includes/SearchEngine.php

   1 <?php
   2 /**
   3  * @defgroup Search Search
   4  *
   5  * @file
   6  * @ingroup Search
   7  */
   8
   9 /**
  10  * Contain a class for special pages
  11  * @ingroup Search
  12  */
  13 class SearchEngine {
  14         var $limit = 10;
  15         var $offset = 0;
  16         var $searchTerms = array();
  17         var $namespaces = array( NS_MAIN );
  18         var $showRedirects = false;
  19
  20         /**
  21          * Perform a full text search query and return a result set.
  22          * If title searches are not supported or disabled, return null.
  23          *
  24          * @param string $term - Raw search term
  25          * @return SearchResultSet
  26          * @access public
  27          * @abstract
  28          */
  29         function searchText( $term ) {
  30                 return null;
  31         }
  32
  33         /**
  34          * Perform a title-only search query and return a result set.
  35          * If title searches are not supported or disabled, return null.
  36          *
  37          * @param string $term - Raw search term
  38          * @return SearchResultSet
  39          * @access public
  40          * @abstract
  41          */
  42         function searchTitle( $term ) {
  43                 return null;
  44         }
  45
  46         /**
  47          * If an exact title match can be find, or a very slightly close match,
  48          * return the title. If no match, returns NULL.
  49          *
  50          * @param string $term
  51          * @return Title
  52          */
  53         public static function getNearMatch( $searchterm ) {
  54                 global $wgContLang;
  55
  56                 $allSearchTerms = array($searchterm);
  57
  58                 if($wgContLang->hasVariants()){
  59                         $allSearchTerms = array_merge($allSearchTerms,$wgContLang->convertLinkToAllVariants($searchterm));
  60                 }
  61
  62                 foreach($allSearchTerms as $term){
  63
  64                         # Exact match? No need to look further.
  65                         $title = Title::newFromText( $term );
  66                         if (is_null($title))
  67                                 return NULL;
  68
  69                         if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal()
  70                              || $title->exists() ) {
  71                                 return $title;
  72                         }
  73
  74                         # Now try all lower case (i.e. first letter capitalized)
  75                         #
  76                         $title = Title::newFromText( $wgContLang->lc( $term ) );
  77                         if ( $title && $title->exists() ) {
  78                                 return $title;
  79                         }
  80
  81                         # Now try capitalized string
  82                         #
  83                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
  84                         if ( $title && $title->exists() ) {
  85                                 return $title;
  86                         }
  87
  88                         # Now try all upper case
  89                         #
  90                         $title = Title::newFromText( $wgContLang->uc( $term ) );
  91                         if ( $title && $title->exists() ) {
  92                                 return $title;
  93                         }
  94
  95                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
  96                         $title = Title::newFromText( $wgContLang->ucwordbreaks($term) );
  97                         if ( $title && $title->exists() ) {
  98                                 return $title;
  99                         }
 100
 101                         // Give hooks a chance at better match variants
 102                         $title = null;
 103                         if( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 104                                 return $title;
 105                         }
 106                 }
 107
 108                 $title = Title::newFromText( $searchterm );
 109
 110                 # Entering an IP address goes to the contributions page
 111                 if ( ( $title->getNamespace() == NS_USER && User::isIP($title->getText() ) )
 112                         || User::isIP( trim( $searchterm ) ) ) {
 113                         return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 114                 }
 115
 116
 117                 # Entering a user goes to the user page whether it's there or not
 118                 if ( $title->getNamespace() == NS_USER ) {
 119                         return $title;
 120                 }
 121
 122                 # Go to images that exist even if there's no local page.
 123                 # There may have been a funny upload, or it may be on a shared
 124                 # file repository such as Wikimedia Commons.
 125                 if( $title->getNamespace() == NS_IMAGE ) {
 126                         $image = wfFindFile( $title );
 127                         if( $image ) {
 128                                 return $title;
 129                         }
 130                 }
 131
 132                 # MediaWiki namespace? Page may be "implied" if not customized.
 133                 # Just return it, with caps forced as the message system likes it.
 134                 if( $title->getNamespace() == NS_MEDIAWIKI ) {
 135                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 136                 }
 137
 138                 # Quoted term? Try without the quotes...
