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 * @param bool $withAllKeyword activate support of the "all:" keyword and its
408 * translations to activate searching on all namespaces.
409 * @param bool $withPrefixSearchExtractNamespaceHook call the PrefixSearchExtractNamespace hook
410 * if classic namespace identification did not match.
411 * @return false|array false if no namespace was extracted, an array
412 * with the parsed query at index 0 and an array of namespaces at index
413 * 1 (or null for all namespaces).
415 * @throws MWException
417 public static function parseNamespacePrefixes(
419 $withAllKeyword = true,
420 $withPrefixSearchExtractNamespaceHook = false
425 if ( strpos( $query, ':' ) === false ) { // nothing to do
428 $extractedNamespace = null;
431 if ( $withAllKeyword ) {
434 $allkeywords[] = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
435 // force all: so that we have a common syntax for all the wikis
436 if ( !in_array( 'all:', $allkeywords ) ) {
437 $allkeywords[] = 'all:';
440 foreach ( $allkeywords as $kw ) {
441 if ( strncmp( $query, $kw, strlen( $kw ) ) == 0 ) {
442 $extractedNamespace = null;
443 $parsed = substr( $query, strlen( $kw ) );
450 if ( !$allQuery && strpos( $query, ':' ) !== false ) {
451 $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
452 $index = $wgContLang->getNsIndex( $prefix );
453 if ( $index !== false ) {
454 $extractedNamespace = [ $index ];
455 $parsed = substr( $query, strlen( $prefix ) +
1 );
456 } elseif ( $withPrefixSearchExtractNamespaceHook ) {
457 $hookNamespaces = [ NS_MAIN
];
459 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$hookNamespaces, &$hookQuery ] );
460 if ( $hookQuery !== $query ) {
461 $parsed = $hookQuery;
462 $extractedNamespace = $hookNamespaces;
471 return [ $parsed, $extractedNamespace ];
475 * Find snippet highlight settings for all users
476 * @return array Contextlines, contextchars
478 public static function userHighlightPrefs() {
479 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
480 $contextchars = 75; // same as above.... :P
481 return [ $contextlines, $contextchars ];
485 * Create or update the search index record for the given page.
486 * Title and text should be pre-processed.
490 * @param string $title
491 * @param string $text
493 function update( $id, $title, $text ) {
498 * Update a search index record's title only.
499 * Title should be pre-processed.
503 * @param string $title
505 function updateTitle( $id, $title ) {
510 * Delete an indexed page
511 * Title should be pre-processed.
514 * @param int $id Page id that was deleted
515 * @param string $title Title of page that was deleted
517 function delete( $id, $title ) {
522 * Get the raw text for updating the index from a content object
523 * Nicer search backends could possibly do something cooler than
524 * just returning raw text
526 * @todo This isn't ideal, we'd really like to have content-specific handling here
527 * @param Title $t Title we're indexing
528 * @param Content|null $c Content of the page to index
531 public function getTextFromContent( Title
$t, Content
$c = null ) {
532 return $c ?
$c->getTextForSearchIndex() : '';
536 * If an implementation of SearchEngine handles all of its own text processing
537 * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
538 * rather silly handling, it should return true here instead.
542 public function textAlreadyUpdatedForIndex() {
547 * Makes search simple string if it was namespaced.
548 * Sets namespaces of the search to namespaces extracted from string.
549 * @param string $search
550 * @return string Simplified search string
552 protected function normalizeNamespaces( $search ) {
553 $queryAndNs = self
::parseNamespacePrefixes( $search, false, true );
554 if ( $queryAndNs !== false ) {
555 $this->setNamespaces( $queryAndNs[1] );
556 return $queryAndNs[0];
562 * Perform an overfetch of completion search results. This allows
563 * determining if another page of results is available.
565 * @param string $search
566 * @return SearchSuggestionSet
568 protected function completionSearchBackendOverfetch( $search ) {
571 return $this->completionSearchBackend( $search );
578 * Perform a completion search.
579 * Does not resolve namespaces and does not check variants.
580 * Search engine implementations may want to override this function.
581 * @param string $search
582 * @return SearchSuggestionSet
584 protected function completionSearchBackend( $search ) {
587 $search = trim( $search );
589 if ( !in_array( NS_SPECIAL
, $this->namespaces
) && // We do not run hook on Special: search
590 !Hooks
::run( 'PrefixSearchBackend',
591 [ $this->namespaces
, $search, $this->limit
, &$results, $this->offset
]
593 // False means hook worked.
