/**
* Basic search engine
*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ * http://www.gnu.org/copyleft/gpl.html
+ *
* @file
* @ingroup Search
*/
* @defgroup Search Search
*/
+use MediaWiki\MediaWikiServices;
+
/**
* Contain a class for special pages
* @ingroup Search
*/
-class SearchEngine {
- var $limit = 10;
- var $offset = 0;
- var $prefix = '';
- var $searchTerms = array();
- var $namespaces = array( NS_MAIN );
- var $showRedirects = false;
-
- function __construct($db = null) {
- if ( $db ) {
- $this->db = $db;
- } else {
- $this->db = wfGetDB( DB_SLAVE );
- }
- }
+abstract class SearchEngine {
+ /** @var string */
+ public $prefix = '';
+
+ /** @var int[]|null */
+ public $namespaces = [ NS_MAIN ];
+
+ /** @var int */
+ protected $limit = 10;
+
+ /** @var int */
+ protected $offset = 0;
+
+ /** @var array|string */
+ protected $searchTerms = [];
+
+ /** @var bool */
+ protected $showSuggestion = true;
+ private $sort = 'relevance';
+
+ /** @var array Feature values */
+ protected $features = [];
+
+ /** @const string profile type for completionSearch */
+ const COMPLETION_PROFILE_TYPE = 'completionSearchProfile';
+
+ /** @const string profile type for query independent ranking features */
+ const FT_QUERY_INDEP_PROFILE_TYPE = 'fulltextQueryIndepProfile';
+
+ /** @const int flag for legalSearchChars: includes all chars allowed in a search query */
+ const CHARS_ALL = 1;
+
+ /** @const int flag for legalSearchChars: includes all chars allowed in a search term */
+ const CHARS_NO_SYNTAX = 2;
/**
* Perform a full text search query and return a result set.
- * If title searches are not supported or disabled, return null.
+ * If full text searches are not supported or disabled, return null.
* STUB
*
- * @param $term String: raw search term
- * @return SearchResultSet
+ * @param string $term Raw search term
+ * @return SearchResultSet|Status|null
*/
function searchText( $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.
+ *
+ * @param string $term Raw search term
+ * @return Status<Title[]>
+ * @since 1.29
+ */
+ function searchArchiveTitle( $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
*
- * @param $term String: raw search term
- * @return SearchResultSet
+ * @param string $term Raw search term
+ * @return SearchResultSet|null
*/
function searchTitle( $term ) {
return null;
}
- /** If this search backend can list/unlist redirects */
- function acceptListRedirects() {
- return true;
+ /**
+ * @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 ) {
+ if ( isset( $this->features[$feature] ) ) {
+ return $this->features[$feature];
+ }
+ return null;
}
/**
* on text to be used for searching or updating search index.
* Default implementation does nothing (simply returns $string).
*
- * @param $string string: String to process
+ * @param string $string String to process
* @return string
*/
public function normalizeText( $string ) {
}
/**
- * Transform search term in cases when parts of the query came as different GET params (when supported)
- * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
+ * Transform search term in cases when parts of the query came as different
+ * GET params (when supported), e.g. for prefix queries:
+ * search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
+ * @param string $term
+ * @return string
*/
- function transformSearchTerm( $term ) {
+ public function transformSearchTerm( $term ) {
return $term;
}
+ /**
+ * Get service class to finding near matches.
+ * @param Config $config Configuration to use for the matcher.
+ * @return SearchNearMatcher
+ */
+ public function getNearMatcher( Config $config ) {
+ global $wgContLang;
+ return new SearchNearMatcher( $config, $wgContLang );
+ }
+
+ /**
+ * Get near matcher for default SearchEngine.
+ * @return SearchNearMatcher
+ */
+ protected static function defaultNearMatcher() {
+ $config = MediaWikiServices::getInstance()->getMainConfig();
+ return MediaWikiServices::getInstance()->newSearchEngine()->getNearMatcher( $config );
+ }
+
/**
* If an exact title match can be found, or a very slightly close match,
* return the title. If no match, returns NULL.
- *
- * @param $searchterm String
+ * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
+ * @param string $searchterm
* @return Title
*/
public static function getNearMatch( $searchterm ) {
- $title = self::getNearMatchInternal( $searchterm );
-
- wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
- return $title;
+ return static::defaultNearMatcher()->getNearMatch( $searchterm );
}
-
+
/**
- * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
+ * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
* SearchResultSet.
- *
- * @param $searchterm string
+ * @deprecated since 1.27; Use SearchEngine::getNearMatcher()
+ * @param string $searchterm
* @return SearchResultSet
*/
public static function getNearMatchResultSet( $searchterm ) {
- return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
+ return static::defaultNearMatcher()->getNearMatchResultSet( $searchterm );
}
/**
- * Really find the title match.
+ * Get chars legal for search
+ * NOTE: usage as static is deprecated and preserved only as BC measure
+ * @param int $type type of search chars (see self::CHARS_ALL
+ * and self::CHARS_NO_SYNTAX). Defaults to CHARS_ALL
+ * @return string
*/
- private static function getNearMatchInternal( $searchterm ) {
- global $wgContLang;
-
- $allSearchTerms = array( $searchterm );
-
- if ( $wgContLang->hasVariants() ) {
- $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
- }
-
- $titleResult = null;
- if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
- return $titleResult;
- }
-
- foreach ( $allSearchTerms as $term ) {
-
- # Exact match? No need to look further.
