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 $prefix = '';
  17         var $searchTerms = array();
  18         var $namespaces = array( NS_MAIN );
  19         var $showRedirects = false;
  20
  21         /**
  22          * Perform a full text search query and return a result set.
  23          * If title searches are not supported or disabled, return null.
  24          * STUB
  25          *
  26          * @param $term String: raw search term
  27          * @return SearchResultSet
  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          * STUB
  37          *
  38          * @param $term String: raw search term
  39          * @return SearchResultSet
  40          */
  41         function searchTitle( $term ) {
  42                 return null;
  43         }
  44
  45         /** If this search backend can list/unlist redirects */
  46         function acceptListRedirects() {
  47                 return true;
  48         }
  49
  50         /**
  51          * Transform search term in cases when parts of the query came as different GET params (when supported)
  52          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
  53          */
  54         function transformSearchTerm( $term ) {
  55                 return $term;
  56         }
  57
  58         /**
  59          * If an exact title match can be find, or a very slightly close match,
  60          * return the title. If no match, returns NULL.
  61          *
  62          * @param $searchterm String
  63          * @return Title
  64          */
  65         public static function getNearMatch( $searchterm ) {
  66                 global $wgContLang;
  67
  68                 $allSearchTerms = array($searchterm);
  69
  70                 if($wgContLang->hasVariants()){
  71                         $allSearchTerms = array_merge($allSearchTerms,$wgContLang->convertLinkToAllVariants($searchterm));
  72                 }
  73
  74                 foreach($allSearchTerms as $term){
  75
  76                         # Exact match? No need to look further.
  77                         $title = Title::newFromText( $term );
  78                         if (is_null($title))
  79                                 return NULL;
  80
  81                         if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal() || $title->exists() ) {
  82                                 return $title;
  83                         }
  84
  85                         # See if it still otherwise has content is some sane sense
  86                         $article = MediaWiki::articleFromTitle( $title );
  87                         if( $article->hasViewableContent() ) {
  88                                 return $title;
  89                         }
  90
  91                         # Now try all lower case (i.e. first letter capitalized)
  92                         #
  93                         $title = Title::newFromText( $wgContLang->lc( $term ) );
  94                         if ( $title && $title->exists() ) {
  95                                 return $title;
  96                         }
  97
  98                         # Now try capitalized string
  99                         #
 100                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 101                         if ( $title && $title->exists() ) {
 102                                 return $title;
 103                         }
 104
 105                         # Now try all upper case
 106                         #
 107                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 108                         if ( $title && $title->exists() ) {
 109                                 return $title;
 110                         }
 111
 112                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 113                         $title = Title::newFromText( $wgContLang->ucwordbreaks($term) );
 114                         if ( $title && $title->exists() ) {
 115                                 return $title;
 116                         }
 117
 118                         // Give hooks a chance at better match variants
 119                         $title = null;
 120                         if( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 121                                 return $title;
 122                         }
 123                 }
 124
 125                 $title = Title::newFromText( $searchterm );
 126
 127                 # Entering an IP address goes to the contributions page
 128                 if ( ( $title->getNamespace() == NS_USER && User::isIP($title->getText() ) )
 129                         || User::isIP( trim( $searchterm ) ) ) {
 130                         return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 131                 }
 132
 133
 134                 # Entering a user goes to the user page whether it's there or not
 135                 if ( $title->getNamespace() == NS_USER ) {
 136                         return $title;
 137                 }
 138
 139                 # Go to images that exist even if there's no local page.
 140                 # There may have been a funny upload, or it may be on a shared
 141                 # file repository such as Wikimedia Commons.
 142                 if( $title->getNamespace() == NS_FILE ) {
 143                         $image = wfFindFile( $title );
 144                         if( $image ) {
 145                                 return $title;
 146                         }
 147                 }
 148
 149                 # MediaWiki namespace? Page may be "implied" if not customized.
 150                 # Just return it, with caps forced as the message system likes it.
 151                 if( $title->getNamespace() == NS_MEDIAWIKI ) {
 152                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 153                 }
 154
 155                 # Quoted term? Try without the quotes...
 156                 $matches = array();
 157                 if( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 158                         return SearchEngine::getNearMatch( $matches[1] );
 159                 }
 160
 161                 return NULL;
 162         }
 163
 164         public static function legalSearchChars() {
 165                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 166         }
 167
 168         /**
 169          * Set the maximum number of results to return
 170          * and how many to skip before returning the first.
