3 * Contain a class for special pages
9 var $searchTerms = array();
10 var $namespaces = array( NS_MAIN
);
11 var $showRedirects = false;
14 * Perform a full text search query and return a result set.
15 * If title searches are not supported or disabled, return null.
17 * @param string $term - Raw search term
18 * @return SearchResultSet
22 function searchText( $term ) {
27 * Perform a title-only search query and return a result set.
28 * If title searches are not supported or disabled, return null.
30 * @param string $term - Raw search term
31 * @return SearchResultSet
35 function searchTitle( $term ) {
40 * If an exact title match can be find, or a very slightly close match,
41 * return the title. If no match, returns NULL.
46 public static function getNearMatch( $searchterm ) {
49 $allSearchTerms = array($searchterm);
51 if($wgContLang->hasVariants()){
52 $allSearchTerms = array_merge($allSearchTerms,$wgContLang->convertLinkToAllVariants($searchterm));
55 foreach($allSearchTerms as $term){
57 # Exact match? No need to look further.
58 $title = Title
::newFromText( $term );
62 if ( $title->getNamespace() == NS_SPECIAL ||
$title->exists() ) {
66 # Now try all lower case (i.e. first letter capitalized)
68 $title = Title
::newFromText( $wgContLang->lc( $term ) );
69 if ( $title->exists() ) {
73 # Now try capitalized string
75 $title = Title
::newFromText( $wgContLang->ucwords( $term ) );
76 if ( $title->exists() ) {
80 # Now try all upper case
82 $title = Title
::newFromText( $wgContLang->uc( $term ) );
83 if ( $title->exists() ) {
87 # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
88 $title = Title
::newFromText( $wgContLang->ucwordbreaks($term) );
89 if ( $title->exists() ) {
93 global $wgCapitalLinks, $wgContLang;
94 if( !$wgCapitalLinks ) {
95 // Catch differs-by-first-letter-case-only
96 $title = Title
::newFromText( $wgContLang->ucfirst( $term ) );
97 if ( $title->exists() ) {
100 $title = Title
::newFromText( $wgContLang->lcfirst( $term ) );
101 if ( $title->exists() ) {
106 // Give hooks a chance at better match variants
108 if( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
113 $title = Title
::newFromText( $searchterm );
115 # Entering an IP address goes to the contributions page
116 if ( ( $title->getNamespace() == NS_USER
&& User
::isIP($title->getText() ) )
117 || User
::isIP( trim( $searchterm ) ) ) {
118 return SpecialPage
::getTitleFor( 'Contributions', $title->getDBkey() );
122 # Entering a user goes to the user page whether it's there or not
123 if ( $title->getNamespace() == NS_USER
) {
127 # Go to images that exist even if there's no local page.
128 # There may have been a funny upload, or it may be on a shared
129 # file repository such as Wikimedia Commons.
130 if( $title->getNamespace() == NS_IMAGE
) {
131 $image = wfFindFile( $title );
137 # MediaWiki namespace? Page may be "implied" if not customized.
138 # Just return it, with caps forced as the message system likes it.
139 if( $title->getNamespace() == NS_MEDIAWIKI
) {
140 return Title
::makeTitle( NS_MEDIAWIKI
, $wgContLang->ucfirst( $title->getText() ) );
143 # Quoted term? Try without the quotes...
145 if( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
146 return SearchEngine
::getNearMatch( $matches[1] );
152 public static function legalSearchChars() {
153 return "A-Za-z_'0-9\\x80-\\xFF\\-";
157 * Set the maximum number of results to return
158 * and how many to skip before returning the first.
164 function setLimitOffset( $limit, $offset = 0 ) {
165 $this->limit
= intval( $limit );
166 $this->offset
= intval( $offset );
170 * Set which namespaces the search should include.
171 * Give an array of namespace index numbers.
173 * @param array $namespaces
176 function setNamespaces( $namespaces ) {
177 $this->namespaces
= $namespaces;
181 * Parse some common prefixes: all (search everything)
184 * @param string $query
186 function replacePrefixes( $query ){
189 if( strpos($query,':') === false )
190 return $query; // nothing to do
193 $allkeyword = wfMsgForContent('searchall').":";
194 if( strncmp($query, $allkeyword, strlen($allkeyword)) == 0 ){
195 $this->namespaces
= null;
196 $parsed = substr($query,strlen($allkeyword));
197 } else if( strpos($query,':') !== false ) {
198 $prefix = substr($query,0,strpos($query,':'));
199 $index = $wgContLang->getNsIndex($prefix);
200 if($index !== false){
201 $this->namespaces
= array($index);
202 $parsed = substr($query,strlen($prefix)+
1);
205 if(trim($parsed) == '')
206 return $query; // prefix was the whole query
212 * Make a list of searchable namespaces and their canonical names.