 139                 $matches = array();
 140                 if( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 141                         return SearchEngine::getNearMatch( $matches[1] );
 142                 }
 143
 144                 return NULL;
 145         }
 146
 147         public static function legalSearchChars() {
 148                 return "A-Za-z_'0-9\\x80-\\xFF\\-";
 149         }
 150
 151         /**
 152          * Set the maximum number of results to return
 153          * and how many to skip before returning the first.
 154          *
 155          * @param int $limit
 156          * @param int $offset
 157          * @access public
 158          */
 159         function setLimitOffset( $limit, $offset = 0 ) {
 160                 $this->limit = intval( $limit );
 161                 $this->offset = intval( $offset );
 162         }
 163
 164         /**
 165          * Set which namespaces the search should include.
 166          * Give an array of namespace index numbers.
 167          *
 168          * @param array $namespaces
 169          * @access public
 170          */
 171         function setNamespaces( $namespaces ) {
 172                 $this->namespaces = $namespaces;
 173         }
 174
 175         /**
 176          * Parse some common prefixes: all (search everything)
 177          * or namespace names
 178          *
 179          * @param string $query
 180          */
 181         function replacePrefixes( $query ){
 182                 global $wgContLang;
 183
 184                 if( strpos($query,':') === false )
 185                         return $query; // nothing to do
 186
 187                 $parsed = $query;
 188                 $allkeyword = wfMsgForContent('searchall').":";
 189                 if( strncmp($query, $allkeyword, strlen($allkeyword)) == 0 ){
 190                         $this->namespaces = null;
 191                         $parsed = substr($query,strlen($allkeyword));
 192                 } else if( strpos($query,':') !== false ) {
 193                         $prefix = substr($query,0,strpos($query,':'));
 194                         $index = $wgContLang->getNsIndex($prefix);
 195                         if($index !== false){
 196                                 $this->namespaces = array($index);
 197                                 $parsed = substr($query,strlen($prefix)+1);
 198                         }
 199                 }
 200                 if(trim($parsed) == '')
 201                         return $query; // prefix was the whole query
 202
 203                 return $parsed;
 204         }
 205
 206         /**
 207          * Make a list of searchable namespaces and their canonical names.
 208          * @return array
 209          */
 210         public static function searchableNamespaces() {
 211                 global $wgContLang;
 212                 $arr = array();
 213                 foreach( $wgContLang->getNamespaces() as $ns => $name ) {
 214                         if( $ns >= NS_MAIN ) {
 215                                 $arr[$ns] = $name;
 216                         }
 217                 }
 218                 return $arr;
 219         }
 220
 221         /**
 222          * Extract default namespaces to search from the given user's
 223          * settings, returning a list of index numbers.
 224          *
 225          * @param User $user
 226          * @return array
 227          * @static
 228          */
 229         public static function userNamespaces( &$user ) {
 230                 $arr = array();
 231                 foreach( SearchEngine::searchableNamespaces() as $ns => $name ) {
 232                         if( $user->getOption( 'searchNs' . $ns ) ) {
 233                                 $arr[] = $ns;
 234                         }
 235                 }
 236                 return $arr;
 237         }
 238
 239         /**
 240          * Find snippet highlight settings for a given user
 241          *
 242          * @param User $user
 243          * @return array contextlines, contextchars
 244          * @static
 245          */
 246         public static function userHighlightPrefs( &$user ){
 247                 //$contextlines = $user->getOption( 'contextlines',  5 );
 248                 //$contextchars = $user->getOption( 'contextchars', 50 );
 249                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 250                 $contextchars = 75; // same as above.... :P
 251                 return array($contextlines, $contextchars);
 252         }
 253
 254         /**
 255          * An array of namespaces indexes to be searched by default
 256          *
 257          * @return array
 258          * @static
 259          */
 260         public static function defaultNamespaces(){
 261                 global $wgNamespacesToBeSearchedDefault;
 262
 263                 return array_keys($wgNamespacesToBeSearchedDefault, true);
 264         }
 265
 266         /**
 267          * Return a 'cleaned up' search string
 268          *
 269          * @return string
 270          * @access public
 271          */
 272         function filter( $text ) {
 273                 $lc = $this->legalSearchChars();
 274                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 275         }
 276         /**
 277          * Load up the appropriate search engine class for the currently
 278          * active database backend, and return a configured instance.
 279          *
 280          * @return SearchEngine
 281          */
 282         public static function create() {
 283                 global $wgSearchType;
 284                 $dbr = wfGetDB( DB_SLAVE );
 285                 if( $wgSearchType ) {
 286                         $class = $wgSearchType;
 287                 } else {
 288                         $class = $dbr->getSearchEngine();
 289                 }
 290                 $search = new $class( $dbr );
 291                 $search->setLimitOffset(0,0);
 292                 return $search;
 293         }
 294
 295         /**
 296          * Create or update the search index record for the given page.