 171          *
 172          * @param $limit Integer
 173          * @param $offset Integer
 174          */
 175         function setLimitOffset( $limit, $offset = 0 ) {
 176                 $this->limit = intval( $limit );
 177                 $this->offset = intval( $offset );
 178         }
 179
 180         /**
 181          * Set which namespaces the search should include.
 182          * Give an array of namespace index numbers.
 183          *
 184          * @param $namespaces Array
 185          */
 186         function setNamespaces( $namespaces ) {
 187                 $this->namespaces = $namespaces;
 188         }
 189
 190         /**
 191          * Parse some common prefixes: all (search everything)
 192          * or namespace names
 193          *
 194          * @param $query String
 195          */
 196         function replacePrefixes( $query ){
 197                 global $wgContLang;
 198
 199                 if( strpos($query,':') === false )
 200                         return $query; // nothing to do
 201
 202                 $parsed = $query;
 203                 $allkeyword = wfMsgForContent('searchall').":";
 204                 if( strncmp($query, $allkeyword, strlen($allkeyword)) == 0 ){
 205                         $this->namespaces = null;
 206                         $parsed = substr($query,strlen($allkeyword));
 207                 } else if( strpos($query,':') !== false ) {
 208                         $prefix = substr($query,0,strpos($query,':'));
 209                         $index = $wgContLang->getNsIndex($prefix);
 210                         if($index !== false){
 211                                 $this->namespaces = array($index);
 212                                 $parsed = substr($query,strlen($prefix)+1);
 213                         }
 214                 }
 215                 if(trim($parsed) == '')
 216                         return $query; // prefix was the whole query
 217
 218                 return $parsed;
 219         }
 220
 221         /**
 222          * Make a list of searchable namespaces and their canonical names.
 223          * @return Array
 224          */
 225         public static function searchableNamespaces() {
 226                 global $wgContLang;
 227                 $arr = array();
 228                 foreach( $wgContLang->getNamespaces() as $ns => $name ) {
 229                         if( $ns >= NS_MAIN ) {
 230                                 $arr[$ns] = $name;
 231                         }
 232                 }
 233                 return $arr;
 234         }
 235
 236         /**
 237          * Extract default namespaces to search from the given user's
 238          * settings, returning a list of index numbers.
 239          *
 240          * @param $user User
 241          * @return Array
 242          */
 243         public static function userNamespaces( $user ) {
 244                 $arr = Preferences::loadOldSearchNs( $user );
 245                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 246
 247                 $arr = array_intersect( $arr, array_keys($searchableNamespaces) ); // Filter
 248
 249                 return $arr;
 250         }
 251
 252         /**
 253          * Find snippet highlight settings for a given user
 254          *
 255          * @param $user User
 256          * @return Array contextlines, contextchars
 257          */
 258         public static function userHighlightPrefs( &$user ){
 259                 //$contextlines = $user->getOption( 'contextlines',  5 );
 260                 //$contextchars = $user->getOption( 'contextchars', 50 );
 261                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 262                 $contextchars = 75; // same as above.... :P
 263                 return array($contextlines, $contextchars);
 264         }
 265
 266         /**
 267          * An array of namespaces indexes to be searched by default
 268          *
 269          * @return Array
 270          */
 271         public static function defaultNamespaces(){
 272                 global $wgNamespacesToBeSearchedDefault;
 273
 274                 return array_keys($wgNamespacesToBeSearchedDefault, true);
 275         }
 276
 277         /**
 278          * Get a list of namespace names useful for showing in tooltips
 279          * and preferences
 280          *
 281          * @param $namespaces Array
 282          */
 283         public static function namespacesAsText( $namespaces ){
 284                 global $wgContLang;
 285
 286                 $formatted = array_map( array($wgContLang,'getFormattedNsText'), $namespaces );
 287                 foreach( $formatted as $key => $ns ){
 288                         if ( empty($ns) )
 289                                 $formatted[$key] = wfMsg( 'blanknamespace' );
 290                 }
 291                 return $formatted;
 292         }
 293
 294         /**
 295          * Return the help namespaces to be shown on Special:Search
 296          *
 297          * @return Array
 298          */
 299         public static function helpNamespaces() {
 300                 global $wgNamespacesToBeSearchedHelp;
 301
 302                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 303         }
 304
 305         /**
 306          * Return a 'cleaned up' search string
 307          *
 308          * @param $text String
 309          * @return String
 310          */
 311         function filter( $text ) {
 312                 $lc = $this->legalSearchChars();
 313                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 314         }
 315         /**
 316          * Load up the appropriate search engine class for the currently
 317          * active database backend, and return a configured instance.
