includes/search/SearchEngine.php

   1 <?php
   2 /**
   3  * Basic search engine
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup Search
  22  */
  23
  24 /**
  25  * @defgroup Search Search
  26  */
  27
  28 /**
  29  * Contain a class for special pages
  30  * @ingroup Search
  31  */
  32 class SearchEngine {
  33         var $limit = 10;
  34         var $offset = 0;
  35         var $prefix = '';
  36         var $searchTerms = array();
  37         var $namespaces = array( NS_MAIN );
  38         var $showRedirects = false;
  39         protected $showSuggestion = true;
  40
  41         /// Feature values
  42         protected $features = array();
  43
  44         /**
  45          * Perform a full text search query and return a result set.
  46          * If title searches are not supported or disabled, return null.
  47          * STUB
  48          *
  49          * @param string $term raw search term
  50          * @return SearchResultSet|Status|null
  51          */
  52         function searchText( $term ) {
  53                 return null;
  54         }
  55
  56         /**
  57          * Perform a title-only search query and return a result set.
  58          * If title searches are not supported or disabled, return null.
  59          * STUB
  60          *
  61          * @param string $term raw search term
  62          * @return SearchResultSet|null
  63          */
  64         function searchTitle( $term ) {
  65                 return null;
  66         }
  67
  68         /**
  69          * @since 1.18
  70          * @param $feature String
  71          * @return Boolean
  72          */
  73         public function supports( $feature ) {
  74                 switch ( $feature ) {
  75                 case 'list-redirects':
  76                 case 'search-update':
  77                         return true;
  78                 case 'title-suffix-filter':
  79                 default:
  80                         return false;
  81                 }
  82         }
  83
  84         /**
  85          * Way to pass custom data for engines
  86          * @since 1.18
  87          * @param $feature String
  88          * @param $data Mixed
  89          * @return bool
  90          */
  91         public function setFeatureData( $feature, $data ) {
  92                 $this->features[$feature] = $data;
  93         }
  94
  95         /**
  96          * When overridden in derived class, performs database-specific conversions
  97          * on text to be used for searching or updating search index.
  98          * Default implementation does nothing (simply returns $string).
  99          *
 100          * @param string $string String to process
 101          * @return string
 102          */
 103         public function normalizeText( $string ) {
 104                 global $wgContLang;
 105
 106                 // Some languages such as Chinese require word segmentation
 107                 return $wgContLang->segmentByWord( $string );
 108         }
 109
 110         /**
 111          * Transform search term in cases when parts of the query came as different GET params (when supported)
 112          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
 113          */
 114         function transformSearchTerm( $term ) {
 115                 return $term;
 116         }
 117
 118         /**
 119          * If an exact title match can be found, or a very slightly close match,
 120          * return the title. If no match, returns NULL.
 121          *
 122          * @param $searchterm String
 123          * @return Title
 124          */
 125         public static function getNearMatch( $searchterm ) {
 126                 $title = self::getNearMatchInternal( $searchterm );
 127
 128                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
 129                 return $title;
 130         }
 131
 132         /**
 133          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 134          * SearchResultSet.
 135          *
 136          * @param $searchterm string
 137          * @return SearchResultSet
 138          */
 139         public static function getNearMatchResultSet( $searchterm ) {
 140                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 141         }
 142
 143         /**
 144          * Really find the title match.
 145          * @return null|Title
 146          */
 147         private static function getNearMatchInternal( $searchterm ) {
 148                 global $wgContLang, $wgEnableSearchContributorsByIP;
 149
 150                 $allSearchTerms = array( $searchterm );
 151
 152                 if ( $wgContLang->hasVariants() ) {
 153                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 154                 }
 155
 156                 $titleResult = null;
 157                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 158                         return $titleResult;
 159                 }
 160
 161                 foreach ( $allSearchTerms as $term ) {
 162
 163                         # Exact match? No need to look further.
 164                         $title = Title::newFromText( $term );
 165                         if ( is_null( $title ) ) {
 166                                 return null;
 167                         }
 168
 169                         # Try files if searching in the Media: namespace
 170                         if ( $title->getNamespace() == NS_MEDIA ) {
 171                                 $title = Title::makeTitle( NS_FILE, $title->getText() );
 172                         }
 173
 174                         if ( $title->isSpecialPage() || $title->isExternal() || $title->exists() ) {
 175                                 return $title;
 176                         }
 177
 178                         # See if it still otherwise has content is some sane sense
 179                         $page = WikiPage::factory( $title );
 180                         if ( $page->hasViewableContent() ) {
 181                                 return $title;
 182                         }
 183
 184                         if ( !wfRunHooks( 'SearchAfterNoDirectMatch', array( $term, &$title ) ) ) {
 185                                 return $title;
 186                         }
 187
 188                         # Now try all lower case (i.e. first letter capitalized)
 189                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 190                         if ( $title && $title->exists() ) {
 191                                 return $title;
 192                         }
 193
 194                         # Now try capitalized string
 195                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 196                         if ( $title && $title->exists() ) {
 197                                 return $title;
 198                         }
 199
 200                         # Now try all upper case
 201                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 202                         if ( $title && $title->exists() ) {
 203                                 return $title;
 204                         }
 205
 206                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 207                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 208                         if ( $title && $title->exists() ) {
 209                                 return $title;
 210                         }
 211
 212                         // Give hooks a chance at better match variants
 213                         $title = null;
 214                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 215                                 return $title;
 216                         }
 217                 }
 218
 219                 $title = Title::newFromText( $searchterm );
 220
 221                 # Entering an IP address goes to the contributions page
 222                 if ( $wgEnableSearchContributorsByIP ) {
 223                         if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 224                                 || User::isIP( trim( $searchterm ) ) ) {
 225                                 return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 226                         }
 227                 }
 228
 229                 # Entering a user goes to the user page whether it's there or not
 230                 if ( $title->getNamespace() == NS_USER ) {
 231                         return $title;
 232                 }
 233
 234                 # Go to images that exist even if there's no local page.
