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.
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.
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
25 * @defgroup Search Search
28 use MediaWiki\MediaWikiServices
;
31 * Contain a class for special pages
34 abstract class SearchEngine
{
35 const DEFAULT_SORT
= 'relevance';
40 /** @var int[]|null */
41 public $namespaces = [ NS_MAIN
];
44 protected $limit = 10;
47 protected $offset = 0;
49 /** @var array|string */
50 protected $searchTerms = [];
53 protected $showSuggestion = true;
54 private $sort = self
::DEFAULT_SORT
;
56 /** @var array Feature values */
57 protected $features = [];
59 /** @const string profile type for completionSearch */
60 const COMPLETION_PROFILE_TYPE
= 'completionSearchProfile';
62 /** @const string profile type for query independent ranking features */
63 const FT_QUERY_INDEP_PROFILE_TYPE
= 'fulltextQueryIndepProfile';
65 /** @const int flag for legalSearchChars: includes all chars allowed in a search query */
68 /** @const int flag for legalSearchChars: includes all chars allowed in a search term */
69 const CHARS_NO_SYNTAX
= 2;
72 * Perform a full text search query and return a result set.
73 * If full text searches are not supported or disabled, return null.
75 * As of 1.32 overriding this function is deprecated. It will
76 * be converted to final in 1.34. Override self::doSearchText().
78 * @param string $term Raw search term
79 * @return SearchResultSet|Status|null
81 public function searchText( $term ) {
82 return $this->maybePaginate( function () use ( $term ) {
83 return $this->doSearchText( $term );
88 * Perform a full text search query and return a result set.
90 * @param string $term Raw search term
91 * @return SearchResultSet|Status|null
94 protected function doSearchText( $term ) {
99 * Perform a title search in the article archive.
100 * NOTE: these results still should be filtered by
101 * matching against PageArchive, permissions checks etc
102 * The results returned by this methods are only sugegstions and
103 * may not end up being shown to the user.
105 * As of 1.32 overriding this function is deprecated. It will
106 * be converted to final in 1.34. Override self::doSearchArchiveTitle().
108 * @param string $term Raw search term
109 * @return Status<Title[]>
112 public function searchArchiveTitle( $term ) {
113 return $this->doSearchArchiveTitle( $term );
117 * Perform a title search in the article archive.
119 * @param string $term Raw search term
120 * @return Status<Title[]>
123 protected function doSearchArchiveTitle( $term ) {
124 return Status
::newGood( [] );
128 * Perform a title-only search query and return a result set.
129 * If title searches are not supported or disabled, return null.
132 * As of 1.32 overriding this function is deprecated. It will
133 * be converted to final in 1.34. Override self::doSearchTitle().
135 * @param string $term Raw search term
136 * @return SearchResultSet|null
138 public function searchTitle( $term ) {
139 return $this->maybePaginate( function () use ( $term ) {
140 return $this->doSearchTitle( $term );
145 * Perform a title-only search query and return a result set.
147 * @param string $term Raw search term
148 * @return SearchResultSet|null
151 protected function doSearchTitle( $term ) {
156 * Performs an overfetch and shrink operation to determine if
157 * the next page is available for search engines that do not
158 * explicitly implement their own pagination.
160 * @param Closure $fn Takes no arguments
161 * @return SearchResultSet|Status<SearchResultSet>|null Result of calling $fn
163 private function maybePaginate( Closure
$fn ) {
164 if ( $this instanceof PaginatingSearchEngine
) {
169 $resultSetOrStatus = $fn();
175 if ( $resultSetOrStatus instanceof SearchResultSet
) {
176 $resultSet = $resultSetOrStatus;
177 } elseif ( $resultSetOrStatus instanceof Status
&&
178 $resultSetOrStatus->getValue() instanceof SearchResultSet
180 $resultSet = $resultSetOrStatus->getValue();
183 $resultSet->shrink( $this->limit
);
186 return $resultSetOrStatus;
191 * @param string $feature
194 public function supports( $feature ) {
195 switch ( $feature ) {
196 case 'search-update':
198 case 'title-suffix-filter':
205 * Way to pass custom data for engines
207 * @param string $feature
210 public function setFeatureData( $feature, $data ) {
211 $this->features
[$feature] = $data;
215 * Way to retrieve custom data set by setFeatureData
216 * or by the engine itself.