- $title = Title::newFromText( $term );
- if ( is_null( $title ) )
- return null;
-
- if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal() || $title->exists() ) {
- return $title;
- }
-
- # See if it still otherwise has content is some sane sense
- $article = MediaWiki::articleFromTitle( $title );
- if ( $article->hasViewableContent() ) {
- return $title;
- }
-
- # Now try all lower case (i.e. first letter capitalized)
- #
- $title = Title::newFromText( $wgContLang->lc( $term ) );
- if ( $title && $title->exists() ) {
- return $title;
- }
-
- # Now try capitalized string
- #
- $title = Title::newFromText( $wgContLang->ucwords( $term ) );
- if ( $title && $title->exists() ) {
- return $title;
- }
-
- # Now try all upper case
- #
- $title = Title::newFromText( $wgContLang->uc( $term ) );
- if ( $title && $title->exists() ) {
- return $title;
- }
-
- # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
- $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
- if ( $title && $title->exists() ) {
- return $title;
- }
-
- // Give hooks a chance at better match variants
- $title = null;
- if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
- return $title;
- }
- }
-
- $title = Title::newFromText( $searchterm );
-
- # Entering an IP address goes to the contributions page
- if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
- || User::isIP( trim( $searchterm ) ) ) {
- return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
- }
-
-
- # Entering a user goes to the user page whether it's there or not
- if ( $title->getNamespace() == NS_USER ) {
- return $title;
- }
-
- # Go to images that exist even if there's no local page.
- # There may have been a funny upload, or it may be on a shared
- # file repository such as Wikimedia Commons.
- if ( $title->getNamespace() == NS_FILE ) {
- $image = wfFindFile( $title );
- if ( $image ) {
- return $title;
- }
- }
-
- # MediaWiki namespace? Page may be "implied" if not customized.
- # Just return it, with caps forced as the message system likes it.
- if ( $title->getNamespace() == NS_MEDIAWIKI ) {
- return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
- }
-
- # Quoted term? Try without the quotes...
- $matches = array();
- if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
- return SearchEngine::getNearMatch( $matches[1] );
- }
-
- return null;
- }
-
- public static function legalSearchChars() {
+ public static 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 $limit Integer
- * @param $offset Integer
+ * @param int $limit
+ * @param int $offset
*/
function setLimitOffset( $limit, $offset = 0 ) {
$this->limit = intval( $limit );
* Set which namespaces the search should include.
* Give an array of namespace index numbers.
*
- * @param $namespaces Array
+ * @param int[]|null $namespaces
*/
function setNamespaces( $namespaces ) {
+ if ( $namespaces ) {
+ // Filter namespaces to only keep valid ones
+ $validNs = $this->searchableNamespaces();
+ $namespaces = array_filter( $namespaces, function ( $ns ) use( $validNs ) {
+ return $ns < 0 || isset( $validNs[$ns] );
+ } );
+ } else {
+ $namespaces = [];
+ }
$this->namespaces = $namespaces;
}
/**
- * Parse some common prefixes: all (search everything)
- * or namespace names
+ * 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 $query String
+ * @param bool $showSuggestion Should the searcher try to build suggestions
*/
- function replacePrefixes( $query ) {
- global $wgContLang;
-
- $parsed = $query;
- if ( strpos( $query, ':' ) === false ) { // nothing to do
- wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
- return $parsed;
- }
-
- $allkeyword = wfMsgForContent( 'searchall' ) . ":";
- if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
- $this->namespaces = null;
- $parsed = substr( $query, strlen( $allkeyword ) );
- } else if ( strpos( $query, ':' ) !== false ) {
- $prefix = substr( $query, 0, strpos( $query, ':' ) );
- $index = $wgContLang->getNsIndex( $prefix );
- if ( $index !== false ) {
- $this->namespaces = array( $index );
- $parsed = substr( $query, strlen( $prefix ) + 1 );
- }
- }
- if ( trim( $parsed ) == '' )
- $parsed = $query; // prefix was the whole query
-
- wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
-
- return $parsed;
+ function setShowSuggestion( $showSuggestion ) {
+ $this->showSuggestion = $showSuggestion;
}
/**
- * Make a list of searchable namespaces and their canonical names.
- * @return Array
+ * Get the valid sort directions. All search engines support 'relevance' but others
+ * might support more. The default in all implementations should be 'relevance.'
+ *
+ * @since 1.25
+ * @return string[] the valid sort directions for setSort
*/
- public static function searchableNamespaces() {
- global $wgContLang;
- $arr = array();
- foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
- if ( $ns >= NS_MAIN ) {
- $arr[$ns] = $name;
- }
- }
-
- wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
- return $arr;
+ public function getValidSorts() {
+ return [ 'relevance' ];
}
/**
- * Extract default namespaces to search from the given user's
- * settings, returning a list of index numbers.
+ * Set the sort direction of the search results. Must be one returned by
+ * SearchEngine::getValidSorts()
*
- * @param $user User
- * @return Array
+ * @since 1.25
+ * @throws InvalidArgumentException
+ * @param string $sort sort direction for query result
*/
- public static function userNamespaces( $user ) {
- global $wgSearchEverythingOnlyLoggedIn;
-
- // get search everything preference, that can be set to be read for logged-in users
- $searcheverything = false;
- if ( ( $wgSearchEverythingOnlyLoggedIn && $user->isLoggedIn() )
- || !$wgSearchEverythingOnlyLoggedIn )
- $searcheverything = $user->getOption( 'searcheverything' );
-
- // searcheverything overrides other options
- if ( $searcheverything )
- return array_keys( SearchEngine::searchableNamespaces() );
-
- $arr = Preferences::loadOldSearchNs( $user );
- $searchableNamespaces = SearchEngine::searchableNamespaces();
-
- $arr = array_intersect( $arr, array_keys( $searchableNamespaces ) ); // Filter
-
- return $arr;
+ 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;
}
/**
- * Find snippet highlight settings for a given user
+ * Get the sort direction of the search results
*
- * @param $user User
- * @return Array contextlines, contextchars
+ * @since 1.25
+ * @return string
*/
- public static function userHighlightPrefs( &$user ) {
- // $contextlines = $user->getOption( 'contextlines', 5 );
- // $contextchars = $user->getOption( 'contextchars', 50 );
- $contextlines = 2; // Hardcode this. Old defaults sucked. :)
- $contextchars = 75; // same as above.... :P
- return array( $contextlines, $contextchars );
+ public function getSort() {
+ return $this->sort;
}
/**
- * An array of namespaces indexes to be searched by default
+ * Parse some common prefixes: all (search everything)
+ * or namespace names and set the list of namespaces
+ * of this class accordingly.