 235                 # There may have been a funny upload, or it may be on a shared
 236                 # file repository such as Wikimedia Commons.
 237                 if ( $title->getNamespace() == NS_FILE ) {
 238                         $image = wfFindFile( $title );
 239                         if ( $image ) {
 240                                 return $title;
 241                         }
 242                 }
 243
 244                 # MediaWiki namespace? Page may be "implied" if not customized.
 245                 # Just return it, with caps forced as the message system likes it.
 246                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 247                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 248                 }
 249
 250                 # Quoted term? Try without the quotes...
 251                 $matches = array();
 252                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 253                         return SearchEngine::getNearMatch( $matches[1] );
 254                 }
 255
 256                 return null;
 257         }
 258
 259         public static function legalSearchChars() {
 260                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 261         }
 262
 263         /**
 264          * Set the maximum number of results to return
 265          * and how many to skip before returning the first.
 266          *
 267          * @param $limit Integer
 268          * @param $offset Integer
 269          */
 270         function setLimitOffset( $limit, $offset = 0 ) {
 271                 $this->limit = intval( $limit );
 272                 $this->offset = intval( $offset );
 273         }
 274
 275         /**
 276          * Set which namespaces the search should include.
 277          * Give an array of namespace index numbers.
 278          *
 279          * @param $namespaces Array
 280          */
 281         function setNamespaces( $namespaces ) {
 282                 $this->namespaces = $namespaces;
 283         }
 284
 285         /**
 286          * Set whether the searcher should try to build a suggestion.  Note: some searchers
 287          * don't support building a suggestion in the first place and others don't respect
 288          * this flag.
 289          *
 290          * @param boolean $showSuggestion should the searcher try to build suggestions
 291          */
 292         function setShowSuggestion( $showSuggestion ) {
 293                 $this->showSuggestion = $showSuggestion;
 294         }
 295
 296         /**
 297          * Parse some common prefixes: all (search everything)
 298          * or namespace names
 299          *
 300          * @param $query String
 301          * @return string
 302          */
 303         function replacePrefixes( $query ) {
 304                 global $wgContLang;
 305
 306                 $parsed = $query;
 307                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 308                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 309                         return $parsed;
 310                 }
 311
 312                 $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
 313                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 314                         $this->namespaces = null;
 315                         $parsed = substr( $query, strlen( $allkeyword ) );
 316                 } elseif ( strpos( $query, ':' ) !== false ) {
 317                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 318                         $index = $wgContLang->getNsIndex( $prefix );
 319                         if ( $index !== false ) {
 320                                 $this->namespaces = array( $index );
 321                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 322                         }
 323                 }
 324                 if ( trim( $parsed ) == '' ) {
 325                         $parsed = $query; // prefix was the whole query
 326                 }
 327
 328                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 329
 330                 return $parsed;
 331         }
 332
 333         /**
 334          * Make a list of searchable namespaces and their canonical names.
 335          * @return Array
 336          */
 337         public static function searchableNamespaces() {
 338                 global $wgContLang;
 339                 $arr = array();
 340                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 341                         if ( $ns >= NS_MAIN ) {
 342                                 $arr[$ns] = $name;
 343                         }
 344                 }
 345
 346                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 347                 return $arr;
 348         }
 349
 350         /**
 351          * Extract default namespaces to search from the given user's
 352          * settings, returning a list of index numbers.
 353          *
 354          * @param $user User
 355          * @return Array
 356          */
 357         public static function userNamespaces( $user ) {
 358                 global $wgSearchEverythingOnlyLoggedIn;
 359
 360                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 361
 362                 // get search everything preference, that can be set to be read for logged-in users
 363                 // it overrides other options
 364                 if ( !$wgSearchEverythingOnlyLoggedIn || $user->isLoggedIn() ) {
 365                         if ( $user->getOption( 'searcheverything' ) ) {
 366                                 return array_keys( $searchableNamespaces );
 367                         }
 368                 }
 369
 370                 $arr = array();
 371                 foreach ( $searchableNamespaces as $ns => $name ) {
 372                         if ( $user->getOption( 'searchNs' . $ns ) ) {
 373                                 $arr[] = $ns;
 374                         }
 375                 }
 376
 377                 return $arr;
 378         }
 379
 380         /**
 381          * Find snippet highlight settings for all users
 382          *
 383          * @return Array contextlines, contextchars
 384          */
 385         public static function userHighlightPrefs() {
 386                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 387                 $contextchars = 75; // same as above.... :P
 388                 return array( $contextlines, $contextchars );
 389         }
 390
 391         /**
 392          * An array of namespaces indexes to be searched by default
 393          *
 394          * @return Array
 395          */
 396         public static function defaultNamespaces() {
 397                 global $wgNamespacesToBeSearchedDefault;
 398
 399                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 400         }
 401
 402         /**
 403          * Get a list of namespace names useful for showing in tooltips
 404          * and preferences
 405          *
 406          * @param $namespaces Array
 407          * @return array
 408          */
 409         public static function namespacesAsText( $namespaces ) {
 410                 global $wgContLang;
 411
 412                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 413                 foreach ( $formatted as $key => $ns ) {
 414                         if ( empty( $ns ) ) {
 415                                 $formatted[$key] = wfMessage( 'blanknamespace' )->text();
 416                         }
 417                 }
 418                 return $formatted;
 419         }
 420
 421         /**
 422          * Return the help namespaces to be shown on Special:Search
 423          *
 424          * @return Array
 425          */
 426         public static function helpNamespaces() {
 427                 global $wgNamespacesToBeSearchedHelp;
 428
 429                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 430         }
 431
 432         /**
 433          * Return a 'cleaned up' search string
 434          *
 435          * @param $text String
 436          * @return String
 437          */
 438         function filter( $text ) {
 439                 $lc = $this->legalSearchChars();
 440                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 441         }
 442         /**
 443          * Load up the appropriate search engine class for the currently
 444          * active database backend, and return a configured instance.
