maybePaginate( function () use ( $term ) { return $this->doSearchText( $term ); } ); } /** * Perform a full text search query and return a result set. * * @param string $term Raw search term * @return ISearchResultSet|Status|null * @since 1.32 */ protected function doSearchText( $term ) { return null; } /** * Perform a title search in the article archive. * NOTE: these results still should be filtered by * matching against PageArchive, permissions checks etc * The results returned by this methods are only sugegstions and * may not end up being shown to the user. * * As of 1.32 overriding this function is deprecated. It will * be converted to final in 1.34. Override self::doSearchArchiveTitle(). * * @param string $term Raw search term * @return Status * @since 1.29 */ public function searchArchiveTitle( $term ) { return $this->doSearchArchiveTitle( $term ); } /** * Perform a title search in the article archive. * * @param string $term Raw search term * @return Status * @since 1.32 */ protected function doSearchArchiveTitle( $term ) { return Status::newGood( [] ); } /** * Perform a title-only search query and return a result set. * If title searches are not supported or disabled, return null. * STUB * * As of 1.32 overriding this function is deprecated. It will * be converted to final in 1.34. Override self::doSearchTitle(). * * @param string $term Raw search term * @return ISearchResultSet|null */ public function searchTitle( $term ) { return $this->maybePaginate( function () use ( $term ) { return $this->doSearchTitle( $term ); } ); } /** * Perform a title-only search query and return a result set. * * @param string $term Raw search term * @return ISearchResultSet|null * @since 1.32 */ protected function doSearchTitle( $term ) { return null; } /** * Performs an overfetch and shrink operation to determine if * the next page is available for search engines that do not * explicitly implement their own pagination. * * @param Closure $fn Takes no arguments * @return ISearchResultSet|Status|null Result of calling $fn */ private function maybePaginate( Closure $fn ) { if ( $this instanceof PaginatingSearchEngine ) { return $fn(); } $this->limit++; try { $resultSetOrStatus = $fn(); } finally { $this->limit--; } $resultSet = null; if ( $resultSetOrStatus instanceof ISearchResultSet ) { $resultSet = $resultSetOrStatus; } elseif ( $resultSetOrStatus instanceof Status && $resultSetOrStatus->getValue() instanceof ISearchResultSet ) { $resultSet = $resultSetOrStatus->getValue(); } if ( $resultSet ) { $resultSet->shrink( $this->limit ); } return $resultSetOrStatus; } /** * @since 1.18 * @param string $feature * @return bool */ public function supports( $feature ) { switch ( $feature ) { case 'search-update': return true; case 'title-suffix-filter': default: return false; } } /** * Way to pass custom data for engines * @since 1.18 * @param string $feature * @param mixed $data */ public function setFeatureData( $feature, $data ) { $this->features[$feature] = $data; } /** * Way to retrieve custom data set by setFeatureData * or by the engine itself. * @since 1.29 * @param string $feature feature name * @return mixed the feature value or null if unset */ public function getFeatureData( $feature ) { return $this->features[$feature] ?? null; } /** * When overridden in derived class, performs database-specific conversions * on text to be used for searching or updating search index. * Default implementation does nothing (simply returns $string). * * @param string $string String to process * @return string */ public function normalizeText( $string ) { // Some languages such as Chinese require word segmentation return MediaWikiServices::getInstance()->getContentLanguage()->segmentByWord( $string ); } /** * Get service class to finding near matches. * @param Config $config Configuration to use for the matcher. * @return SearchNearMatcher */ public function getNearMatcher( Config $config ) { return new SearchNearMatcher( $config, MediaWikiServices::getInstance()->getContentLanguage() ); } /** * Get near matcher for