{$top}
\n" ); - - # For powersearch - $a2l = ''; - $akk = array_keys( $this->addToQuery ); - foreach ( $akk AS $ak ) { - $a2l .= "&{$ak}={$this->addToQuery[$ak]}" ; + return $arr; + } + + /** + * Extract default namespaces to search from the given user's + * settings, returning a list of index numbers. + * + * @param User $user + * @return array + * @static + */ + public static function userNamespaces( &$user ) { + $arr = array(); + foreach( SearchEngine::searchableNamespaces() as $ns => $name ) { + if( $user->getOption( 'searchNs' . $ns ) ) { + $arr[] = $ns; + } } - - $prevnext = wfViewPrevNext( $offset, $limit, '', - 'search=' . wfUrlencode( $this->filteredText ) . $a2l ); - $wgOut->addHTML( "{$prevnext}
\n" ); - $wgOut->addHTML( $powersearch ); + return array_keys($wgNamespacesToBeSearchedDefault, true); } - function legalSearchChars() { - $lc = "A-Za-z_'0-9\\x80-\\xFF\\-"; - return $lc; + /** + * Return a 'cleaned up' search string + * + * @return string + * @access public + */ + function filter( $text ) { + $lc = $this->legalSearchChars(); + return trim( preg_replace( "/[^{$lc}]/", " ", $text ) ); } - - function parseQuery() { - global $wgDBmysql4; - if( $wgDBmysql4 ) { - # Use cleaner boolean search if available - return $this->parseQuery4(); + /** + * Load up the appropriate search engine class for the currently + * active database backend, and return a configured instance. + * + * @fixme Ask the database class for his default search class + * instead of knowing about every backend here. + * @return SearchEngine + */ + public static function create() { + global $wgDBtype, $wgSearchType; + if( $wgSearchType ) { + $class = $wgSearchType; + } elseif( $wgDBtype == 'mysql' ) { + $class = 'SearchMySQL'; + } else if ( $wgDBtype == 'postgres' ) { + $class = 'SearchPostgres'; + } else if ( $wgDBtype == 'oracle' ) { + $class = 'SearchOracle'; } else { - # Fall back to ugly hack with multiple search clauses - return $this->parseQuery3(); + $class = 'SearchEngineDummy'; } + $search = new $class( wfGetDB( DB_SLAVE ) ); + $search->setLimitOffset(0,0); + return $search; } - - function parseQuery3() { - global $wgDBminWordLen, $wgLang; - - # on non mysql4 database: get list of words we don't want to search for - require_once( 'FulltextStoplist.php' ); - - $lc = SearchEngine::legalSearchChars() . '()'; - $q = preg_replace( "/([()])/", " \\1 ", $this->filteredText ); - $q = preg_replace( "/\\s+/", " ", $q ); - $w = explode( ' ', trim( $q ) ); - - $last = $cond = ''; - foreach ( $w as $word ) { - $word = $wgLang->stripForSearch( $word ); - if ( 'and' == $word || 'or' == $word || 'not' == $word - || '(' == $word || ')' == $word ) { - $cond .= ' ' . strtoupper( $word ); - $last = ''; - } else if ( strlen( $word ) < $wgDBminWordLen ) { - continue; - } else if ( FulltextStoplist::inList( $word ) ) { - continue; - } else { - if ( '' != $last ) { $cond .= ' AND'; } - $cond .= " (MATCH (##field##) AGAINST ('" . - $this->db->strencode( $word ). "'))"; - $last = $word; - array_push( $this->searchTerms, "\\b" . $word . "\\b" ); - } - } - if ( 0 == count( $this->searchTerms ) ) { - return MW_SEARCH_BAD_QUERY; - } - $this->titleCond = '(' . str_replace( '##field##', - 'si_title', $cond ) . ' )'; + /** + * Create or update the search index record for the given page. + * Title and text should be pre-processed. + * + * @param int $id + * @param string $title + * @param string $text + * @abstract + */ + function update( $id, $title, $text ) { + // no-op + } - $this->textCond = '(' . str_replace( '##field##', - 'si_text', $cond ) . ' AND (cur_is_redirect=0) )'; - - return MW_SEARCH_OK; + /** + * Update a search index record's title only. + * Title should be pre-processed. + * + * @param int $id + * @param string $title + * @abstract + */ + function updateTitle( $id, $title ) { + // no-op } - function parseQuery4() { - global $wgLang; - $lc = SearchEngine::legalSearchChars(); - $searchon = ''; - $this->searchTerms = array(); - - # FIXME: This doesn't handle parenthetical expressions. - if( preg_match_all( '/([-+<>~]?)(([' . $lc . ']+)(\*?)|"[^"]*")/', - $this->filteredText, $m, PREG_SET_ORDER ) ) { - foreach( $m as $terms ) { - if( $searchon !== '' ) $searchon .= ' '; - if( $this->strictMatching && ($terms[1] == '') ) { - $terms[1] = '+'; - } - $searchon .= $terms[1] . $wgLang->stripForSearch( $terms[2] ); - if( !empty( $terms[3] ) ) { - $regexp = preg_quote( $terms[3] ); - if( $terms[4] ) $regexp .= "[0-9A-Za-z_]+"; - } else { - $regexp = preg_quote( str_replace( '"', '', $terms[2] ) ); - } - $this->searchTerms[] = $regexp; - } - wfDebug( "Would search with '$searchon'\n" ); - wfDebug( "Match with /\b" . implode( '\b|\b', $this->searchTerms ) . "\b/\n" ); - } else { - wfDebug( "Can't understand search query '{$this->filteredText}'\n" ); + /** + * Get OpenSearch suggestion template + * + * @return string + * @static + */ + public static function getOpenSearchTemplate() { + global $wgOpenSearchTemplate, $wgServer, $wgScriptPath; + if($wgOpenSearchTemplate) + return $wgOpenSearchTemplate; + else{ + $ns = implode(',',SearchEngine::defaultNamespaces()); + if(!