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          * If this search backend can list/unlist redirects
  70          * @deprecated since 1.18 Call supports( 'list-redirects' );
  71          * @return bool
  72          */
  73         function acceptListRedirects() {
  74                 wfDeprecated( __METHOD__, '1.18' );
  75                 return $this->supports( 'list-redirects' );
  76         }
  77
  78         /**
  79          * @since 1.18
  80          * @param $feature String
  81          * @return Boolean
  82          */
  83         public function supports( $feature ) {
  84                 switch ( $feature ) {
  85                 case 'list-redirects':
  86                 case 'search-update':
  87                         return true;
  88                 case 'title-suffix-filter':
  89                 default:
  90                         return false;
  91                 }
  92         }
  93
  94         /**
  95          * Way to pass custom data for engines
  96          * @since 1.18
  97          * @param $feature String
  98          * @param $data Mixed
  99          * @return bool
 100          */
 101         public function setFeatureData( $feature, $data ) {
 102                 $this->features[$feature] = $data;
 103         }
 104
 105         /**
 106          * When overridden in derived class, performs database-specific conversions
 107          * on text to be used for searching or updating search index.
 108          * Default implementation does nothing (simply returns $string).
 109          *
 110          * @param string $string String to process
 111          * @return string
 112          */
 113         public function normalizeText( $string ) {
 114                 global $wgContLang;
 115
 116                 // Some languages such as Chinese require word segmentation
 117                 return $wgContLang->segmentByWord( $string );
 118         }
 119
 120         /**
 121          * Transform search term in cases when parts of the query came as different GET params (when supported)
 122          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
 123          */
 124         function transformSearchTerm( $term ) {
 125                 return $term;
 126         }
 127
 128         /**
 129          * If an exact title match can be found, or a very slightly close match,
 130          * return the title. If no match, returns NULL.
 131          *
 132          * @param $searchterm String
 133          * @return Title
 134          */
 135         public static function getNearMatch( $searchterm ) {
 136                 $title = self::getNearMatchInternal( $searchterm );
 137
 138                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
 139                 return $title;
 140         }
 141
 142         /**
 143          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 144          * SearchResultSet.
 145          *
 146          * @param $searchterm string
 147          * @return SearchResultSet
 148          */
 149         public static function getNearMatchResultSet( $searchterm ) {
 150                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 151         }
 152
 153         /**
 154          * Really find the title match.
 155          * @return null|Title
 156          */
 157         private static function getNearMatchInternal( $searchterm ) {
 158                 global $wgContLang, $wgEnableSearchContributorsByIP;
 159
 160                 $allSearchTerms = array( $searchterm );
 161
 162                 if ( $wgContLang->hasVariants() ) {
 163                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 164                 }
 165
 166                 $titleResult = null;
 167                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 168                         return $titleResult;
 169                 }
 170
 171                 foreach ( $allSearchTerms as $term ) {
 172
 173                         # Exact match? No need to look further.
 174                         $title = Title::newFromText( $term );
 175                         if ( is_null( $title ) ) {
 176                                 return null;
 177                         }
 178
 179                         # Try files if searching in the Media: namespace
 180                         if ( $title->getNamespace() == NS_MEDIA ) {
 181                                 $title = Title::makeTitle( NS_FILE, $title->getText() );
 182                         }
 183
 184                         if ( $title->isSpecialPage() || $title->isExternal() || $title->exists() ) {
 185                                 return $title;
 186                         }
 187
 188                         # See if it still otherwise has content is some sane sense
 189                         $page = WikiPage::factory( $title );
 190                         if ( $page->hasViewableContent() ) {
 191                                 return $title;
 192                         }
 193
 194                         if ( !wfRunHooks( 'SearchAfterNoDirectMatch', array( $term, &$title ) ) ) {
 195                                 return $title;
 196                         }
 197
 198                         # Now try all lower case (i.e. first letter capitalized)
 199                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 200                         if ( $title && $title->exists() ) {
 201                                 return $title;
 202                         }
 203
 204                         # Now try capitalized string
 205                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 206                         if ( $title && $title->exists() ) {
 207                                 return $title;
 208                         }
 209
 210                         # Now try all upper case
 211                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 212                         if ( $title && $title->exists() ) {
 213                                 return $title;
 214                         }
 215
 216                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 217                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 218                         if ( $title && $title->exists() ) {
 219                                 return $title;
 220                         }
 221
 222                         // Give hooks a chance at better match variants
 223                         $title = null;
 224                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 225                                 return $title;
 226                         }
 227                 }
 228
 229                 $title = Title::newFromText( $searchterm );
 230
 231                 # Entering an IP address goes to the contributions page
 232                 if ( $wgEnableSearchContributorsByIP ) {
 233                         if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 234                                 || User::isIP( trim( $searchterm ) ) ) {
 235                                 return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 236                         }
 237                 }
 238
 239                 # Entering a user goes to the user page whether it's there or not
 240                 if ( $title->getNamespace() == NS_USER ) {
 241                         return $title;
 242                 }
 243
 244                 # Go to images that exist even if there's no local page.