 445          *
 446          * @param String $type Type of search backend, if not the default
 447          * @return SearchEngine
 448          */
 449         public static function create( $type = null ) {
 450                 global $wgSearchType;
 451                 $dbr = null;
 452
 453                 $alternatives = self::getSearchTypes();
 454
 455                 if ( $type && in_array( $type, $alternatives ) ) {
 456                         $class = $type;
 457                 } elseif ( $wgSearchType !== null ) {
 458                         $class = $wgSearchType;
 459                 } else {
 460                         $dbr = wfGetDB( DB_SLAVE );
 461                         $class = $dbr->getSearchEngine();
 462                 }
 463
 464                 $search = new $class( $dbr );
 465                 return $search;
 466         }
 467
 468         /**
 469          * Return the search engines we support. If only $wgSearchType
 470          * is set, it'll be an array of just that one item.
 471          *
 472          * @return array
 473          */
 474         public static function getSearchTypes() {
 475                 global $wgSearchType, $wgSearchTypeAlternatives;
 476
 477                 $alternatives = $wgSearchTypeAlternatives ?: array();
 478                 array_unshift( $alternatives, $wgSearchType );
 479
 480                 return $alternatives;
 481         }
 482
 483         /**
 484          * Create or update the search index record for the given page.
 485          * Title and text should be pre-processed.
 486          * STUB
 487          *
 488          * @param $id Integer
 489          * @param $title String
 490          * @param $text String
 491          */
 492         function update( $id, $title, $text ) {
 493                 // no-op
 494         }
 495
 496         /**
 497          * Update a search index record's title only.
 498          * Title should be pre-processed.
 499          * STUB
 500          *
 501          * @param $id Integer
 502          * @param $title String
 503          */
 504         function updateTitle( $id, $title ) {
 505                 // no-op
 506         }
 507
 508         /**
 509          * Delete an indexed page
 510          * Title should be pre-processed.
 511          * STUB
 512          *
 513          * @param Integer $id Page id that was deleted
 514          * @param String $title Title of page that was deleted
 515          */
 516         function delete( $id, $title ) {
 517                 // no-op
 518         }
 519
 520         /**
 521          * Get OpenSearch suggestion template
 522          *
 523          * @return String
 524          */
 525         public static function getOpenSearchTemplate() {
 526                 global $wgOpenSearchTemplate, $wgCanonicalServer;
 527                 if ( $wgOpenSearchTemplate ) {
 528                         return $wgOpenSearchTemplate;
 529                 } else {
 530                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 531                         if ( !$ns ) {
 532                                 $ns = "0";
 533                         }
 534                         return $wgCanonicalServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
 535                 }
 536         }
 537
 538         /**
 539          * Get the raw text for updating the index from a content object
 540          * Nicer search backends could possibly do something cooler than
 541          * just returning raw text
 542          *
 543          * @todo This isn't ideal, we'd really like to have content-specific handling here
 544          * @param Title $t Title we're indexing
 545          * @param Content $c Content of the page to index
 546          * @return string
 547          */
 548         public function getTextFromContent( Title $t, Content $c = null ) {
 549                 return $c ? $c->getTextForSearchIndex() : '';
 550         }
 551
 552         /**
 553          * If an implementation of SearchEngine handles all of its own text processing
 554          * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
 555          * rather silly handling, it should return true here instead.
 556          *
 557          * @return bool
 558          */
 559         public function textAlreadyUpdatedForIndex() {
 560                 return false;
 561         }
 562 }
 563
 564 /**
 565  * @ingroup Search
 566  */
 567 class SearchResultSet {
 568         /**
 569          * Fetch an array of regular expression fragments for matching
 570          * the search terms as parsed by this engine in a text extract.
 571          * STUB
 572          *
 573          * @return Array
 574          */
 575         function termMatches() {
 576                 return array();
 577         }
 578
 579         function numRows() {
 580                 return 0;
 581         }
 582
 583         /**
 584          * Return true if results are included in this result set.
 585          * STUB
 586          *
 587          * @return Boolean
 588          */
 589         function hasResults() {
 590                 return false;
 591         }
 592
 593         /**
 594          * Some search modes return a total hit count for the query
 595          * in the entire article database. This may include pages
 596          * in namespaces that would not be matched on the given
 597          * settings.
