452 $arr = array();
453 foreach( $diffs as $lang => $title ) {
454 $arr[] = array(
455 'll_from' => $this->mId,
456 'll_lang' => $lang,
457 'll_title' => $title
458 );
459 }
460 return $arr;
461 }
462
463 /**
464 * Get an array of page property insertions
465 */
466 function getPropertyInsertions( $existing = array() ) {
467 $diffs = array_diff_assoc( $this->mProperties, $existing );
468 $arr = array();
469 foreach ( $diffs as $name => $value ) {
470 $arr[] = array(
471 'pp_page' => $this->mId,
472 'pp_propname' => $name,
473 'pp_value' => $value,
474 );
475 }
476 return $arr;
477 }
478
479
480 /**
481 * Given an array of existing links, returns those links which are not in $this
482 * and thus should be deleted.
483 * @private
484 */
485 function getLinkDeletions( $existing ) {
486 $del = array();
487 foreach ( $existing as $ns => $dbkeys ) {
488 if ( isset( $this->mLinks[$ns] ) ) {
489 $del[$ns] = array_diff_key( $existing[$ns], $this->mLinks[$ns] );
490 } else {
491 $del[$ns] = $existing[$ns];
492 }
493 }
494 return $del;
495 }
496
497 /**
498 * Given an array of existing templates, returns those templates which are not in $this
499 * and thus should be deleted.
500 * @private
501 */
502 function getTemplateDeletions( $existing ) {
503 $del = array();
504 foreach ( $existing as $ns => $dbkeys ) {
505 if ( isset( $this->mTemplates[$ns] ) ) {
506 $del[$ns] = array_diff_key( $existing[$ns], $this->mTemplates[$ns] );
507 } else {
508 $del[$ns] = $existing[$ns];
509 }
510 }
511 return $del;
512 }
513
514 /**
515 * Given an array of existing images, returns those images which are not in $this
516 * and thus should be deleted.
517 * @private
518 */
519 function getImageDeletions( $existing ) {
520 return array_diff_key( $existing, $this->mImages );
521 }
522
523 /**
524 * Given an array of existing external links, returns those links which are not
525 * in $this and thus should be deleted.
526 * @private
527 */
528 function getExternalDeletions( $existing ) {
529 return array_diff_key( $existing, $this->mExternals );
530 }
531
532 /**
533 * Given an array of existing categories, returns those categories which are not in $this
534 * and thus should be deleted.
535 * @private
536 */
537 function getCategoryDeletions( $existing ) {
538 return array_diff_assoc( $existing, $this->mCategories );
539 }
540
541 /**
542 * Given an array of existing interlanguage links, returns those links which are not
543 * in $this and thus should be deleted.
544 * @private
545 */
546 function getInterlangDeletions( $existing ) {
547 return array_diff_assoc( $existing, $this->mInterlangs );
548 }
549
550 /**
551 * Get array of properties which should be deleted.