 245                 # There may have been a funny upload, or it may be on a shared
 246                 # file repository such as Wikimedia Commons.
 247                 if ( $title->getNamespace() == NS_FILE ) {
 248                         $image = wfFindFile( $title );
 249                         if ( $image ) {
 250                                 return $title;
 251                         }
 252                 }
 253
 254                 # MediaWiki namespace? Page may be "implied" if not customized.
 255                 # Just return it, with caps forced as the message system likes it.
 256                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 257                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 258                 }
 259
 260                 # Quoted term? Try without the quotes...
 261                 $matches = array();
 262                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 263                         return SearchEngine::getNearMatch( $matches[1] );
 264                 }
 265
 266                 return null;
 267         }
 268
 269         public static function legalSearchChars() {
 270                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 271         }
 272
 273         /**
 274          * Set the maximum number of results to return
 275          * and how many to skip before returning the first.
 276          *
 277          * @param $limit Integer
 278          * @param $offset Integer
 279          */
 280         function setLimitOffset( $limit, $offset = 0 ) {
 281                 $this->limit = intval( $limit );
 282                 $this->offset = intval( $offset );
 283         }
 284
 285         /**
 286          * Set which namespaces the search should include.
 287          * Give an array of namespace index numbers.
 288          *
 289          * @param $namespaces Array
 290          */
 291         function setNamespaces( $namespaces ) {
 292                 $this->namespaces = $namespaces;
 293         }
 294
 295         /**
 296          * Set whether the searcher should try to build a suggestion.  Note: some searchers
 297          * don't support building a suggestion in the first place and others don't respect
 298          * this flag.
 299          *
 300          * @param boolean $showSuggestion should the searcher try to build suggestions
 301          */
 302         function setShowSuggestion( $showSuggestion ) {
 303                 $this->showSuggestion = $showSuggestion;
 304         }
 305
 306         /**
 307          * Parse some common prefixes: all (search everything)
 308          * or namespace names
 309          *
 310          * @param $query String
 311          * @return string
 312          */
 313         function replacePrefixes( $query ) {
 314                 global $wgContLang;
 315
 316                 $parsed = $query;
 317                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 318                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 319                         return $parsed;
 320                 }
 321
 322                 $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
 323                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 324                         $this->namespaces = null;
 325                         $parsed = substr( $query, strlen( $allkeyword ) );
 326                 } elseif ( strpos( $query, ':' ) !== false ) {
 327                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 328                         $index = $wgContLang->getNsIndex( $prefix );
 329                         if ( $index !== false ) {
 330                                 $this->namespaces = array( $index );
 331                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 332                         }
 333                 }
 334                 if ( trim( $parsed ) == '' ) {
 335                         $parsed = $query; // prefix was the whole query
 336                 }
 337
 338                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 339
 340                 return $parsed;
 341         }
 342
 343         /**
 344          * Make a list of searchable namespaces and their canonical names.
 345          * @return Array
 346          */
 347         public static function searchableNamespaces() {
 348                 global $wgContLang;
 349                 $arr = array();
 350                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 351                         if ( $ns >= NS_MAIN ) {
 352                                 $arr[$ns] = $name;
 353                         }
 354                 }
 355
 356                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 357                 return $arr;
 358         }
 359
 360         /**
 361          * Extract default namespaces to search from the given user's
 362          * settings, returning a list of index numbers.