 318          *
 319          * @return SearchEngine
 320          */
 321         public static function create() {
 322                 global $wgSearchType;
 323                 $dbr = wfGetDB( DB_SLAVE );
 324                 if( $wgSearchType ) {
 325                         $class = $wgSearchType;
 326                 } else {
 327                         $class = $dbr->getSearchEngine();
 328                 }
 329                 $search = new $class( $dbr );
 330                 $search->setLimitOffset(0,0);
 331                 return $search;
 332         }
 333
 334         /**
 335          * Create or update the search index record for the given page.
 336          * Title and text should be pre-processed.
 337          * STUB
 338          *
 339          * @param $id Integer
 340          * @param $title String
 341          * @param $text String
 342          */
 343         function update( $id, $title, $text ) {
 344                 // no-op
 345         }
 346
 347         /**
 348          * Update a search index record's title only.
 349          * Title should be pre-processed.
 350          * STUB
 351          *
 352          * @param $id Integer
 353          * @param $title String
 354          */
 355         function updateTitle( $id, $title ) {
 356                 // no-op
 357         }
 358
 359         /**
 360          * Get OpenSearch suggestion template
 361          *
 362          * @return String
 363          */
 364         public static function getOpenSearchTemplate() {
 365                 global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
 366                 if( $wgOpenSearchTemplate )     {
 367                         return $wgOpenSearchTemplate;
 368                 } else {
 369                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 370                         if( !$ns ) $ns = "0";
 371                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace='.$ns;
 372                 }
 373         }
 374
 375         /**
 376          * Get internal MediaWiki Suggest template
 377          *
 378          * @return String
 379          */
 380         public static function getMWSuggestTemplate() {
 381                 global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
 382                 if($wgMWSuggestTemplate)
 383                         return $wgMWSuggestTemplate;
 384                 else
 385                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
 386         }
 387 }
 388
 389 /**
 390  * @ingroup Search
 391  */
 392 class SearchResultSet {
 393         /**
 394          * Fetch an array of regular expression fragments for matching
 395          * the search terms as parsed by this engine in a text extract.
 396          * STUB
 397          *
 398          * @return Array
 399          */
 400         function termMatches() {
 401                 return array();
 402         }
 403
 404         function numRows() {
 405                 return 0;
 406         }
 407
 408         /**
 409          * Return true if results are included in this result set.
 410          * STUB
 411          *
 412          * @return Boolean
 413          */
 414         function hasResults() {
 415                 return false;
 416         }
 417
 418         /**
 419          * Some search modes return a total hit count for the query
 420          * in the entire article database. This may include pages
 421          * in namespaces that would not be matched on the given
 422          * settings.
 423          *
 424          * Return null if no total hits number is supported.
 425          *
 426          * @return Integer
 427          */
 428         function getTotalHits() {
 429                 return null;
 430         }
 431
 432         /**
 433          * Some search modes return a suggested alternate term if there are
 434          * no exact hits. Returns true if there is one on this set.
 435          *
 436          * @return Boolean
 437          */
 438         function hasSuggestion() {
 439                 return false;
 440         }
 441
 442         /**
 443          * @return String: suggested query, null if none
 444          */
 445         function getSuggestionQuery(){
 446                 return null;
 447         }
 448
 449         /**
 450          * @return String: HTML highlighted suggested query, '' if none
 451          */
 452         function getSuggestionSnippet(){
 453                 return '';
 454         }
 455
 456         /**
 457          * Return information about how and from where the results were fetched,
 458          * should be useful for diagnostics and debugging
 459          *
 460          * @return String
 461          */
 462         function getInfo() {
 463                 return null;
 464         }
 465
 466         /**
 467          * Return a result set of hits on other (multiple) wikis associated with this one
 468          *
 469          * @return SearchResultSet
 470          */
 471         function getInterwikiResults() {
 472                 return null;
 473         }
 474
 475         /**
 476          * Check if there are results on other wikis
 477          *
 478          * @return Boolean
 479          */
 480         function hasInterwikiResults() {
 481                 return $this->getInterwikiResults() != null;
 482         }
 483
 484
 485         /**
 486          * Fetches next search result, or false.
 487          * STUB
 488          *
 489          * @return SearchResult
 490          */
 491         function next() {
 492                 return false;
 493         }
 494
 495         /**
 496          * Frees the result set, if applicable.
 497          */
 498         function free() {
 499                 // ...