552 * @private
553 */
554 function getPropertyDeletions( $existing ) {
555 return array_diff_assoc( $existing, $this->mProperties );
556 }
557
558 /**
559 * Get an array of existing links, as a 2-D array
560 * @private
561 */
562 function getExistingLinks() {
563 $res = $this->mDb->select( 'pagelinks', array( 'pl_namespace', 'pl_title' ),
564 array( 'pl_from' => $this->mId ), __METHOD__, $this->mOptions );
565 $arr = array();
566 while ( $row = $this->mDb->fetchObject( $res ) ) {
567 if ( !isset( $arr[$row->pl_namespace] ) ) {
568 $arr[$row->pl_namespace] = array();
569 }
570 $arr[$row->pl_namespace][$row->pl_title] = 1;
571 }
572 $this->mDb->freeResult( $res );
573 return $arr;
574 }
575
576 /**
577 * Get an array of existing templates, as a 2-D array
578 * @private
579 */
580 function getExistingTemplates() {
581 $res = $this->mDb->select( 'templatelinks', array( 'tl_namespace', 'tl_title' ),
582 array( 'tl_from' => $this->mId ), __METHOD__, $this->mOptions );
583 $arr = array();
584 while ( $row = $this->mDb->fetchObject( $res ) ) {
585 if ( !isset( $arr[$row->tl_namespace] ) ) {
586 $arr[$row->tl_namespace] = array();
587 }
588 $arr[$row->tl_namespace][$row->tl_title] = 1;
589 }
590 $this->mDb->freeResult( $res );
591 return $arr;
592 }
593
594 /**
595 * Get an array of existing images, image names in the keys
596 * @private
597 */
598 function getExistingImages() {
599 $res = $this->mDb->select( 'imagelinks', array( 'il_to' ),
600 array( 'il_from' => $this->mId ), __METHOD__, $this->mOptions );
601 $arr = array();
602 while ( $row = $this->mDb->fetchObject( $res ) ) {
603 $arr[$row->il_to] = 1;
604 }
605 $this->mDb->freeResult( $res );
606 return $arr;
607 }
608
609 /**
610 * Get an array of existing external links, URLs in the keys
611 * @private
612 */
613 function getExistingExternals() {
614 $res = $this->mDb->select( 'externallinks', array( 'el_to' ),
615 array( 'el_from' => $this->mId ), __METHOD__, $this->mOptions );
616 $arr = array();
617 while ( $row = $this->mDb->fetchObject( $res ) ) {
618 $arr[$row->el_to] = 1;
619 }
620 $this->mDb->freeResult( $res );
621 return $arr;
622 }
623
624 /**
625 * Get an array of existing categories, with the name in the key and sort key in the value.
626 * @private
627 */
628 function getExistingCategories() {
629 $res = $this->mDb->select( 'categorylinks', array( 'cl_to', 'cl_sortkey' ),
630 array( 'cl_from' => $this->mId ), __METHOD__, $this->mOptions );
631 $arr = array();
632 while ( $row = $this->mDb->fetchObject( $res ) ) {
633 $arr[$row->cl_to] = $row->cl_sortkey;
634 }
635 $this->mDb->freeResult( $res );
636 return $arr;
637 }
638
639 /**
640 * Get an array of existing interlanguage links, with the language code in the key and the
641 * title in the value.
642 * @private
643 */
644 function getExistingInterlangs() {
645 $res = $this->mDb->select( 'langlinks', array( 'll_lang', 'll_title' ),
646 array( 'll_from' => $this->mId ), __METHOD__, $this->mOptions );
647 $arr = array();
648 while ( $row = $this->mDb->fetchObject( $res ) ) {
649 $arr[$row->ll_lang] = $row->ll_title;
650 }
651 return $arr;
652 }
653
654 /**
655 * Get an array of existing categories, with the name in the key and sort key in the value.
656 * @private
657 */
658 function getExistingProperties() {
659 $res = $this->mDb->select( 'page_props', array( 'pp_propname', 'pp_value' ),
660 array( 'pp_page' => $this->mId ), __METHOD__, $this->mOptions );
661 $arr = array();
662 while ( $row = $this->mDb->fetchObject( $res ) ) {
663 $arr[$row->pp_propname] = $row->pp_value;
664 }
665 $this->mDb->freeResult( $res );
666 return $arr;
667 }
668
669
670 /**
671 * Return the title object of the page being updated
672 */
673 function getTitle() {
674 return $this->mTitle;
675 }
676
677 /**
678 * Return the list of images used as generated by the parser
679 */
680 public function getImages() {
681 return $this->mImages;
682 }
683
684 /**
685 * Invalidate any necessary link lists related to page property changes
686 */
687 function invalidateProperties( $changed ) {
688 global $wgPagePropLinkInvalidations;
689
690 foreach ( $changed as $name => $value ) {
691 if ( isset( $wgPagePropLinkInvalidations[$name] ) ) {
692 $inv = $wgPagePropLinkInvalidations[$name];
693 if ( !is_array( $inv ) ) {
694 $inv = array( $inv );
695 }
696 foreach ( $inv as $table ) {
697 $update = new HTMLCacheUpdate( $this->mTitle, $table );
698 $update->doUpdate();
699 }
700 }
701 }
702 }
703 }