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 if ( isset( $this->features
[$feature] ) ) {
223 return $this->features
[$feature];
229 * When overridden in derived class, performs database-specific conversions
230 * on text to be used for searching or updating search index.
231 * Default implementation does nothing (simply returns $string).
233 * @param string $string String to process
236 public function normalizeText( $string ) {
237 // Some languages such as Chinese require word segmentation
238 return MediaWikiServices
::getInstance()->getContentLanguage()->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 ) {
260 return new SearchNearMatcher( $config,
261 MediaWikiServices
::getInstance()->getContentLanguage() );
265 * Get near matcher for default SearchEngine.
266 * @return SearchNearMatcher
268 protected static function defaultNearMatcher() {
269 $services = MediaWikiServices
::getInstance();
270 $config = $services->getMainConfig();
271 return $services->newSearchEngine()->getNearMatcher( $config );
275 * If an exact title match can be found, or a very slightly close match,
276 * return the title. If no match, returns NULL.
277 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
278 * @param string $searchterm
281 public static function getNearMatch( $searchterm ) {
282 return static::defaultNearMatcher()->getNearMatch( $searchterm );
286 * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
288 * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
289 * @param string $searchterm
290 * @return SearchResultSet
292 public static function getNearMatchResultSet( $searchterm ) {
293 return static::defaultNearMatcher()->getNearMatchResultSet( $searchterm );
297 * Get chars legal for search
298 * NOTE: usage as static is deprecated and preserved only as BC measure
299 * @param int $type type of search chars (see self::CHARS_ALL
300 * and self::CHARS_NO_SYNTAX). Defaults to CHARS_ALL
303 public static function legalSearchChars( $type = self
::CHARS_ALL
) {
304 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
308 * Set the maximum number of results to return
309 * and how many to skip before returning the first.
314 function setLimitOffset( $limit, $offset = 0 ) {
315 $this->limit
= intval( $limit );
316 $this->offset
= intval( $offset );
320 * Set which namespaces the search should include.
321 * Give an array of namespace index numbers.
323 * @param int[]|null $namespaces
325 function setNamespaces( $namespaces ) {
327 // Filter namespaces to only keep valid ones
328 $validNs = $this->searchableNamespaces();
329 $namespaces = array_filter( $namespaces, function ( $ns ) use( $validNs ) {
330 return $ns < 0 ||
isset( $validNs[$ns] );
335 $this->namespaces
= $namespaces;
339 * Set whether the searcher should try to build a suggestion. Note: some searchers
340 * don't support building a suggestion in the first place and others don't respect
343 * @param bool $showSuggestion Should the searcher try to build suggestions
345 function setShowSuggestion( $showSuggestion ) {
346 $this->showSuggestion
= $showSuggestion;
350 * Get the valid sort directions. All search engines support 'relevance' but others
351 * might support more. The default in all implementations must be 'relevance.'
354 * @return string[] the valid sort directions for setSort
356 public function getValidSorts() {
357 return [ self
::DEFAULT_SORT
];
361 * Set the sort direction of the search results. Must be one returned by
362 * SearchEngine::getValidSorts()
365 * @throws InvalidArgumentException
366 * @param string $sort sort direction for query result
368 public function setSort( $sort ) {
369 if ( !in_array( $sort, $this->getValidSorts() ) ) {
370 throw new InvalidArgumentException( "Invalid sort: $sort. " .
371 "Must be one of: " . implode( ', ', $this->getValidSorts() ) );
377 * Get the sort direction of the search results
382 public function getSort() {
387 * Parse some common prefixes: all (search everything)
388 * or namespace names and set the list of namespaces
389 * of this class accordingly.
391 * @deprecated since 1.32; should be handled internally by the search engine
392 * @param string $query
395 function replacePrefixes( $query ) {
400 * Parse some common prefixes: all (search everything)
403 * @param string $query
404 * @param bool $withAllKeyword activate support of the "all:" keyword and its
405 * translations to activate searching on all namespaces.
406 * @param bool $withPrefixSearchExtractNamespaceHook call the PrefixSearchExtractNamespace hook
407 * if classic namespace identification did not match.
408 * @return false|array false if no namespace was extracted, an array
409 * with the parsed query at index 0 and an array of namespaces at index
410 * 1 (or null for all namespaces).