218 * @param string $feature feature name
219 * @return mixed the feature value or null if unset
221 public function getFeatureData( $feature ) {
222 return $this->features
[$feature] ??
null;
226 * When overridden in derived class, performs database-specific conversions
227 * on text to be used for searching or updating search index.
228 * Default implementation does nothing (simply returns $string).
230 * @param string $string String to process
233 public function normalizeText( $string ) {
234 // Some languages such as Chinese require word segmentation
235 return MediaWikiServices
::getInstance()->getContentLanguage()->segmentByWord( $string );
239 * Transform search term in cases when parts of the query came as different
240 * GET params (when supported), e.g. for prefix queries:
241 * search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
242 * @param string $term
244 * @deprecated since 1.32 this should now be handled internally by the
247 public function transformSearchTerm( $term ) {
252 * Get service class to finding near matches.
253 * @param Config $config Configuration to use for the matcher.
254 * @return SearchNearMatcher
256 public function getNearMatcher( Config
$config ) {
257 return new SearchNearMatcher( $config,
258 MediaWikiServices
::getInstance()->getContentLanguage() );
262 * Get near matcher for default SearchEngine.
263 * @return SearchNearMatcher
265 protected static function defaultNearMatcher() {
266 $services = MediaWikiServices
::getInstance();
267 $config = $services->getMainConfig();
268 return $services->newSearchEngine()->getNearMatcher( $config );
272 * If an exact title match can be found, or a very slightly close match,
273 * return the title. If no match, returns NULL.
274 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
275 * @param string $searchterm
278 public static function getNearMatch( $searchterm ) {
279 return static::defaultNearMatcher()->getNearMatch( $searchterm );
283 * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
285 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
286 * @param string $searchterm
287 * @return SearchResultSet
289 public static function getNearMatchResultSet( $searchterm ) {
290 return static::defaultNearMatcher()->getNearMatchResultSet( $searchterm );
294 * Get chars legal for search
295 * NOTE: usage as static is deprecated and preserved only as BC measure
296 * @param int $type type of search chars (see self::CHARS_ALL
297 * and self::CHARS_NO_SYNTAX). Defaults to CHARS_ALL
300 public static function legalSearchChars( $type = self
::CHARS_ALL
) {
301 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
305 * Set the maximum number of results to return
306 * and how many to skip before returning the first.
311 function setLimitOffset( $limit, $offset = 0 ) {
312 $this->limit
= intval( $limit );
313 $this->offset
= intval( $offset );
317 * Set which namespaces the search should include.
318 * Give an array of namespace index numbers.
320 * @param int[]|null $namespaces
322 function setNamespaces( $namespaces ) {
324 // Filter namespaces to only keep valid ones
325 $validNs = $this->searchableNamespaces();
326 $namespaces = array_filter( $namespaces, function ( $ns ) use( $validNs ) {
327 return $ns < 0 ||
isset( $validNs[$ns] );
332 $this->namespaces
= $namespaces;
336 * Set whether the searcher should try to build a suggestion. Note: some searchers
337 * don't support building a suggestion in the first place and others don't respect
340 * @param bool $showSuggestion Should the searcher try to build suggestions
342 function setShowSuggestion( $showSuggestion ) {
343 $this->showSuggestion
= $showSuggestion;
347 * Get the valid sort directions. All search engines support 'relevance' but others
348 * might support more. The default in all implementations must be 'relevance.'
351 * @return string[] the valid sort directions for setSort
353 public function getValidSorts() {
354 return [ self
::DEFAULT_SORT
];
358 * Set the sort direction of the search results. Must be one returned by
359 * SearchEngine::getValidSorts()
362 * @throws InvalidArgumentException
363 * @param string $sort sort direction for query result
365 public function setSort( $sort ) {
366 if ( !in_array( $sort, $this->getValidSorts() ) ) {
367 throw new InvalidArgumentException( "Invalid sort: $sort. " .
368 "Must be one of: " . implode( ', ', $this->getValidSorts() ) );
374 * Get the sort direction of the search results
379 public function getSort() {
384 * Parse some common prefixes: all (search everything)
385 * or namespace names and set the list of namespaces
386 * of this class accordingly.
388 * @deprecated since 1.32; should be handled internally by the search engine
389 * @param string $query
392 function replacePrefixes( $query ) {
397 * Parse some common prefixes: all (search everything)
400 * @param string $query
401 * @param bool $withAllKeyword activate support of the "all:" keyword and its
402 * translations to activate searching on all namespaces.
403 * @param bool $withPrefixSearchExtractNamespaceHook call the PrefixSearchExtractNamespace hook
404 * if classic namespace identification did not match.
