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
{
38 /** @var int[]|null */
39 public $namespaces = [ NS_MAIN
];
42 protected $limit = 10;
45 protected $offset = 0;
47 /** @var array|string */
48 protected $searchTerms = [];
51 protected $showSuggestion = true;
52 private $sort = 'relevance';
54 /** @var array Feature values */
55 protected $features = [];
57 /** @const string profile type for completionSearch */
58 const COMPLETION_PROFILE_TYPE
= 'completionSearchProfile';
60 /** @const string profile type for query independent ranking features */
61 const FT_QUERY_INDEP_PROFILE_TYPE
= 'fulltextQueryIndepProfile';
63 /** @const int flag for legalSearchChars: includes all chars allowed in a search query */
66 /** @const int flag for legalSearchChars: includes all chars allowed in a search term */
67 const CHARS_NO_SYNTAX
= 2;
70 * Perform a full text search query and return a result set.
71 * If full text searches are not supported or disabled, return null.
73 * As of 1.32 overriding this function is deprecated. It will
74 * be converted to final in 1.34. Override self::doSearchText().
76 * @param string $term Raw search term
77 * @return SearchResultSet|Status|null
79 public function searchText( $term ) {
80 return $this->maybePaginate( function () use ( $term ) {
81 return $this->doSearchText( $term );
86 * Perform a full text search query and return a result set.
88 * @param string $term Raw search term
89 * @return SearchResultSet|Status|null
92 protected function doSearchText( $term ) {
97 * Perform a title search in the article archive.
98 * NOTE: these results still should be filtered by
99 * matching against PageArchive, permissions checks etc
100 * The results returned by this methods are only sugegstions and
101 * may not end up being shown to the user.
103 * As of 1.32 overriding this function is deprecated. It will
104 * be converted to final in 1.34. Override self::doSearchArchiveTitle().
106 * @param string $term Raw search term
107 * @return Status<Title[]>
110 public function searchArchiveTitle( $term ) {
111 return $this->doSearchArchiveTitle( $term );
115 * Perform a title search in the article archive.
117 * @param string $term Raw search term
118 * @return Status<Title[]>
121 protected function doSearchArchiveTitle( $term ) {
122 return Status
::newGood( [] );
126 * Perform a title-only search query and return a result set.
127 * If title searches are not supported or disabled, return null.
130 * As of 1.32 overriding this function is deprecated. It will
131 * be converted to final in 1.34. Override self::doSearchTitle().
133 * @param string $term Raw search term
134 * @return SearchResultSet|null
136 public function searchTitle( $term ) {
137 return $this->maybePaginate( function () use ( $term ) {
138 return $this->doSearchTitle( $term );
143 * Perform a title-only search query and return a result set.
145 * @param string $term Raw search term
146 * @return SearchResultSet|null
149 protected function doSearchTitle( $term ) {
154 * Performs an overfetch and shrink operation to determine if
155 * the next page is available for search engines that do not
156 * explicitly implement their own pagination.
158 * @param Closure $fn Takes no arguments
159 * @return SearchResultSet|Status<SearchResultSet>|null Result of calling $fn
161 private function maybePaginate( Closure
$fn ) {
162 if ( $this instanceof PaginatingSearchEngine
) {
167 $resultSetOrStatus = $fn();
173 if ( $resultSetOrStatus instanceof SearchResultSet
) {
174 $resultSet = $resultSetOrStatus;
175 } elseif ( $resultSetOrStatus instanceof Status
&&
176 $resultSetOrStatus->getValue() instanceof SearchResultSet
178 $resultSet = $resultSetOrStatus->getValue();
181 $resultSet->shrink( $this->limit
);
184 return $resultSetOrStatus;
189 * @param string $feature
192 public function supports( $feature ) {
193 switch ( $feature ) {
194 case 'search-update':
196 case 'title-suffix-filter':
203 * Way to pass custom data for engines
205 * @param string $feature
208 public function setFeatureData( $feature, $data ) {
209 $this->features
[$feature] = $data;
213 * Way to retrieve custom data set by setFeatureData
214 * or by the engine itself.
216 * @param string $feature feature name
217 * @return mixed the feature value or null if unset
219 public function getFeatureData( $feature ) {
220 if ( isset( $this->features
[$feature] ) ) {
221 return $this->features
[$feature];
227 * When overridden in derived class, performs database-specific conversions
228 * on text to be used for searching or updating search index.
229 * Default implementation does nothing (simply returns $string).
231 * @param string $string String to process
234 public function normalizeText( $string ) {
237 // Some languages such as Chinese require word segmentation
238 return $wgContLang->segmentByWord( $string );
242 * Transform search term in cases when parts of the query came as different
243 * GET params (when supported), e.g. for prefix queries:
244 * search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
245 * @param string $term
248 public function transformSearchTerm( $term ) {
253 * Get service class to finding near matches.