*
- * @return Array
+ * @param string $query
+ * @return string
*/
- public static function defaultNamespaces() {
- global $wgNamespacesToBeSearchedDefault;
-
- return array_keys( $wgNamespacesToBeSearchedDefault, true );
+ function replacePrefixes( $query ) {
+ $queryAndNs = self::parseNamespacePrefixes( $query );
+ if ( $queryAndNs === false ) {
+ return $query;
+ }
+ $this->namespaces = $queryAndNs[1];
+ return $queryAndNs[0];
}
/**
- * Get a list of namespace names useful for showing in tooltips
- * and preferences
+ * Parse some common prefixes: all (search everything)
+ * or namespace names
*
- * @param $namespaces Array
+ * @param string $query
+ * @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).
*/
- public static function namespacesAsText( $namespaces ) {
+ public static function parseNamespacePrefixes( $query ) {
global $wgContLang;
- $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
- foreach ( $formatted as $key => $ns ) {
- if ( empty( $ns ) )
- $formatted[$key] = wfMsg( 'blanknamespace' );
+ $parsed = $query;
+ if ( strpos( $query, ':' ) === false ) { // nothing to do
+ return false;
}
- return $formatted;
- }
+ $extractedNamespace = null;
- /**
- * Return the help namespaces to be shown on Special:Search
- *
- * @return Array
- */
- public static function helpNamespaces() {
- global $wgNamespacesToBeSearchedHelp;
+ $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
+ if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
+ $extractedNamespace = null;
+ $parsed = substr( $query, strlen( $allkeyword ) );
+ } elseif ( strpos( $query, ':' ) !== false ) {
+ // TODO: should we unify with PrefixSearch::extractNamespace ?
+ $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
+ $index = $wgContLang->getNsIndex( $prefix );
+ if ( $index !== false ) {
+ $extractedNamespace = [ $index ];
+ $parsed = substr( $query, strlen( $prefix ) + 1 );
+ } else {
+ return false;
+ }
+ }
- return array_keys( $wgNamespacesToBeSearchedHelp, true );
- }
+ if ( trim( $parsed ) == '' ) {
+ $parsed = $query; // prefix was the whole query
+ }
- /**
- * Return a 'cleaned up' search string
- *
- * @param $text String
- * @return String
- */
- function filter( $text ) {
- $lc = $this->legalSearchChars();
- return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
+ return [ $parsed, $extractedNamespace ];
}
+
/**
- * Load up the appropriate search engine class for the currently
- * active database backend, and return a configured instance.
- *
- * @return SearchEngine
+ * Find snippet highlight settings for all users
+ * @return array Contextlines, contextchars
*/
- public static function create() {
- global $wgSearchType;
- $dbr = null;
- if ( $wgSearchType ) {
- $class = $wgSearchType;
- } else {
- $dbr = wfGetDB( DB_SLAVE );
- $class = $dbr->getSearchEngine();
- }
- $search = new $class( $dbr );
- $search->setLimitOffset( 0, 0 );
- return $search;
+ public static function userHighlightPrefs() {
+ $contextlines = 2; // Hardcode this. Old defaults sucked. :)
+ $contextchars = 75; // same as above.... :P
+ return [ $contextlines, $contextchars ];
}
/**
* Title and text should be pre-processed.
* STUB
*
- * @param $id Integer
- * @param $title String
- * @param $text String
+ * @param int $id
+ * @param string $title
+ * @param string $text
*/
function update( $id, $title, $text ) {
// no-op
* Title should be pre-processed.
* STUB
*
- * @param $id Integer
- * @param $title String
+ * @param int $id
+ * @param string $title
*/
function updateTitle( $id, $title ) {
// no-op
}
/**
- * Get OpenSearch suggestion template
- *
- * @return String
- */
- public static function getOpenSearchTemplate() {
- global $wgOpenSearchTemplate, $wgServer;
- if ( $wgOpenSearchTemplate ) {
- return $wgOpenSearchTemplate;
- } else {
- $ns = implode( '|', SearchEngine::defaultNamespaces() );
- if ( !$ns ) $ns = "0";
- return $wgServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
- }
- }
-
- /**
- * Get internal MediaWiki Suggest template
- *
- * @return String
- */
- public static function getMWSuggestTemplate() {
- global $wgMWSuggestTemplate, $wgServer;
- if ( $wgMWSuggestTemplate )
- return $wgMWSuggestTemplate;
- else
- return $wgServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
- }
-}
-
-/**
- * @ingroup Search
- */
-class SearchResultSet {
- /**
- * Fetch an array of regular expression fragments for matching
- * the search terms as parsed by this engine in a text extract.
+ * Delete an indexed page
+ * Title should be pre-processed.
* STUB
*
- * @return Array
+ * @param int $id Page id that was deleted
+ * @param string $title Title of page that was deleted
*/
- function termMatches() {
- return array();
- }
-
- function numRows() {
- return 0;
+ function delete( $id, $title ) {
+ // no-op
}
/**
- * Return true if results are included in this result set.
- * STUB
+ * Get OpenSearch suggestion template
*
- * @return Boolean
+ * @deprecated since 1.25
+ * @return string
*/
- function hasResults() {
- return false;
+ public static function getOpenSearchTemplate() {
+ wfDeprecated( __METHOD__, '1.25' );
+ return ApiOpenSearch::getOpenSearchTemplate( 'application/x-suggestions+json' );
}
/**
- * Some search modes return a total hit count for the query
- * in the entire article database. This may include pages
- * in namespaces that would not be matched on the given
- * settings.
