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