254 * @param Config $config Configuration to use for the matcher.
255 * @return SearchNearMatcher
257 public function getNearMatcher( Config
$config ) {
259 return new SearchNearMatcher( $config, $wgContLang );
263 * Get near matcher for default SearchEngine.
264 * @return SearchNearMatcher
266 protected static function defaultNearMatcher() {
267 $config = MediaWikiServices
::getInstance()->getMainConfig();
268 return MediaWikiServices
::getInstance()->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 should be 'relevance.'
351 * @return string[] the valid sort directions for setSort
353 public function getValidSorts() {
354 return [ 'relevance' ];
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 * @param string $query
391 function replacePrefixes( $query ) {
392 $queryAndNs = self
::parseNamespacePrefixes( $query );
393 if ( $queryAndNs === false ) {
396 $this->namespaces
= $queryAndNs[1];
397 return $queryAndNs[0];
401 * Parse some common prefixes: all (search everything)
404 * @param string $query
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 public static function parseNamespacePrefixes( $query ) {
413 if ( strpos( $query, ':' ) === false ) { // nothing to do
416 $extractedNamespace = null;
419 $allkeywords[] = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
420 // force all: so that we have a common syntax for all the wikis
421 if ( !in_array( 'all:', $allkeywords ) ) {
422 $allkeywords[] = 'all:';
426 foreach ( $allkeywords as $kw ) {
427 if ( strncmp( $query, $kw, strlen( $kw ) ) == 0 ) {
428 $extractedNamespace = null;
429 $parsed = substr( $query, strlen( $kw ) );
435 if ( !$allQuery && strpos( $query, ':' ) !== false ) {
436 // TODO: should we unify with PrefixSearch::extractNamespace ?
437 $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
438 $index = $wgContLang->getNsIndex( $prefix );
439 if ( $index !== false ) {
440 $extractedNamespace = [ $index ];
441 $parsed = substr( $query, strlen( $prefix ) +
1 );
447 if ( trim( $parsed ) == '' ) {
448 $parsed = $query; // prefix was the whole query
451 return [ $parsed, $extractedNamespace ];
455 * Find snippet highlight settings for all users
456 * @return array Contextlines, contextchars
458 public static function userHighlightPrefs() {
459 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
460 $contextchars = 75; // same as above.... :P
461 return [ $contextlines, $contextchars ];
465 * Create or update the search index record for the given page.
466 * Title and text should be pre-processed.
470 * @param string $title
471 * @param string $text
473 function update( $id, $title, $text ) {
478 * Update a search index record's title only.
479 * Title should be pre-processed.
483 * @param string $title
485 function updateTitle( $id, $title ) {
490 * Delete an indexed page
491 * Title should be pre-processed.
494 * @param int $id Page id that was deleted
495 * @param string $title Title of page that was deleted
497 function delete( $id, $title ) {
502 * Get the raw text for updating the index from a content object
503 * Nicer search backends could possibly do something cooler than
504 * just returning raw text
506 * @todo This isn't ideal, we'd really like to have content-specific handling here
507 * @param Title $t Title we're indexing
508 * @param Content $c Content of the page to index
511 public function getTextFromContent( Title
$t, Content
$c = null ) {
512 return $c ?
$c->getTextForSearchIndex() : '';
516 * If an implementation of SearchEngine handles all of its own text processing
517 * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
518 * rather silly handling, it should return true here instead.
522 public function textAlreadyUpdatedForIndex() {
527 * Makes search simple string if it was namespaced.
528 * Sets namespaces of the search to namespaces extracted from string.
529 * @param string $search
530 * @return string Simplified search string
532 protected function normalizeNamespaces( $search ) {
533 // Find a Title which is not an interwiki and is in NS_MAIN
534 $title = Title
::newFromText( $search );
535 $ns = $this->namespaces
;
536 if ( $title && !$title->isExternal() ) {
537 $ns = [ $title->getNamespace() ];
538 $search = $title->getText();
539 if ( $ns[0] == NS_MAIN
) {
540 $ns = $this->namespaces
; // no explicit prefix, use default namespaces
541 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
544 $title = Title
::newFromText( $search . 'Dummy' );
545 if ( $title && $title->getText() == 'Dummy'
546 && $title->getNamespace() != NS_MAIN
547 && !$title->isExternal()
549 $ns = [ $title->getNamespace() ];
552 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
556 $ns = array_map( function ( $space ) {
557 return $space == NS_MEDIA ? NS_FILE
: $space;
560 $this->setNamespaces( $ns );
565 * Perform an overfetch of completion search results. This allows
566 * determining if another page of results is available.
568 * @param string $search
569 * @return SearchSuggestionSet
571 protected function completionSearchBackendOverfetch( $search ) {
574 return $this->completionSearchBackend( $search );
581 * Perform a completion search.
