SearchEngine.php

Go to the documentation of this file.

<?php
class SearchEngine {
        var $limit = 10;
        var $offset = 0;
        var $prefix = '';
        var $searchTerms = array();
        var $namespaces = array( NS_MAIN );
        var $showRedirects = false;

        protected $features = array();

        protected $db;

        function __construct($db = null) {
                if ( $db ) {
                        $this->db = $db;
                } else {
                        $this->db = wfGetDB( DB_SLAVE );
                }
        }

        function searchText( $term ) {
                return null;
        }

        function searchTitle( $term ) {
                return null;
        }

        function acceptListRedirects() {
                wfDeprecated( __METHOD__, '1.18' );
                return $this->supports( 'list-redirects' );
        }

        public function supports( $feature ) {
                switch( $feature ) {
                case 'list-redirects':
                        return true;
                case 'title-suffix-filter':
                default:
                        return false;
                }
        }

        public function setFeatureData( $feature, $data ) {
                $this->features[$feature] = $data;
        }

        public function normalizeText( $string ) {
                global $wgContLang;

                // Some languages such as Chinese require word segmentation
                return $wgContLang->segmentByWord( $string );
        }

        function transformSearchTerm( $term ) {
                return $term;
        }

        public static function getNearMatch( $searchterm ) {
                $title = self::getNearMatchInternal( $searchterm );

                wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
                return $title;
        }

        public static function getNearMatchResultSet( $searchterm ) {
                return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
        }

        private static function getNearMatchInternal( $searchterm ) {
                global $wgContLang, $wgEnableSearchContributorsByIP;

                $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->isSpecialPage() || $title->isExternal() || $title->exists() ) {
                                return $title;
                        }

                        # See if it still otherwise has content is some sane sense
                        $page = WikiPage::factory( $title );
                        if ( $page->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 ( $wgEnableSearchContributorsByIP ) {
                        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() {
                return "A-Za-z_'.0-9\\x80-\\xFF\\-";
        }

        function setLimitOffset( $limit, $offset = 0 ) {
                $this->limit = intval( $limit );
                $this->offset = intval( $offset );
        }

        function setNamespaces( $namespaces ) {
                $this->namespaces = $namespaces;
        }

        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 ) );
                } elseif ( 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;
        }

        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 static function userNamespaces( $user ) {
                global $wgSearchEverythingOnlyLoggedIn;

                $searchableNamespaces = SearchEngine::searchableNamespaces();

                // get search everything preference, that can be set to be read for logged-in users
                // it overrides other options
                if ( !$wgSearchEverythingOnlyLoggedIn || $user->isLoggedIn() ) {
                        if ( $user->getOption( 'searcheverything' ) ) {
                                return array_keys( $searchableNamespaces );
                        }
                }

                $arr = array();
                foreach ( $searchableNamespaces as $ns => $name ) {
                        if ( $user->getOption( 'searchNs' . $ns ) ) {
                                $arr[] = $ns;
                        }
                }

                return $arr;
        }

        public static function userHighlightPrefs() {
                $contextlines = 2; // Hardcode this. Old defaults sucked. :)
                $contextchars = 75; // same as above.... :P
                return array( $contextlines, $contextchars );
        }

        public static function defaultNamespaces() {
                global $wgNamespacesToBeSearchedDefault;

                return array_keys( $wgNamespacesToBeSearchedDefault, true );
        }

        public static function namespacesAsText( $namespaces ) {
                global $wgContLang;

                $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
                foreach ( $formatted as $key => $ns ) {
                        if ( empty( $ns ) )
                                $formatted[$key] = wfMsg( 'blanknamespace' );
                }
                return $formatted;
        }

        public static function helpNamespaces() {
                global $wgNamespacesToBeSearchedHelp;

                return array_keys( $wgNamespacesToBeSearchedHelp, true );
        }

        function filter( $text ) {
                $lc = $this->legalSearchChars();
                return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
        }
        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;
        }

        function update( $id, $title, $text ) {
                // no-op
        }

        function updateTitle( $id, $title ) {
                // no-op
        }

        public static function getOpenSearchTemplate() {
                global $wgOpenSearchTemplate, $wgCanonicalServer;
                if ( $wgOpenSearchTemplate ) {
                        return $wgOpenSearchTemplate;
                } else {
                        $ns = implode( '|', SearchEngine::defaultNamespaces() );
                        if ( !$ns ) {
                                $ns = "0";
                        }
                        return $wgCanonicalServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
                }
        }