 363          *
 364          * @param $user User
 365          * @return Array
 366          */
 367         public static function userNamespaces( $user ) {
 368                 global $wgSearchEverythingOnlyLoggedIn;
 369
 370                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 371
 372                 // get search everything preference, that can be set to be read for logged-in users
 373                 // it overrides other options
 374                 if ( !$wgSearchEverythingOnlyLoggedIn || $user->isLoggedIn() ) {
 375                         if ( $user->getOption( 'searcheverything' ) ) {
 376                                 return array_keys( $searchableNamespaces );
 377                         }
 378                 }
 379
 380                 $arr = array();
 381                 foreach ( $searchableNamespaces as $ns => $name ) {
 382                         if ( $user->getOption( 'searchNs' . $ns ) ) {
 383                                 $arr[] = $ns;
 384                         }
 385                 }
 386
 387                 return $arr;
 388         }
 389
 390         /**
 391          * Find snippet highlight settings for all users
 392          *
 393          * @return Array contextlines, contextchars
 394          */
 395         public static function userHighlightPrefs() {
 396                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 397                 $contextchars = 75; // same as above.... :P
 398                 return array( $contextlines, $contextchars );
 399         }
 400
 401         /**
 402          * An array of namespaces indexes to be searched by default
 403          *
 404          * @return Array
 405          */
 406         public static function defaultNamespaces() {
 407                 global $wgNamespacesToBeSearchedDefault;
 408
 409                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 410         }
 411
 412         /**
 413          * Get a list of namespace names useful for showing in tooltips
 414          * and preferences
 415          *
 416          * @param $namespaces Array
 417          * @return array
 418          */
 419         public static function namespacesAsText( $namespaces ) {
 420                 global $wgContLang;
 421
 422                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 423                 foreach ( $formatted as $key => $ns ) {
 424                         if ( empty( $ns ) ) {
 425                                 $formatted[$key] = wfMessage( 'blanknamespace' )->text();
 426                         }
 427                 }
 428                 return $formatted;
 429         }
 430
 431         /**
 432          * Return the help namespaces to be shown on Special:Search
 433          *
 434          * @return Array
 435          */
 436         public static function helpNamespaces() {
 437                 global $wgNamespacesToBeSearchedHelp;
 438
 439                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 440         }
 441
 442         /**
 443          * Return a 'cleaned up' search string
 444          *
 445          * @param $text String
 446          * @return String
 447          */
 448         function filter( $text ) {
 449                 $lc = $this->legalSearchChars();
 450                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 451         }
 452         /**
 453          * Load up the appropriate search engine class for the currently
 454          * active database backend, and return a configured instance.
 455          *
 456          * @param String $type Type of search backend, if not the default
 457          * @return SearchEngine
 458          */
 459         public static function create( $type = null ) {
 460                 global $wgSearchType;
 461                 $dbr = null;
 462
 463                 $alternatives = self::getSearchTypes();
 464
 465                 if ( $type && in_array( $type, $alternatives ) ) {
 466                         $class = $type;
 467                 } elseif ( $wgSearchType !== null ) {
 468                         $class = $wgSearchType;
 469                 } else {
 470                         $dbr = wfGetDB( DB_SLAVE );
 471                         $class = $dbr->getSearchEngine();
 472                 }
 473
 474                 $search = new $class( $dbr );
 475                 return $search;
 476         }
 477
 478         /**
 479          * Return the search engines we support. If only $wgSearchType
 480          * is set, it'll be an array of just that one item.
 481          *
 482          * @return array
 483          */
 484         public static function getSearchTypes() {
 485                 global $wgSearchType, $wgSearchTypeAlternatives;
 486
 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         protected $mRevision = null;
 752
 753         /**
 754          * @var File
 755          */
 756         protected $mImage = null;
 757
 758         /**
 759          * @var Title
 760          */
 761         protected $mTitle;
 762
 763         /**
 764          * @var String
 765          */
 766         protected $mText;
 767
 768         /**
 769          * Return a new SearchResult and initializes it with a title.
