+class SearchHighlighter {
+ var $mCleanWikitext = true;
+
+ function SearchHighlighter($cleanupWikitext = true){
+ $this->mCleanWikitext = $cleanupWikitext;
+ }
+
+ /**
+ * Default implementation of wikitext highlighting
+ *
+ * @param string $text
+ * @param array $terms Terms to highlight (unescaped)
+ * @param int $contextlines
+ * @param int $contextchars
+ * @return string
+ */
+ public function highlightText( $text, $terms, $contextlines, $contextchars ) {
+ global $wgLang, $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;
+ }
+ }
+ $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
+ $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;