412 * @throws MWException
414 public static function parseNamespacePrefixes(
416 $withAllKeyword = true,
417 $withPrefixSearchExtractNamespaceHook = false
420 if ( strpos( $query, ':' ) === false ) { // nothing to do
423 $extractedNamespace = null;
426 if ( $withAllKeyword ) {
429 $allkeywords[] = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
430 // force all: so that we have a common syntax for all the wikis
431 if ( !in_array( 'all:', $allkeywords ) ) {
432 $allkeywords[] = 'all:';
435 foreach ( $allkeywords as $kw ) {
436 if ( strncmp( $query, $kw, strlen( $kw ) ) == 0 ) {
437 $extractedNamespace = null;
438 $parsed = substr( $query, strlen( $kw ) );
445 if ( !$allQuery && strpos( $query, ':' ) !== false ) {
446 $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
447 $index = MediaWikiServices
::getInstance()->getContentLanguage()->getNsIndex( $prefix );
448 if ( $index !== false ) {
449 $extractedNamespace = [ $index ];
450 $parsed = substr( $query, strlen( $prefix ) +
1 );
451 } elseif ( $withPrefixSearchExtractNamespaceHook ) {
452 $hookNamespaces = [ NS_MAIN
];
454 Hooks
::run( 'PrefixSearchExtractNamespace', [ &$hookNamespaces, &$hookQuery ] );
455 if ( $hookQuery !== $query ) {
456 $parsed = $hookQuery;
457 $extractedNamespace = $hookNamespaces;
466 return [ $parsed, $extractedNamespace ];
470 * Find snippet highlight settings for all users
471 * @return array Contextlines, contextchars
473 public static function userHighlightPrefs() {
474 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
475 $contextchars = 75; // same as above.... :P
476 return [ $contextlines, $contextchars ];
480 * Create or update the search index record for the given page.
481 * Title and text should be pre-processed.
485 * @param string $title
486 * @param string $text
488 function update( $id, $title, $text ) {
493 * Update a search index record's title only.
494 * Title should be pre-processed.
498 * @param string $title
500 function updateTitle( $id, $title ) {
505 * Delete an indexed page
506 * Title should be pre-processed.
509 * @param int $id Page id that was deleted
510 * @param string $title Title of page that was deleted
512 function delete( $id, $title ) {
517 * Get the raw text for updating the index from a content object
518 * Nicer search backends could possibly do something cooler than
519 * just returning raw text
521 * @todo This isn't ideal, we'd really like to have content-specific handling here
522 * @param Title $t Title we're indexing
523 * @param Content|null $c Content of the page to index
526 public function getTextFromContent( Title
$t, Content
$c = null ) {
527 return $c ?
$c->getTextForSearchIndex() : '';
531 * If an implementation of SearchEngine handles all of its own text processing
532 * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
533 * rather silly handling, it should return true here instead.
537 public function textAlreadyUpdatedForIndex() {
542 * Makes search simple string if it was namespaced.
543 * Sets namespaces of the search to namespaces extracted from string.
544 * @param string $search
545 * @return string Simplified search string
547 protected function normalizeNamespaces( $search ) {
548 $queryAndNs = self
::parseNamespacePrefixes( $search, false, true );
549 if ( $queryAndNs !== false ) {
550 $this->setNamespaces( $queryAndNs[1] );
551 return $queryAndNs[0];
557 * Perform an overfetch of completion search results. This allows
558 * determining if another page of results is available.
560 * @param string $search
561 * @return SearchSuggestionSet
563 protected function completionSearchBackendOverfetch( $search ) {
566 return $this->completionSearchBackend( $search );
573 * Perform a completion search.
574 * Does not resolve namespaces and does not check variants.
575 * Search engine implementations may want to override this function.
576 * @param string $search
577 * @return SearchSuggestionSet
579 protected function completionSearchBackend( $search ) {
582 $search = trim( $search );
584 if ( !in_array( NS_SPECIAL
, $this->namespaces
) && // We do not run hook on Special: search
585 !Hooks
::run( 'PrefixSearchBackend',
586 [ $this->namespaces
, $search, $this->limit
, &$results, $this->offset
]
588 // False means hook worked.
589 // FIXME: Yes, the API is weird. That's why it is going to be deprecated.
591 return SearchSuggestionSet
::fromStrings( $results );
593 // Hook did not do the job, use default simple search
594 $results = $this->simplePrefixSearch( $search );
595 return SearchSuggestionSet
::fromTitles( $results );
600 * Perform a completion search.