 297          * Title and text should be pre-processed.
 298          *
 299          * @param int $id
 300          * @param string $title
 301          * @param string $text
 302          * @abstract
 303          */
 304         function update( $id, $title, $text ) {
 305                 // no-op
 306         }
 307
 308         /**
 309          * Update a search index record's title only.
 310          * Title should be pre-processed.
 311          *
 312          * @param int $id
 313          * @param string $title
 314          * @abstract
 315          */
 316         function updateTitle( $id, $title ) {
 317                 // no-op
 318         }
 319
 320         /**
 321          * Get OpenSearch suggestion template
 322          *
 323          * @return string
 324          * @static
 325          */
 326         public static function getOpenSearchTemplate() {
 327                 global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
 328                 if($wgOpenSearchTemplate)
 329                         return $wgOpenSearchTemplate;
 330                 else{
 331                         $ns = implode(',',SearchEngine::defaultNamespaces());
 332                         if(!$ns) $ns = "0";
 333                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace='.$ns;
 334                 }
 335         }
 336
 337         /**
 338          * Get internal MediaWiki Suggest template
 339          *
 340          * @return string
 341          * @static
 342          */
 343         public static function getMWSuggestTemplate() {
 344                 global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
 345                 if($wgMWSuggestTemplate)
 346                         return $wgMWSuggestTemplate;
 347                 else
 348                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}';
 349         }
 350 }
 351
 352 /**
 353  * @ingroup Search
 354  */
 355 class SearchResultSet {
 356         /**
 357          * Fetch an array of regular expression fragments for matching
 358          * the search terms as parsed by this engine in a text extract.
 359          *
 360          * @return array
 361          * @access public
 362          * @abstract
 363          */
 364         function termMatches() {
 365                 return array();
 366         }
 367
 368         function numRows() {
 369                 return 0;
 370         }
 371
 372         /**
 373          * Return true if results are included in this result set.
 374          * @return bool
 375          * @abstract
 376          */
 377         function hasResults() {
 378                 return false;
 379         }
 380
 381         /**
 382          * Some search modes return a total hit count for the query
 383          * in the entire article database. This may include pages
 384          * in namespaces that would not be matched on the given
 385          * settings.
 386          *
 387          * Return null if no total hits number is supported.
 388          *
 389          * @return int
 390          * @access public
 391          */
 392         function getTotalHits() {
 393                 return null;
 394         }
 395
 396         /**
 397          * Some search modes return a suggested alternate term if there are
 398          * no exact hits. Returns true if there is one on this set.
 399          *
 400          * @return bool
 401          * @access public
 402          */
 403         function hasSuggestion() {
 404                 return false;
 405         }
 406
 407         /**
 408          * @return string suggested query, null if none
 409          */
 410         function getSuggestionQuery(){
 411                 return null;
 412         }
 413
 414         /**
 415          * @return string highlighted suggested query, '' if none
 416          */
 417         function getSuggestionSnippet(){
 418                 return '';
 419         }
 420
 421         /**
 422          * Return information about how and from where the results were fetched,
 423          * should be useful for diagnostics and debugging
 424          *
 425          * @return string
 426          */
 427         function getInfo() {
 428                 return null;
 429         }
 430
 431         /**
 432          * Return a result set of hits on other (multiple) wikis associated with this one
 433          *
 434          * @return SearchResultSet
 435          */
 436         function getInterwikiResults() {
 437                 return null;
 438         }
 439
 440         /**
 441          * Check if there are results on other wikis
 442          *
 443          * @return boolean
 444          */
 445         function hasInterwikiResults() {
 446                 return $this->getInterwikiResults() != null;
 447         }
 448
 449
 450         /**
 451          * Fetches next search result, or false.
 452          * @return SearchResult
 453          * @access public
 454          * @abstract
 455          */
 456         function next() {
 457                 return false;
 458         }
 459
 460         /**
 461          * Frees the result set, if applicable.
 462          * @ access public
 463          */
 464         function free() {
 465                 // ...