582 * Does not resolve namespaces and does not check variants.
583 * Search engine implementations may want to override this function.
584 * @param string $search
585 * @return SearchSuggestionSet
587 protected function completionSearchBackend( $search ) {
590 $search = trim( $search );
592 if ( !in_array( NS_SPECIAL
, $this->namespaces
) && // We do not run hook on Special: search
593 !Hooks
::run( 'PrefixSearchBackend',
594 [ $this->namespaces
, $search, $this->limit
, &$results, $this->offset
]
596 // False means hook worked.
597 // FIXME: Yes, the API is weird. That's why it is going to be deprecated.
599 return SearchSuggestionSet
::fromStrings( $results );
601 // Hook did not do the job, use default simple search
602 $results = $this->simplePrefixSearch( $search );
603 return SearchSuggestionSet
::fromTitles( $results );
608 * Perform a completion search.
609 * @param string $search
610 * @return SearchSuggestionSet
612 public function completionSearch( $search ) {
613 if ( trim( $search ) === '' ) {
614 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
616 $search = $this->normalizeNamespaces( $search );
617 $suggestions = $this->completionSearchBackendOverfetch( $search );
618 return $this->processCompletionResults( $search, $suggestions );
622 * Perform a completion search with variants.
623 * @param string $search
624 * @return SearchSuggestionSet
626 public function completionSearchWithVariants( $search ) {
627 if ( trim( $search ) === '' ) {
628 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
630 $search = $this->normalizeNamespaces( $search );
632 $results = $this->completionSearchBackendOverfetch( $search );
633 $fallbackLimit = 1 +
$this->limit
- $results->getSize();
634 if ( $fallbackLimit > 0 ) {
637 $fallbackSearches = $wgContLang->autoConvertToAllVariants( $search );
638 $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
640 foreach ( $fallbackSearches as $fbs ) {
641 $this->setLimitOffset( $fallbackLimit );
642 $fallbackSearchResult = $this->completionSearch( $fbs );
643 $results->appendAll( $fallbackSearchResult );
644 $fallbackLimit -= $fallbackSearchResult->getSize();
645 if ( $fallbackLimit <= 0 ) {
650 return $this->processCompletionResults( $search, $results );
654 * Extract titles from completion results
655 * @param SearchSuggestionSet $completionResults
658 public function extractTitles( SearchSuggestionSet
$completionResults ) {
659 return $completionResults->map( function ( SearchSuggestion
$sugg ) {
660 return $sugg->getSuggestedTitle();
665 * Process completion search results.
666 * Resolves the titles and rescores.
667 * @param string $search
668 * @param SearchSuggestionSet $suggestions
669 * @return SearchSuggestionSet
671 protected function processCompletionResults( $search, SearchSuggestionSet
$suggestions ) {
672 // We over-fetched to determine pagination. Shrink back down if we have extra results
673 // and mark if pagination is possible
674 $suggestions->shrink( $this->limit
);
676 $search = trim( $search );
677 // preload the titles with LinkBatch
678 $lb = new LinkBatch( $suggestions->map( function ( SearchSuggestion
$sugg ) {
679 return $sugg->getSuggestedTitle();
681 $lb->setCaller( __METHOD__
);
684 $diff = $suggestions->filter( function ( SearchSuggestion
$sugg ) {
685 return $sugg->getSuggestedTitle()->isKnown();
688 MediaWikiServices
::getInstance()->getStatsdDataFactory()
689 ->updateCount( 'search.completion.missing', $diff );
692 $results = $suggestions->map( function ( SearchSuggestion
$sugg ) {
693 return $sugg->getSuggestedTitle()->getPrefixedText();
696 if ( $this->offset
=== 0 ) {
697 // Rescore results with an exact title match
698 // NOTE: in some cases like cross-namespace redirects
699 // (frequently used as shortcuts e.g. WP:WP on huwiki) some
700 // backends like Cirrus will return no results. We should still
701 // try an exact title match to workaround this limitation
702 $rescorer = new SearchExactMatchRescorer();
703 $rescoredResults = $rescorer->rescore( $search, $this->namespaces
, $results, $this->limit
);
705 // No need to rescore if offset is not 0
706 // The exact match must have been returned at position 0
708 $rescoredResults = $results;
711 if ( count( $rescoredResults ) > 0 ) {
712 $found = array_search( $rescoredResults[0], $results );
713 if ( $found === false ) {
714 // If the first result is not in the previous array it
715 // means that we found a new exact match
716 $exactMatch = SearchSuggestion
::fromTitle( 0, Title
::newFromText( $rescoredResults[0] ) );
717 $suggestions->prepend( $exactMatch );
718 $suggestions->shrink( $this->limit
);
720 // if the first result is not the same we need to rescore
722 $suggestions->rescore( $found );
731 * Simple prefix search for subpages.