+ * 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
*
- * Return null if no total hits number is supported.
- *
- * @return Integer
+ * @todo This isn't ideal, we'd really like to have content-specific handling here
+ * @param Title $t Title we're indexing
+ * @param Content $c Content of the page to index
+ * @return string
*/
- function getTotalHits() {
- return null;
+ public function getTextFromContent( Title $t, Content $c = null ) {
+ return $c ? $c->getTextForSearchIndex() : '';
}
/**
- * Some search modes return a suggested alternate term if there are
- * no exact hits. Returns true if there is one on this set.
+ * 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 Boolean
+ * @return bool
*/
- function hasSuggestion() {
+ public function textAlreadyUpdatedForIndex() {
return false;
}
/**
- * @return String: suggested query, null if none
+ * 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
*/
- function getSuggestionQuery() {
- return null;
- }
-
- /**
- * @return String: HTML highlighted suggested query, '' if none
- */
- function getSuggestionSnippet() {
- return '';
- }
-
- /**
- * Return information about how and from where the results were fetched,
- * should be useful for diagnostics and debugging
- *
- * @return String
- */
- function getInfo() {
- return null;
- }
-
- /**
- * Return a result set of hits on other (multiple) wikis associated with this one
- *
- * @return SearchResultSet
- */
- function getInterwikiResults() {
- return null;
- }
+ protected function normalizeNamespaces( $search ) {
+ // Find a Title which is not an interwiki and is in NS_MAIN
+ $title = Title::newFromText( $search );
+ $ns = $this->namespaces;
+ if ( $title && !$title->isExternal() ) {
+ $ns = [ $title->getNamespace() ];
+ $search = $title->getText();
+ if ( $ns[0] == NS_MAIN ) {
+ $ns = $this->namespaces; // no explicit prefix, use default namespaces
+ Hooks::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
+ }
+ } else {
+ $title = Title::newFromText( $search . 'Dummy' );
+ if ( $title && $title->getText() == 'Dummy'
+ && $title->getNamespace() != NS_MAIN
+ && !$title->isExternal()
+ ) {
+ $ns = [ $title->getNamespace() ];
+ $search = '';
+ } else {
+ Hooks::run( 'PrefixSearchExtractNamespace', [ &$ns, &$search ] );
+ }
+ }
- /**
- * Check if there are results on other wikis
- *
- * @return Boolean
- */
- function hasInterwikiResults() {
- return $this->getInterwikiResults() != null;
- }
+ $ns = array_map( function ( $space ) {
+ return $space == NS_MEDIA ? NS_FILE : $space;
+ }, $ns );
- /**
- * Fetches next search result, or false.
- * STUB
- *
- * @return SearchResult
- */
- function next() {
- return false;
+ $this->setNamespaces( $ns );
+ return $search;
}
/**
- * Frees the result set, if applicable.
+ * 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
*/
- function free() {
- // ...
- }
-}
-
-/**
- * This class is used for different SQL-based search engines shipped with MediaWiki
- */
-class SqlSearchResultSet extends SearchResultSet {
- function __construct( $resultSet, $terms ) {
- $this->mResultSet = $resultSet;
- $this->mTerms = $terms;
- }
-
- function termMatches() {
- return $this->mTerms;
- }
+ protected function completionSearchBackend( $search ) {
+ $results = [];
- function numRows() {
- if ( $this->mResultSet === false )
- return false;
-
- return $this->mResultSet->numRows();
- }
+ $search = trim( $search );
- function next() {
- if ( $this->mResultSet === false )
- return false;
-
- $row = $this->mResultSet->fetchObject();
- if ( $row === false )
- return false;
-
- return SearchResult::newFromRow( $row );
- }
+ 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.
- function free() {
- if ( $this->mResultSet === false )
- return false;
-
- $this->mResultSet->free();
+ return SearchSuggestionSet::fromStrings( $results );
+ } else {
+ // Hook did not do the job, use default simple search
+ $results = $this->simplePrefixSearch( $search );
+ return SearchSuggestionSet::fromTitles( $results );
+ }
}
-}
-
-/**
- * @ingroup Search
- */
-class SearchResultTooMany {
- # # Some search engines may bail out if too many matches are found
-}
-
-
-/**
- * @todo Fixme: This class is horribly factored. It would probably be better to
- * have a useful base class to which you pass some standard information, then
- * let the fancy self-highlighters extend that.
- * @ingroup Search
- */
-class SearchResult {
- var $mRevision = null;
- var $mImage = null;
/**
- * Return a new SearchResult and initializes it with a title.
- *
- * @param $title Title
- * @return SearchResult
+ * Perform a completion search.
+ * @param string $search
+ * @return SearchSuggestionSet
*/
- public static function newFromTitle( $title ) {
- $result = new self();
- $result->initFromTitle( $title );
- return $result;
- }
- /**
- * Return a new SearchResult and initializes it with a row.
- *
- * @param $row object
- * @return SearchResult
- */
- public static function newFromRow( $row ) {
- $result = new self();
- $result->initFromRow( $row );
- return $result;
- }
-
- public function __construct( $row = null ) {
- if ( !is_null( $row ) ) {
- // Backwards compatibility with pre-1.17 callers
- $this->initFromRow( $row );
+ public function completionSearch( $search ) {
+ if ( trim( $search ) === '' ) {
+ return SearchSuggestionSet::emptySuggestionSet(); // Return empty result
}
+ $search = $this->normalizeNamespaces( $search );
+ return $this->processCompletionResults( $search, $this->completionSearchBackend( $search ) );
}
-
- /**
- * Initialize from a database row. Makes a Title and passes that to
- * initFromTitle.
- *
- * @param $row object
- */
- protected function initFromRow( $row ) {
- $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
- }
-
+
/**
- * Initialize from a Title and if possible initializes a corresponding
- * Revision and File.
- *
- * @param $title Title
+ * Perform a completion search with variants.