 770          *
 771          * @param $title Title
 772          * @return SearchResult
 773          */
 774         public static function newFromTitle( $title ) {
 775                 $result = new self();
 776                 $result->initFromTitle( $title );
 777                 return $result;
 778         }
 779         /**
 780          * Return a new SearchResult and initializes it with a row.
 781          *
 782          * @param $row object
 783          * @return SearchResult
 784          */
 785         public static function newFromRow( $row ) {
 786                 $result = new self();
 787                 $result->initFromRow( $row );
 788                 return $result;
 789         }
 790
 791         public function __construct( $row = null ) {
 792                 if ( !is_null( $row ) ) {
 793                         // Backwards compatibility with pre-1.17 callers
 794                         $this->initFromRow( $row );
 795                 }
 796         }
 797
 798         /**
 799          * Initialize from a database row. Makes a Title and passes that to
 800          * initFromTitle.
 801          *
 802          * @param $row object
 803          */
 804         protected function initFromRow( $row ) {
 805                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 806         }
 807
 808         /**
 809          * Initialize from a Title and if possible initializes a corresponding
 810          * Revision and File.
 811          *
 812          * @param $title Title
 813          */
 814         protected function initFromTitle( $title ) {
 815                 $this->mTitle = $title;
 816                 if ( !is_null( $this->mTitle ) ) {
 817                         $id = false;
 818                         wfRunHooks( 'SearchResultInitFromTitle', array( $title, &$id ) );
 819                         $this->mRevision = Revision::newFromTitle(
 820                                 $this->mTitle, $id, Revision::READ_NORMAL );
 821                         if ( $this->mTitle->getNamespace() === NS_FILE ) {
 822                                 $this->mImage = wfFindFile( $this->mTitle );
 823                         }
 824                 }
 825         }
 826
 827         /**
 828          * Check if this is result points to an invalid title
 829          *
 830          * @return Boolean
 831          */
 832         function isBrokenTitle() {
 833                 return is_null( $this->mTitle );
 834         }
 835
 836         /**
 837          * Check if target page is missing, happens when index is out of date
 838          *
 839          * @return Boolean
 840          */
 841         function isMissingRevision() {
 842                 return !$this->mRevision && !$this->mImage;
 843         }
 844
 845         /**
 846          * @return Title
 847          */
 848         function getTitle() {
 849                 return $this->mTitle;
 850         }
 851
 852         /**
 853          * Get the file for this page, if one exists
 854          * @return File|null
 855          */
 856         function getFile() {
 857                 return $this->mImage;
 858         }
 859
 860         /**
 861          * @return float|null if not supported
 862          */
 863         function getScore() {
 864                 return null;
 865         }
 866
 867         /**
 868          * Lazy initialization of article text from DB
 869          */
 870         protected function initText() {
 871                 if ( !isset( $this->mText ) ) {
 872                         if ( $this->mRevision != null ) {
 873                                 $this->mText = SearchEngine::create()
 874                                         ->getTextFromContent( $this->mTitle, $this->mRevision->getContent() );
 875                         } else { // TODO: can we fetch raw wikitext for commons images?
 876                                 $this->mText = '';
 877                         }
 878                 }
 879         }
 880
 881         /**
 882          * @param array $terms terms to highlight
 883          * @return String: highlighted text snippet, null (and not '') if not supported
 884          */
 885         function getTextSnippet( $terms ) {
 886                 global $wgAdvancedSearchHighlighting;
 887                 $this->initText();
 888
 889                 // TODO: make highliter take a content object. Make ContentHandler a factory for SearchHighliter.