215 public static function searchableNamespaces() {
218 foreach( $wgContLang->getNamespaces() as $ns => $name ) {
219 if( $ns >= NS_MAIN
) {
227 * Extract default namespaces to search from the given user's
228 * settings, returning a list of index numbers.
234 public static function userNamespaces( &$user ) {
236 foreach( SearchEngine
::searchableNamespaces() as $ns => $name ) {
237 if( $user->getOption( 'searchNs' . $ns ) ) {
245 * Find snippet highlight settings for a given user
248 * @return array contextlines, contextchars
251 public static function userHighlightPrefs( &$user ){
252 //$contextlines = $user->getOption( 'contextlines', 5 );
253 //$contextchars = $user->getOption( 'contextchars', 50 );
254 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
255 $contextchars = 75; // same as above.... :P
256 return array($contextlines, $contextchars);
260 * An array of namespaces indexes to be searched by default
265 public static function defaultNamespaces(){
266 global $wgNamespacesToBeSearchedDefault;
268 return array_keys($wgNamespacesToBeSearchedDefault, true);
272 * Return a 'cleaned up' search string
277 function filter( $text ) {
278 $lc = $this->legalSearchChars();
279 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
282 * Load up the appropriate search engine class for the currently
283 * active database backend, and return a configured instance.
285 * @return SearchEngine
287 public static function create() {
288 global $wgDBtype, $wgSearchType;
289 if( $wgSearchType ) {
290 $class = $wgSearchType;
291 } elseif( $wgDBtype == 'mysql' ) {
292 $class = 'SearchMySQL';
293 } else if ( $wgDBtype == 'postgres' ) {
294 $class = 'SearchPostgres';
295 } else if ( $wgDBtype == 'oracle' ) {
296 $class = 'SearchOracle';
298 $class = 'SearchEngineDummy';
300 $search = new $class( wfGetDB( DB_SLAVE
) );
301 $search->setLimitOffset(0,0);
306 * Create or update the search index record for the given page.
307 * Title and text should be pre-processed.
310 * @param string $title
311 * @param string $text
314 function update( $id, $title, $text ) {
319 * Update a search index record's title only.
320 * Title should be pre-processed.
323 * @param string $title
326 function updateTitle( $id, $title ) {
331 * Get OpenSearch suggestion template
336 public static function getOpenSearchTemplate() {
337 global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
338 if($wgOpenSearchTemplate)
339 return $wgOpenSearchTemplate;
341 $ns = implode(',',SearchEngine
::defaultNamespaces());
343 return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace='.$ns;
348 * Get internal MediaWiki Suggest template
353 public static function getMWSuggestTemplate() {
354 global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
355 if($wgMWSuggestTemplate)
356 return $wgMWSuggestTemplate;
358 return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}';
366 class SearchResultSet
{
368 * Fetch an array of regular expression fragments for matching
369 * the search terms as parsed by this engine in a text extract.
375 function termMatches() {
384 * Return true if results are included in this result set.
388 function hasResults() {
393 * Some search modes return a total hit count for the query
394 * in the entire article database. This may include pages
395 * in namespaces that would not be matched on the given
398 * Return null if no total hits number is supported.
403 function getTotalHits() {
408 * Some search modes return a suggested alternate term if there are
409 * no exact hits. Returns true if there is one on this set.
414 function hasSuggestion() {
419 * @return string suggested query, null if none
421 function getSuggestionQuery(){
426 * @return string highlighted suggested query, '' if none
428 function getSuggestionSnippet(){
433 * Return information about how and from where the results were fetched,
434 * should be useful for diagnostics and debugging
443 * Return a result set of hits on other (multiple) wikis associated with this one
445 * @return SearchResultSet
447 function getInterwikiResults() {
452 * Check if there are results on other wikis
456 function hasInterwikiResults() {
457 return $this->getInterwikiResults() != null;
462 * Fetches next search result, or false.