+ * @param string $search
+ * @return SearchSuggestionSet
*/
- protected function initFromTitle( $title ) {
- $this->mTitle = $title;
- if ( !is_null( $this->mTitle ) ) {
- $this->mRevision = Revision::newFromTitle( $this->mTitle );
- if ( $this->mTitle->getNamespace() === NS_FILE )
- $this->mImage = wfFindFile( $this->mTitle );
+ public function completionSearchWithVariants( $search ) {
+ if ( trim( $search ) === '' ) {
+ return SearchSuggestionSet::emptySuggestionSet(); // Return empty result
}
- }
+ $search = $this->normalizeNamespaces( $search );
- /**
- * Check if this is result points to an invalid title
- *
- * @return Boolean
- */
- function isBrokenTitle() {
- if ( is_null( $this->mTitle ) )
- return true;
- return false;
- }
+ $results = $this->completionSearchBackend( $search );
+ $fallbackLimit = $this->limit - $results->getSize();
+ if ( $fallbackLimit > 0 ) {
+ global $wgContLang;
- /**
- * Check if target page is missing, happens when index is out of date
- *
- * @return Boolean
- */
- function isMissingRevision() {
- return !$this->mRevision && !$this->mImage;
- }
+ $fallbackSearches = $wgContLang->autoConvertToAllVariants( $search );
+ $fallbackSearches = array_diff( array_unique( $fallbackSearches ), [ $search ] );
- /**
- * @return Title
- */
- function getTitle() {
- return $this->mTitle;
+ foreach ( $fallbackSearches as $fbs ) {
+ $this->setLimitOffset( $fallbackLimit );
+ $fallbackSearchResult = $this->completionSearch( $fbs );
+ $results->appendAll( $fallbackSearchResult );
+ $fallbackLimit -= count( $fallbackSearchResult );
+ if ( $fallbackLimit <= 0 ) {
+ break;
+ }
+ }
+ }
+ return $this->processCompletionResults( $search, $results );
}
/**
- * @return Double or null if not supported
+ * Extract titles from completion results
+ * @param SearchSuggestionSet $completionResults
+ * @return Title[]
*/
- function getScore() {
- return null;
+ public function extractTitles( SearchSuggestionSet $completionResults ) {
+ return $completionResults->map( function ( SearchSuggestion $sugg ) {
+ return $sugg->getSuggestedTitle();
+ } );
}
/**
- * Lazy initialization of article text from DB
+ * Process completion search results.
+ * Resolves the titles and rescores.
+ * @param string $search
+ * @param SearchSuggestionSet $suggestions
+ * @return SearchSuggestionSet
*/
- protected function initText() {
- if ( !isset( $this->mText ) ) {
- if ( $this->mRevision != null )
- $this->mText = $this->mRevision->getText();
- else // TODO: can we fetch raw wikitext for commons images?
- $this->mText = '';
+ protected function processCompletionResults( $search, SearchSuggestionSet $suggestions ) {
+ $search = trim( $search );
+ // preload the titles with LinkBatch
+ $titles = $suggestions->map( function ( SearchSuggestion $sugg ) {
+ return $sugg->getSuggestedTitle();
+ } );
+ $lb = new LinkBatch( $titles );
+ $lb->setCaller( __METHOD__ );
+ $lb->execute();
+ $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 );
+ }
+ }
}
- }
- /**
- * @param $terms Array: terms to highlight
- * @return String: highlighted text snippet, null (and not '') if not supported
- */
- function getTextSnippet( $terms ) {
- global $wgUser, $wgAdvancedSearchHighlighting;
- $this->initText();
- list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
- $h = new SearchHighlighter();
- if ( $wgAdvancedSearchHighlighting )
- return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
- else
- return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
+ return $suggestions;
}
/**
- * @param $terms Array: terms to highlight
- * @return String: highlighted title, '' if not supported
+ * Simple prefix search for subpages.
+ * @param string $search
+ * @return Title[]
*/
- function getTitleSnippet( $terms ) {
- return '';
- }
+ public function defaultPrefixSearch( $search ) {
+ if ( trim( $search ) === '' ) {
+ return [];
+ }
- /**
- * @param $terms Array: terms to highlight
- * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
- */
- function getRedirectSnippet( $terms ) {
- return '';
+ $search = $this->normalizeNamespaces( $search );
+ return $this->simplePrefixSearch( $search );
}
/**
- * @return Title object for the redirect to this page, null if none or not supported
+ * Call out to simple search backend.
+ * Defaults to TitlePrefixSearch.
+ * @param string $search
+ * @return Title[]
*/
- function getRedirectTitle() {
- return null;
+ protected function simplePrefixSearch( $search ) {
+ // Use default database prefix search
+ $backend = new TitlePrefixSearch;
+ return $backend->defaultSearchBackend( $this->namespaces, $search, $this->limit, $this->offset );
}
/**
- * @return string highlighted relevant section name, null if none or not supported
+ * Make a list of searchable namespaces and their canonical names.
+ * @deprecated since 1.27; use SearchEngineConfig::searchableNamespaces()
+ * @return array
*/
- function getSectionSnippet() {
- return '';
+ public static function searchableNamespaces() {
+ return MediaWikiServices::getInstance()->getSearchEngineConfig()->searchableNamespaces();
}
/**
- * @return Title object (pagename+fragment) for the section, null if none or not supported
+ * Extract default namespaces to search from the given user's
+ * settings, returning a list of index numbers.
