Merge "Revert "Use display name in category page subheadings if provided""
[lhc/web/wiklou.git] / includes / search / SearchEngine.php
1 <?php
2 /**
3 * Basic search engine
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 * @file
21 * @ingroup Search
22 */
23
24 /**
25 * @defgroup Search Search
26 */
27
28 use MediaWiki\MediaWikiServices;
29
30 /**
31 * Contain a class for special pages
32 * @ingroup Search
33 */
34 abstract class SearchEngine {
35 /** @var string */
36 public $prefix = '';
37
38 /** @var int[]|null */
39 public $namespaces = [ NS_MAIN ];
40
41 /** @var int */
42 protected $limit = 10;
43
44 /** @var int */
45 protected $offset = 0;
46
47 /** @var array|string */
48 protected $searchTerms = [];
49
50 /** @var bool */
51 protected $showSuggestion = true;
52 private $sort = 'relevance';
53
54 /** @var array Feature values */
55 protected $features = [];
56
57 /** @const string profile type for completionSearch */
58 const COMPLETION_PROFILE_TYPE = 'completionSearchProfile';
59
60 /** @const string profile type for query independent ranking features */
61 const FT_QUERY_INDEP_PROFILE_TYPE = 'fulltextQueryIndepProfile';
62
63 /**
64 * Perform a full text search query and return a result set.
65 * If full text searches are not supported or disabled, return null.
66 * STUB
67 *
68 * @param string $term Raw search term
69 * @return SearchResultSet|Status|null
70 */
71 function searchText( $term ) {
72 return null;
73 }
74
75 /**
76 * Perform a title-only search query and return a result set.
77 * If title searches are not supported or disabled, return null.
78 * STUB
79 *
80 * @param string $term Raw search term
81 * @return SearchResultSet|null
82 */
83 function searchTitle( $term ) {
84 return null;
85 }
86
87 /**
88 * @since 1.18
89 * @param string $feature
90 * @return bool
91 */
92 public function supports( $feature ) {
93 switch ( $feature ) {
94 case 'search-update':
95 return true;
96 case 'title-suffix-filter':
97 default:
98 return false;
99 }
100 }
101
102 /**
103 * Way to pass custom data for engines
104 * @since 1.18
105 * @param string $feature
106 * @param mixed $data
107 * @return bool
108 */
109 public function setFeatureData( $feature, $data ) {
110 $this->features[$feature] = $data;
111 }
112
113 /**
114 * When overridden in derived class, performs database-specific conversions
115 * on text to be used for searching or updating search index.
116 * Default implementation does nothing (simply returns $string).
117 *
118 * @param string $string String to process
119 * @return string
120 */
121 public function normalizeText( $string ) {
122 global $wgContLang;
123
124 // Some languages such as Chinese require word segmentation
125 return $wgContLang->segmentByWord( $string );
126 }
127
128 /**
129 * Transform search term in cases when parts of the query came as different
130 * GET params (when supported), e.g. for prefix queries:
131 * search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
132 * @param string $term
133 * @return string
134 */
135 public function transformSearchTerm( $term ) {
136 return $term;
137 }
138
139 /**
140 * Get service class to finding near matches.
141 * @param Config $config Configuration to use for the matcher.
142 * @return SearchNearMatcher
143 */
144 public function getNearMatcher( Config $config ) {
145 global $wgContLang;
146 return new SearchNearMatcher( $config, $wgContLang );
147 }
148
149 /**
150 * Get near matcher for default SearchEngine.
151 * @return SearchNearMatcher
152 */
153 protected static function defaultNearMatcher() {
154 $config = MediaWikiServices::getInstance()->getMainConfig();
155 return MediaWikiServices::getInstance()->newSearchEngine()->getNearMatcher( $config );
156 }
157
158 /**
159 * If an exact title match can be found, or a very slightly close match,
160 * return the title. If no match, returns NULL.
161 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
162 * @param string $searchterm
163 * @return Title
164 */
165 public static function getNearMatch( $searchterm ) {
166 return static::defaultNearMatcher()->getNearMatch( $searchterm );
167 }
168
169 /**
170 * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
171 * SearchResultSet.
172 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
173 * @param string $searchterm
174 * @return SearchResultSet
175 */
176 public static function getNearMatchResultSet( $searchterm ) {
177 return static::defaultNearMatcher()->getNearMatchResultSet( $searchterm );
178 }
179
180 /**
181 * Get chars legal for search.
182 * NOTE: usage as static is deprecated and preserved only as BC measure
183 * @return string
184 */
185 public static function legalSearchChars() {
186 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
187 }
188
189 /**
190 * Set the maximum number of results to return
191 * and how many to skip before returning the first.