 500         }
 501 }
 502
 503
 504 /**
 505  * @ingroup Search
 506  */
 507 class SearchResultTooMany {
 508         ## Some search engines may bail out if too many matches are found
 509 }
 510
 511
 512 /**
 513  * @todo Fixme: This class is horribly factored. It would probably be better to
 514  * have a useful base class to which you pass some standard information, then
 515  * let the fancy self-highlighters extend that.
 516  * @ingroup Search
 517  */
 518 class SearchResult {
 519         var $mRevision = null;
 520         var $mImage = null;
 521
 522         function __construct( $row ) {
 523                 $this->mTitle = Title::makeTitle( $row->page_namespace, $row->page_title );
 524                 if( !is_null($this->mTitle) ){
 525                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 526                         if( $this->mTitle->getNamespace() === NS_FILE )
 527                                 $this->mImage = wfFindFile( $this->mTitle );
 528                 }
 529         }
 530
 531         /**
 532          * Check if this is result points to an invalid title
 533          *
 534          * @return Boolean
 535          */
 536         function isBrokenTitle(){
 537                 if( is_null($this->mTitle) )
 538                         return true;
 539                 return false;
 540         }
 541
 542         /**
 543          * Check if target page is missing, happens when index is out of date
 544          *
 545          * @return Boolean
 546          */
 547         function isMissingRevision(){
 548                 return !$this->mRevision && !$this->mImage;
 549         }
 550
 551         /**
 552          * @return Title
 553          */
 554         function getTitle() {
 555                 return $this->mTitle;
 556         }
 557
 558         /**
 559          * @return Double or null if not supported
 560          */
 561         function getScore() {
 562                 return null;
 563         }
 564
 565         /**
 566          * Lazy initialization of article text from DB
 567          */
 568         protected function initText(){
 569                 if( !isset($this->mText) ){
 570                         if($this->mRevision != null)
 571                                 $this->mText = $this->mRevision->getText();
 572                         else // TODO: can we fetch raw wikitext for commons images?
 573                                 $this->mText = '';
 574
 575                 }
 576         }
 577
 578         /**
 579          * @param $terms Array: terms to highlight
 580          * @return String: highlighted text snippet, null (and not '') if not supported
 581          */
 582         function getTextSnippet($terms){
 583                 global $wgUser, $wgAdvancedSearchHighlighting;
 584                 $this->initText();
 585                 list($contextlines,$contextchars) = SearchEngine::userHighlightPrefs($wgUser);
 586                 $h = new SearchHighlighter();
 587                 if( $wgAdvancedSearchHighlighting )
 588                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 589                 else
 590                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 591         }
 592
 593         /**
 594          * @param $terms Array: terms to highlight
 595          * @return String: highlighted title, '' if not supported
 596          */
 597         function getTitleSnippet($terms){
 598                 return '';
 599         }
 600
 601         /**
 602          * @param $terms Array: terms to highlight
 603          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 604          */
 605         function getRedirectSnippet($terms){
 606                 return '';
 607         }
 608
 609         /**
 610          * @return Title object for the redirect to this page, null if none or not supported
 611          */
 612         function getRedirectTitle(){
 613                 return null;
 614         }
 615
 616         /**
 617          * @return string highlighted relevant section name, null if none or not supported
 618          */
 619         function getSectionSnippet(){
 620                 return '';
 621         }
 622
 623         /**
 624          * @return Title object (pagename+fragment) for the section, null if none or not supported
 625          */
 626         function getSectionTitle(){
 627                 return null;
 628         }
 629
 630         /**
 631          * @return String: timestamp
 632          */
 633         function getTimestamp(){
 634                 if( $this->mRevision )
 635                         return $this->mRevision->getTimestamp();
 636                 else if( $this->mImage )
 637                         return $this->mImage->getTimestamp();
 638                 return '';
 639         }
 640
 641         /**
 642          * @return Integer: number of words
 643          */
 644         function getWordCount(){
 645                 $this->initText();
 646                 return str_word_count( $this->mText );
 647         }
 648
 649         /**
 650          * @return Integer: size in bytes
 651          */
 652         function getByteSize(){
 653                 $this->initText();
 654                 return strlen( $this->mText );
 655         }
 656
 657         /**
 658          * @return Boolean if hit has related articles
 659          */
 660         function hasRelated(){
 661                 return false;
 662         }
 663
 664         /**
 665          * @return String: interwiki prefix of the title (return iw even if title is broken)
 666          */
 667         function getInterwikiPrefix(){
 668                 return '';
 669         }
 670 }
 671
 672 /**
 673  * Highlight bits of wikitext
 674  *
 675  * @ingroup Search
 676  */
 677 class SearchHighlighter {
 678         var $mCleanWikitext = true;
 679
 680         function SearchHighlighter($cleanupWikitext = true){
 681                 $this->mCleanWikitext = $cleanupWikitext;