 466         }
 467 }
 468
 469
 470 /**
 471  * @ingroup Search
 472  */
 473 class SearchResultTooMany {
 474         ## Some search engines may bail out if too many matches are found
 475 }
 476
 477
 478 /**
 479  * @fixme This class is horribly factored. It would probably be better to have
 480  * a useful base class to which you pass some standard information, then let
 481  * the fancy self-highlighters extend that.
 482  * @ingroup Search
 483  */
 484 class SearchResult {
 485         var $mRevision = null;
 486
 487         function SearchResult( $row ) {
 488                 $this->mTitle = Title::makeTitle( $row->page_namespace, $row->page_title );
 489                 if( !is_null($this->mTitle) )
 490                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 491         }
 492
 493         /**
 494          * Check if this is result points to an invalid title
 495          *
 496          * @return boolean
 497          * @access public
 498          */
 499         function isBrokenTitle(){
 500                 if( is_null($this->mTitle) )
 501                         return true;
 502                 return false;
 503         }
 504
 505         /**
 506          * Check if target page is missing, happens when index is out of date
 507          *
 508          * @return boolean
 509          * @access public
 510          */
 511         function isMissingRevision(){
 512                 if( !$this->mRevision )
 513                         return true;
 514                 return false;
 515         }
 516
 517         /**
 518          * @return Title
 519          * @access public
 520          */
 521         function getTitle() {
 522                 return $this->mTitle;
 523         }
 524
 525         /**
 526          * @return double or null if not supported
 527          */
 528         function getScore() {
 529                 return null;
 530         }
 531
 532         /**
 533          * Lazy initialization of article text from DB
 534          */
 535         protected function initText(){
 536                 if( !isset($this->mText) ){
 537                         $this->mText = $this->mRevision->getText();
 538                 }
 539         }
 540
 541         /**
 542          * @param array $terms terms to highlight
 543          * @return string highlighted text snippet, null (and not '') if not supported
 544          */
 545         function getTextSnippet($terms){
 546                 global $wgUser, $wgAdvancedSearchHighlighting;
 547                 $this->initText();
 548                 list($contextlines,$contextchars) = SearchEngine::userHighlightPrefs($wgUser);
 549                 $h = new SearchHighlighter();
 550                 if( $wgAdvancedSearchHighlighting )
 551                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 552                 else
 553                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 554         }
 555
 556         /**
 557          * @param array $terms terms to highlight
 558          * @return string highlighted title, '' if not supported
 559          */
 560         function getTitleSnippet($terms){
 561                 return '';
 562         }
 563
 564         /**
 565          * @param array $terms terms to highlight
 566          * @return string highlighted redirect name (redirect to this page), '' if none or not supported
 567          */
 568         function getRedirectSnippet($terms){
 569                 return '';
 570         }
 571
 572         /**
 573          * @return Title object for the redirect to this page, null if none or not supported
 574          */
 575         function getRedirectTitle(){
 576                 return null;
 577         }
 578
 579         /**
 580          * @return string highlighted relevant section name, null if none or not supported
 581          */
 582         function getSectionSnippet(){
 583                 return '';
 584         }
 585
 586         /**
 587          * @return Title object (pagename+fragment) for the section, null if none or not supported
 588          */
 589         function getSectionTitle(){
 590                 return null;
 591         }
 592
 593         /**
 594          * @return string timestamp
 595          */
 596         function getTimestamp(){
 597                 return $this->mRevision->getTimestamp();
 598         }
 599
 600         /**
 601          * @return int number of words
 602          */
 603         function getWordCount(){
 604                 $this->initText();
 605                 return str_word_count( $this->mText );
 606         }
 607
 608         /**
 609          * @return int size in bytes
 610          */
 611         function getByteSize(){
 612                 $this->initText();
 613                 return strlen( $this->mText );
 614         }
 615
 616         /**
 617          * @return boolean if hit has related articles
 618          */
 619         function hasRelated(){
 620                 return false;
 621         }
 622
 623         /**
 624          * @return interwiki prefix of the title (return iw even if title is broken)
 625          */
 626         function getInterwikiPrefix(){
 627                 return '';
 628         }
 629 }
 630
 631 /**
 632  * Highlight bits of wikitext
 633  *
 634  * @ingroup Search
 635  */
 636 class SearchHighlighter {
 637         var $mCleanWikitext = true;
 638
 639         function SearchHighlighter($cleanupWikitext = true){
 640                 $this->mCleanWikitext = $cleanupWikitext;
 641         }
 642
 643         /**
 644          * Default implementation of wikitext highlighting
 645          *
 646          * @param string $text
 647          * @param array $terms Terms to highlight (unescaped)
 648          * @param int $contextlines
 649          * @param int $contextchars
 650          * @return string
 651          */