 598          *
 599          * Return null if no total hits number is supported.
 600          *
 601          * @return Integer
 602          */
 603         function getTotalHits() {
 604                 return null;
 605         }
 606
 607         /**
 608          * Some search modes return a suggested alternate term if there are
 609          * no exact hits. Returns true if there is one on this set.
 610          *
 611          * @return Boolean
 612          */
 613         function hasSuggestion() {
 614                 return false;
 615         }
 616
 617         /**
 618          * @return String: suggested query, null if none
 619          */
 620         function getSuggestionQuery() {
 621                 return null;
 622         }
 623
 624         /**
 625          * @return String: HTML highlighted suggested query, '' if none
 626          */
 627         function getSuggestionSnippet() {
 628                 return '';
 629         }
 630
 631         /**
 632          * Return information about how and from where the results were fetched,
 633          * should be useful for diagnostics and debugging
 634          *
 635          * @return String
 636          */
 637         function getInfo() {
 638                 return null;
 639         }
 640
 641         /**
 642          * Return a result set of hits on other (multiple) wikis associated with this one
 643          *
 644          * @return SearchResultSet
 645          */
 646         function getInterwikiResults() {
 647                 return null;
 648         }
 649
 650         /**
 651          * Check if there are results on other wikis
 652          *
 653          * @return Boolean
 654          */
 655         function hasInterwikiResults() {
 656                 return $this->getInterwikiResults() != null;
 657         }
 658
 659         /**
 660          * Fetches next search result, or false.
 661          * STUB
 662          *
 663          * @return SearchResult
 664          */
 665         function next() {
 666                 return false;
 667         }
 668
 669         /**
 670          * Frees the result set, if applicable.
 671          */
 672         function free() {
 673                 // ...
 674         }
 675 }
 676
 677 /**
 678  * This class is used for different SQL-based search engines shipped with MediaWiki
 679  */
 680 class SqlSearchResultSet extends SearchResultSet {
 681
 682         protected $mResultSet;
 683
 684         function __construct( $resultSet, $terms ) {
 685                 $this->mResultSet = $resultSet;
 686                 $this->mTerms = $terms;
 687         }
 688
 689         function termMatches() {
 690                 return $this->mTerms;
 691         }
 692
 693         function numRows() {
 694                 if ( $this->mResultSet === false ) {
 695                         return false;
 696                 }
 697
 698                 return $this->mResultSet->numRows();
 699         }
 700
 701         function next() {
 702                 if ( $this->mResultSet === false ) {
 703                         return false;
 704                 }
 705
 706                 $row = $this->mResultSet->fetchObject();
 707                 if ( $row === false ) {
 708                         return false;
 709                 }
 710
 711                 return SearchResult::newFromRow( $row );
 712         }
 713
 714         function free() {
 715                 if ( $this->mResultSet === false ) {
 716                         return false;
 717                 }
 718
 719                 $this->mResultSet->free();
 720         }
 721 }
 722
 723 /**
 724  * @ingroup Search
 725  */
 726 class SearchResultTooMany {
 727         # # Some search engines may bail out if too many matches are found
 728 }
 729
 730 /**
 731  * @todo FIXME: This class is horribly factored. It would probably be better to
 732  * have a useful base class to which you pass some standard information, then
 733  * let the fancy self-highlighters extend that.
 734  * @ingroup Search
 735  */
 736 class SearchResult {
 737
 738         /**
 739          * @var Revision
 740          */
 741         protected $mRevision = null;
 742
 743         /**
 744          * @var File
 745          */
 746         protected $mImage = null;
 747
 748         /**
 749          * @var Title
 750          */
 751         protected $mTitle;
 752
 753         /**
 754          * @var String
 755          */
 756         protected $mText;
 757
 758         /**
 759          * Return a new SearchResult and initializes it with a title.
 760          *
 761          * @param $title Title
 762          * @return SearchResult
 763          */
 764         public static function newFromTitle( $title ) {
 765                 $result = new self();
 766                 $result->initFromTitle( $title );
 767                 return $result;
 768         }
 769         /**
 770          * Return a new SearchResult and initializes it with a row.
 771          *
 772          * @param $row object
 773          * @return SearchResult
 774          */
 775         public static function newFromRow( $row ) {
 776                 $result = new self();
 777                 $result->initFromRow( $row );
 778                 return $result;
 779         }
 780
 781         public function __construct( $row = null ) {
 782                 if ( !is_null( $row ) ) {
 783                         // Backwards compatibility with pre-1.17 callers
 784                         $this->initFromRow( $row );
 785                 }
 786         }
 787
 788         /**
 789          * Initialize from a database row. Makes a Title and passes that to
 790          * initFromTitle.
 791          *
 792          * @param $row object
 793          */
 794         protected function initFromRow( $row ) {
 795                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 796         }
 797
 798         /**
 799          * Initialize from a Title and if possible initializes a corresponding
 800          * Revision and File.