192 *
193 * @param int $limit
194 * @param int $offset
195 */
196 function setLimitOffset( $limit, $offset = 0 ) {
197 $this->limit = intval( $limit );
198 $this->offset = intval( $offset );
199 }
200
201 /**
202 * Set which namespaces the search should include.
203 * Give an array of namespace index numbers.
204 *
205 * @param int[]|null $namespaces
206 */
207 function setNamespaces( $namespaces ) {
208 if ( $namespaces ) {
209 // Filter namespaces to only keep valid ones
210 $validNs = $this->searchableNamespaces();
211 $namespaces = array_filter( $namespaces, function( $ns ) use( $validNs ) {
212 return $ns < 0 || isset( $validNs[$ns] );
213 } );
214 } else {
215 $namespaces = [];
216 }
217 $this->namespaces = $namespaces;
218 }
219
220 /**
221 * Set whether the searcher should try to build a suggestion. Note: some searchers
222 * don't support building a suggestion in the first place and others don't respect
223 * this flag.
224 *
225 * @param bool $showSuggestion Should the searcher try to build suggestions
226 */
227 function setShowSuggestion( $showSuggestion ) {
228 $this->showSuggestion = $showSuggestion;
229 }
230
231 /**
232 * Get the valid sort directions. All search engines support 'relevance' but others
233 * might support more. The default in all implementations should be 'relevance.'
234 *
235 * @since 1.25
236 * @return array(string) the valid sort directions for setSort
237 */
238 public function getValidSorts() {
239 return [ 'relevance' ];
240 }
241
242 /**
243 * Set the sort direction of the search results. Must be one returned by
244 * SearchEngine::getValidSorts()
245 *
246 * @since 1.25
247 * @throws InvalidArgumentException
248 * @param string $sort sort direction for query result
249 */
250 public function setSort( $sort ) {
251 if ( !in_array( $sort, $this->getValidSorts() ) ) {
252 throw new InvalidArgumentException( "Invalid sort: $sort. " .
253 "Must be one of: " . implode( ', ', $this->getValidSorts() ) );
254 }
255 $this->sort = $sort;
256 }
257
258 /**
259 * Get the sort direction of the search results
260 *
261 * @since 1.25
262 * @return string
263 */
264 public function getSort() {
265 return $this->sort;
266 }
267
268 /**
269 * Parse some common prefixes: all (search everything)
270 * or namespace names and set the list of namespaces
271 * of this class accordingly.
272 *
273 * @param string $query
274 * @return string
275 */
276 function replacePrefixes( $query ) {
277 $queryAndNs = self::parseNamespacePrefixes( $query );
278 if ( $queryAndNs === false ) {
279 return $query;
280 }
281 $this->namespaces = $queryAndNs[1];
282 return $queryAndNs[0];
283 }
284
285 /**
286 * Parse some common prefixes: all (search everything)
287 * or namespace names
288 *
289 * @param string $query
290 * @return false|array false if no namespace was extracted, an array
291 * with the parsed query at index 0 and an array of namespaces at index
292 * 1 (or null for all namespaces).
293 */
294 public static function parseNamespacePrefixes( $query ) {
295 global $wgContLang;
296
297 $parsed = $query;
298 if ( strpos( $query, ':' ) === false ) { // nothing to do
299 return false;
300 }
301 $extractedNamespace = null;
302
303 $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
304 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
305 $extractedNamespace = null;
306 $parsed = substr( $query, strlen( $allkeyword ) );
307 } elseif ( strpos( $query, ':' ) !== false ) {
308 // TODO: should we unify with PrefixSearch::extractNamespace ?
309 $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
310 $index = $wgContLang->getNsIndex( $prefix );
311 if ( $index !== false ) {
312 $extractedNamespace = [ $index ];
313 $parsed = substr( $query, strlen( $prefix ) + 1 );
314 } else {
315 return false;
316 }
317 }
318
319 if ( trim( $parsed ) == '' ) {
320 $parsed = $query; // prefix was the whole query
321 }
322
323 return [ $parsed, $extractedNamespace ];
324 }
325
326 /**
327 * Find snippet highlight settings for all users
328 * @return array Contextlines, contextchars
329 */
330 public static function userHighlightPrefs() {
331 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
332 $contextchars = 75; // same as above.... :P
333 return [ $contextlines, $contextchars ];
334 }
335
336 /**
337 * Create or update the search index record for the given page.
338 * Title and text should be pre-processed.