default SearchEngine. * @return SearchNearMatcher */ protected static function defaultNearMatcher() { $services = MediaWikiServices::getInstance(); $config = $services->getMainConfig(); return $services->newSearchEngine()->getNearMatcher( $config ); } /** * Get chars legal for search * @param int $type type of search chars (see self::CHARS_ALL * and self::CHARS_NO_SYNTAX). Defaults to CHARS_ALL * @return string */ public function legalSearchChars( $type = self::CHARS_ALL ) { return "A-Za-z_'.0-9\\x80-\\xFF\\-"; } /** * Set the maximum number of results to return * and how many to skip before returning the first. * * @param int $limit * @param int $offset */ function setLimitOffset( $limit, $offset = 0 ) { $this->limit = intval( $limit ); $this->offset = intval( $offset ); } /** * Set which namespaces the search should include. * Give an array of namespace index numbers. * * @param int[]|null $namespaces */ function setNamespaces( $namespaces ) { if ( $namespaces ) { // Filter namespaces to only keep valid ones $validNs = MediaWikiServices::getInstance()->getSearchEngineConfig()->searchableNamespaces(); $namespaces = array_filter( $namespaces, function ( $ns ) use( $validNs ) { return $ns < 0 || isset( $validNs[$ns] ); } ); } else { $namespaces = []; } $this->namespaces = $namespaces; } /** * Set whether the searcher should try to build a suggestion. Note: some searchers * don't support building a suggestion in the first place and others don't respect * this flag. * * @param bool $showSuggestion Should the searcher try to build suggestions */ function setShowSuggestion( $showSuggestion ) { $this->showSuggestion = $showSuggestion; } /** * Get the valid sort directions. All search engines support 'relevance' but others * might support more. The default in all implementations must be 'relevance.' * * @since 1.25 * @return string[] the valid sort directions for setSort */ public function getValidSorts() { return [ self::DEFAULT_SORT ]; } /** * Set the sort direction of the search results. Must be one returned by * SearchEngine::getValidSorts() * * @since 1.25 * @throws InvalidArgumentException * @param string $sort sort direction for query result */ public function setSort( $sort ) { if ( !in_array( $sort, $this->getValidSorts() ) ) { throw new InvalidArgumentException( "Invalid sort: $sort. " . "Must be one of: " . implode( ', ', $this->getValidSorts() ) ); } $this->sort = $sort; } /** * Get the sort direction of the search results * * @since 1.25 * @return string */ public function getSort() { return $this->sort; } /** * Parse some common prefixes: all (search everything) * or namespace names and set the list of namespaces * of this class accordingly. * * @deprecated since 1.32; should be handled internally by the search engine * @param string $query * @return string */ function replacePrefixes( $query ) { return $query; } /** * Parse some common prefixes: all (search everything) * or namespace names * * @param string $query * @param bool $withAllKeyword activate support of the "all:" keyword and its * translations to activate searching on all namespaces. * @param bool $withPrefixSearchExtractNamespaceHook call the PrefixSearchExtractNamespace hook * if classic namespace identification did not match. * @return false|array false if no namespace was extracted, an array * with the parsed query at index 0 and an array of namespaces at index * 1 (or null for all namespaces). * @throws FatalError * @throws MWException */ public static function parseNamespacePrefixes( $query, $withAllKeyword = true, $withPrefixSearchExtractNamespaceHook = false ) { $parsed = $query; if ( strpos( $query, ':' ) === false ) { // nothing to do return false; } $extractedNamespace = null; $allQuery = false; if ( $withAllKeyword ) { $allkeywords = []; $allkeywords[] = wfMessage( 'searchall' )->inContentLanguage()->text() . ":"; // force all: so that we have a common syntax for all the wikis if ( !in_array( 'all:', $allkeywords ) ) { $allkeywords[] = 'all:'; } foreach ( $allkeywords as $kw ) { if ( strncmp( $query, $kw, strlen( $kw ) ) == 0 ) { $extractedNamespace = null; $parsed = substr( $query, strlen( $kw ) ); $allQuery = true; break; } } } if ( !