+ * @deprecated since 1.27; use SearchEngineConfig::userNamespaces()
+ * @param user $user
+ * @return array
*/
- function getSectionTitle() {
- return null;
+ public static function userNamespaces( $user ) {
+ return MediaWikiServices::getInstance()->getSearchEngineConfig()->userNamespaces( $user );
}
/**
- * @return String: timestamp
+ * An array of namespaces indexes to be searched by default
+ * @deprecated since 1.27; use SearchEngineConfig::defaultNamespaces()
+ * @return array
*/
- function getTimestamp() {
- if ( $this->mRevision )
- return $this->mRevision->getTimestamp();
- else if ( $this->mImage )
- return $this->mImage->getTimestamp();
- return '';
+ public static function defaultNamespaces() {
+ return MediaWikiServices::getInstance()->getSearchEngineConfig()->defaultNamespaces();
}
/**
- * @return Integer: number of words
+ * Get a list of namespace names useful for showing in tooltips
+ * and preferences
+ * @deprecated since 1.27; use SearchEngineConfig::namespacesAsText()
+ * @param array $namespaces
+ * @return array
*/
- function getWordCount() {
- $this->initText();
- return str_word_count( $this->mText );
+ public static function namespacesAsText( $namespaces ) {
+ return MediaWikiServices::getInstance()->getSearchEngineConfig()->namespacesAsText( $namespaces );
}
/**
- * @return Integer: size in bytes
+ * Load up the appropriate search engine class for the currently
+ * active database backend, and return a configured instance.
+ * @deprecated since 1.27; Use SearchEngineFactory::create
+ * @param string $type Type of search backend, if not the default
+ * @return SearchEngine
*/
- function getByteSize() {
- $this->initText();
- return strlen( $this->mText );
+ public static function create( $type = null ) {
+ return MediaWikiServices::getInstance()->getSearchEngineFactory()->create( $type );
}
/**
- * @return Boolean if hit has related articles
+ * Return the search engines we support. If only $wgSearchType
+ * is set, it'll be an array of just that one item.
+ * @deprecated since 1.27; use SearchEngineConfig::getSearchTypes()
+ * @return array
*/
- function hasRelated() {
- return false;
+ public static function getSearchTypes() {
+ return MediaWikiServices::getInstance()->getSearchEngineConfig()->getSearchTypes();
}
/**
- * @return String: interwiki prefix of the title (return iw even if title is broken)
+ * 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
*/
- function getInterwikiPrefix() {
- return '';
+ public function getProfiles( $profileType, User $user = null ) {
+ return null;
}
-}
-/**
- * A SearchResultSet wrapper for SearchEngine::getNearMatch
- */
-class SearchNearMatchResultSet extends SearchResultSet {
- private $fetched = false;
+
/**
- * @param $match mixed Title if matched, else null
+ * Create a search field definition.
+ * Specific search engines should override this method to create search fields.
+ * @param string $name
+ * @param int $type One of the types in SearchIndexField::INDEX_TYPE_*
+ * @return SearchIndexField
+ * @since 1.28
*/
- public function __construct( $match ) {
- $this->result = $match;
- }
- public function hasResult() {
- return (bool)$this->result;
- }
- public function numRows() {
- return $this->hasResults() ? 1 : 0;
- }
- public function next() {
- if ( $this->fetched || !$this->result ) {
- return false;
- }
- $this->fetched = true;
- return SearchResult::newFromTitle( $this->result );
- }
-}
-
-/**
- * Highlight bits of wikitext
- *
- * @ingroup Search
- */
-class SearchHighlighter {
- var $mCleanWikitext = true;
-
- function __construct( $cleanupWikitext = true ) {
- $this->mCleanWikitext = $cleanupWikitext;
+ public function makeSearchFieldMapping( $name, $type ) {
+ return new NullIndexField();
}
/**
- * Default implementation of wikitext highlighting
- *
- * @param $text String
- * @param $terms Array: terms to highlight (unescaped)
- * @param $contextlines Integer
- * @param $contextchars Integer
- * @return String
+ * Get fields for search index
+ * @since 1.28
+ * @return SearchIndexField[] Index field definitions for all content handlers
*/
- public function highlightText( $text, $terms, $contextlines, $contextchars ) {
- global $wgContLang;
- global $wgSearchHighlightBoundaries;
- $fname = __METHOD__;
-
- if ( $text == '' )
- return '';
-
- // spli text into text + templates/links/tables
- $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
- // first capture group is for detecting nested templates/links/tables/references
- $endPatterns = array(
- 1 => '/(\{\{)|(\}\})/', // template
- 2 => '/(\[\[)|(\]\])/', // image
- 3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
-
- // FIXME: this should prolly be a hook or something
- if ( function_exists( 'wfCite' ) ) {
- $spat .= '|(<ref>)'; // references via cite extension
- $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
- }
- $spat .= '/';
- $textExt = array(); // text extracts
- $otherExt = array(); // other extracts
- wfProfileIn( "$fname-split" );
- $start = 0;
- $textLen = strlen( $text );
- $count = 0; // sequence number to maintain ordering
- while ( $start < $textLen ) {
- // find start of template/image/table
- if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
- $epat = '';
- foreach ( $matches as $key => $val ) {
- if ( $key > 0 && $val[1] != - 1 ) {
- if ( $key == 2 ) {
- // see if this is an image link
- $ns = substr( $val[0], 2, - 1 );
- if ( $wgContLang->getNsIndex( $ns ) != NS_FILE )
- break;
-
- }
- $epat = $endPatterns[$key];
- $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
- $start = $val[1];
- break;
- }
- }
- if ( $epat ) {
- // find end (and detect any nested elements)
- $level = 0;
- $offset = $start + 1;
- $found = false;
- while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