463 * @return SearchResult
472 * Frees the result set, if applicable.
484 class SearchResultTooMany
{
485 ## Some search engines may bail out if too many matches are found
493 var $mRevision = null;
495 function SearchResult( $row ) {
496 $this->mTitle
= Title
::makeTitle( $row->page_namespace
, $row->page_title
);
497 if( !is_null($this->mTitle
) )
498 $this->mRevision
= Revision
::newFromTitle( $this->mTitle
);
502 * Check if this is result points to an invalid title
507 function isBrokenTitle(){
508 if( is_null($this->mTitle
) )
514 * Check if target page is missing, happens when index is out of date
519 function isMissingRevision(){
520 if( !$this->mRevision
)
529 function getTitle() {
530 return $this->mTitle
;
534 * @return double or null if not supported
536 function getScore() {
541 * Lazy initialization of article text from DB
543 protected function initText(){
544 if( !isset($this->mText
) ){
545 $this->mText
= $this->mRevision
->getText();
550 * @param array $terms Terms to highlight (unescaped)
551 * @return string highlighted text snippet, null (and not '') if not supported
553 function getTextSnippet($terms){
556 list($contextlines,$contextchars) = SearchEngine
::userHighlightPrefs($wgUser);
557 $h = new SearchHighlighter();
558 return $h->highlightText( $this->mText
, $terms, $contextlines, $contextchars);
562 * @param array $terms terms to highlight
563 * @return string highlighted title, '' if not supported
565 function getTitleSnippet($terms){
570 * @param array $terms terms to highlight
571 * @return string highlighted redirect name (redirect to this page), '' if none or not supported
573 function getRedirectSnippet($terms){
578 * @return Title object for the redirect to this page, null if none or not supported
580 function getRedirectTitle(){
585 * @return string highlighted relevant section name, null if none or not supported
587 function getSectionSnippet(){
592 * @return Title object (pagename+fragment) for the section, null if none or not supported
594 function getSectionTitle(){
599 * @return string timestamp
601 function getTimestamp(){
602 return $this->mRevision
->getTimestamp();
606 * @return int number of words
608 function getWordCount(){
610 return str_word_count( $this->mText
);
614 * @return int size in bytes
616 function getByteSize(){
618 return strlen( $this->mText
);
622 * @return boolean if hit has related articles
624 function hasRelated(){
629 * @return interwiki prefix of the title (return iw even if title is broken)
631 function getInterwikiPrefix(){
637 * Highlight bits of wikitext
641 class SearchHighlighter
{
642 var $mCleanWikitext = true;
644 function SearchHighlighter($cleanupWikitext = true){
645 $this->mCleanWikitext
= $cleanupWikitext;
649 * Default implementation of wikitext highlighting
651 * @param string $text
652 * @param array $terms Terms to highlight (unescaped)
653 * @param int $contextlines
654 * @param int $contextchars
657 public function highlightText( $text, $terms, $contextlines, $contextchars ) {
658 global $wgLang, $wgContLang;
664 // spli text into text + templates/links/tables
665 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)/";
666 // first capture group is for detecting nested templates/links/tables
667 $endPatterns = array(
668 1 => '/(\{\{)|(\}\})/', // template
669 2 => '/(\[\[)|(\]\])/', // image
670 3 => "/(\n\\{\\|)|(\n\\|\\})/"); // table
671 $textExt = array(); // text extracts
672 $otherExt = array(); // other extracts
673 wfProfileIn( "$fname-split" );
675 $textLen = strlen($text);
676 $count = 0; // sequence number to maintain ordering
677 while( $start < $textLen ){
678 // find start of template/image/table
679 if( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE
, $start ) ){
681 foreach($matches as $key => $val){
682 if($key > 0 && $val[1] != -1){
684 // see if this is an image link
685 $ns = substr($val[0],2,-1);
686 if( $wgContLang->getNsIndex($ns) != NS_IMAGE
)
690 $epat = $endPatterns[$key];
691 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
697 // find end (and detect any nested elements)
699 $offset = $start +
1;
701 while( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE
, $offset ) ){
702 if( array_key_exists(2,$endMatches) ){
705 $len = strlen($endMatches[2][0]);
706 $off = $endMatches[2][1];
707 $this->splitAndAdd( $otherExt, $count,
708 substr( $text, $start, $off +
$len - $start ) );
709 $start = $off +
$len;
713 // end of nested element
720 $offset = $endMatches[0][1] +
strlen($endMatches[0][0]);
723 // couldn't find appropriate closing tag, skip
724 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen($matches[0][0]) ) );
725 $start +
= strlen($matches[0][0]);
730 // else: add as text extract
731 $this->splitAndAdd( $textExt, $count, substr($text,$start) );
735 $all = $textExt +
$otherExt; // these have disjunct key sets
737 wfProfileOut( "$fname-split" );
740 foreach( $terms as $index => $term ) {
741 $terms[$index] = preg_quote( $term, '/' );
742 // manually do upper/lowercase stuff for utf-8 since PHP won't do it
743 if(preg_match('/[\x80-\xff]/', $term) ){
744 $terms[$index] = preg_replace_callback('/./us',array($this,'caseCallback'),$terms[$index]);
749 $anyterm = implode( '|', $terms );
750 $phrase = implode('[, .:;\(\)"\'\-\+]+', $terms );
752 // FIXME: a hack to scale contextchars, a correct solution
753 // would be to have contextchars actually be char and not byte
754 // length, and do proper utf-8 substrings and lengths everywhere,
755 // but PHP is making that very hard and unclean to implement :(
756 $scale = strlen($anyterm) / mb_strlen($anyterm);
757 $contextchars = intval( $contextchars * $scale );
759 $pat1 = '/('.$phrase.')/ui';
760 $pat2 = '/('.$anyterm.')/ui';
762 wfProfileIn( "$fname-extract" );
764 $left = $contextlines;
768 // match whole query on text
769 $this->process($pat1, $textExt, $left, $contextchars, $snippets, $offsets);
770 // match whole query on templates/tables/images
771 $this->process($pat1, $otherExt, $left, $contextchars, $snippets, $offsets);
772 // match any words on text
773 $this->process($pat2, $textExt, $left, $contextchars, $snippets, $offsets);
774 // match any words on templates/tables/images
775 $this->process($pat2, $otherExt, $left, $contextchars, $snippets, $offsets);
779 $first = array_keys($textExt);
780 if( isset($first[0]))
785 // add extra chars to each snippet to make snippets constant size
787 if( count( $snippets ) == 0){
788 // couldn't find the target words, just show beginning of article
789 $targetchars = $contextchars * $contextlines;
790 $snippets[$first] = '';
791 $offsets[$first] = 0;
793 // if begin of the article contains the whole phrase, show only that !!
794 if( array_key_exists($first,$snippets) && preg_match($pat1,$snippets[$first])
795 && $offsets[$first] < $contextchars * 2 ){
796 $snippets = array ($first => $snippets[$first]);
799 // calc by how much to extend existing snippets
800 $targetchars = intval( ($contextchars * $contextlines) / count ( $snippets ) );
803 foreach($snippets as $index => $line){
804 $extended[$index] = $line;
805 $len = strlen($line);
806 if( $len < $targetchars - 20 ){
807 // complete this line
808 if($len < strlen( $all[$index] )){
809 $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index]+
$targetchars, $offsets[$index]);
810 $len = strlen( $extended[$index] );
815 while( $len < $targetchars - 20
816 && array_key_exists($add,$all)
817 && !array_key_exists($add,$snippets) ){
819 $tt = "\n".$this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
820 $extended[$add] = $tt;
821 $len +
= strlen( $tt );
827 $snippets = array_map('htmlspecialchars', $extended);
830 foreach($snippets as $index => $line){
832 $extract .= $line; // first line
833 elseif($last+
1 == $index && $offsets[$last]+
strlen($snippets[$last]) >= strlen($all[$last]))
834 $extract .= " ".$line; // continous lines
836 $extract .= '<b> ... </b>' . $line;
841 $extract .= '<b> ... </b>';
844 $pat3 = '/(' . $anyterm . ")/ui";
845 $extract = preg_replace( $pat3,
846 "<span class='searchmatch'>\\1</span>", $extract );
848 wfProfileOut( "$fname-extract" );
854 * Split text into lines and add it to extracts array
856 * @param array $extracts index -> $line
858 * @param string $text
860 function splitAndAdd(&$extracts, &$count, $text){
861 $split = explode( "\n", $this->mCleanWikitext?
