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
247 * @deprecated since 1.32 this should now be handled internally by the
250 public function transformSearchTerm( $term ) {
255 * Get service class to finding near matches.
256 * @param Config $config Configuration to use for the matcher.
257 * @return SearchNearMatcher
259 public function getNearMatcher( Config
$config ) {
261 return new SearchNearMatcher( $config, $wgContLang );
265 * Get near matcher for default SearchEngine.
266 * @return SearchNearMatcher
268 protected static function defaultNearMatcher() {
269 $config = MediaWikiServices
::getInstance()->getMainConfig();
270 return MediaWikiServices
::getInstance()->newSearchEngine()->getNearMatcher( $config );
274 * If an exact title match can be found, or a very slightly close match,
275 * return the title. If no match, returns NULL.
276 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
277 * @param string $searchterm
280 public static function getNearMatch( $searchterm ) {
281 return static::defaultNearMatcher()->getNearMatch( $searchterm );
285 * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
287 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
288 * @param string $searchterm
289 * @return SearchResultSet
291 public static function getNearMatchResultSet( $searchterm ) {
292 return static::defaultNearMatcher()->getNearMatchResultSet( $searchterm );
296 * Get chars legal for search
297 * NOTE: usage as static is deprecated and preserved only as BC measure
298 * @param int $type type of search chars (see self::CHARS_ALL
299 * and self::CHARS_NO_SYNTAX). Defaults to CHARS_ALL
302 public static function legalSearchChars( $type = self
::CHARS_ALL
) {
303 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
307 * Set the maximum number of results to return
308 * and how many to skip before returning the first.
313 function setLimitOffset( $limit, $offset = 0 ) {
314 $this->limit
= intval( $limit );
315 $this->offset
= intval( $offset );
319 * Set which namespaces the search should include.
320 * Give an array of namespace index numbers.
322 * @param int[]|null $namespaces
324 function setNamespaces( $namespaces ) {
326 // Filter namespaces to only keep valid ones
327 $validNs = $this->searchableNamespaces();
328 $namespaces = array_filter( $namespaces, function ( $ns ) use( $validNs ) {
329 return $ns < 0 ||
isset( $validNs[$ns] );
334 $this->namespaces
= $namespaces;
338 * Set whether the searcher should try to build a suggestion. Note: some searchers
339 * don't support building a suggestion in the first place and others don't respect
342 * @param bool $showSuggestion Should the searcher try to build suggestions
344 function setShowSuggestion( $showSuggestion ) {
345 $this->showSuggestion
= $showSuggestion;
349 * Get the valid sort directions. All search engines support 'relevance' but others
350 * might support more. The default in all implementations should be 'relevance.'
353 * @return string[] the valid sort directions for setSort
355 public function getValidSorts() {
356 return [ 'relevance' ];
360 * Set the sort direction of the search results. Must be one returned by
361 * SearchEngine::getValidSorts()
364 * @throws InvalidArgumentException
365 * @param string $sort sort direction for query result
367 public function setSort( $sort ) {
368 if ( !in_array( $sort, $this->getValidSorts() ) ) {
369 throw new InvalidArgumentException( "Invalid sort: $sort. " .
370 "Must be one of: " . implode( ', ', $this->getValidSorts() ) );
376 * Get the sort direction of the search results
381 public function getSort() {
386 * Parse some common prefixes: all (search everything)
387 * or namespace names and set the list of namespaces
388 * of this class accordingly.
390 * @param string $query
393 function replacePrefixes( $query ) {
394 $queryAndNs = self
::parseNamespacePrefixes( $query );
395 if ( $queryAndNs === false ) {
398 $this->namespaces
= $queryAndNs[1];
399 return $queryAndNs[0];
403 * Parse some common prefixes: all (search everything)
406 * @param string $query
407 * @return false|array false if no namespace was extracted, an array
408 * with the parsed query at index 0 and an array of namespaces at index
409 * 1 (or null for all namespaces).
411 public static function parseNamespacePrefixes( $query ) {
415 if ( strpos( $query, ':' ) === false ) { // nothing to do
418 $extractedNamespace = null;
421 $allkeywords[] = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
422 // force all: so that we have a common syntax for all the wikis
423 if ( !in_array( 'all:', $allkeywords ) ) {
424 $allkeywords[] = 'all:';
428 foreach ( $allkeywords as $kw ) {
429 if ( strncmp( $query, $kw, strlen( $kw ) ) == 0 ) {
430 $extractedNamespace = null;
431 $parsed = substr( $query, strlen( $kw ) );
437 if ( !$allQuery && strpos( $query, ':' ) !== false ) {
438 // TODO: should we unify with PrefixSearch::extractNamespace ?