405 * @return false|array false if no namespace was extracted, an array
406 * with the parsed query at index 0 and an array of namespaces at index
407 * 1 (or null for all namespaces).
409 * @throws MWException
411 public static function parseNamespacePrefixes(
413 $withAllKeyword = true,
414 $withPrefixSearchExtractNamespaceHook = false
417 if ( strpos( $query, ':' ) === false ) { // nothing to do
420 $extractedNamespace = null;
423 if ( $withAllKeyword ) {
426 $allkeywords[] = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
427 // force all: so that we have a common syntax for all the wikis
428 if ( !in_array( 'all:', $allkeywords ) ) {
429 $allkeywords[] = 'all:';
432 foreach ( $allkeywords as $kw ) {
433 if ( strncmp( $query, $kw, strlen( $kw ) ) == 0 ) {
434 $extractedNamespace = null;
435 $parsed = substr( $query, strlen( $kw ) );
442 if ( !$allQuery && strpos( $query, ':' ) !== false ) {
443 $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
444 $index = MediaWikiServices
::getInstance()->getContentLanguage()->getNsIndex( $prefix );
445 if ( $index !== false ) {
446 $extractedNamespace = [ $index ];
447 $parsed = substr( $query, strlen( $prefix ) +
1 );
448 } elseif ( $withPrefixSearchExtractNamespaceHook ) {
449 $hookNamespaces = [ NS_MAIN
];
451 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$hookNamespaces, &$hookQuery ] );
452 if ( $hookQuery !== $query ) {
453 $parsed = $hookQuery;
454 $extractedNamespace = $hookNamespaces;
463 return [ $parsed, $extractedNamespace ];
467 * Find snippet highlight settings for all users
468 * @return array Contextlines, contextchars
470 public static function userHighlightPrefs() {
471 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
472 $contextchars = 75; // same as above.... :P
473 return [ $contextlines, $contextchars ];
477 * Create or update the search index record for the given page.
478 * Title and text should be pre-processed.
482 * @param string $title
483 * @param string $text
485 function update( $id, $title, $text ) {
490 * Update a search index record's title only.
491 * Title should be pre-processed.
495 * @param string $title
497 function updateTitle( $id, $title ) {
502 * Delete an indexed page
503 * Title should be pre-processed.
506 * @param int $id Page id that was deleted
507 * @param string $title Title of page that was deleted
509 function delete( $id, $title ) {
514 * Get the raw text for updating the index from a content object
515 * Nicer search backends could possibly do something cooler than
516 * just returning raw text
518 * @todo This isn't ideal, we'd really like to have content-specific handling here
519 * @param Title $t Title we're indexing
520 * @param Content|null $c Content of the page to index
523 public function getTextFromContent( Title
$t, Content
$c = null ) {
524 return $c ?
$c->getTextForSearchIndex() : '';
528 * If an implementation of SearchEngine handles all of its own text processing
529 * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
530 * rather silly handling, it should return true here instead.
534 public function textAlreadyUpdatedForIndex() {
539 * Makes search simple string if it was namespaced.
540 * Sets namespaces of the search to namespaces extracted from string.
541 * @param string $search
542 * @return string Simplified search string
544 protected function normalizeNamespaces( $search ) {
545 $queryAndNs = self
::parseNamespacePrefixes( $search, false, true );
546 if ( $queryAndNs !== false ) {
547 $this->setNamespaces( $queryAndNs[1] );
548 return $queryAndNs[0];
554 * Perform an overfetch of completion search results. This allows
555 * determining if another page of results is available.
557 * @param string $search
558 * @return SearchSuggestionSet
560 protected function completionSearchBackendOverfetch( $search ) {
563 return $this->completionSearchBackend( $search );
570 * Perform a completion search.
571 * Does not resolve namespaces and does not check variants.
572 * Search engine implementations may want to override this function.
573 * @param string $search
574 * @return SearchSuggestionSet
576 protected function completionSearchBackend( $search ) {
579 $search = trim( $search );
581 if ( !in_array( NS_SPECIAL
, $this->namespaces
) && // We do not run hook on Special: search
582 !Hooks
::run( 'PrefixSearchBackend',
583 [ $this->namespaces
, $search, $this->limit
, &$results, $this->offset
]
585 // False means hook worked.
586 // FIXME: Yes, the API is weird. That's why it is going to be deprecated.