594 // FIXME: Yes, the API is weird. That's why it is going to be deprecated.
596 return SearchSuggestionSet
::fromStrings( $results );
598 // Hook did not do the job, use default simple search
599 $results = $this->simplePrefixSearch( $search );
600 return SearchSuggestionSet
::fromTitles( $results );
605 * Perform a completion search.
606 * @param string $search
607 * @return SearchSuggestionSet
609 public function completionSearch( $search ) {
610 if ( trim( $search ) === '' ) {
611 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
613 $search = $this->normalizeNamespaces( $search );
614 $suggestions = $this->completionSearchBackendOverfetch( $search );
615 return $this->processCompletionResults( $search, $suggestions );
619 * Perform a completion search with variants.
620 * @param string $search
621 * @return SearchSuggestionSet
623 public function completionSearchWithVariants( $search ) {
624 if ( trim( $search ) === '' ) {
625 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
627 $search = $this->normalizeNamespaces( $search );
629 $results = $this->completionSearchBackendOverfetch( $search );
630 $fallbackLimit = 1 +
$this->limit
- $results->getSize();
631 if ( $fallbackLimit > 0 ) {
634 $fallbackSearches = $wgContLang->autoConvertToAllVariants( $search );
635 $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
637 foreach ( $fallbackSearches as $fbs ) {
638 $this->setLimitOffset( $fallbackLimit );
639 $fallbackSearchResult = $this->completionSearch( $fbs );
640 $results->appendAll( $fallbackSearchResult );
641 $fallbackLimit -= $fallbackSearchResult->getSize();
642 if ( $fallbackLimit <= 0 ) {
647 return $this->processCompletionResults( $search, $results );
651 * Extract titles from completion results
652 * @param SearchSuggestionSet $completionResults
655 public function extractTitles( SearchSuggestionSet
$completionResults ) {
656 return $completionResults->map( function ( SearchSuggestion
$sugg ) {
657 return $sugg->getSuggestedTitle();
662 * Process completion search results.
663 * Resolves the titles and rescores.
664 * @param string $search
665 * @param SearchSuggestionSet $suggestions
666 * @return SearchSuggestionSet
668 protected function processCompletionResults( $search, SearchSuggestionSet
$suggestions ) {
669 // We over-fetched to determine pagination. Shrink back down if we have extra results
670 // and mark if pagination is possible
671 $suggestions->shrink( $this->limit
);
673 $search = trim( $search );
674 // preload the titles with LinkBatch
675 $lb = new LinkBatch( $suggestions->map( function ( SearchSuggestion
$sugg ) {
676 return $sugg->getSuggestedTitle();
678 $lb->setCaller( __METHOD__
);
681 $diff = $suggestions->filter( function ( SearchSuggestion
$sugg ) {
682 return $sugg->getSuggestedTitle()->isKnown();
685 MediaWikiServices
::getInstance()->getStatsdDataFactory()
686 ->updateCount( 'search.completion.missing', $diff );
689 $results = $suggestions->map( function ( SearchSuggestion
$sugg ) {
690 return $sugg->getSuggestedTitle()->getPrefixedText();
693 if ( $this->offset
=== 0 ) {
694 // Rescore results with an exact title match
695 // NOTE: in some cases like cross-namespace redirects
696 // (frequently used as shortcuts e.g. WP:WP on huwiki) some
697 // backends like Cirrus will return no results. We should still
698 // try an exact title match to workaround this limitation
699 $rescorer = new SearchExactMatchRescorer();
700 $rescoredResults = $rescorer->rescore( $search, $this->namespaces
, $results, $this->limit
);
702 // No need to rescore if offset is not 0
703 // The exact match must have been returned at position 0
705 $rescoredResults = $results;
708 if ( count( $rescoredResults ) > 0 ) {
709 $found = array_search( $rescoredResults[0], $results );
710 if ( $found === false ) {
711 // If the first result is not in the previous array it
712 // means that we found a new exact match
713 $exactMatch = SearchSuggestion
::fromTitle( 0, Title
::newFromText( $rescoredResults[0] ) );
714 $suggestions->prepend( $exactMatch );
715 $suggestions->shrink( $this->limit
);
717 // if the first result is not the same we need to rescore
719 $suggestions->rescore( $found );
728 * Simple prefix search for subpages.
729 * @param string $search
732 public function defaultPrefixSearch( $search ) {
733 if ( trim( $search ) === '' ) {
737 $search = $this->normalizeNamespaces( $search );
738 return $this->simplePrefixSearch( $search );
742 * Call out to simple search backend.