 652         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 653                 global $wgLang, $wgContLang;
 654                 global $wgSearchHighlightBoundaries;
 655                 $fname = __METHOD__;
 656
 657                 if($text == '')
 658                         return '';
 659
 660                 // spli text into text + templates/links/tables
 661                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 662                 // first capture group is for detecting nested templates/links/tables/references
 663                 $endPatterns = array(
 664                         1 => '/(\{\{)|(\}\})/', // template
 665                         2 => '/(\[\[)|(\]\])/', // image
 666                         3 => "/(\n\\{\\|)|(\n\\|\\})/"); // table
 667
 668                 // FIXME: this should prolly be a hook or something
 669                 if(function_exists('wfCite')){
 670                         $spat .= '|(<ref>)'; // references via cite extension
 671                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 672                 }
 673                 $spat .= '/';
 674                 $textExt = array(); // text extracts
 675                 $otherExt = array();  // other extracts
 676                 wfProfileIn( "$fname-split" );
 677                 $start = 0;
 678                 $textLen = strlen($text);
 679                 $count = 0; // sequence number to maintain ordering
 680                 while( $start < $textLen ){
 681                         // find start of template/image/table
 682                         if( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ){
 683                                 $epat = '';
 684                                 foreach($matches as $key => $val){
 685                                         if($key > 0 && $val[1] != -1){
 686                                                 if($key == 2){
 687                                                         // see if this is an image link
 688                                                         $ns = substr($val[0],2,-1);
 689                                                         if( $wgContLang->getNsIndex($ns) != NS_IMAGE )
 690                                                                 break;
 691
 692                                                 }
 693                                                 $epat = $endPatterns[$key];
 694                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 695                                                 $start = $val[1];
 696                                                 break;
 697                                         }
 698                                 }
 699                                 if( $epat ){
 700                                         // find end (and detect any nested elements)
 701                                         $level = 0;
 702                                         $offset = $start + 1;
 703                                         $found = false;
 704                                         while( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ){
 705                                                 if( array_key_exists(2,$endMatches) ){
 706                                                         // found end
 707                                                         if($level == 0){
 708                                                                 $len = strlen($endMatches[2][0]);
 709                                                                 $off = $endMatches[2][1];
 710                                                                 $this->splitAndAdd( $otherExt, $count,
 711                                                                         substr( $text, $start, $off + $len  - $start ) );
 712                                                                 $start = $off + $len;
 713                                                                 $found = true;
 714                                                                 break;
 715                                                         } else{
 716                                                                 // end of nested element
 717                                                                 $level -= 1;
 718                                                         }
 719                                                 } else{
 720                                                         // nested
 721                                                         $level += 1;
 722                                                 }
 723                                                 $offset = $endMatches[0][1] + strlen($endMatches[0][0]);
 724                                         }
 725                                         if( ! $found ){
 726                                                 // couldn't find appropriate closing tag, skip
 727                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen($matches[0][0]) ) );
 728                                                 $start += strlen($matches[0][0]);
 729                                         }
 730                                         continue;
 731                                 }
 732                         }
 733                         // else: add as text extract
 734                         $this->splitAndAdd( $textExt, $count, substr($text,$start) );
 735                         break;
 736                 }
 737
 738                 $all = $textExt + $otherExt; // these have disjunct key sets
 739
 740                 wfProfileOut( "$fname-split" );
 741
 742                 // prepare regexps
 743                 foreach( $terms as $index => $term ) {
 744                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 745                         if(preg_match('/[\x80-\xff]/', $term) ){