$ns) $ns = "0"; + return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace='.$ns; } - - $searchon = $this->db->strencode( $searchon ); - $this->titleCond = " MATCH(si_title) AGAINST('$searchon' IN BOOLEAN MODE)"; - $this->textCond = " (MATCH(si_text) AGAINST('$searchon' IN BOOLEAN MODE) AND cur_is_redirect=0)"; - return MW_SEARCH_OK; } + + /** + * Get internal MediaWiki Suggest template + * + * @return string + * @static + */ + public static function getMWSuggestTemplate() { + global $wgMWSuggestTemplate, $wgServer, $wgScriptPath; + if($wgMWSuggestTemplate) + return $wgMWSuggestTemplate; + else + return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}'; + } +} - function &getMatches( $cond, $limit, $offset = 0 ) { - $searchindex = $this->db->tableName( 'searchindex' ); - $cur = $this->db->tableName( 'cur' ); - $searchnamespaces = $this->queryNamespaces(); - $redircond = $this->searchRedirects(); - - $sql = "SELECT cur_id,cur_namespace,cur_title," . - "cur_text FROM $cur,$searchindex " . - "WHERE cur_id=si_page AND {$cond} " . - "{$searchnamespaces} {$redircond} " . - $this->db->limitResult( $limit, $offset ); - - $res = $this->db->query( $sql, 'SearchEngine::getMatches' ); - $matches = array(); - while ( $row = $this->db->fetchObject( $res ) ) { - $matches[] = $row; - } - $this->db->freeResult( $res ); - - return $matches; +/** + * @ingroup Search + */ +class SearchResultSet { + /** + * Fetch an array of regular expression fragments for matching + * the search terms as parsed by this engine in a text extract. + * + * @return array + * @access public + * @abstract + */ + function termMatches() { + return array(); } - function showMatches( &$matches, $offset, $msgEmpty, $msgFound ) { - global $wgOut; - if ( 0 == count( $matches ) ) { - $wgOut->addHTML( "' . wfMsg('nogomatch', $editurl ) . "
\n" ); - - # Try a fuzzy title search - $anyhit = false; - global $wgDisableFuzzySearch; - if(! $wgDisableFuzzySearch ){ - foreach( array(NS_MAIN, NS_PROJECT, NS_USER, NS_IMAGE, NS_MEDIAWIKI) as $namespace){ - $anyhit |= SearchEngine::doFuzzyTitleSearch( $this->rawText, $namespace ); + 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( ! $anyhit ){ - return $this->showResults(); + 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); } - } - - /** - * @static - */ - function doFuzzyTitleSearch( $search, $namespace ){ - global $wgLang, $wgOut; - $this->setupPage(); + // 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 + $targetchars = $contextchars * $contextlines; + $snippets[$first] = ''; + $offsets[$first] = 0; + } 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]); + } + + // 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++; + } + } + } - $sstr = ucfirst($search); - $sstr = str_replace(' ', '_', $sstr); - $fuzzymatches = SearchEngine::fuzzyTitles( $sstr, $namespace ); - $fuzzymatches = array_slice($fuzzymatches, 0, 10); - $slen = strlen( $search ); - $wikitext = ''; - foreach($fuzzymatches as $res){ - $t = str_replace('_', ' ', $res[1]); - $tfull = $wgLang->getNsText( $namespace ) . ":$t|$t"; - if( $namespace == NS_MAIN ) - $tfull = "$t"; - $distance = $res[0]; - $closeness = (strlen( $search ) - $distance) / strlen( $search ); - $percent = intval( $closeness * 100 ) . '%'; - $stars = str_repeat('*', ceil(5 * $closeness) ); - $wikitext .= "* [[$tfull]] $percent ($stars)\n"; + //$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 .= ' ... ' . $line; + + $last = $index; } - if( $wikitext ){ - if( $namespace != NS_MAIN ) - $wikitext = '=== ' . $wgLang->getNsText( $namespace ) . " ===\n" . $wikitext; - $wgOut->addWikiText( $wikitext ); - return true; + if( $extract ) + $extract .= ' ... '; + + $processed = array(); + foreach($terms as $term){ + if( ! isset($processed[$term]) ){ + $pat3 = "/$patPre(".$term.")$patPost/ui"; // highlight word + $extract = preg_replace( $pat3, + "\\1\\2\\3", $extract ); + $processed[$term] = true; + } } - return false; + + wfProfileOut( "$fname-extract" ); + + return $extract; } - + /** - * @static + * Split text into lines and add it to extracts array + * + * @param array $extracts index -> $line + * @param int $count + * @param string $text */ - function fuzzyTitles( $sstr, $namespace = NS_MAIN ){ - $span = 0.10; // weed on title length before doing levenshtein. - $tolerance = 0.35; // allowed percentage of erronous characters - $slen = strlen($sstr); - $tolerance_count = ceil($tolerance * $slen); - $spanabs = ceil($slen * (1 + $span)) - $slen; - # print "Word: $sstr, len = $slen, range = [$min, $max], tolerance_count = $tolerance_count