743 * Defaults to TitlePrefixSearch.
744 * @param string $search
747 protected function simplePrefixSearch( $search ) {
748 // Use default database prefix search
749 $backend = new TitlePrefixSearch
;
750 return $backend->defaultSearchBackend( $this->namespaces
, $search, $this->limit
, $this->offset
);
754 * Make a list of searchable namespaces and their canonical names.
755 * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
758 public static function searchableNamespaces() {
759 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->searchableNamespaces();
763 * Extract default namespaces to search from the given user's
764 * settings, returning a list of index numbers.
765 * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
769 public static function userNamespaces( $user ) {
770 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
774 * An array of namespaces indexes to be searched by default
775 * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
778 public static function defaultNamespaces() {
779 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->defaultNamespaces();
783 * Get a list of namespace names useful for showing in tooltips
785 * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
786 * @param array $namespaces
789 public static function namespacesAsText( $namespaces ) {
790 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
794 * Load up the appropriate search engine class for the currently
795 * active database backend, and return a configured instance.
796 * @deprecated since 1.27; Use SearchEngineFactory::create
797 * @param string $type Type of search backend, if not the default
798 * @return SearchEngine
800 public static function create( $type = null ) {
801 return MediaWikiServices
::getInstance()->getSearchEngineFactory()->create( $type );
805 * Return the search engines we support. If only $wgSearchType
806 * is set, it'll be an array of just that one item.
807 * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
810 public static function getSearchTypes() {
811 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->getSearchTypes();
815 * Get a list of supported profiles.
816 * Some search engine implementations may expose specific profiles to fine-tune
818 * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName )
819 * The array returned by this function contains the following keys:
820 * - name: the profile name to use with setFeatureData
821 * - desc-message: the i18n description
822 * - default: set to true if this profile is the default
825 * @param string $profileType the type of profiles
826 * @param User|null $user the user requesting the list of profiles
827 * @return array|null the list of profiles or null if none available
829 public function getProfiles( $profileType, User
$user = null ) {
834 * Create a search field definition.
835 * Specific search engines should override this method to create search fields.
836 * @param string $name
837 * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
838 * @return SearchIndexField
841 public function makeSearchFieldMapping( $name, $type ) {
842 return new NullIndexField();
846 * Get fields for search index
848 * @return SearchIndexField[] Index field definitions for all content handlers
850 public function getSearchIndexFields() {
851 $models = ContentHandler
::getContentModels();
853 $seenHandlers = new SplObjectStorage();
854 foreach ( $models as $model ) {
856 $handler = ContentHandler
::getForModelID( $model );
858 catch ( MWUnknownContentModelException
$e ) {
859 // If we can find no handler, ignore it
862 // Several models can have the same handler, so avoid processing it repeatedly
863 if ( $seenHandlers->contains( $handler ) ) {
864 // We already did this one
867 $seenHandlers->attach( $handler );
868 $handlerFields = $handler->getFieldsForSearchIndex( $this );
869 foreach ( $handlerFields as $fieldName => $fieldData ) {
870 if ( empty( $fields[$fieldName] ) ) {
871 $fields[$fieldName] = $fieldData;
873 // TODO: do we allow some clashes with the same type or reject all of them?
874 $mergeDef = $fields[$fieldName]->merge( $fieldData );
876 throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" );
878 $fields[$fieldName] = $mergeDef;
882 // Hook to allow extensions to produce search mapping fields
883 Hooks
::run( 'SearchIndexFields', [ &$fields, $this ] );
888 * Augment search results with extra data.
890 * @param SearchResultSet $resultSet
892 public function augmentSearchResults( SearchResultSet
$resultSet ) {
895 Hooks
::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
896 if ( !$setAugmentors && !$rowAugmentors ) {
901 // Convert row augmentors to set augmentor
902 foreach ( $rowAugmentors as $name => $row ) {
903 if ( isset( $setAugmentors[$name] ) ) {
904 throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" );
906 $setAugmentors[$name] = new PerRowAugmentor( $row );
909 foreach ( $setAugmentors as $name => $augmentor ) {
910 $data = $augmentor->augmentAll( $resultSet );
912 $resultSet->setAugmentedData( $name, $data );
919 * Dummy class to be used when non-supported Database engine is present.
920 * @todo FIXME: Dummy class should probably try something at least mildly useful,
921 * such as a LIKE search through titles.
924 class SearchEngineDummy
extends SearchEngine
{