- if ( array_key_exists( 2, $endMatches ) ) {
- // found end
- if ( $level == 0 ) {
- $len = strlen( $endMatches[2][0] );
- $off = $endMatches[2][1];
- $this->splitAndAdd( $otherExt, $count,
- substr( $text, $start, $off + $len - $start ) );
- $start = $off + $len;
- $found = true;
- break;
- } else {
- // end of nested element
- $level -= 1;
- }
- } else {
- // nested
- $level += 1;
- }
- $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
- }
- if ( ! $found ) {
- // couldn't find appropriate closing tag, skip
- $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
- $start += strlen( $matches[0][0] );
- }
- continue;
- }
- }
- // else: add as text extract
- $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
- break;
- }
-
- $all = $textExt + $otherExt; // these have disjunct key sets
-
- wfProfileOut( "$fname-split" );
-
- // prepare regexps
- foreach ( $terms as $index => $term ) {
- // manually do upper/lowercase stuff for utf-8 since PHP won't do it
- if ( preg_match( '/[\x80-\xff]/', $term ) ) {
- $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
- } else {
- $terms[$index] = $term;
+ public function getSearchIndexFields() {
+ $models = ContentHandler::getContentModels();
+ $fields = [];
+ $seenHandlers = new SplObjectStorage();
+ foreach ( $models as $model ) {
+ try {
+ $handler = ContentHandler::getForModelID( $model );
}
- }
- $anyterm = implode( '|', $terms );
- $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
-
- // FIXME: a hack to scale contextchars, a correct solution
- // would be to have contextchars actually be char and not byte
- // length, and do proper utf-8 substrings and lengths everywhere,
- // but PHP is making that very hard and unclean to implement :(
- $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
- $contextchars = intval( $contextchars * $scale );
-
- $patPre = "(^|$wgSearchHighlightBoundaries)";
- $patPost = "($wgSearchHighlightBoundaries|$)";
-
- $pat1 = "/(" . $phrase . ")/ui";
- $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
-
- wfProfileIn( "$fname-extract" );
-
- $left = $contextlines;
-
- $snippets = array();
- $offsets = array();
-
- // show beginning only if it contains all words
- $first = 0;
- $firstText = '';
- foreach ( $textExt as $index => $line ) {
- if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
- $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
- $first = $index;
- break;
- }
- }
- if ( $firstText ) {
- $succ = true;
- // check if first text contains all terms
- foreach ( $terms as $term ) {
- if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
- $succ = false;
- break;
- }
- }
- if ( $succ ) {
- $snippets[$first] = $firstText;
- $offsets[$first] = 0;
- }
- }
- if ( ! $snippets ) {
- // match whole query on text
- $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
- // match whole query on templates/tables/images
- $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
- // match any words on text
- $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
- // match any words on templates/tables/images
- $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
-
- ksort( $snippets );
- }
-
- // add extra chars to each snippet to make snippets constant size
- $extended = array();
- if ( count( $snippets ) == 0 ) {
- // couldn't find the target words, just show beginning of article
- if ( array_key_exists( $first, $all ) ) {
- $targetchars = $contextchars * $contextlines;
- $snippets[$first] = '';
- $offsets[$first] = 0;
+ catch ( MWUnknownContentModelException $e ) {
+ // If we can find no handler, ignore it
+ continue;
}
- } else {
- // if begin of the article contains the whole phrase, show only that !!
- if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
- && $offsets[$first] < $contextchars * 2 ) {
- $snippets = array ( $first => $snippets[$first] );
+ // Several models can have the same handler, so avoid processing it repeatedly
+ if ( $seenHandlers->contains( $handler ) ) {
+ // We already did this one
+ continue;
}
-
- // calc by how much to extend existing snippets
- $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
- }
-
- foreach ( $snippets as $index => $line ) {
- $extended[$index] = $line;
- $len = strlen( $line );
- if ( $len < $targetchars - 20 ) {
- // complete this line
- if ( $len < strlen( $all[$index] ) ) {
- $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
- $len = strlen( $extended[$index] );
- }
-
- // add more lines
- $add = $index + 1;
- while ( $len < $targetchars - 20
- && array_key_exists( $add, $all )
- && !array_key_exists( $add, $snippets ) ) {
- $offsets[$add] = 0;
- $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
- $extended[$add] = $tt;
- $len += strlen( $tt );
- $add++;
+ $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;
}
}
}
-
- // $snippets = array_map('htmlspecialchars', $extended);
- $snippets = $extended;
- $last = - 1;
- $extract = '';
- foreach ( $snippets as $index => $line ) {
- if ( $last == - 1 )
- $extract .= $line; // first line
- elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) )
- $extract .= " " . $line; // continous lines
- else
- $extract .= '<b> ... </b>' . $line;
-
- $last = $index;
- }
- if ( $extract )
- $extract .= '<b> ... </b>';
-
- $processed = array();
- foreach ( $terms as $term ) {
- if ( ! isset( $processed[$term] ) ) {
- $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
- $extract = preg_replace( $pat3,
- "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
- $processed[$term] = true;
- }
- }
-
- wfProfileOut( "$fname-extract" );
-
- return $extract;
+ // Hook to allow extensions to produce search mapping fields
+ Hooks::run( 'SearchIndexFields', [ &$fields, $this ] );
+ return $fields;
}
/**
- * Split text into lines and add it to extracts array
+ * Augment search results with extra data.