 890                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs();
 891                 $h = new SearchHighlighter();
 892                 if ( $wgAdvancedSearchHighlighting ) {
 893                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 894                 } else {
 895                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 896                 }
 897         }
 898
 899         /**
 900          * @param array $terms terms to highlight
 901          * @return String: highlighted title, '' if not supported
 902          */
 903         function getTitleSnippet( $terms ) {
 904                 return '';
 905         }
 906
 907         /**
 908          * @param array $terms terms to highlight
 909          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 910          */
 911         function getRedirectSnippet( $terms ) {
 912                 return '';
 913         }
 914
 915         /**
 916          * @return Title object for the redirect to this page, null if none or not supported
 917          */
 918         function getRedirectTitle() {
 919                 return null;
 920         }
 921
 922         /**
 923          * @return string highlighted relevant section name, null if none or not supported
 924          */
 925         function getSectionSnippet() {
 926                 return '';
 927         }
 928
 929         /**
 930          * @return Title object (pagename+fragment) for the section, null if none or not supported
 931          */
 932         function getSectionTitle() {
 933                 return null;
 934         }
 935
 936         /**
 937          * @return String: timestamp
 938          */
 939         function getTimestamp() {
 940                 if ( $this->mRevision ) {
 941                         return $this->mRevision->getTimestamp();
 942                 } elseif ( $this->mImage ) {
 943                         return $this->mImage->getTimestamp();
 944                 }
 945                 return '';
 946         }
 947
 948         /**
 949          * @return Integer: number of words
 950          */
 951         function getWordCount() {
 952                 $this->initText();
 953                 return str_word_count( $this->mText );
 954         }
 955
 956         /**
 957          * @return Integer: size in bytes
 958          */
 959         function getByteSize() {
 960                 $this->initText();
 961                 return strlen( $this->mText );
 962         }
 963
 964         /**
 965          * @return Boolean if hit has related articles
 966          */
 967         function hasRelated() {
 968                 return false;
 969         }
 970
 971         /**
 972          * @return String: interwiki prefix of the title (return iw even if title is broken)
 973          */
 974         function getInterwikiPrefix() {
 975                 return '';
 976         }
 977
 978         /**
 979          * Did this match file contents (eg: PDF/DJVU)?
 980          */
 981         function isFileMatch() {
 982                 return false;
 983         }
 984 }
 985 /**
 986  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 987  */
 988 class SearchNearMatchResultSet extends SearchResultSet {
 989         private $fetched = false;
 990         /**
 991          * @param $match mixed Title if matched, else null
 992          */
 993         public function __construct( $match ) {
 994                 $this->result = $match;
 995         }
 996         public function hasResult() {
 997                 return (bool)$this->result;
 998         }
 999         public function numRows() {
1000                 return $this->hasResults() ? 1 : 0;
1001         }
1002         public function next() {
1003                 if ( $this->fetched || !$this->result ) {
1004                         return false;
1005                 }
1006                 $this->fetched = true;
1007                 return SearchResult::newFromTitle( $this->result );
1008         }
1009 }
1010
1011 /**
1012  * Highlight bits of wikitext
1013  *
1014  * @ingroup Search
1015  */
1016 class SearchHighlighter {
1017         var $mCleanWikitext = true;
1018
1019         function __construct( $cleanupWikitext = true ) {
1020                 $this->mCleanWikitext = $cleanupWikitext;
1021         }
1022
1023         /**
1024          * Default implementation of wikitext highlighting
1025          *
1026          * @param $text String
1027          * @param array $terms terms to highlight (unescaped)
1028          * @param $contextlines Integer
1029          * @param $contextchars Integer
1030          * @return String
1031          */
1032         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
1033                 global $wgContLang;
1034                 global $wgSearchHighlightBoundaries;
1035                 $fname = __METHOD__;
1036
1037                 if ( $text == '' ) {
1038                         return '';
1039                 }
1040
1041                 // spli text into text + templates/links/tables
1042                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
1043                 // first capture group is for detecting nested templates/links/tables/references
1044                 $endPatterns = array(
1045                         1 => '/(\{\{)|(\}\})/', // template
1046                         2 => '/(\[\[)|(\]\])/', // image
1047                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
1048
1049                 // @todo FIXME: This should prolly be a hook or something
1050                 if ( function_exists( 'wfCite' ) ) {
1051                         $spat .= '|(<ref>)'; // references via cite extension
1052                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
1053                 }
1054                 $spat .= '/';
1055                 $textExt = array(); // text extracts
1056                 $otherExt = array(); // other extracts
1057                 wfProfileIn( "$fname-split" );
1058                 $start = 0;
1059                 $textLen = strlen( $text );
1060                 $count = 0; // sequence number to maintain ordering
1061                 while ( $start < $textLen ) {
1062                         // find start of template/image/table
1063                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
1064                                 $epat = '';
1065                                 foreach ( $matches as $key => $val ) {
1066                                         if ( $key > 0 && $val[1] != - 1 ) {
1067                                                 if ( $key == 2 ) {
1068                                                         // see if this is an image link
1069                                                         $ns = substr( $val[0], 2, - 1 );
1070                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE ) {
1071                                                                 break;