 682         }
 683
 684         /**
 685          * Default implementation of wikitext highlighting
 686          *
 687          * @param $text String
 688          * @param $terms Array: terms to highlight (unescaped)
 689          * @param $contextlines Integer
 690          * @param $contextchars Integer
 691          * @return String
 692          */
 693         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 694                 global $wgLang, $wgContLang;
 695                 global $wgSearchHighlightBoundaries;
 696                 $fname = __METHOD__;
 697
 698                 if($text == '')
 699                         return '';
 700
 701                 // spli text into text + templates/links/tables
 702                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 703                 // first capture group is for detecting nested templates/links/tables/references
 704                 $endPatterns = array(
 705                         1 => '/(\{\{)|(\}\})/', // template
 706                         2 => '/(\[\[)|(\]\])/', // image
 707                         3 => "/(\n\\{\\|)|(\n\\|\\})/"); // table
 708
 709                 // FIXME: this should prolly be a hook or something
 710                 if(function_exists('wfCite')){
 711                         $spat .= '|(<ref>)'; // references via cite extension
 712                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 713                 }
 714                 $spat .= '/';
 715                 $textExt = array(); // text extracts
 716                 $otherExt = array();  // other extracts
 717                 wfProfileIn( "$fname-split" );
 718                 $start = 0;
 719                 $textLen = strlen($text);
 720                 $count = 0; // sequence number to maintain ordering
 721                 while( $start < $textLen ){
 722                         // find start of template/image/table
 723                         if( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ){
 724                                 $epat = '';
 725                                 foreach($matches as $key => $val){
 726                                         if($key > 0 && $val[1] != -1){
 727                                                 if($key == 2){
 728                                                         // see if this is an image link
 729                                                         $ns = substr($val[0],2,-1);
 730                                                         if( $wgContLang->getNsIndex($ns) != NS_FILE )
 731                                                                 break;
 732
 733                                                 }
 734                                                 $epat = $endPatterns[$key];
 735                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 736                                                 $start = $val[1];
 737                                                 break;
 738                                         }
 739                                 }
 740                                 if( $epat ){
 741                                         // find end (and detect any nested elements)
 742                                         $level = 0;
 743                                         $offset = $start + 1;
 744                                         $found = false;
 745                                         while( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ){
 746                                                 if( array_key_exists(2,$endMatches) ){
 747                                                         // found end
 748                                                         if($level == 0){
 749                                                                 $len = strlen($endMatches[2][0]);
 750                                                                 $off = $endMatches[2][1];
 751                                                                 $this->splitAndAdd( $otherExt, $count,
 752                                                                         substr( $text, $start, $off + $len  - $start ) );
 753                                                                 $start = $off + $len;
 754                                                                 $found = true;
 755                                                                 break;
 756                                                         } else{
 757                                                                 // end of nested element
 758                                                                 $level -= 1;
 759                                                         }
 760                                                 } else{
 761                                                         // nested
 762                                                         $level += 1;
 763                                                 }
 764                                                 $offset = $endMatches[0][1] + strlen($endMatches[0][0]);
 765                                         }
 766                                         if( ! $found ){
 767                                                 // couldn't find appropriate closing tag, skip
 768                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen($matches[0][0]) ) );
 769                                                 $start += strlen($matches[0][0]);
 770                                         }
 771                                         continue;
 772                                 }
 773                         }
 774                         // else: add as text extract
 775                         $this->splitAndAdd( $textExt, $count, substr($text,$start) );
 776                         break;
 777                 }
 778
 779                 $all = $textExt + $otherExt; // these have disjunct key sets
 780
 781                 wfProfileOut( "$fname-split" );