 801          *
 802          * @param $title Title
 803          */
 804         protected function initFromTitle( $title ) {
 805                 $this->mTitle = $title;
 806                 if ( !is_null( $this->mTitle ) ) {
 807                         $id = false;
 808                         wfRunHooks( 'SearchResultInitFromTitle', array( $title, &$id ) );
 809                         $this->mRevision = Revision::newFromTitle(
 810                                 $this->mTitle, $id, Revision::READ_NORMAL );
 811                         if ( $this->mTitle->getNamespace() === NS_FILE ) {
 812                                 $this->mImage = wfFindFile( $this->mTitle );
 813                         }
 814                 }
 815         }
 816
 817         /**
 818          * Check if this is result points to an invalid title
 819          *
 820          * @return Boolean
 821          */
 822         function isBrokenTitle() {
 823                 return is_null( $this->mTitle );
 824         }
 825
 826         /**
 827          * Check if target page is missing, happens when index is out of date
 828          *
 829          * @return Boolean
 830          */
 831         function isMissingRevision() {
 832                 return !$this->mRevision && !$this->mImage;
 833         }
 834
 835         /**
 836          * @return Title
 837          */
 838         function getTitle() {
 839                 return $this->mTitle;
 840         }
 841
 842         /**
 843          * Get the file for this page, if one exists
 844          * @return File|null
 845          */
 846         function getFile() {
 847                 return $this->mImage;
 848         }
 849
 850         /**
 851          * @return float|null if not supported
 852          */
 853         function getScore() {
 854                 return null;
 855         }
 856
 857         /**
 858          * Lazy initialization of article text from DB
 859          */
 860         protected function initText() {
 861                 if ( !isset( $this->mText ) ) {
 862                         if ( $this->mRevision != null ) {
 863                                 $this->mText = SearchEngine::create()
 864                                         ->getTextFromContent( $this->mTitle, $this->mRevision->getContent() );
 865                         } else { // TODO: can we fetch raw wikitext for commons images?
 866                                 $this->mText = '';
 867                         }
 868                 }
 869         }
 870
 871         /**
 872          * @param array $terms terms to highlight
 873          * @return String: highlighted text snippet, null (and not '') if not supported
 874          */
 875         function getTextSnippet( $terms ) {
 876                 global $wgAdvancedSearchHighlighting;
 877                 $this->initText();
 878
 879                 // TODO: make highliter take a content object. Make ContentHandler a factory for SearchHighliter.
 880                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs();
 881                 $h = new SearchHighlighter();
 882                 if ( $wgAdvancedSearchHighlighting ) {
 883                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 884                 } else {
 885                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 886                 }
 887         }
 888
 889         /**
 890          * @param array $terms terms to highlight
 891          * @return String: highlighted title, '' if not supported
 892          */
 893         function getTitleSnippet( $terms ) {
 894                 return '';
 895         }
 896
 897         /**
 898          * @param array $terms terms to highlight
 899          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 900          */
 901         function getRedirectSnippet( $terms ) {
 902                 return '';
 903         }
 904
 905         /**
 906          * @return Title object for the redirect to this page, null if none or not supported
 907          */
 908         function getRedirectTitle() {
 909                 return null;
 910         }
 911
 912         /**
 913          * @return string highlighted relevant section name, null if none or not supported
 914          */
 915         function getSectionSnippet() {
 916                 return '';
 917         }
 918
 919         /**
 920          * @return Title object (pagename+fragment) for the section, null if none or not supported
 921          */
 922         function getSectionTitle() {
 923                 return null;
 924         }
 925
 926         /**
 927          * @return String: timestamp
 928          */
 929         function getTimestamp() {
 930                 if ( $this->mRevision ) {
 931                         return $this->mRevision->getTimestamp();
 932                 } elseif ( $this->mImage ) {
 933                         return $this->mImage->getTimestamp();
 934                 }
 935                 return '';
 936         }
 937
 938         /**
 939          * @return Integer: number of words
 940          */
 941         function getWordCount() {
 942                 $this->initText();
 943                 return str_word_count( $this->mText );
 944         }
 945
 946         /**
 947          * @return Integer: size in bytes
 948          */
 949         function getByteSize() {
 950                 $this->initText();
 951                 return strlen( $this->mText );
 952         }
 953
 954         /**
 955          * @return Boolean if hit has related articles
 956          */
 957         function hasRelated() {
 958                 return false;
 959         }
 960
 961         /**
 962          * @return String: interwiki prefix of the title (return iw even if title is broken)
 963          */
 964         function getInterwikiPrefix() {
 965                 return '';
 966         }
 967
 968         /**
 969          * Did this match file contents (eg: PDF/DJVU)?