        public static function getMWSuggestTemplate() {
                global $wgMWSuggestTemplate, $wgServer;
                if ( $wgMWSuggestTemplate )
                        return $wgMWSuggestTemplate;
                else
                        return $wgServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
        }
}

class SearchResultSet {
        function termMatches() {
                return array();
        }

        function numRows() {
                return 0;
        }

        function hasResults() {
                return false;
        }

        function getTotalHits() {
                return null;
        }

        function hasSuggestion() {
                return false;
        }

        function getSuggestionQuery() {
                return null;
        }

        function getSuggestionSnippet() {
                return '';
        }

        function getInfo() {
                return null;
        }

        function getInterwikiResults() {
                return null;
        }

        function hasInterwikiResults() {
                return $this->getInterwikiResults() != null;
        }

        function next() {
                return false;
        }

        function free() {
                // ...
        }
}

class SqlSearchResultSet extends SearchResultSet {

        protected $mResultSet;

        function __construct( $resultSet, $terms ) {
                $this->mResultSet = $resultSet;
                $this->mTerms = $terms;
        }

        function termMatches() {
                return $this->mTerms;
        }

        function numRows() {
                if ( $this->mResultSet === false )
                        return false;

                return $this->mResultSet->numRows();
        }

        function next() {
                if ( $this->mResultSet === false )
                        return false;

                $row = $this->mResultSet->fetchObject();
                if ( $row === false )
                        return false;

                return SearchResult::newFromRow( $row );
        }

        function free() {
                if ( $this->mResultSet === false )
                        return false;

                $this->mResultSet->free();
        }
}

class SearchResultTooMany {
        # # Some search engines may bail out if too many matches are found
}


class SearchResult {

        var $mRevision = null;
        var $mImage = null;

        var $mTitle;

        var $mText;

        public static function newFromTitle( $title ) {
                $result = new self();
                $result->initFromTitle( $title );
                return $result;
        }
        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 );
                }
        }

        protected function initFromRow( $row ) {
                $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
        }

        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 );
                }
        }

        function isBrokenTitle() {
                if ( is_null( $this->mTitle ) )
                        return true;
                return false;
        }

        function isMissingRevision() {
                return !$this->mRevision && !$this->mImage;
        }

        function getTitle() {
                return $this->mTitle;
        }

        function getScore() {
                return null;
        }

        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 = '';

                }
        }

        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 );
        }

        function getTitleSnippet( $terms ) {
                return '';
        }

        function getRedirectSnippet( $terms ) {
                return '';
        }

        function getRedirectTitle() {
                return null;
        }

        function getSectionSnippet() {
                return '';
        }

        function getSectionTitle() {
                return null;
        }

        function getTimestamp() {
                if ( $this->mRevision )
                        return $this->mRevision->getTimestamp();
                elseif ( $this->mImage )
                        return $this->mImage->getTimestamp();
                return '';
        }

        function getWordCount() {
                $this->initText();
                return str_word_count( $this->mText );
        }

        function getByteSize() {
                $this->initText();
                return strlen( $this->mText );
        }

        function hasRelated() {
                return false;
        }

        function getInterwikiPrefix() {
                return '';
        }
}
class SearchNearMatchResultSet extends SearchResultSet {
        private $fetched = false;
        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 );
        }
}

class SearchHighlighter {
        var $mCleanWikitext = true;

        function __construct( $cleanupWikitext = true ) {
                $this->mCleanWikitext = $cleanupWikitext;
        }

        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

                // @todo 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;
                        }
                }
                $anyterm = implode( '|', $terms );
                $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );

                // @todo 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;
                        }
                } 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++;
                                }
                        }
                }

                // $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;
        }

        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;
                }
        }

        function caseCallback( $matches ) {
                global $wgContLang;
                if ( strlen( $matches[0] ) > 1 ) {
                        return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
                } else {
                        return $matches[0];
                }
        }

        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 '';
                }
        }

        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] );
                        }
                        return $point;

                }
        }

        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;
                }
        }

        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;
        }

        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];

        }

        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;
                        // truncate function changes ... to relevant i18n message.
                        $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );

                        if ( count( $m ) < 3 ) {
                                $post = '';
                        } else {
                                $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
                        }

                        $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;
        }

}

class SearchEngineDummy extends SearchEngine {
        // no-op
}