588 return SearchSuggestionSet
::fromStrings( $results );
590 // Hook did not do the job, use default simple search
591 $results = $this->simplePrefixSearch( $search );
592 return SearchSuggestionSet
::fromTitles( $results );
597 * Perform a completion search.
598 * @param string $search
599 * @return SearchSuggestionSet
601 public function completionSearch( $search ) {
602 if ( trim( $search ) === '' ) {
603 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
605 $search = $this->normalizeNamespaces( $search );
606 $suggestions = $this->completionSearchBackendOverfetch( $search );
607 return $this->processCompletionResults( $search, $suggestions );
611 * Perform a completion search with variants.
612 * @param string $search
613 * @return SearchSuggestionSet
615 public function completionSearchWithVariants( $search ) {
616 if ( trim( $search ) === '' ) {
617 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
619 $search = $this->normalizeNamespaces( $search );
621 $results = $this->completionSearchBackendOverfetch( $search );
622 $fallbackLimit = 1 +
$this->limit
- $results->getSize();
623 if ( $fallbackLimit > 0 ) {
624 $fallbackSearches = MediaWikiServices
::getInstance()->getContentLanguage()->
625 autoConvertToAllVariants( $search );
626 $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
628 foreach ( $fallbackSearches as $fbs ) {
629 $this->setLimitOffset( $fallbackLimit );
630 $fallbackSearchResult = $this->completionSearch( $fbs );
631 $results->appendAll( $fallbackSearchResult );
632 $fallbackLimit -= $fallbackSearchResult->getSize();
633 if ( $fallbackLimit <= 0 ) {
638 return $this->processCompletionResults( $search, $results );
642 * Extract titles from completion results
643 * @param SearchSuggestionSet $completionResults
646 public function extractTitles( SearchSuggestionSet
$completionResults ) {
647 return $completionResults->map( function ( SearchSuggestion
$sugg ) {
648 return $sugg->getSuggestedTitle();
653 * Process completion search results.
654 * Resolves the titles and rescores.
655 * @param string $search
656 * @param SearchSuggestionSet $suggestions
657 * @return SearchSuggestionSet
659 protected function processCompletionResults( $search, SearchSuggestionSet
$suggestions ) {
660 // We over-fetched to determine pagination. Shrink back down if we have extra results
661 // and mark if pagination is possible
662 $suggestions->shrink( $this->limit
);
664 $search = trim( $search );
665 // preload the titles with LinkBatch
666 $lb = new LinkBatch( $suggestions->map( function ( SearchSuggestion
$sugg ) {
667 return $sugg->getSuggestedTitle();
669 $lb->setCaller( __METHOD__
);
672 $diff = $suggestions->filter( function ( SearchSuggestion
$sugg ) {
673 return $sugg->getSuggestedTitle()->isKnown();
676 MediaWikiServices
::getInstance()->getStatsdDataFactory()
677 ->updateCount( 'search.completion.missing', $diff );
680 $results = $suggestions->map( function ( SearchSuggestion
$sugg ) {
681 return $sugg->getSuggestedTitle()->getPrefixedText();
684 if ( $this->offset
=== 0 ) {
685 // Rescore results with an exact title match
686 // NOTE: in some cases like cross-namespace redirects
687 // (frequently used as shortcuts e.g. WP:WP on huwiki) some
688 // backends like Cirrus will return no results. We should still
689 // try an exact title match to workaround this limitation
690 $rescorer = new SearchExactMatchRescorer();
691 $rescoredResults = $rescorer->rescore( $search, $this->namespaces
, $results, $this->limit
);
693 // No need to rescore if offset is not 0
694 // The exact match must have been returned at position 0
696 $rescoredResults = $results;
699 if ( count( $rescoredResults ) > 0 ) {
700 $found = array_search( $rescoredResults[0], $results );
701 if ( $found === false ) {
702 // If the first result is not in the previous array it
703 // means that we found a new exact match
704 $exactMatch = SearchSuggestion
::fromTitle( 0, Title
::newFromText( $rescoredResults[0] ) );
705 $suggestions->prepend( $exactMatch );
706 $suggestions->shrink( $this->limit
);
708 // if the first result is not the same we need to rescore
710 $suggestions->rescore( $found );
719 * Simple prefix search for subpages.
720 * @param string $search
723 public function defaultPrefixSearch( $search ) {
724 if ( trim( $search ) === '' ) {
728 $search = $this->normalizeNamespaces( $search );
729 return $this->simplePrefixSearch( $search );
733 * Call out to simple search backend.