 970          */
 971         function isFileMatch() {
 972                 return false;
 973         }
 974 }
 975 /**
 976  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 977  */
 978 class SearchNearMatchResultSet extends SearchResultSet {
 979         private $fetched = false;
 980         /**
 981          * @param $match mixed Title if matched, else null
 982          */
 983         public function __construct( $match ) {
 984                 $this->result = $match;
 985         }
 986         public function hasResult() {
 987                 return (bool)$this->result;
 988         }
 989         public function numRows() {
 990                 return $this->hasResults() ? 1 : 0;
 991         }
 992         public function next() {
 993                 if ( $this->fetched || !$this->result ) {
 994                         return false;
 995                 }
 996                 $this->fetched = true;
 997                 return SearchResult::newFromTitle( $this->result );
 998         }
 999 }
1000
1001 /**
1002  * Highlight bits of wikitext
1003  *
1004  * @ingroup Search
1005  */
1006 class SearchHighlighter {
1007         var $mCleanWikitext = true;
1008
1009         function __construct( $cleanupWikitext = true ) {
1010                 $this->mCleanWikitext = $cleanupWikitext;
1011         }
1012
1013         /**
1014          * Default implementation of wikitext highlighting
1015          *
1016          * @param $text String
1017          * @param array $terms terms to highlight (unescaped)
1018          * @param $contextlines Integer
1019          * @param $contextchars Integer
1020          * @return String
1021          */
1022         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
1023                 global $wgContLang;
1024                 global $wgSearchHighlightBoundaries;
1025                 $fname = __METHOD__;
1026
1027                 if ( $text == '' ) {
1028                         return '';
1029                 }
1030
1031                 // spli text into text + templates/links/tables
1032                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
1033                 // first capture group is for detecting nested templates/links/tables/references
1034                 $endPatterns = array(
1035                         1 => '/(\{\{)|(\}\})/', // template
1036                         2 => '/(\[\[)|(\]\])/', // image
1037                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
1038
1039                 // @todo FIXME: This should prolly be a hook or something
1040                 if ( function_exists( 'wfCite' ) ) {
1041                         $spat .= '|(<ref>)'; // references via cite extension
1042                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
1043                 }
1044                 $spat .= '/';
1045                 $textExt = array(); // text extracts
1046                 $otherExt = array(); // other extracts
1047                 wfProfileIn( "$fname-split" );
1048                 $start = 0;
1049                 $textLen = strlen( $text );
1050                 $count = 0; // sequence number to maintain ordering
1051                 while ( $start < $textLen ) {
1052                         // find start of template/image/table
1053                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
1054                                 $epat = '';
1055                                 foreach ( $matches as $key => $val ) {
1056                                         if ( $key > 0 && $val[1] != - 1 ) {
1057                                                 if ( $key == 2 ) {
1058                                                         // see if this is an image link
1059                                                         $ns = substr( $val[0], 2, - 1 );
1060                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE ) {
1061                                                                 break;
1062                                                         }
1063
1064                                                 }
1065                                                 $epat = $endPatterns[$key];
1066                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
1067                                                 $start = $val[1];
1068                                                 break;
1069                                         }
1070                                 }
1071                                 if ( $epat ) {
1072                                         // find end (and detect any nested elements)
1073                                         $level = 0;
1074                                         $offset = $start + 1;
1075                                         $found = false;
1076                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
1077                                                 if ( array_key_exists( 2, $endMatches ) ) {
1078                                                         // found end
1079                                                         if ( $level == 0 ) {
1080                                                                 $len = strlen( $endMatches[2][0] );
1081                                                                 $off = $endMatches[2][1];
1082                                                                 $this->splitAndAdd( $otherExt, $count,
1083                                                                         substr( $text, $start, $off + $len - $start ) );
1084                                                                 $start = $off + $len;
1085                                                                 $found = true;
1086                                                                 break;
1087                                                         } else {
1088                                                                 // end of nested element
1089                                                                 $level -= 1;
1090                                                         }
1091                                                 } else {
1092                                                         // nested
1093                                                         $level += 1;
1094                                                 }
1095                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1096                                         }
1097                                         if ( ! $found ) {
1098                                                 // couldn't find appropriate closing tag, skip
1099                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1100                                                 $start += strlen( $matches[0][0] );
1101                                         }
1102                                         continue;
1103                                 }
1104                         }
1105                         // else: add as text extract
1106                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1107                         break;
1108                 }
1109
1110                 $all = $textExt + $otherExt; // these have disjunct key sets
1111
1112                 wfProfileOut( "$fname-split" );
1113
1114                 // prepare regexps
1115                 foreach ( $terms as $index => $term ) {
1116                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1117                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1118                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1119                         } else {
1120                                 $terms[$index] = $term;
1121                         }
1122                 }
1123                 $anyterm = implode( '|', $terms );
1124                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1125
1126                 // @todo FIXME: A hack to scale contextchars, a correct solution
1127                 // would be to have contextchars actually be char and not byte
1128                 // length, and do proper utf-8 substrings and lengths everywhere,
1129                 // but PHP is making that very hard and unclean to implement :(
1130                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1131                 $contextchars = intval( $contextchars * $scale );
1132
1133                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1134                 $patPost = "($wgSearchHighlightBoundaries|$)";
1135
1136                 $pat1 = "/(" . $phrase . ")/ui";
1137                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1138
1139                 wfProfileIn( "$fname-extract" );
1140
1141                 $left = $contextlines;
1142
1143                 $snippets = array();
1144                 $offsets = array();
1145
1146                 // show beginning only if it contains all words
1147                 $first = 0;
1148                 $firstText = '';
1149                 foreach ( $textExt as $index => $line ) {
1150                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1151                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1152                                 $first = $index;
1153                                 break;
1154                         }
1155                 }
1156                 if ( $firstText ) {
1157                         $succ = true;
1158                         // check if first text contains all terms
1159                         foreach ( $terms as $term ) {
1160                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1161                                         $succ = false;
1162                                         break;
1163                                 }
1164                         }
1165                         if ( $succ ) {
1166                                 $snippets[$first] = $firstText;
1167                                 $offsets[$first] = 0;
1168                         }
1169                 }
1170                 if ( ! $snippets ) {
1171                         // match whole query on text
1172                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1173                         // match whole query on templates/tables/images
1174                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1175                         // match any words on text
1176                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1177                         // match any words on templates/tables/images
1178                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1179
1180                         ksort( $snippets );
1181                 }
1182
1183                 // add extra chars to each snippet to make snippets constant size
1184                 $extended = array();
1185                 if ( count( $snippets ) == 0 ) {
1186                         // couldn't find the target words, just show beginning of article
1187                         if ( array_key_exists( $first, $all ) ) {
1188                                 $targetchars = $contextchars * $contextlines;
1189                                 $snippets[$first] = '';
1190                                 $offsets[$first] = 0;
1191                         }
1192                 } else {
1193                         // if begin of the article contains the whole phrase, show only that !!