339 * STUB
340 *
341 * @param int $id
342 * @param string $title
343 * @param string $text
344 */
345 function update( $id, $title, $text ) {
346 // no-op
347 }
348
349 /**
350 * Update a search index record's title only.
351 * Title should be pre-processed.
352 * STUB
353 *
354 * @param int $id
355 * @param string $title
356 */
357 function updateTitle( $id, $title ) {
358 // no-op
359 }
360
361 /**
362 * Delete an indexed page
363 * Title should be pre-processed.
364 * STUB
365 *
366 * @param int $id Page id that was deleted
367 * @param string $title Title of page that was deleted
368 */
369 function delete( $id, $title ) {
370 // no-op
371 }
372
373 /**
374 * Get OpenSearch suggestion template
375 *
376 * @deprecated since 1.25
377 * @return string
378 */
379 public static function getOpenSearchTemplate() {
380 wfDeprecated( __METHOD__, '1.25' );
381 return ApiOpenSearch::getOpenSearchTemplate( 'application/x-suggestions+json' );
382 }
383
384 /**
385 * Get the raw text for updating the index from a content object
386 * Nicer search backends could possibly do something cooler than
387 * just returning raw text
388 *
389 * @todo This isn't ideal, we'd really like to have content-specific handling here
390 * @param Title $t Title we're indexing
391 * @param Content $c Content of the page to index
392 * @return string
393 */
394 public function getTextFromContent( Title $t, Content $c = null ) {
395 return $c ? $c->getTextForSearchIndex() : '';
396 }
397
398 /**
399 * If an implementation of SearchEngine handles all of its own text processing
400 * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
401 * rather silly handling, it should return true here instead.
402 *
403 * @return bool
404 */
405 public function textAlreadyUpdatedForIndex() {
406 return false;
407 }
408
409 /**
410 * Makes search simple string if it was namespaced.
411 * Sets namespaces of the search to namespaces extracted from string.
412 * @param string $search
413 * @return string Simplified search string
414 */
415 protected function normalizeNamespaces( $search ) {
416 // Find a Title which is not an interwiki and is in NS_MAIN
417 $title = Title::newFromText( $search );
418 $ns = $this->namespaces;
419 if ( $title && !$title->isExternal() ) {
420 $ns = [ $title->getNamespace() ];
421 $search = $title->getText();
422 if ( $ns[0] == NS_MAIN ) {
423 $ns = $this->namespaces; // no explicit prefix, use default namespaces
424 Hooks::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
425 }
426 } else {
427 $title = Title::newFromText( $search . 'Dummy' );
428 if ( $title && $title->getText() == 'Dummy'
429 && $title->getNamespace() != NS_MAIN
430 && !$title->isExternal() )
431 {
432 $ns = [ $title->getNamespace() ];
433 $search = '';
434 } else {
435 Hooks::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
436 }
437 }
438
439 $ns = array_map( function( $space ) {
440 return $space == NS_MEDIA ? NS_FILE : $space;
441 }, $ns );
442
443 $this->setNamespaces( $ns );
444 return $search;
445 }
446
447 /**
448 * Perform a completion search.
449 * Does not resolve namespaces and does not check variants.
450 * Search engine implementations may want to override this function.
451 * @param string $search
452 * @return SearchSuggestionSet
453 */
454 protected function completionSearchBackend( $search ) {
455 $results = [];
456
457 $search = trim( $search );
458
459 if ( !in_array( NS_SPECIAL, $this->namespaces ) && // We do not run hook on Special: search
460 !Hooks::run( 'PrefixSearchBackend',
461 [ $this->namespaces, $search, $this->limit, &$results, $this->offset ]
462 ) ) {
463 // False means hook worked.
464 // FIXME: Yes, the API is weird. That's why it is going to be deprecated.
465
466 return SearchSuggestionSet::fromStrings( $results );
467 } else {
468 // Hook did not do the job, use default simple search
469 $results = $this->simplePrefixSearch( $search );
470 return SearchSuggestionSet::fromTitles( $results );
471 }
472 }
473
474 /**
475 * Perform a completion search.
476 * @param string $search
477 * @return SearchSuggestionSet
478 */
479 public function completionSearch( $search ) {
480 if ( trim( $search ) === '' ) {
481 return SearchSuggestionSet::emptySuggestionSet(); // Return empty result
482 }
483 $search = $this->normalizeNamespaces( $search );
484 return $this->processCompletionResults( $search, $this->completionSearchBackend( $search ) );
485 }
486
487 /**
488 * Perform a completion search with variants.