$allQuery && strpos( $query, ':' ) !== false ) { $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) ); $index = MediaWikiServices::getInstance()->getContentLanguage()->getNsIndex( $prefix ); if ( $index !== false ) { $extractedNamespace = [ $index ]; $parsed = substr( $query, strlen( $prefix ) + 1 ); } elseif ( $withPrefixSearchExtractNamespaceHook ) { $hookNamespaces = [ NS_MAIN ]; $hookQuery = $query; Hooks::run( 'PrefixSearchExtractNamespace', [ &$hookNamespaces, &$hookQuery ] ); if ( $hookQuery !== $query ) { $parsed = $hookQuery; $extractedNamespace = $hookNamespaces; } else { return false; } } else { return false; } } return [ $parsed, $extractedNamespace ]; } /** * Find snippet highlight settings for all users * @return array Contextlines, contextchars * @deprecated in 1.34 use the SearchHighlighter constants directly * @see SearchHighlighter::DEFAULT_CONTEXT_CHARS * @see SearchHighlighter::DEFAULT_CONTEXT_LINES */ public static function userHighlightPrefs() { $contextlines = SearchHighlighter::DEFAULT_CONTEXT_LINES; $contextchars = SearchHighlighter::DEFAULT_CONTEXT_CHARS; return [ $contextlines, $contextchars ]; } /** * Create or update the search index record for the given page. * Title and text should be pre-processed. * STUB * * @param int $id * @param string $title * @param string $text */ function update( $id, $title, $text ) { // no-op } /** * Update a search index record's title only. * Title should be pre-processed. * STUB * * @param int $id * @param string $title */ function updateTitle( $id, $title ) { // no-op } /** * Delete an indexed page * Title should be pre-processed. * STUB * * @param int $id Page id that was deleted * @param string $title Title of page that was deleted */ function delete( $id, $title ) { // no-op } /** * Get the raw text for updating the index from a content object * Nicer search backends could possibly do something cooler than * just returning raw text * * @todo This isn't ideal, we'd really like to have content-specific handling here * @param Title $t Title we're indexing * @param Content|null $c Content of the page to index * @return string * @deprecated since 1.34 use Content::getTextForSearchIndex directly */ public function getTextFromContent( Title $t, Content $c = null ) { return $c ? $c->getTextForSearchIndex() : ''; } /** * If an implementation of SearchEngine handles all of its own text processing * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s * rather silly handling, it should return true here instead. * * @return bool * @deprecated since 1.34 no longer needed since getTextFromContent is being deprecated */ public function textAlreadyUpdatedForIndex() { return false; } /** * Makes search simple string if it was namespaced. * Sets namespaces of the search to namespaces extracted from string. * @param string $search * @return string Simplified search string */ protected function normalizeNamespaces( $search ) { $queryAndNs = self::parseNamespacePrefixes( $search, false, true ); if ( $queryAndNs !== false ) { $this->setNamespaces( $queryAndNs[1] ); return $queryAndNs[0]; } return $search; } /** * Perform an overfetch of completion search results. This allows * determining if another page of results is available. * * @param string $search * @return SearchSuggestionSet */ protected function completionSearchBackendOverfetch( $search ) { $this->limit++; try { return $this->completionSearchBackend( $search ); } finally { $this->limit--; } } /** * Perform a completion search. * Does not resolve namespaces and does not check variants. * Search engine implementations may want to override this function. * @param string $search * @return SearchSuggestionSet */ protected function completionSearchBackend( $search ) { $results = []; $search = trim( $search ); if ( !in_array( NS_SPECIAL, $this->namespaces ) && // We do not run hook on Special: search !Hooks::run( 'PrefixSearchBackend', [ $this->namespaces, $search, $this->limit, &$results, $this->offset ] ) ) { // False means hook worked. // FIXME: Yes, the API is weird. That's why it is going to be deprecated. return SearchSuggestionSet::fromStrings( $results ); } else { // Hook did not do the job, use default simple search $results = $this->simplePrefixSearch( $search ); return SearchSuggestionSet::fromTitles( $results ); } } /** * Perform a completion search. * @param string $search * @return SearchSuggestionSet */ public function completionSearch( $search ) { if ( trim( $search ) === '' ) { return SearchSuggestionSet::emptySuggestionSet(); // Return empty result } $search = $this->normalizeNamespaces( $search ); $suggestions = $this->completionSearchBackendOverfetch( $search ); return $this->processCompletionResults( $search, $suggestions ); } /** * Perform a completion search with variants. * @param string $search * @return SearchSuggestionSet */ public function completionSearchWithVariants( $search ) { if ( trim( $search ) === '' ) { return SearchSuggestionSet::emptySuggestionSet(); // Return empty result } $search = $this->normalizeNamespaces( $search ); $results = $this->completionSearchBackendOverfetch( $search ); $fallbackLimit = 1 + $this->limit - $results->getSize(); if ( $fallbackLimit > 0 ) { $fallbackSearches = MediaWikiServices::getInstance()->getContentLanguage()-> autoConvertToAllVariants( $search ); $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] ); foreach ( $fallbackSearches as $fbs ) { $this->setLimitOffset( $fallbackLimit ); $fallbackSearchResult = $this->completionSearch( $fbs ); $results->appendAll( $fallbackSearchResult ); $fallbackLimit -= $fallbackSearchResult->getSize(); if ( $fallbackLimit <= 0 ) { break; } } } return $this->processCompletionResults( $search, $results ); } /** * Extract titles from completion results * @param SearchSuggestionSet $completionResults * @return Title[] */ public function extractTitles( SearchSuggestionSet $completionResults ) { return $completionResults->map( function ( SearchSuggestion $sugg ) { return $sugg->getSuggestedTitle(); } ); } /** * Process completion search results. * Resolves the titles and rescores. * @param string $search * @param SearchSuggestionSet $suggestions * @return SearchSuggestionSet */ protected function processCompletionResults( $search, SearchSuggestionSet $suggestions ) { // We over-fetched to determine pagination. Shrink back down if we have extra results // and mark if pagination is possible $suggestions->shrink( $this->limit ); $search = trim( $search ); // preload the titles with LinkBatch $lb = new LinkBatch( $suggestions->map( function ( SearchSuggestion $sugg ) { return $sugg->getSuggestedTitle(); } ) ); $lb->setCaller( __METHOD__ ); $lb->execute(); $diff = $suggestions->filter( function ( SearchSuggestion $sugg ) { return $sugg->getSuggestedTitle()->isKnown(); } ); if ( $diff > 0 ) { MediaWikiServices::getInstance()->getStatsdDataFactory() ->updateCount( 'search.completion.missing', $diff ); } $results = $suggestions->map( function ( SearchSuggestion $sugg ) { return $sugg->getSuggestedTitle()->getPrefixedText(); } ); if ( $this->offset === 0 ) { // Rescore results with an exact title match // NOTE: in some cases like cross-namespace redirects // (frequently used as shortcuts e.g. WP:WP on huwiki) some // backends like Cirrus will return no results. We should still // try an exact title match to workaround this limitation $rescorer = new SearchExactMatchRescorer(); $rescoredResults = $rescorer->rescore( $search, $this->namespaces, $results, $this->limit ); } else { // No need to rescore if offset is not 0 // The exact match must have been returned at position 0 // if it existed. $rescoredResults = $results; } if ( count( $rescoredResults ) > 0 ) { $found = array_search( $rescoredResults[0], $results ); if ( $found === false ) { // If the first result is not in the previous array it // means that we found a new exact match $exactMatch = SearchSuggestion::fromTitle( 0, Title::newFromText( $rescoredResults[0] ) ); $suggestions->prepend( $exactMatch ); $suggestions->shrink( $this->limit ); } else { // if the first result is not the same we need to rescore if ( $found > 0 ) { $suggestions->rescore( $found ); } } } return $suggestions; } /** * Simple prefix search for subpages. * @param string $search * @return Title[] */ public function defaultPrefixSearch( $search ) { if ( trim( $search ) === '' ) { return []; } $search = $this->normalizeNamespaces( $search ); return $this->simplePrefixSearch( $search ); } /** * Call out to simple search backend. * Defaults to TitlePrefixSearch. * @param string $search * @return Title[] */ protected function simplePrefixSearch( $search ) { // Use default database prefix search $backend = new TitlePrefixSearch; return $backend->defaultSearchBackend( $this->namespaces, $search, $this->limit, $this->offset ); } /** * Get a list of supported profiles. * Some search engine implementations may expose specific profiles to fine-tune * its behaviors. * The profile can be passed as a feature data with setFeatureData( $profileType, $profileName ) * The array returned by this function contains the following keys: * - name: the profile name to use with setFeatureData * - desc-message: the i18n description * - default: set to true if this profile is the default * * @since 1.28 * @param string $profileType the type of profiles * @param User|null $user the user requesting the list of profiles * @return array|null the list of profiles or null if none available * @phan-return null|array{name:string,desc-message:string,default?:bool} */ public function getProfiles( $profileType, User $user = null ) { return null; } /** * Create a search field definition. * Specific search engines should override this method to create search fields. * @param string $name * @param string $type One of the types in SearchIndexField::INDEX_TYPE_* * @return SearchIndexField * @since 1.28 */ public function makeSearchFieldMapping( $name, $type ) { return new NullIndexField(); } /** * Get fields for search index * @since 1.28 * @return SearchIndexField[] Index field definitions for all content handlers */ public function getSearchIndexFields() { $models = ContentHandler::getContentModels(); $fields = []; $seenHandlers = new SplObjectStorage(); foreach ( $models as $model ) { try { $handler = ContentHandler::getForModelID( $model ); } catch ( MWUnknownContentModelException $e ) { // If we can find no handler, ignore it continue; } // Several models can have the same handler, so avoid processing it repeatedly if ( $seenHandlers->contains( $handler ) ) { // We already did this one continue; } $seenHandlers->attach( $handler ); $handlerFields = $handler->getFieldsForSearchIndex( $this ); foreach ( $handlerFields as $fieldName => $fieldData ) { if ( empty( $fields[$fieldName] ) ) { $fields[$fieldName] = $fieldData; } else { // TODO: do we allow some clashes with the same type or reject all of them? $mergeDef = $fields[$fieldName]->merge( $fieldData ); if ( !$mergeDef ) { throw new InvalidArgumentException( "Duplicate field $fieldName for model $model" ); } $fields[$fieldName] = $mergeDef; } } } // Hook to allow extensions to produce search mapping fields Hooks::run( 'SearchIndexFields', [ &$fields, $this ] ); return $fields; } /** * Augment search results with extra data. * * @param ISearchResultSet $resultSet */ public function augmentSearchResults( ISearchResultSet $resultSet ) { $setAugmentors = []; $rowAugmentors = []; Hooks::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] ); if ( !$setAugmentors && !$rowAugmentors ) { // We're done here return; } // Convert row augmentors to set augmentor foreach ( $rowAugmentors as $name => $row ) { if ( isset( $setAugmentors[$name] ) ) { throw new InvalidArgumentException( "Both row and set augmentors are defined for $name" ); } $setAugmentors[$name] = new PerRowAugmentor( $row ); } /** * @var string $name * @var ResultSetAugmentor $augmentor */ foreach ( $setAugmentors as $name => $augmentor ) { $data = $augmentor->augmentAll( $resultSet ); if ( $data ) { $resultSet->setAugmentedData( $name, $data ); } } } }