 782
 783                 // prepare regexps
 784                 foreach( $terms as $index => $term ) {
 785                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 786                         if(preg_match('/[\x80-\xff]/', $term) ){
 787                                 $terms[$index] = preg_replace_callback('/./us',array($this,'caseCallback'),$terms[$index]);
 788                         } else {
 789                                 $terms[$index] = $term;
 790                         }
 791                 }
 792                 $anyterm = implode( '|', $terms );
 793                 $phrase = implode("$wgSearchHighlightBoundaries+", $terms );
 794
 795                 // FIXME: a hack to scale contextchars, a correct solution
 796                 // would be to have contextchars actually be char and not byte
 797                 // length, and do proper utf-8 substrings and lengths everywhere,
 798                 // but PHP is making that very hard and unclean to implement :(
 799                 $scale = strlen($anyterm) / mb_strlen($anyterm);
 800                 $contextchars = intval( $contextchars * $scale );
 801
 802                 $patPre = "(^|$wgSearchHighlightBoundaries)";
 803                 $patPost = "($wgSearchHighlightBoundaries|$)";
 804
 805                 $pat1 = "/(".$phrase.")/ui";
 806                 $pat2 = "/$patPre(".$anyterm.")$patPost/ui";
 807
 808                 wfProfileIn( "$fname-extract" );
 809
 810                 $left = $contextlines;
 811
 812                 $snippets = array();
 813                 $offsets = array();
 814
 815                 // show beginning only if it contains all words
 816                 $first = 0;
 817                 $firstText = '';
 818                 foreach($textExt as $index => $line){
 819                         if(strlen($line)>0 && $line[0] != ';' && $line[0] != ':'){
 820                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
 821                                 $first = $index;
 822                                 break;
 823                         }
 824                 }
 825                 if( $firstText ){
 826                         $succ = true;
 827                         // check if first text contains all terms
 828                         foreach($terms as $term){
 829                                 if( ! preg_match("/$patPre".$term."$patPost/ui", $firstText) ){
 830                                         $succ = false;
 831                                         break;
 832                                 }
 833                         }
 834                         if( $succ ){
 835                                 $snippets[$first] = $firstText;
 836                                 $offsets[$first] = 0;
 837                         }
 838                 }
 839                 if( ! $snippets ) {
 840                         // match whole query on text
 841                         $this->process($pat1, $textExt, $left, $contextchars, $snippets, $offsets);
 842                         // match whole query on templates/tables/images
 843                         $this->process($pat1, $otherExt, $left, $contextchars, $snippets, $offsets);
 844                         // match any words on text
 845                         $this->process($pat2, $textExt, $left, $contextchars, $snippets, $offsets);
 846                         // match any words on templates/tables/images
 847                         $this->process($pat2, $otherExt, $left, $contextchars, $snippets, $offsets);
 848
 849                         ksort($snippets);
 850                 }
 851
 852                 // add extra chars to each snippet to make snippets constant size
 853                 $extended = array();
 854                 if( count( $snippets ) == 0){
 855                         // couldn't find the target words, just show beginning of article
 856                         $targetchars = $contextchars * $contextlines;
 857                         $snippets[$first] = '';
 858                         $offsets[$first] = 0;
 859                 } else{
 860                         // if begin of the article contains the whole phrase, show only that !!
 861                         if( array_key_exists($first,$snippets) && preg_match($pat1,$snippets[$first])
 862                             && $offsets[$first] < $contextchars * 2 ){
 863                                 $snippets = array ($first => $snippets[$first]);
 864                         }
 865
 866                         // calc by how much to extend existing snippets
 867                         $targetchars = intval( ($contextchars * $contextlines) / count ( $snippets ) );
 868                 }
 869
 870                 foreach($snippets as $index => $line){
 871                         $extended[$index] = $line;
 872                         $len = strlen($line);
 873                         if( $len < $targetchars - 20 ){
 874                                 // complete this line
 875                                 if($len < strlen( $all[$index] )){
 876                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index]+$targetchars, $offsets[$index]);
 877                                         $len = strlen( $extended[$index] );
 878                                 }
 879
 880                                 // add more lines
 881                                 $add = $index + 1;
 882                                 while( $len < $targetchars - 20
 883                                        && array_key_exists($add,$all)
 884                                        && !array_key_exists($add,$snippets) ){
 885                                     $offsets[$add] = 0;
 886                                     $tt = "\n".$this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
 887                                         $extended[$add] = $tt;
 888                                         $len += strlen( $tt );
 889                                         $add++;