489 * @param string $search
490 * @return SearchSuggestionSet
491 */
492 public function completionSearchWithVariants( $search ) {
493 if ( trim( $search ) === '' ) {
494 return SearchSuggestionSet::emptySuggestionSet(); // Return empty result
495 }
496 $search = $this->normalizeNamespaces( $search );
497
498 $results = $this->completionSearchBackend( $search );
499 $fallbackLimit = $this->limit - $results->getSize();
500 if ( $fallbackLimit > 0 ) {
501 global $wgContLang;
502
503 $fallbackSearches = $wgContLang->autoConvertToAllVariants( $search );
504 $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
505
506 foreach ( $fallbackSearches as $fbs ) {
507 $this->setLimitOffset( $fallbackLimit );
508 $fallbackSearchResult = $this->completionSearch( $fbs );
509 $results->appendAll( $fallbackSearchResult );
510 $fallbackLimit -= count( $fallbackSearchResult );
511 if ( $fallbackLimit <= 0 ) {
512 break;
513 }
514 }
515 }
516 return $this->processCompletionResults( $search, $results );
517 }
518
519 /**
520 * Extract titles from completion results
521 * @param SearchSuggestionSet $completionResults
522 * @return Title[]
523 */
524 public function extractTitles( SearchSuggestionSet $completionResults ) {
525 return $completionResults->map( function( SearchSuggestion $sugg ) {
526 return $sugg->getSuggestedTitle();
527 } );
528 }
529
530 /**
531 * Process completion search results.
532 * Resolves the titles and rescores.
533 * @param SearchSuggestionSet $suggestions
534 * @return SearchSuggestionSet
535 */
536 protected function processCompletionResults( $search, SearchSuggestionSet $suggestions ) {
537 $search = trim( $search );
538 // preload the titles with LinkBatch
539 $titles = $suggestions->map( function( SearchSuggestion $sugg ) {
540 return $sugg->getSuggestedTitle();
541 } );
542 $lb = new LinkBatch( $titles );
543 $lb->setCaller( __METHOD__ );
544 $lb->execute();
545
546 $results = $suggestions->map( function( SearchSuggestion $sugg ) {
547 return $sugg->getSuggestedTitle()->getPrefixedText();
548 } );
549
550 if ( $this->offset === 0 ) {
551 // Rescore results with an exact title match
552 // NOTE: in some cases like cross-namespace redirects
553 // (frequently used as shortcuts e.g. WP:WP on huwiki) some
554 // backends like Cirrus will return no results. We should still
555 // try an exact title match to workaround this limitation
556 $rescorer = new SearchExactMatchRescorer();
557 $rescoredResults = $rescorer->rescore( $search, $this->namespaces, $results, $this->limit );
558 } else {
559 // No need to rescore if offset is not 0
560 // The exact match must have been returned at position 0
561 // if it existed.
562 $rescoredResults = $results;
563 }
564
565 if ( count( $rescoredResults ) > 0 ) {
566 $found = array_search( $rescoredResults[0], $results );
567 if ( $found === false ) {
568 // If the first result is not in the previous array it
569 // means that we found a new exact match
570 $exactMatch = SearchSuggestion::fromTitle( 0, Title::newFromText( $rescoredResults[0] ) );
571 $suggestions->prepend( $exactMatch );
572 $suggestions->shrink( $this->limit );
573 } else {
574 // if the first result is not the same we need to rescore
575 if ( $found > 0 ) {
576 $suggestions->rescore( $found );
577 }
578 }
579 }
580
581 return $suggestions;
582 }
583
584 /**
585 * Simple prefix search for subpages.
586 * @param string $search
587 * @return Title[]
588 */
589 public function defaultPrefixSearch( $search ) {
590 if ( trim( $search ) === '' ) {
591 return [];
592 }
593
594 $search = $this->normalizeNamespaces( $search );
595 return $this->simplePrefixSearch( $search );
596 }
597
598 /**
599 * Call out to simple search backend.
600 * Defaults to TitlePrefixSearch.
601 * @param string $search
602 * @return Title[]
603 */
604 protected function simplePrefixSearch( $search ) {
605 // Use default database prefix search
606 $backend = new TitlePrefixSearch;
607 return $backend->defaultSearchBackend( $this->namespaces, $search, $this->limit, $this->offset );
608 }
609
610 /**
611 * Make a list of searchable namespaces and their canonical names.
612 * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
613 * @return array
614 */
615 public static function searchableNamespaces() {
616 return MediaWikiServices::getInstance()->getSearchEngineConfig()->searchableNamespaces();
617 }
618
619 /**
620 * Extract default namespaces to search from the given user's
621 * settings, returning a list of index numbers.