1194                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1195                                 && $offsets[$first] < $contextchars * 2 ) {
1196                                 $snippets = array( $first => $snippets[$first] );
1197                         }
1198
1199                         // calc by how much to extend existing snippets
1200                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1201                 }
1202
1203                 foreach ( $snippets as $index => $line ) {
1204                         $extended[$index] = $line;
1205                         $len = strlen( $line );
1206                         if ( $len < $targetchars - 20 ) {
1207                                 // complete this line
1208                                 if ( $len < strlen( $all[$index] ) ) {
1209                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1210                                         $len = strlen( $extended[$index] );
1211                                 }
1212
1213                                 // add more lines
1214                                 $add = $index + 1;
1215                                 while ( $len < $targetchars - 20
1216                                                 && array_key_exists( $add, $all )
1217                                                 && !array_key_exists( $add, $snippets ) ) {
1218                                         $offsets[$add] = 0;
1219                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1220                                         $extended[$add] = $tt;
1221                                         $len += strlen( $tt );
1222                                         $add++;
1223                                 }
1224                         }
1225                 }
1226
1227                 // $snippets = array_map( 'htmlspecialchars', $extended );
1228                 $snippets = $extended;
1229                 $last = - 1;
1230                 $extract = '';
1231                 foreach ( $snippets as $index => $line ) {
1232                         if ( $last == - 1 ) {
1233                                 $extract .= $line; // first line
1234                         } elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) ) {
1235                                 $extract .= " " . $line; // continous lines
1236                         } else {
1237                                 $extract .= '<b> ... </b>' . $line;
1238                         }
1239
1240                         $last = $index;
1241                 }
1242                 if ( $extract ) {
1243                         $extract .= '<b> ... </b>';
1244                 }
1245
1246                 $processed = array();
1247                 foreach ( $terms as $term ) {
1248                         if ( ! isset( $processed[$term] ) ) {
1249                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1250                                 $extract = preg_replace( $pat3,
1251                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1252                                 $processed[$term] = true;
1253                         }
1254                 }
1255
1256                 wfProfileOut( "$fname-extract" );
1257
1258                 return $extract;
1259         }
1260
1261         /**
1262          * Split text into lines and add it to extracts array
1263          *
1264          * @param array $extracts index -> $line
1265          * @param $count Integer
1266          * @param $text String
1267          */
1268         function splitAndAdd( &$extracts, &$count, $text ) {
1269                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1270                 foreach ( $split as $line ) {
1271                         $tt = trim( $line );
1272                         if ( $tt ) {
1273                                 $extracts[$count++] = $tt;
1274                         }
1275                 }
1276         }
1277
1278         /**
1279          * Do manual case conversion for non-ascii chars
1280          *
1281          * @param $matches Array
1282          * @return string
1283          */
1284         function caseCallback( $matches ) {
1285                 global $wgContLang;
1286                 if ( strlen( $matches[0] ) > 1 ) {
1287                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1288                 } else {
1289                         return $matches[0];
1290                 }
1291         }
1292
1293         /**
1294          * Extract part of the text from start to end, but by
1295          * not chopping up words
1296          * @param $text String
1297          * @param $start Integer
1298          * @param $end Integer
1299          * @param $posStart Integer: (out) actual start position
1300          * @param $posEnd Integer: (out) actual end position
1301          * @return String
1302          */
1303         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1304                 if ( $start != 0 ) {
1305                         $start = $this->position( $text, $start, 1 );
1306                 }
1307                 if ( $end >= strlen( $text ) ) {
1308                         $end = strlen( $text );
1309                 } else {
1310                         $end = $this->position( $text, $end );
1311                 }
1312
1313                 if ( !is_null( $posStart ) ) {
1314                         $posStart = $start;
1315                 }
1316                 if ( !is_null( $posEnd ) ) {
1317                         $posEnd = $end;
1318                 }
1319
1320                 if ( $end > $start ) {
1321                         return substr( $text, $start, $end - $start );
1322                 } else {
1323                         return '';
1324                 }
1325         }
1326
1327         /**
1328          * Find a nonletter near a point (index) in the text
1329          *
1330          * @param $text String
1331          * @param $point Integer
1332          * @param $offset Integer: offset to found index
1333          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1334          */
1335         function position( $text, $point, $offset = 0 ) {
1336                 $tolerance = 10;
1337                 $s = max( 0, $point - $tolerance );
1338                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1339                 $m = array();
1340                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1341                         return $m[0][1] + $s + $offset;
1342                 } else {
1343                         // check if point is on a valid first UTF8 char
1344                         $char = ord( $text[$point] );
1345                         while ( $char >= 0x80 && $char < 0xc0 ) {
1346                                 // skip trailing bytes
1347                                 $point++;
1348                                 if ( $point >= strlen( $text ) ) {
1349                                         return strlen( $text );
1350                                 }
1351                                 $char = ord( $text[$point] );
1352                         }
1353                         return $point;
1354
1355                 }
1356         }
1357