601 * @param string $search
602 * @return SearchSuggestionSet
604 public function completionSearch( $search ) {
605 if ( trim( $search ) === '' ) {
606 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
608 $search = $this->normalizeNamespaces( $search );
609 $suggestions = $this->completionSearchBackendOverfetch( $search );
610 return $this->processCompletionResults( $search, $suggestions );
614 * Perform a completion search with variants.
615 * @param string $search
616 * @return SearchSuggestionSet
618 public function completionSearchWithVariants( $search ) {
619 if ( trim( $search ) === '' ) {
620 return SearchSuggestionSet
::emptySuggestionSet(); // Return empty result
622 $search = $this->normalizeNamespaces( $search );
624 $results = $this->completionSearchBackendOverfetch( $search );
625 $fallbackLimit = 1 +
$this->limit
- $results->getSize();
626 if ( $fallbackLimit > 0 ) {
627 $fallbackSearches = MediaWikiServices
::getInstance()->getContentLanguage()->
628 autoConvertToAllVariants( $search );
629 $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
631 foreach ( $fallbackSearches as $fbs ) {
632 $this->setLimitOffset( $fallbackLimit );
633 $fallbackSearchResult = $this->completionSearch( $fbs );
634 $results->appendAll( $fallbackSearchResult );
635 $fallbackLimit -= $fallbackSearchResult->getSize();
636 if ( $fallbackLimit <= 0 ) {
641 return $this->processCompletionResults( $search, $results );
645 * Extract titles from completion results
646 * @param SearchSuggestionSet $completionResults
649 public function extractTitles( SearchSuggestionSet
$completionResults ) {
650 return $completionResults->map( function ( SearchSuggestion
$sugg ) {
651 return $sugg->getSuggestedTitle();
656 * Process completion search results.
657 * Resolves the titles and rescores.
658 * @param string $search
659 * @param SearchSuggestionSet $suggestions
660 * @return SearchSuggestionSet
662 protected function processCompletionResults( $search, SearchSuggestionSet
$suggestions ) {
663 // We over-fetched to determine pagination. Shrink back down if we have extra results
664 // and mark if pagination is possible
665 $suggestions->shrink( $this->limit
);
667 $search = trim( $search );
668 // preload the titles with LinkBatch
669 $lb = new LinkBatch( $suggestions->map( function ( SearchSuggestion
$sugg ) {
670 return $sugg->getSuggestedTitle();
672 $lb->setCaller( __METHOD__
);
675 $diff = $suggestions->filter( function ( SearchSuggestion
$sugg ) {
676 return $sugg->getSuggestedTitle()->isKnown();
679 MediaWikiServices
::getInstance()->getStatsdDataFactory()
680 ->updateCount( 'search.completion.missing', $diff );
683 $results = $suggestions->map( function ( SearchSuggestion
$sugg ) {
684 return $sugg->getSuggestedTitle()->getPrefixedText();
687 if ( $this->offset
=== 0 ) {
688 // Rescore results with an exact title match
689 // NOTE: in some cases like cross-namespace redirects
690 // (frequently used as shortcuts e.g. WP:WP on huwiki) some
691 // backends like Cirrus will return no results. We should still
692 // try an exact title match to workaround this limitation
693 $rescorer = new SearchExactMatchRescorer();
694 $rescoredResults = $rescorer->rescore( $search, $this->namespaces
, $results, $this->limit
);
696 // No need to rescore if offset is not 0
697 // The exact match must have been returned at position 0
699 $rescoredResults = $results;
702 if ( count( $rescoredResults ) > 0 ) {
703 $found = array_search( $rescoredResults[0], $results );
704 if ( $found === false ) {
705 // If the first result is not in the previous array it
706 // means that we found a new exact match
707 $exactMatch = SearchSuggestion
::fromTitle( 0, Title
::newFromText( $rescoredResults[0] ) );
708 $suggestions->prepend( $exactMatch );
709 $suggestions->shrink( $this->limit
);
711 // if the first result is not the same we need to rescore
713 $suggestions->rescore( $found );
722 * Simple prefix search for subpages.
723 * @param string $search
726 public function defaultPrefixSearch( $search ) {
727 if ( trim( $search ) === '' ) {
731 $search = $this->normalizeNamespaces( $search );
732 return $this->simplePrefixSearch( $search );
736 * Call out to simple search backend.
737 * Defaults to TitlePrefixSearch.