439 $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
440 $index = $wgContLang->getNsIndex( $prefix );
441 if ( $index !== false ) {
442 $extractedNamespace = [ $index ];
443 $parsed = substr( $query, strlen( $prefix ) +
1 );
449 if ( trim( $parsed ) == '' ) {
450 $parsed = $query; // prefix was the whole query
453 return [ $parsed, $extractedNamespace ];
457 * Find snippet highlight settings for all users
458 * @return array Contextlines, contextchars
460 public static function userHighlightPrefs() {
461 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
462 $contextchars = 75; // same as above.... :P
463 return [ $contextlines, $contextchars ];
467 * Create or update the search index record for the given page.
468 * Title and text should be pre-processed.
472 * @param string $title
473 * @param string $text
475 function update( $id, $title, $text ) {
480 * Update a search index record's title only.
481 * Title should be pre-processed.
485 * @param string $title
487 function updateTitle( $id, $title ) {
492 * Delete an indexed page
493 * Title should be pre-processed.
496 * @param int $id Page id that was deleted
497 * @param string $title Title of page that was deleted
499 function delete( $id, $title ) {
504 * Get the raw text for updating the index from a content object
505 * Nicer search backends could possibly do something cooler than
506 * just returning raw text
508 * @todo This isn't ideal, we'd really like to have content-specific handling here
509 * @param Title $t Title we're indexing
510 * @param Content|null $c Content of the page to index
513 public function getTextFromContent( Title
$t, Content
$c = null ) {
514 return $c ?
$c->getTextForSearchIndex() : '';
518 * If an implementation of SearchEngine handles all of its own text processing
519 * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
520 * rather silly handling, it should return true here instead.
524 public function textAlreadyUpdatedForIndex() {
529 * Makes search simple string if it was namespaced.
530 * Sets namespaces of the search to namespaces extracted from string.
531 * @param string $search
532 * @return string Simplified search string
534 protected function normalizeNamespaces( $search ) {
535 // Find a Title which is not an interwiki and is in NS_MAIN
536 $title = Title
::newFromText( $search );
537 $ns = $this->namespaces
;
538 if ( $title && !$title->isExternal() ) {
539 $ns = [ $title->getNamespace() ];
540 $search = $title->getText();
541 if ( $ns[0] == NS_MAIN
) {
542 $ns = $this->namespaces
; // no explicit prefix, use default namespaces
543 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
546 $title = Title
::newFromText( $search . 'Dummy' );
547 if ( $title && $title->getText() == 'Dummy'
548 && $title->getNamespace() != NS_MAIN
549 && !$title->isExternal()
551 $ns = [ $title->getNamespace() ];
554 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
558 $ns = array_map( function ( $space ) {
559 return $space == NS_MEDIA ? NS_FILE
: $space;
562 $this->setNamespaces( $ns );
567 * Perform an overfetch of completion search results. This allows
568 * determining if another page of results is available.
570 * @param string $search
571 * @return SearchSuggestionSet
573 protected function completionSearchBackendOverfetch( $search ) {
576 return $this->completionSearchBackend( $search );
583 * Perform a completion search.
584 * Does not resolve namespaces and does not check variants.
585 * Search engine implementations may want to override this function.
586 * @param string $search
587 * @return SearchSuggestionSet
589 protected function completionSearchBackend( $search ) {
592 $search = trim( $search );
594 if ( !in_array( NS_SPECIAL
, $this->namespaces
) && // We do not run hook on Special: search
595 !Hooks
::run( 'PrefixSearchBackend',
596 [ $this->namespaces
, $search, $this->limit
, &$results, $this->offset
]
598 // False means hook worked.
599 // FIXME: Yes, the API is weird. That's why it is going to be deprecated.
601 return SearchSuggestionSet
::fromStrings( $results );
603 // Hook did not do the job, use default simple search
604 $results = $this->simplePrefixSearch( $search );
605 return SearchSuggestionSet
::fromTitles( $results );
610 * Perform a completion search.