*
- * @param $extracts Array: index -> $line
- * @param $count Integer
- * @param $text String
+ * @param SearchResultSet $resultSet
*/
- function splitAndAdd( &$extracts, &$count, $text ) {
- $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
- foreach ( $split as $line ) {
- $tt = trim( $line );
- if ( $tt )
- $extracts[$count++] = $tt;
- }
- }
+ public function augmentSearchResults( SearchResultSet $resultSet ) {
+ $setAugmentors = [];
+ $rowAugmentors = [];
+ Hooks::run( "SearchResultsAugment", [ &$setAugmentors, &$rowAugmentors ] );
- /**
- * Do manual case conversion for non-ascii chars
- *
- * @param $matches Array
- */
- function caseCallback( $matches ) {
- global $wgContLang;
- if ( strlen( $matches[0] ) > 1 ) {
- return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
- } else
- return $matches[0];
- }
-
- /**
- * Extract part of the text from start to end, but by
- * not chopping up words
- * @param $text String
- * @param $start Integer
- * @param $end Integer
- * @param $posStart Integer: (out) actual start position
- * @param $posEnd Integer: (out) actual end position
- * @return String
- */
- function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
- if ( $start != 0 )
- $start = $this->position( $text, $start, 1 );
- if ( $end >= strlen( $text ) )
- $end = strlen( $text );
- else
- $end = $this->position( $text, $end );
-
- if ( !is_null( $posStart ) )
- $posStart = $start;
- if ( !is_null( $posEnd ) )
- $posEnd = $end;
-
- if ( $end > $start )
- return substr( $text, $start, $end - $start );
- else
- return '';
- }
+ if ( !$setAugmentors && !$rowAugmentors ) {
+ // We're done here
+ return;
+ }
- /**
- * Find a nonletter near a point (index) in the text
- *
- * @param $text String
- * @param $point Integer
- * @param $offset Integer: offset to found index
- * @return Integer: nearest nonletter index, or beginning of utf8 char if none
- */
- function position( $text, $point, $offset = 0 ) {
- $tolerance = 10;
- $s = max( 0, $point - $tolerance );
- $l = min( strlen( $text ), $point + $tolerance ) - $s;
- $m = array();
- if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
- return $m[0][1] + $s + $offset;
- } else {
- // check if point is on a valid first UTF8 char
- $char = ord( $text[$point] );
- while ( $char >= 0x80 && $char < 0xc0 ) {
- // skip trailing bytes
- $point++;
- if ( $point >= strlen( $text ) )
- return strlen( $text );
- $char = ord( $text[$point] );
+ // 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" );
}
- return $point;
-
+ $setAugmentors[$name] = new PerRowAugmentor( $row );
}
- }
- /**
- * Search extracts for a pattern, and return snippets
- *
- * @param $pattern String: regexp for matching lines
- * @param $extracts Array: extracts to search
- * @param $linesleft Integer: number of extracts to make
- * @param $contextchars Integer: length of snippet
- * @param $out Array: map for highlighted snippets
- * @param $offsets Array: map of starting points of snippets
- * @protected
- */
- function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
- if ( $linesleft == 0 )
- return; // nothing to do
- foreach ( $extracts as $index => $line ) {
- if ( array_key_exists( $index, $out ) )
- continue; // this line already highlighted
-
- $m = array();
- if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
- continue;
-
- $offset = $m[0][1];
- $len = strlen( $m[0][0] );
- if ( $offset + $len < $contextchars )
- $begin = 0;
- elseif ( $len > $contextchars )
- $begin = $offset;
- else
- $begin = $offset + intval( ( $len - $contextchars ) / 2 );
-
- $end = $begin + $contextchars;
-
- $posBegin = $begin;
- // basic snippet from this line
- $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
- $offsets[$index] = $posBegin;
- $linesleft--;
- if ( $linesleft == 0 )
- return;
+ foreach ( $setAugmentors as $name => $augmentor ) {
+ $data = $augmentor->augmentAll( $resultSet );
+ if ( $data ) {
+ $resultSet->setAugmentedData( $name, $data );
+ }
}
}
-
- /**
- * Basic wikitext removal
- * @protected
- */
- function removeWiki( $text ) {
- $fname = __METHOD__;
- wfProfileIn( $fname );
-
- // $text = preg_replace("/'{2,5}/", "", $text);
- // $text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
- // $text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
- // $text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
- // $text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
- // $text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
- $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
- $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
- $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
- $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
- // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
- $text = preg_replace( "/<\/?[^>]+>/", "", $text );
- $text = preg_replace( "/'''''/", "", $text );
- $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
- $text = preg_replace( "/''/", "", $text );
-
- wfProfileOut( $fname );
- return $text;
- }
-
- /**
- * callback to replace [[target|caption]] kind of links, if
- * the target is category or image, leave it
- *
- * @param $matches Array
- */
- function linkReplace( $matches ) {
- $colon = strpos( $matches[1], ':' );
- if ( $colon === false )
- return $matches[2]; // replace with caption
- global $wgContLang;
- $ns = substr( $matches[1], 0, $colon );
- $index = $wgContLang->getNsIndex( $ns );
- if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) )
- return $matches[0]; // return the whole thing
- else
- return $matches[2];
-
- }
-
- /**
- * Simple & fast snippet extraction, but gives completely unrelevant
- * snippets
- *
- * @param $text String
- * @param $terms Array
- * @param $contextlines Integer
- * @param $contextchars Integer
- * @return String
- */
- public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
- global $wgContLang;
- $fname = __METHOD__;
-
- $lines = explode( "\n", $text );
-
- $terms = implode( '|', $terms );
- $max = intval( $contextchars ) + 1;
- $pat1 = "/(.*)($terms)(.{0,$max})/i";
-
- $lineno = 0;
-
- $extract = "";
- wfProfileIn( "$fname-extract" );
- foreach ( $lines as $line ) {
- if ( 0 == $contextlines ) {
- break;
- }
- ++$lineno;
- $m = array();
- if ( ! preg_match( $pat1, $line, $m ) ) {
- continue;
- }
- --$contextlines;
- $pre = $wgContLang->truncate( $m[1], - $contextchars );
-
- if ( count( $m ) < 3 ) {
- $post = '';
- } else {
- $post = $wgContLang->truncate( $m[3], $contextchars );
- }
-
- $found = $m[2];
-
- $line = htmlspecialchars( $pre . $found . $post );
- $pat2 = '/(' . $terms . ")/i";
- $line = preg_replace( $pat2,
- "<span class='searchmatch'>\\1</span>", $line );
-
- $extract .= "${line}\n";
- }
- wfProfileOut( "$fname-extract" );
-
- return $extract;
- }
-
}
/**
* Dummy class to be used when non-supported Database engine is present.
- * @todo Fixme: dummy class should probably try something at least mildly useful,
+ * @todo FIXME: Dummy class should probably try something at least mildly useful,
* such as a LIKE search through titles.
* @ingroup Search
*/