 890                                 }
 891                         }
 892                 }
 893
 894                 //$snippets = array_map('htmlspecialchars', $extended);
 895                 $snippets = $extended;
 896                 $last = -1;
 897                 $extract = '';
 898                 foreach($snippets as $index => $line){
 899                         if($last == -1)
 900                                 $extract .= $line; // first line
 901                         elseif($last+1 == $index && $offsets[$last]+strlen($snippets[$last]) >= strlen($all[$last]))
 902                                 $extract .= " ".$line; // continous lines
 903                         else
 904                                 $extract .= '<b> ... </b>' . $line;
 905
 906                         $last = $index;
 907                 }
 908                 if( $extract )
 909                         $extract .= '<b> ... </b>';
 910
 911                 $processed = array();
 912                 foreach($terms as $term){
 913                         if( ! isset($processed[$term]) ){
 914                                 $pat3 = "/$patPre(".$term.")$patPost/ui"; // highlight word
 915                                 $extract = preg_replace( $pat3,
 916                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
 917                                 $processed[$term] = true;
 918                         }
 919                 }
 920
 921                 wfProfileOut( "$fname-extract" );
 922
 923                 return $extract;
 924         }
 925
 926         /**
 927          * Split text into lines and add it to extracts array
 928          *
 929          * @param $extracts Array: index -> $line
 930          * @param $count Integer
 931          * @param $text String
 932          */
 933         function splitAndAdd(&$extracts, &$count, $text){
 934                 $split = explode( "\n", $this->mCleanWikitext? $this->removeWiki($text) : $text );
 935                 foreach($split as $line){
 936                         $tt = trim($line);
 937                         if( $tt )
 938                                 $extracts[$count++] = $tt;
 939                 }
 940         }
 941
 942         /**
 943          * Do manual case conversion for non-ascii chars
 944          *
 945          * @param $matches Array
 946          */
 947         function caseCallback($matches){
 948                 global $wgContLang;
 949                 if( strlen($matches[0]) > 1 ){
 950                         return '['.$wgContLang->lc($matches[0]).$wgContLang->uc($matches[0]).']';
 951                 } else
 952                         return $matches[0];
 953         }
 954
 955         /**
 956          * Extract part of the text from start to end, but by
 957          * not chopping up words
 958          * @param $text String
 959          * @param $start Integer
 960          * @param $end Integer
 961          * @param $posStart Integer: (out) actual start position
 962          * @param $posEnd Integer: (out) actual end position
 963          * @return String
 964          */
 965         function extract($text, $start, $end, &$posStart = null, &$posEnd = null ){
 966                 global $wgContLang;
 967
 968                 if( $start != 0)
 969                         $start = $this->position( $text, $start, 1 );
 970                 if( $end >= strlen($text) )
 971                         $end = strlen($text);
 972                 else
 973                         $end = $this->position( $text, $end );
 974
 975                 if(!is_null($posStart))
 976                         $posStart = $start;
 977                 if(!is_null($posEnd))
 978                         $posEnd = $end;
 979
 980                 if($end > $start)
 981                         return substr($text, $start, $end-$start);
 982                 else
 983                         return '';
 984         }
 985
 986         /**
 987          * Find a nonletter near a point (index) in the text
 988          *
 989          * @param $text String
 990          * @param $point Integer
 991          * @param $offset Integer: offset to found index
 992          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
 993          */
 994         function position($text, $point, $offset=0 ){
 995                 $tolerance = 10;
 996                 $s = max( 0, $point - $tolerance );
 997                 $l = min( strlen($text), $point + $tolerance ) - $s;
 998                 $m = array();
 999                 if( preg_match('/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr($text,$s,$l), $m, PREG_OFFSET_CAPTURE ) ){
1000                         return $m[0][1] + $s + $offset;
1001                 } else{
1002                         // check if point is on a valid first UTF8 char
1003                         $char = ord( $text[$point] );
1004                         while( $char >= 0x80 && $char < 0xc0 ) {
1005                                 // skip trailing bytes
1006                                 $point++;
1007                                 if($point >= strlen($text))
1008                                         return strlen($text);
1009                                 $char = ord( $text[$point] );
1010                         }
1011                         return $point;
1012
1013                 }
1014         }
1015
1016         /**
1017          * Search extracts for a pattern, and return snippets
1018          *
1019          * @param $pattern String: regexp for matching lines
1020          * @param $extracts Array: extracts to search
1021          * @param $linesleft Integer: number of extracts to make
1022          * @param $contextchars Integer: length of snippet
1023          * @param $out Array: map for highlighted snippets
1024          * @param $offsets Array: map of starting points of snippets
1025          * @protected
1026          */
1027         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ){