1358         /**
1359          * Search extracts for a pattern, and return snippets
1360          *
1361          * @param string $pattern regexp for matching lines
1362          * @param array $extracts extracts to search
1363          * @param $linesleft Integer: number of extracts to make
1364          * @param $contextchars Integer: length of snippet
1365          * @param array $out map for highlighted snippets
1366          * @param array $offsets map of starting points of snippets
1367          * @protected
1368          */
1369         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1370                 if ( $linesleft == 0 ) {
1371                         return; // nothing to do
1372                 }
1373                 foreach ( $extracts as $index => $line ) {
1374                         if ( array_key_exists( $index, $out ) ) {
1375                                 continue; // this line already highlighted
1376                         }
1377
1378                         $m = array();
1379                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) ) {
1380                                 continue;
1381                         }
1382
1383                         $offset = $m[0][1];
1384                         $len = strlen( $m[0][0] );
1385                         if ( $offset + $len < $contextchars ) {
1386                                 $begin = 0;
1387                         } elseif ( $len > $contextchars ) {
1388                                 $begin = $offset;
1389                         } else {
1390                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1391                         }
1392
1393                         $end = $begin + $contextchars;
1394
1395                         $posBegin = $begin;
1396                         // basic snippet from this line
1397                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1398                         $offsets[$index] = $posBegin;
1399                         $linesleft--;
1400                         if ( $linesleft == 0 ) {
1401                                 return;
1402                         }
1403                 }
1404         }
1405
1406         /**
1407          * Basic wikitext removal
1408          * @protected
1409          * @return mixed
1410          */
1411         function removeWiki( $text ) {
1412                 $fname = __METHOD__;
1413                 wfProfileIn( $fname );
1414
1415                 // $text = preg_replace( "/'{2,5}/", "", $text );
1416                 // $text = preg_replace( "/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text );
1417                 // $text = preg_replace( "/\[\[([^]|]+)\]\]/", "\\1", $text );
1418                 // $text = preg_replace( "/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text );
1419                 // $text = preg_replace( "/\\{\\|(.*?)\\|\\}/", "", $text );
1420                 // $text = preg_replace( "/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text );
1421                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1422                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1423                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1424                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1425                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1426                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1427                 $text = preg_replace( "/'''''/", "", $text );
1428                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1429                 $text = preg_replace( "/''/", "", $text );
1430
1431                 wfProfileOut( $fname );
1432                 return $text;
1433         }
1434
1435         /**
1436          * callback to replace [[target|caption]] kind of links, if
1437          * the target is category or image, leave it
1438          *
1439          * @param $matches Array
1440          */
1441         function linkReplace( $matches ) {
1442                 $colon = strpos( $matches[1], ':' );
1443                 if ( $colon === false ) {
1444                         return $matches[2]; // replace with caption
1445                 }
1446                 global $wgContLang;
1447                 $ns = substr( $matches[1], 0, $colon );
1448                 $index = $wgContLang->getNsIndex( $ns );
1449                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) ) {
1450                         return $matches[0]; // return the whole thing
1451                 } else {
1452                         return $matches[2];
1453                 }
1454         }
1455
1456         /**
1457          * Simple & fast snippet extraction, but gives completely unrelevant
1458          * snippets
1459          *
1460          * @param $text String
1461          * @param $terms Array
1462          * @param $contextlines Integer
1463          * @param $contextchars Integer
1464          * @return String
1465          */
1466         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1467                 global $wgContLang;
1468                 $fname = __METHOD__;
1469
1470                 $lines = explode( "\n", $text );
1471
1472                 $terms = implode( '|', $terms );
1473                 $max = intval( $contextchars ) + 1;
1474                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
1475
1476                 $lineno = 0;
1477
1478                 $extract = "";
1479                 wfProfileIn( "$fname-extract" );
1480                 foreach ( $lines as $line ) {
1481                         if ( 0 == $contextlines ) {
1482                                 break;
1483                         }
1484                         ++$lineno;
1485                         $m = array();
1486                         if ( ! preg_match( $pat1, $line, $m ) ) {
1487                                 continue;
1488                         }
1489                         --$contextlines;
1490                         // truncate function changes ... to relevant i18n message.
1491                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1492
1493                         if ( count( $m ) < 3 ) {
1494                                 $post = '';
1495                         } else {
1496                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1497                         }
1498
1499                         $found = $m[2];
1500
1501                         $line = htmlspecialchars( $pre . $found . $post );
1502                         $pat2 = '/(' . $terms . ")/i";
1503                         $line = preg_replace( $pat2, "<span class='searchmatch'>\\1</span>", $line );
1504
1505                         $extract .= "${line}\n";
1506                 }
1507                 wfProfileOut( "$fname-extract" );
1508
1509                 return $extract;
1510         }
1511
1512 }
1513
1514 /**
1515  * Dummy class to be used when non-supported Database engine is present.
1516  * @todo FIXME: Dummy class should probably try something at least mildly useful,
1517  * such as a LIKE search through titles.
1518  * @ingroup Search
1519  */
1520 class SearchEngineDummy extends SearchEngine {
1521         // no-op
1522 }