732 * @param string $search
735 public function defaultPrefixSearch( $search ) {
736 if ( trim( $search ) === '' ) {
740 $search = $this->normalizeNamespaces( $search );
741 return $this->simplePrefixSearch( $search );
745 * Call out to simple search backend.
746 * Defaults to TitlePrefixSearch.
747 * @param string $search
750 protected function simplePrefixSearch( $search ) {
751 // Use default database prefix search
752 $backend = new TitlePrefixSearch
;
753 return $backend->defaultSearchBackend( $this->namespaces
, $search, $this->limit
, $this->offset
);
757 * Make a list of searchable namespaces and their canonical names.
758 * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
761 public static function searchableNamespaces() {
762 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->searchableNamespaces();
766 * Extract default namespaces to search from the given user's
767 * settings, returning a list of index numbers.
768 * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
772 public static function userNamespaces( $user ) {
773 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
777 * An array of namespaces indexes to be searched by default
778 * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
781 public static function defaultNamespaces() {
782 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->defaultNamespaces();
786 * Get a list of namespace names useful for showing in tooltips
788 * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
789 * @param array $namespaces
792 public static function namespacesAsText( $namespaces ) {
793 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
797 * Load up the appropriate search engine class for the currently
798 * active database backend, and return a configured instance.
799 * @deprecated since 1.27; Use SearchEngineFactory::create
800 * @param string $type Type of search backend, if not the default
801 * @return SearchEngine
803 public static function create( $type = null ) {
804 return MediaWikiServices
::getInstance()->getSearchEngineFactory()->create( $type );
808 * Return the search engines we support. If only $wgSearchType
809 * is set, it'll be an array of just that one item.
810 * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
813 public static function getSearchTypes() {
814 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->getSearchTypes();
818 * Get a list of supported profiles.
819 * Some search engine implementations may expose specific profiles to fine-tune
821 * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName )
822 * The array returned by this function contains the following keys:
823 * - name: the profile name to use with setFeatureData
824 * - desc-message: the i18n description
825 * - default: set to true if this profile is the default
828 * @param string $profileType the type of profiles
829 * @param User|null $user the user requesting the list of profiles
830 * @return array|null the list of profiles or null if none available
832 public function getProfiles( $profileType, User
$user = null ) {
837 * Create a search field definition.
838 * Specific search engines should override this method to create search fields.
839 * @param string $name
840 * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
841 * @return SearchIndexField
844 public function makeSearchFieldMapping( $name, $type ) {
845 return new NullIndexField();
849 * Get fields for search index
851 * @return SearchIndexField[] Index field definitions for all content handlers
853 public function getSearchIndexFields() {
854 $models = ContentHandler
::getContentModels();
856 $seenHandlers = new SplObjectStorage();
857 foreach ( $models as $model ) {
859 $handler = ContentHandler
::getForModelID( $model );
861 catch ( MWUnknownContentModelException
$e ) {
862 // If we can find no handler, ignore it
865 // Several models can have the same handler, so avoid processing it repeatedly
866 if ( $seenHandlers->contains( $handler ) ) {
867 // We already did this one
870 $seenHandlers->attach( $handler );
871 $handlerFields = $handler->getFieldsForSearchIndex( $this );
872 foreach ( $handlerFields as $fieldName => $fieldData ) {
873 if ( empty( $fields[$fieldName] ) ) {
874 $fields[$fieldName] = $fieldData;
876 // TODO: do we allow some clashes with the same type or reject all of them?
877 $mergeDef = $fields[$fieldName]->merge( $fieldData );
879 throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" );
881 $fields[$fieldName] = $mergeDef;
885 // Hook to allow extensions to produce search mapping fields
886 Hooks
::run( 'SearchIndexFields', [ &$fields, $this ] );
891 * Augment search results with extra data.
893 * @param SearchResultSet $resultSet
895 public function augmentSearchResults( SearchResultSet
$resultSet ) {
898 Hooks
::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
899 if ( !$setAugmentors && !$rowAugmentors ) {
904 // Convert row augmentors to set augmentor
905 foreach ( $rowAugmentors as $name => $row ) {
906 if ( isset( $setAugmentors[$name] ) ) {
907 throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" );
909 $setAugmentors[$name] = new PerRowAugmentor( $row );
912 foreach ( $setAugmentors as $name => $augmentor ) {
913 $data = $augmentor->augmentAll( $resultSet );
915 $resultSet->setAugmentedData( $name, $data );
922 * Dummy class to be used when non-supported Database engine is present.
923 * @todo FIXME: Dummy class should probably try something at least mildly useful,
924 * such as a LIKE search through titles.
927 class SearchEngineDummy
extends SearchEngine
{