 746                                 $terms[$index] = preg_replace_callback('/./us',array($this,'caseCallback'),$terms[$index]);
 747                         } else {
 748                                 $terms[$index] = $term;
 749                         }
 750                 }
 751                 $anyterm = implode( '|', $terms );
 752                 $phrase = implode("$wgSearchHighlightBoundaries+", $terms );
 753
 754                 // FIXME: a hack to scale contextchars, a correct solution
 755                 // would be to have contextchars actually be char and not byte
 756                 // length, and do proper utf-8 substrings and lengths everywhere,
 757                 // but PHP is making that very hard and unclean to implement :(
 758                 $scale = strlen($anyterm) / mb_strlen($anyterm);
 759                 $contextchars = intval( $contextchars * $scale );
 760
 761                 $patPre = "(^|$wgSearchHighlightBoundaries)";
 762                 $patPost = "($wgSearchHighlightBoundaries|$)";
 763
 764                 $pat1 = "/(".$phrase.")/ui";
 765                 $pat2 = "/$patPre(".$anyterm.")$patPost/ui";
 766
 767                 wfProfileIn( "$fname-extract" );
 768
 769                 $left = $contextlines;
 770
 771                 $snippets = array();
 772                 $offsets = array();
 773
 774                 // show beginning only if it contains all words
 775                 $first = 0;
 776                 $firstText = '';
 777                 foreach($textExt as $index => $line){
 778                         if(strlen($line)>0 && $line[0] != ';' && $line[0] != ':'){
 779                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
 780                                 $first = $index;
 781                                 break;
 782                         }
 783                 }
 784                 if( $firstText ){
 785                         $succ = true;
 786                         // check if first text contains all terms
 787                         foreach($terms as $term){
 788                                 if( ! preg_match("/$patPre".$term."$patPost/ui", $firstText) ){
 789                                         $succ = false;
 790                                         break;
 791                                 }
 792                         }
 793                         if( $succ ){
 794                                 $snippets[$first] = $firstText;
 795                                 $offsets[$first] = 0;
 796                         }
 797                 }
 798                 if( ! $snippets ) {
 799                         // match whole query on text
 800                         $this->process($pat1, $textExt, $left, $contextchars, $snippets, $offsets);
 801                         // match whole query on templates/tables/images
 802                         $this->process($pat1, $otherExt, $left, $contextchars, $snippets, $offsets);
 803                         // match any words on text
 804                         $this->process($pat2, $textExt, $left, $contextchars, $snippets, $offsets);
 805                         // match any words on templates/tables/images
 806                         $this->process($pat2, $otherExt, $left, $contextchars, $snippets, $offsets);
 807
 808                         ksort($snippets);
 809                 }
 810
 811                 // add extra chars to each snippet to make snippets constant size
 812                 $extended = array();
 813                 if( count( $snippets ) == 0){
 814                         // couldn't find the target words, just show beginning of article
 815                         $targetchars = $contextchars * $contextlines;
 816                         $snippets[$first] = '';
 817                         $offsets[$first] = 0;
 818                 } else{
 819                         // if begin of the article contains the whole phrase, show only that !!
 820                         if( array_key_exists($first,$snippets) && preg_match($pat1,$snippets[$first])
 821                             && $offsets[$first] < $contextchars * 2 ){
 822                                 $snippets = array ($first => $snippets[$first]);
 823                         }
 824
 825                         // calc by how much to extend existing snippets
 826                         $targetchars = intval( ($contextchars * $contextlines) / count ( $snippets ) );
 827                 }
 828
 829                 foreach($snippets as $index => $line){
 830                         $extended[$index] = $line;
 831                         $len = strlen($line);
 832                         if( $len < $targetchars - 20 ){
 833                                 // complete this line
 834                                 if($len < strlen( $all[$index] )){
 835                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index]+$targetchars, $offsets[$index]);
 836                                         $len = strlen( $extended[$index] );
 837                                 }
 838
 839                                 // add more lines
 840                                 $add = $index + 1;
 841                                 while( $len < $targetchars - 20
 842                                        && array_key_exists($add,$all)
 843                                        && !array_key_exists($add,$snippets) ){
 844                                     $offsets[$add] = 0;
 845                                     $tt = "\n".$this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
 846                                         $extended[$add] = $tt;
 847                                         $len += strlen( $tt );
 848                                         $add++;
 849                                 }
 850                         }
 851                 }
 852
 853                 //$snippets = array_map('htmlspecialchars', $extended);