1028                 if($linesleft == 0)
1029                         return; // nothing to do
1030                 foreach($extracts as $index => $line){
1031                         if( array_key_exists($index,$out) )
1032                                 continue; // this line already highlighted
1033
1034                         $m = array();
1035                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1036                                 continue;
1037
1038                         $offset = $m[0][1];
1039                         $len = strlen($m[0][0]);
1040                         if($offset + $len < $contextchars)
1041                                 $begin = 0;
1042                         elseif( $len > $contextchars)
1043                                 $begin = $offset;
1044                         else
1045                                 $begin = $offset + intval( ($len - $contextchars) / 2 );
1046
1047                         $end = $begin + $contextchars;
1048
1049                         $posBegin = $begin;
1050                         // basic snippet from this line
1051                         $out[$index] = $this->extract($line,$begin,$end,$posBegin);
1052                         $offsets[$index] = $posBegin;
1053                         $linesleft--;
1054                         if($linesleft == 0)
1055                                 return;
1056                 }
1057         }
1058
1059         /**
1060          * Basic wikitext removal
1061          * @protected
1062          */
1063         function removeWiki($text) {
1064                 $fname = __METHOD__;
1065                 wfProfileIn( $fname );
1066
1067                 //$text = preg_replace("/'{2,5}/", "", $text);
1068                 //$text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1069                 //$text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1070                 //$text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1071                 //$text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1072                 //$text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1073                 $text = preg_replace("/\\{\\{([^|]+?)\\}\\}/", "", $text);
1074                 $text = preg_replace("/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text);
1075                 $text = preg_replace("/\\[\\[([^|]+?)\\]\\]/", "\\1", $text);
1076                 $text = preg_replace_callback("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array($this,'linkReplace'), $text);
1077                 //$text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1078                 $text = preg_replace("/<\/?[^>]+>/", "", $text);
1079                 $text = preg_replace("/'''''/", "", $text);
1080                 $text = preg_replace("/('''|<\/?[iIuUbB]>)/", "", $text);
1081                 $text = preg_replace("/''/", "", $text);
1082
1083                 wfProfileOut( $fname );
1084                 return $text;
1085         }
1086
1087         /**
1088          * callback to replace [[target|caption]] kind of links, if
1089          * the target is category or image, leave it
1090          *
1091          * @param $matches Array
1092          */
1093         function linkReplace($matches){
1094                 $colon = strpos( $matches[1], ':' );
1095                 if( $colon === false )
1096                         return $matches[2]; // replace with caption
1097                 global $wgContLang;
1098                 $ns = substr( $matches[1], 0, $colon );
1099                 $index = $wgContLang->getNsIndex($ns);
1100                 if( $index !== false && ($index == NS_FILE || $index == NS_CATEGORY) )
1101                         return $matches[0]; // return the whole thing
1102                 else
1103                         return $matches[2];
1104
1105         }
1106
1107         /**
1108      * Simple & fast snippet extraction, but gives completely unrelevant
1109      * snippets
1110      *
1111      * @param $text String
1112      * @param $terms Array
1113      * @param $contextlines Integer
1114      * @param $contextchars Integer
1115      * @return String
1116      */
1117     public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1118         global $wgLang, $wgContLang;
1119         $fname = __METHOD__;
1120
1121         $lines = explode( "\n", $text );
1122
1123         $terms = implode( '|', $terms );
1124         $max = intval( $contextchars ) + 1;
1125         $pat1 = "/(.*)($terms)(.{0,$max})/i";
1126
1127         $lineno = 0;
1128
1129         $extract = "";
1130         wfProfileIn( "$fname-extract" );
1131         foreach ( $lines as $line ) {
1132             if ( 0 == $contextlines ) {
1133                 break;
1134             }
1135             ++$lineno;
1136             $m = array();
1137             if ( ! preg_match( $pat1, $line, $m ) ) {
1138                 continue;
1139             }
1140             --$contextlines;
1141             $pre = $wgContLang->truncate( $m[1], -$contextchars );
1142
1143             if ( count( $m ) < 3 ) {
1144                 $post = '';
1145             } else {
1146                 $post = $wgContLang->truncate( $m[3], $contextchars );
1147             }
1148
1149             $found = $m[2];
1150
1151             $line = htmlspecialchars( $pre . $found . $post );
1152             $pat2 = '/(' . $terms . ")/i";
1153             $line = preg_replace( $pat2,
1154               "<span class='searchmatch'>\\1</span>", $line );
1155
1156             $extract .= "${line}\n";
1157         }
1158         wfProfileOut( "$fname-extract" );
1159
1160         return $extract;
1161     }
1162
1163 }
1164
1165 /**
1166  * Dummy class to be used when non-supported Database engine is present.
1167  * @todo Fixme: dummy class should probably try something at least mildly useful,
1168  * such as a LIKE search through titles.
1169  * @ingroup Search
1170  */
1171 class SearchEngineDummy extends SearchEngine {
1172         // no-op
1173 }