738 * @param string $search
741 protected function simplePrefixSearch( $search ) {
742 // Use default database prefix search
743 $backend = new TitlePrefixSearch
;
744 return $backend->defaultSearchBackend( $this->namespaces
, $search, $this->limit
, $this->offset
);
748 * Make a list of searchable namespaces and their canonical names.
749 * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
752 public static function searchableNamespaces() {
753 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->searchableNamespaces();
757 * Extract default namespaces to search from the given user's
758 * settings, returning a list of index numbers.
759 * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
763 public static function userNamespaces( $user ) {
764 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
768 * An array of namespaces indexes to be searched by default
769 * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
772 public static function defaultNamespaces() {
773 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->defaultNamespaces();
777 * Get a list of namespace names useful for showing in tooltips
779 * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
780 * @param array $namespaces
783 public static function namespacesAsText( $namespaces ) {
784 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
788 * Load up the appropriate search engine class for the currently
789 * active database backend, and return a configured instance.
790 * @deprecated since 1.27; Use SearchEngineFactory::create
791 * @param string $type Type of search backend, if not the default
792 * @return SearchEngine
794 public static function create( $type = null ) {
795 return MediaWikiServices
::getInstance()->getSearchEngineFactory()->create( $type );
799 * Return the search engines we support. If only $wgSearchType
800 * is set, it'll be an array of just that one item.
801 * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
804 public static function getSearchTypes() {
805 return MediaWikiServices
::getInstance()->getSearchEngineConfig()->getSearchTypes();
809 * Get a list of supported profiles.
810 * Some search engine implementations may expose specific profiles to fine-tune
812 * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName )
813 * The array returned by this function contains the following keys:
814 * - name: the profile name to use with setFeatureData
815 * - desc-message: the i18n description
816 * - default: set to true if this profile is the default
819 * @param string $profileType the type of profiles
820 * @param User|null $user the user requesting the list of profiles
821 * @return array|null the list of profiles or null if none available
823 public function getProfiles( $profileType, User
$user = null ) {
828 * Create a search field definition.
829 * Specific search engines should override this method to create search fields.
830 * @param string $name
831 * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
832 * @return SearchIndexField
835 public function makeSearchFieldMapping( $name, $type ) {
836 return new NullIndexField();
840 * Get fields for search index
842 * @return SearchIndexField[] Index field definitions for all content handlers
844 public function getSearchIndexFields() {
845 $models = ContentHandler
::getContentModels();
847 $seenHandlers = new SplObjectStorage();
848 foreach ( $models as $model ) {
850 $handler = ContentHandler
::getForModelID( $model );
852 catch ( MWUnknownContentModelException
$e ) {
853 // If we can find no handler, ignore it
856 // Several models can have the same handler, so avoid processing it repeatedly
857 if ( $seenHandlers->contains( $handler ) ) {
858 // We already did this one
861 $seenHandlers->attach( $handler );
862 $handlerFields = $handler->getFieldsForSearchIndex( $this );
863 foreach ( $handlerFields as $fieldName => $fieldData ) {
864 if ( empty( $fields[$fieldName] ) ) {
865 $fields[$fieldName] = $fieldData;
867 // TODO: do we allow some clashes with the same type or reject all of them?
868 $mergeDef = $fields[$fieldName]->merge( $fieldData );
870 throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" );
872 $fields[$fieldName] = $mergeDef;
876 // Hook to allow extensions to produce search mapping fields
877 Hooks
::run( 'SearchIndexFields', [ &$fields, $this ] );
882 * Augment search results with extra data.
884 * @param SearchResultSet $resultSet
886 public function augmentSearchResults( SearchResultSet
$resultSet ) {
889 Hooks
::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
890 if ( !$setAugmentors && !$rowAugmentors ) {
895 // Convert row augmentors to set augmentor
896 foreach ( $rowAugmentors as $name => $row ) {
897 if ( isset( $setAugmentors[$name] ) ) {
898 throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" );
900 $setAugmentors[$name] = new PerRowAugmentor( $row );
903 foreach ( $setAugmentors as $name => $augmentor ) {
904 $data = $augmentor->augmentAll( $resultSet );
906 $resultSet->setAugmentedData( $name, $data );
913 * Dummy class to be used when non-supported Database engine is present.
914 * @todo FIXME: Dummy class should probably try something at least mildly useful,
915 * such as a LIKE search through titles.
918 class SearchEngineDummy
extends SearchEngine
{