611 * @param string $search
612 * @return SearchSuggestionSet
614 public function completionSearch( $search ) {
615 if ( trim( $search ) === '' ) {
616 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
618 $search = $this->normalizeNamespaces( $search );
619 $suggestions = $this->completionSearchBackendOverfetch( $search );
620 return $this->processCompletionResults( $search, $suggestions );
624 * Perform a completion search with variants.
625 * @param string $search
626 * @return SearchSuggestionSet
628 public function completionSearchWithVariants( $search ) {
629 if ( trim( $search ) === '' ) {
630 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
632 $search = $this->normalizeNamespaces( $search );
634 $results = $this->completionSearchBackendOverfetch( $search );
635 $fallbackLimit = 1 +
$this->limit
- $results->getSize();
636 if ( $fallbackLimit > 0 ) {
639 $fallbackSearches = $wgContLang->autoConvertToAllVariants( $search );
640 $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
642 foreach ( $fallbackSearches as $fbs ) {
643 $this->setLimitOffset( $fallbackLimit );
644 $fallbackSearchResult = $this->completionSearch( $fbs );
645 $results->appendAll( $fallbackSearchResult );
646 $fallbackLimit -= $fallbackSearchResult->getSize();
647 if ( $fallbackLimit <= 0 ) {
652 return $this->processCompletionResults( $search, $results );
656 * Extract titles from completion results
657 * @param SearchSuggestionSet $completionResults
660 public function extractTitles( SearchSuggestionSet
$completionResults ) {
661 return $completionResults->map( function ( SearchSuggestion
$sugg ) {
662 return $sugg->getSuggestedTitle();
667 * Process completion search results.
668 * Resolves the titles and rescores.
669 * @param string $search
670 * @param SearchSuggestionSet $suggestions
671 * @return SearchSuggestionSet
673 protected function processCompletionResults( $search, SearchSuggestionSet
$suggestions ) {
674 // We over-fetched to determine pagination. Shrink back down if we have extra results
675 // and mark if pagination is possible
676 $suggestions->shrink( $this->limit
);
678 $search = trim( $search );
679 // preload the titles with LinkBatch
680 $lb = new LinkBatch( $suggestions->map( function ( SearchSuggestion
$sugg ) {
681 return $sugg->getSuggestedTitle();
683 $lb->setCaller( __METHOD__
);
686 $diff = $suggestions->filter( function ( SearchSuggestion
$sugg ) {
687 return $sugg->getSuggestedTitle()->isKnown();
690 MediaWikiServices
::getInstance()->getStatsdDataFactory()
691 ->updateCount( 'search.completion.missing', $diff );
694 $results = $suggestions->map( function ( SearchSuggestion
$sugg ) {
695 return $sugg->getSuggestedTitle()->getPrefixedText();
698 if ( $this->offset
=== 0 ) {
699 // Rescore results with an exact title match
700 // NOTE: in some cases like cross-namespace redirects
701 // (frequently used as shortcuts e.g. WP:WP on huwiki) some
702 // backends like Cirrus will return no results. We should still
703 // try an exact title match to workaround this limitation
704 $rescorer = new SearchExactMatchRescorer();
705 $rescoredResults = $rescorer->rescore( $search, $this->namespaces
, $results, $this->limit
);
707 // No need to rescore if offset is not 0
708 // The exact match must have been returned at position 0
710 $rescoredResults = $results;
713 if ( count( $rescoredResults ) > 0 ) {
714 $found = array_search( $rescoredResults[0], $results );
715 if ( $found === false ) {
716 // If the first result is not in the previous array it
717 // means that we found a new exact match
718 $exactMatch = SearchSuggestion
::fromTitle( 0, Title
::newFromText( $rescoredResults[0] ) );
719 $suggestions->prepend( $exactMatch );
720 $suggestions->shrink( $this->limit
);
722 // if the first result is not the same we need to rescore
724 $suggestions->rescore( $found );
733 * Simple prefix search for subpages.
734 * @param string $search
737 public function defaultPrefixSearch( $search ) {
738 if ( trim( $search ) === '' ) {
742 $search = $this->normalizeNamespaces( $search );
743 return $this->simplePrefixSearch( $search );
747 * Call out to simple search backend.
748 * Defaults to TitlePrefixSearch.