 854                 $snippets = $extended;
 855                 $last = -1;
 856                 $extract = '';
 857                 foreach($snippets as $index => $line){
 858                         if($last == -1)
 859                                 $extract .= $line; // first line
 860                         elseif($last+1 == $index && $offsets[$last]+strlen($snippets[$last]) >= strlen($all[$last]))
 861                                 $extract .= " ".$line; // continous lines
 862                         else
 863                                 $extract .= '<b> ... </b>' . $line;
 864
 865                         $last = $index;
 866                 }
 867                 if( $extract )
 868                         $extract .= '<b> ... </b>';
 869
 870                 $processed = array();
 871                 foreach($terms as $term){
 872                         if( ! isset($processed[$term]) ){
 873                                 $pat3 = "/$patPre(".$term.")$patPost/ui"; // highlight word
 874                                 $extract = preg_replace( $pat3,
 875                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
 876                                 $processed[$term] = true;
 877                         }
 878                 }
 879
 880                 wfProfileOut( "$fname-extract" );
 881
 882                 return $extract;
 883         }
 884
 885         /**
 886          * Split text into lines and add it to extracts array
 887          *
 888          * @param array $extracts index -> $line
 889          * @param int $count
 890          * @param string $text
 891          */
 892         function splitAndAdd(&$extracts, &$count, $text){
 893                 $split = explode( "\n", $this->mCleanWikitext? $this->removeWiki($text) : $text );
 894                 foreach($split as $line){
 895                         $tt = trim($line);
 896                         if( $tt )
 897                                 $extracts[$count++] = $tt;
 898                 }
 899         }
 900
 901         /**
 902          * Do manual case conversion for non-ascii chars
 903          *
 904          * @param unknown_type $matches
 905          */
 906         function caseCallback($matches){
 907                 global $wgContLang;
 908                 if( strlen($matches[0]) > 1 ){
 909                         return '['.$wgContLang->lc($matches[0]).$wgContLang->uc($matches[0]).']';
 910                 } else
 911                         return $matches[0];
 912         }
 913
 914         /**
 915          * Extract part of the text from start to end, but by
 916          * not chopping up words
 917          * @param string $text
 918          * @param int $start
 919          * @param int $end
 920          * @param int $posStart (out) actual start position
 921          * @param int $posEnd (out) actual end position
 922          * @return string
 923          */
 924         function extract($text, $start, $end, &$posStart = null, &$posEnd = null ){
 925                 global $wgContLang;
 926
 927                 if( $start != 0)
 928                         $start = $this->position( $text, $start, 1 );
 929                 if( $end >= strlen($text) )
 930                         $end = strlen($text);
 931                 else
 932                         $end = $this->position( $text, $end );
 933
 934                 if(!is_null($posStart))
 935                         $posStart = $start;
 936                 if(!is_null($posEnd))
 937                         $posEnd = $end;
 938
 939                 if($end > $start)
 940                         return substr($text, $start, $end-$start);
 941                 else
 942                         return '';
 943         }
 944
 945         /**
 946          * Find a nonletter near a point (index) in the text
 947          *
 948          * @param string $text
 949          * @param int $point
 950          * @param int $offset to found index
 951          * @return int nearest nonletter index, or beginning of utf8 char if none
 952          */
 953         function position($text, $point, $offset=0 ){
 954                 $tolerance = 10;
 955                 $s = max( 0, $point - $tolerance );
 956                 $l = min( strlen($text), $point + $tolerance ) - $s;
 957                 $m = array();
 958                 if( preg_match('/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr($text,$s,$l), $m, PREG_OFFSET_CAPTURE ) ){
 959                         return $m[0][1] + $s + $offset;
 960                 } else{
 961                         // check if point is on a valid first UTF8 char
 962                         $char = ord( $text[$point] );
 963                         while( $char >= 0x80 && $char < 0xc0 ) {
 964                                 // skip trailing bytes
 965                                 $point++;
 966                                 if($point >= strlen($text))
 967                                         return strlen($text);
 968                                 $char = ord( $text[$point] );
 969                         }
 970                         return $point;
 971
 972                 }
 973         }
 974
 975         /**
 976          * Search extracts for a pattern, and return snippets
 977          *
 978          * @param string $pattern regexp for matching lines
 979          * @param array $extracts extracts to search
 980          * @param int $linesleft number of extracts to make
 981          * @param int $contextchars length of snippet
 982          * @param array $out map for highlighted snippets
 983          * @param array $offsets map of starting points of snippets
 984          * @protected
 985          */
 986         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ){
 987                 if($linesleft == 0)
 988                         return; // nothing to do