734 * Defaults to TitlePrefixSearch.
735 * @param string $search
738 protected function simplePrefixSearch( $search ) {
739 // Use default database prefix search
740 $backend = new TitlePrefixSearch
;
741 return $backend->defaultSearchBackend( $this->namespaces
, $search, $this->limit
, $this->offset
);
745 * Make a list of searchable namespaces and their canonical names.
746 * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
749 public static function searchableNamespaces() {
750 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->searchableNamespaces();
754 * Extract default namespaces to search from the given user's
755 * settings, returning a list of index numbers.
756 * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
760 public static function userNamespaces( $user ) {
761 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
765 * An array of namespaces indexes to be searched by default
766 * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
769 public static function defaultNamespaces() {
770 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->defaultNamespaces();
774 * Get a list of namespace names useful for showing in tooltips
776 * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
777 * @param array $namespaces
780 public static function namespacesAsText( $namespaces ) {
781 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
785 * Load up the appropriate search engine class for the currently
786 * active database backend, and return a configured instance.
787 * @deprecated since 1.27; Use SearchEngineFactory::create
788 * @param string $type Type of search backend, if not the default
789 * @return SearchEngine
791 public static function create( $type = null ) {
792 return MediaWikiServices
::getInstance()->getSearchEngineFactory()->create( $type );
796 * Return the search engines we support. If only $wgSearchType
797 * is set, it'll be an array of just that one item.
798 * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
801 public static function getSearchTypes() {
802 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->getSearchTypes();
806 * Get a list of supported profiles.
807 * Some search engine implementations may expose specific profiles to fine-tune
809 * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName )
810 * The array returned by this function contains the following keys:
811 * - name: the profile name to use with setFeatureData
812 * - desc-message: the i18n description
813 * - default: set to true if this profile is the default
816 * @param string $profileType the type of profiles
817 * @param User|null $user the user requesting the list of profiles
818 * @return array|null the list of profiles or null if none available
820 public function getProfiles( $profileType, User
$user = null ) {
825 * Create a search field definition.
826 * Specific search engines should override this method to create search fields.
827 * @param string $name
828 * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
829 * @return SearchIndexField
832 public function makeSearchFieldMapping( $name, $type ) {
833 return new NullIndexField();
837 * Get fields for search index
839 * @return SearchIndexField[] Index field definitions for all content handlers
841 public function getSearchIndexFields() {
842 $models = ContentHandler
::getContentModels();
844 $seenHandlers = new SplObjectStorage();
845 foreach ( $models as $model ) {
847 $handler = ContentHandler
::getForModelID( $model );
849 catch ( MWUnknownContentModelException
$e ) {
850 // If we can find no handler, ignore it
853 // Several models can have the same handler, so avoid processing it repeatedly
854 if ( $seenHandlers->contains( $handler ) ) {
855 // We already did this one
858 $seenHandlers->attach( $handler );
859 $handlerFields = $handler->getFieldsForSearchIndex( $this );
860 foreach ( $handlerFields as $fieldName => $fieldData ) {
861 if ( empty( $fields[$fieldName] ) ) {
862 $fields[$fieldName] = $fieldData;
864 // TODO: do we allow some clashes with the same type or reject all of them?
865 $mergeDef = $fields[$fieldName]->merge( $fieldData );
867 throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" );
869 $fields[$fieldName] = $mergeDef;
873 // Hook to allow extensions to produce search mapping fields
874 Hooks
::run( 'SearchIndexFields', [ &$fields, $this ] );
879 * Augment search results with extra data.
881 * @param SearchResultSet $resultSet
883 public function augmentSearchResults( SearchResultSet
$resultSet ) {
886 Hooks
::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
887 if ( !$setAugmentors && !$rowAugmentors ) {
892 // Convert row augmentors to set augmentor
893 foreach ( $rowAugmentors as $name => $row ) {
894 if ( isset( $setAugmentors[$name] ) ) {
895 throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" );
897 $setAugmentors[$name] = new PerRowAugmentor( $row );
900 foreach ( $setAugmentors as $name => $augmentor ) {
901 $data = $augmentor->augmentAll( $resultSet );
903 $resultSet->setAugmentedData( $name, $data );
910 * Dummy class to be used when non-supported Database engine is present.
911 * @todo FIXME: Dummy class should probably try something at least mildly useful,
912 * such as a LIKE search through titles.
915 class SearchEngineDummy
extends SearchEngine
{