622 * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
623 * @param user $user
624 * @return array
625 */
626 public static function userNamespaces( $user ) {
627 return MediaWikiServices::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
628 }
629
630 /**
631 * An array of namespaces indexes to be searched by default
632 * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
633 * @return array
634 */
635 public static function defaultNamespaces() {
636 return MediaWikiServices::getInstance()->getSearchEngineConfig()->defaultNamespaces();
637 }
638
639 /**
640 * Get a list of namespace names useful for showing in tooltips
641 * and preferences
642 * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
643 * @param array $namespaces
644 * @return array
645 */
646 public static function namespacesAsText( $namespaces ) {
647 return MediaWikiServices::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
648 }
649
650 /**
651 * Load up the appropriate search engine class for the currently
652 * active database backend, and return a configured instance.
653 * @deprecated since 1.27; Use SearchEngineFactory::create
654 * @param string $type Type of search backend, if not the default
655 * @return SearchEngine
656 */
657 public static function create( $type = null ) {
658 return MediaWikiServices::getInstance()->getSearchEngineFactory()->create( $type );
659 }
660
661 /**
662 * Return the search engines we support. If only $wgSearchType
663 * is set, it'll be an array of just that one item.
664 * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
665 * @return array
666 */
667 public static function getSearchTypes() {
668 return MediaWikiServices::getInstance()->getSearchEngineConfig()->getSearchTypes();
669 }
670
671 /**
672 * Get a list of supported profiles.
673 * Some search engine implementations may expose specific profiles to fine-tune
674 * its behaviors.
675 * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName )
676 * The array returned by this function contains the following keys:
677 * - name: the profile name to use with setFeatureData
678 * - desc-message: the i18n description
679 * - default: set to true if this profile is the default
680 *
681 * @since 1.28
682 * @param string $profileType the type of profiles
683 * @param User|null $user the user requesting the list of profiles
684 * @return array|null the list of profiles or null if none available
685 */
686 public function getProfiles( $profileType, User $user = null ) {
687 return null;
688 }
689
690 /**
691 * Create a search field definition.
692 * Specific search engines should override this method to create search fields.
693 * @param string $name
694 * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
695 * @return SearchIndexField
696 * @since 1.28
697 */
698 public function makeSearchFieldMapping( $name, $type ) {
699 return new NullIndexField();
700 }
701
702 /**
703 * Get fields for search index
704 * @since 1.28
705 * @return SearchIndexField[] Index field definitions for all content handlers
706 */
707 public function getSearchIndexFields() {
708 $models = ContentHandler::getContentModels();
709 $fields = [];
710 foreach ( $models as $model ) {
711 $handler = ContentHandler::getForModelID( $model );
712 $handlerFields = $handler->getFieldsForSearchIndex( $this );
713 foreach ( $handlerFields as $fieldName => $fieldData ) {
714 if ( empty( $fields[$fieldName] ) ) {
715 $fields[$fieldName] = $fieldData;
716 } else {
717 // TODO: do we allow some clashes with the same type or reject all of them?
718 $mergeDef = $fields[$fieldName]->merge( $fieldData );
719 if ( !$mergeDef ) {
720 throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" );
721 }
722 $fields[$fieldName] = $mergeDef;
723 }
724 }
725 }
726 // Hook to allow extensions to produce search mapping fields
727 Hooks::run( 'SearchIndexFields', [ &$fields, $this ] );
728 return $fields;
729 }
730
731 /**
732 * Augment search results with extra data.
733 *
734 * @param SearchResultSet $resultSet
735 */
736 public function augmentSearchResults( SearchResultSet $resultSet ) {
737 $setAugmentors = [];
738 $rowAugmentors = [];
739 Hooks::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
740
741 if ( !$setAugmentors && !$rowAugmentors ) {
742 // We're done here
743 return;
744 }
745
746 // Convert row augmentors to set augmentor
747 foreach ( $rowAugmentors as $name => $row ) {
748 if ( isset( $setAugmentors[$name] ) ) {
749 throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" );
750 }
751 $setAugmentors[$name] = new PerRowAugmentor( $row );
752 }
753
754 foreach ( $setAugmentors as $name => $augmentor ) {
755 $data = $augmentor->augmentAll( $resultSet );
756 if ( $data ) {
757 $resultSet->setAugmentedData( $name, $data );
758 }
759 }
760 }
761 }
762
763 /**
764 * Dummy class to be used when non-supported Database engine is present.
765 * @todo FIXME: Dummy class should probably try something at least mildly useful,
766 * such as a LIKE search through titles.
767 * @ingroup Search
768 */
769 class SearchEngineDummy extends SearchEngine {
770 // no-op
771 }