 989                 foreach($extracts as $index => $line){
 990                         if( array_key_exists($index,$out) )
 991                                 continue; // this line already highlighted
 992
 993                         $m = array();
 994                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
 995                                 continue;
 996
 997                         $offset = $m[0][1];
 998                         $len = strlen($m[0][0]);
 999                         if($offset + $len < $contextchars)
1000                                 $begin = 0;
1001                         elseif( $len > $contextchars)
1002                                 $begin = $offset;
1003                         else
1004                                 $begin = $offset + intval( ($len - $contextchars) / 2 );
1005
1006                         $end = $begin + $contextchars;
1007
1008                         $posBegin = $begin;
1009                         // basic snippet from this line
1010                         $out[$index] = $this->extract($line,$begin,$end,$posBegin);
1011                         $offsets[$index] = $posBegin;
1012                         $linesleft--;
1013                         if($linesleft == 0)
1014                                 return;
1015                 }
1016         }
1017
1018         /**
1019          * Basic wikitext removal
1020          * @protected
1021          */
1022         function removeWiki($text) {
1023                 $fname = __METHOD__;
1024                 wfProfileIn( $fname );
1025
1026                 //$text = preg_replace("/'{2,5}/", "", $text);
1027                 //$text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1028                 //$text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1029                 //$text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1030                 //$text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1031                 //$text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1032                 $text = preg_replace("/\\{\\{([^|]+?)\\}\\}/", "", $text);
1033                 $text = preg_replace("/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text);
1034                 $text = preg_replace("/\\[\\[([^|]+?)\\]\\]/", "\\1", $text);
1035                 $text = preg_replace_callback("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array($this,'linkReplace'), $text);
1036                 //$text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1037                 $text = preg_replace("/<\/?[^>]+>/", "", $text);
1038                 $text = preg_replace("/'''''/", "", $text);
1039                 $text = preg_replace("/('''|<\/?[iIuUbB]>)/", "", $text);
1040                 $text = preg_replace("/''/", "", $text);
1041
1042                 wfProfileOut( $fname );
1043                 return $text;
1044         }
1045
1046         /**
1047          * callback to replace [[target|caption]] kind of links, if
1048          * the target is category or image, leave it
1049          *
1050          * @param array $matches
1051          */
1052         function linkReplace($matches){
1053                 $colon = strpos( $matches[1], ':' );
1054                 if( $colon === false )
1055                         return $matches[2]; // replace with caption
1056                 global $wgContLang;
1057                 $ns = substr( $matches[1], 0, $colon );
1058                 $index = $wgContLang->getNsIndex($ns);
1059                 if( $index !== false && ($index == NS_IMAGE || $index == NS_CATEGORY) )
1060                         return $matches[0]; // return the whole thing
1061                 else
1062                         return $matches[2];
1063
1064         }
1065
1066         /**
1067      * Simple & fast snippet extraction, but gives completely unrelevant
1068      * snippets
1069      *
1070      * @param string $text
1071      * @param array $terms
1072      * @param int $contextlines
1073      * @param int $contextchars
1074      * @return string
1075      */
1076     public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1077         global $wgLang, $wgContLang;
1078         $fname = __METHOD__;
1079
1080         $lines = explode( "\n", $text );
1081
1082         $terms = implode( '|', $terms );
1083         $max = intval( $contextchars ) + 1;
1084         $pat1 = "/(.*)($terms)(.{0,$max})/i";
1085
1086         $lineno = 0;
1087
1088         $extract = "";
1089         wfProfileIn( "$fname-extract" );
1090         foreach ( $lines as $line ) {
1091             if ( 0 == $contextlines ) {
1092                 break;
1093             }
1094             ++$lineno;
1095             $m = array();
1096             if ( ! preg_match( $pat1, $line, $m ) ) {
1097                 continue;
1098             }
1099             --$contextlines;
1100             $pre = $wgContLang->truncate( $m[1], -$contextchars, ' ... ' );
1101
1102             if ( count( $m ) < 3 ) {
1103                 $post = '';
1104             } else {
1105                 $post = $wgContLang->truncate( $m[3], $contextchars, ' ... ' );
1106             }
1107
1108             $found = $m[2];
1109
1110             $line = htmlspecialchars( $pre . $found . $post );
1111             $pat2 = '/(' . $terms . ")/i";
1112             $line = preg_replace( $pat2,
1113               "<span class='searchmatch'>\\1</span>", $line );
1114
1115             $extract .= "${line}\n";
1116         }
1117         wfProfileOut( "$fname-extract" );
1118
1119         return $extract;
1120     }
1121
1122 }
1123
1124 /**
1125  * Dummy class to be used when non-supported Database engine is present.
1126  * @fixme Dummy class should probably try something at least mildly useful,
1127  * such as a LIKE search through titles.
1128  * @ingroup Search
1129  */
1130 class SearchEngineDummy extends SearchEngine {
1131         // no-op
1132 }