749 * @param string $search
752 protected function simplePrefixSearch( $search ) {
753 // Use default database prefix search
754 $backend = new TitlePrefixSearch
;
755 return $backend->defaultSearchBackend( $this->namespaces
, $search, $this->limit
, $this->offset
);
759 * Make a list of searchable namespaces and their canonical names.
760 * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
763 public static function searchableNamespaces() {
764 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->searchableNamespaces();
768 * Extract default namespaces to search from the given user's
769 * settings, returning a list of index numbers.
770 * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
774 public static function userNamespaces( $user ) {
775 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
779 * An array of namespaces indexes to be searched by default
780 * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
783 public static function defaultNamespaces() {
784 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->defaultNamespaces();
788 * Get a list of namespace names useful for showing in tooltips
790 * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
791 * @param array $namespaces
794 public static function namespacesAsText( $namespaces ) {
795 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
799 * Load up the appropriate search engine class for the currently
800 * active database backend, and return a configured instance.
801 * @deprecated since 1.27; Use SearchEngineFactory::create
802 * @param string $type Type of search backend, if not the default
803 * @return SearchEngine
805 public static function create( $type = null ) {
806 return MediaWikiServices
::getInstance()->getSearchEngineFactory()->create( $type );
810 * Return the search engines we support. If only $wgSearchType
811 * is set, it'll be an array of just that one item.
812 * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
815 public static function getSearchTypes() {
816 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->getSearchTypes();
820 * Get a list of supported profiles.
821 * Some search engine implementations may expose specific profiles to fine-tune
823 * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName )
824 * The array returned by this function contains the following keys:
825 * - name: the profile name to use with setFeatureData
826 * - desc-message: the i18n description
827 * - default: set to true if this profile is the default
830 * @param string $profileType the type of profiles
831 * @param User|null $user the user requesting the list of profiles
832 * @return array|null the list of profiles or null if none available
834 public function getProfiles( $profileType, User
$user = null ) {
839 * Create a search field definition.
840 * Specific search engines should override this method to create search fields.
841 * @param string $name
842 * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
843 * @return SearchIndexField
846 public function makeSearchFieldMapping( $name, $type ) {
847 return new NullIndexField();
851 * Get fields for search index
853 * @return SearchIndexField[] Index field definitions for all content handlers
855 public function getSearchIndexFields() {
856 $models = ContentHandler
::getContentModels();
858 $seenHandlers = new SplObjectStorage();
859 foreach ( $models as $model ) {
861 $handler = ContentHandler
::getForModelID( $model );
863 catch ( MWUnknownContentModelException
$e ) {
864 // If we can find no handler, ignore it
867 // Several models can have the same handler, so avoid processing it repeatedly
868 if ( $seenHandlers->contains( $handler ) ) {
869 // We already did this one
872 $seenHandlers->attach( $handler );
873 $handlerFields = $handler->getFieldsForSearchIndex( $this );
874 foreach ( $handlerFields as $fieldName => $fieldData ) {
875 if ( empty( $fields[$fieldName] ) ) {
876 $fields[$fieldName] = $fieldData;
878 // TODO: do we allow some clashes with the same type or reject all of them?
879 $mergeDef = $fields[$fieldName]->merge( $fieldData );
881 throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" );
883 $fields[$fieldName] = $mergeDef;
887 // Hook to allow extensions to produce search mapping fields
888 Hooks
::run( 'SearchIndexFields', [ &$fields, $this ] );
893 * Augment search results with extra data.
895 * @param SearchResultSet $resultSet
897 public function augmentSearchResults( SearchResultSet
$resultSet ) {
900 Hooks
::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
901 if ( !$setAugmentors && !$rowAugmentors ) {
906 // Convert row augmentors to set augmentor
907 foreach ( $rowAugmentors as $name => $row ) {
908 if ( isset( $setAugmentors[$name] ) ) {
909 throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" );
911 $setAugmentors[$name] = new PerRowAugmentor( $row );
914 foreach ( $setAugmentors as $name => $augmentor ) {
915 $data = $augmentor->augmentAll( $resultSet );
917 $resultSet->setAugmentedData( $name, $data );
924 * Dummy class to be used when non-supported Database engine is present.
925 * @todo FIXME: Dummy class should probably try something at least mildly useful,
926 * such as a LIKE search through titles.
929 class SearchEngineDummy
extends SearchEngine
{