$this->removeWiki($text) : $text );
862 foreach($split as $line){
865 $extracts[$count++
] = $tt;
870 * Do manual case conversion for non-ascii chars
872 * @param unknown_type $matches
874 function caseCallback($matches){
876 if( strlen($matches[0]) > 1 ){
877 return '['.$wgContLang->lc($matches[0]).$wgContLang->uc($matches[0]).']';
883 * Extract part of the text from start to end, but by
884 * not chopping up words
885 * @param string $text
888 * @param int $posStart (out) actual start position
889 * @param int $posEnd (out) actual end position
892 function extract($text, $start, $end, &$posStart = null, &$posEnd = null ){
896 $start = $this->position( $text, $start, 1 );
897 if( $end >= strlen($text) )
898 $end = strlen($text);
900 $end = $this->position( $text, $end );
902 if(!is_null($posStart))
904 if(!is_null($posEnd))
908 return substr($text, $start, $end-$start);
914 * Find a nonletter near a point (index) in the text
916 * @param string $text
918 * @param int $offset to found index
919 * @return int nearest nonletter index, or beginning of utf8 char if none
921 function position($text, $point, $offset=0 ){
923 $s = max( 0, $point - $tolerance );
924 $l = min( strlen($text), $point +
$tolerance ) - $s;
926 if( preg_match('/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr($text,$s,$l), $m, PREG_OFFSET_CAPTURE
) ){
927 return $m[0][1] +
$s +
$offset;
929 // check if point is on a valid first UTF8 char
930 $char = ord( $text[$point] );
931 while( $char >= 0x80 && $char < 0xc0 ) {
932 // skip trailing bytes
934 if($point >= strlen($text))
935 return strlen($text);
936 $char = ord( $text[$point] );
944 * Search extracts for a pattern, and return snippets
946 * @param string $pattern regexp for matching lines
947 * @param array $extracts extracts to search
948 * @param int $linesleft number of extracts to make
949 * @param int $contextchars length of snippet
950 * @param array $out map for highlighted snippets
951 * @param array $offsets map of starting points of snippets
954 function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ){
956 return; // nothing to do
957 foreach($extracts as $index => $line){
958 if( array_key_exists($index,$out) )
959 continue; // this line already highlighted
962 if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE
) )
966 $len = strlen($m[0][0]);
967 if($offset +
$len < $contextchars)
969 elseif( $len > $contextchars)
972 $begin = $offset +
intval( ($len - $contextchars) / 2 );
974 $end = $begin +
$contextchars;
977 // basic snippet from this line
978 $out[$index] = $this->extract($line,$begin,$end,$posBegin);
979 $offsets[$index] = $posBegin;
987 * Basic wikitext removal
990 function removeWiki($text) {
992 wfProfileIn( $fname );
994 //$text = preg_replace("/'{2,5}/", "", $text);
995 //$text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
996 //$text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
997 //$text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
998 //$text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
999 //$text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1000 $text = preg_replace("/\\{\\{([^|]+?)\\}\\}/", "", $text);
1001 $text = preg_replace("/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text);
1002 $text = preg_replace("/\\[\\[([^|]+?)\\]\\]/", "\\1", $text);
1003 $text = preg_replace_callback("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array($this,'linkReplace'), $text);
1004 //$text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1005 $text = preg_replace("/<\/?[^>]+>/", "", $text);
1006 $text = preg_replace("/'''''/", "", $text);
1007 $text = preg_replace("/('''|<\/?[iIuUbB]>)/", "", $text);
1008 $text = preg_replace("/''/", "", $text);
1010 wfProfileOut( $fname );
1015 * callback to replace [[target|caption]] kind of links, if
1016 * the target is category or image, leave it
1018 * @param array $matches
1020 function linkReplace($matches){
1021 $colon = strpos( $matches[1], ':' );
1022 if( $colon === false )
1023 return $matches[2]; // replace with caption
1025 $ns = substr( $matches[1], 0, $colon );
1026 $index = $wgContLang->getNsIndex($ns);
1027 if( $index !== false && ($index == NS_IMAGE ||
$index == NS_CATEGORY
) )
1028 return $matches[0]; // return the whole thing
1036 * @addtogroup Search
1038 class SearchEngineDummy
{
1039 function search( $term ) {
1042 function setLimitOffset($l, $o) {}
1043 function legalSearchChars() {}
1044 function update() {}
1045 function setnamespaces() {}
1046 function searchtitle() {}
1047 function searchtext() {}