1072                                                         }
1073
1074                                                 }
1075                                                 $epat = $endPatterns[$key];
1076                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
1077                                                 $start = $val[1];
1078                                                 break;
1079                                         }
1080                                 }
1081                                 if ( $epat ) {
1082                                         // find end (and detect any nested elements)
1083                                         $level = 0;
1084                                         $offset = $start + 1;
1085                                         $found = false;
1086                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
1087                                                 if ( array_key_exists( 2, $endMatches ) ) {
1088                                                         // found end
1089                                                         if ( $level == 0 ) {
1090                                                                 $len = strlen( $endMatches[2][0] );
1091                                                                 $off = $endMatches[2][1];
1092                                                                 $this->splitAndAdd( $otherExt, $count,
1093                                                                         substr( $text, $start, $off + $len - $start ) );
1094                                                                 $start = $off + $len;
1095                                                                 $found = true;
1096                                                                 break;
1097                                                         } else {
1098                                                                 // end of nested element
1099                                                                 $level -= 1;
1100                                                         }
1101                                                 } else {
1102                                                         // nested
1103                                                         $level += 1;
1104                                                 }
1105                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1106                                         }
1107                                         if ( ! $found ) {
1108                                                 // couldn't find appropriate closing tag, skip
1109                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1110                                                 $start += strlen( $matches[0][0] );
1111                                         }
1112                                         continue;
1113                                 }
1114                         }
1115                         // else: add as text extract
1116                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1117                         break;
1118                 }
1119
1120                 $all = $textExt + $otherExt; // these have disjunct key sets
1121
1122                 wfProfileOut( "$fname-split" );
1123
1124                 // prepare regexps
1125                 foreach ( $terms as $index => $term ) {
1126                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1127                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1128                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1129                         } else {
1130                                 $terms[$index] = $term;
1131                         }
1132                 }
1133                 $anyterm = implode( '|', $terms );
1134                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1135
1136                 // @todo FIXME: A hack to scale contextchars, a correct solution
1137                 // would be to have contextchars actually be char and not byte
1138                 // length, and do proper utf-8 substrings and lengths everywhere,
1139                 // but PHP is making that very hard and unclean to implement :(
1140                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1141                 $contextchars = intval( $contextchars * $scale );
1142
1143                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1144                 $patPost = "($wgSearchHighlightBoundaries|$)";
1145
1146                 $pat1 = "/(" . $phrase . ")/ui";
1147                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1148
1149                 wfProfileIn( "$fname-extract" );
1150
1151                 $left = $contextlines;
1152
1153                 $snippets = array();
1154                 $offsets = array();
1155
1156                 // show beginning only if it contains all words
1157                 $first = 0;
1158                 $firstText = '';
1159                 foreach ( $textExt as $index => $line ) {
1160                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1161                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1162                                 $first = $index;
1163                                 break;
1164                         }
1165                 }
1166                 if ( $firstText ) {
1167                         $succ = true;
1168                         // check if first text contains all terms
1169                         foreach ( $terms as $term ) {
1170                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1171                                         $succ = false;
1172                                         break;
1173                                 }
1174                         }
1175                         if ( $succ ) {
1176                                 $snippets[$first] = $firstText;
1177                                 $offsets[$first] = 0;
1178                         }
1179                 }
1180                 if ( ! $snippets ) {
1181                         // match whole query on text
1182                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1183                         // match whole query on templates/tables/images
1184                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1185                         // match any words on text
1186                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1187                         // match any words on templates/tables/images
1188                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1189
1190                         ksort( $snippets );
1191                 }
1192
1193                 // add extra chars to each snippet to make snippets constant size
1194                 $extended = array();
1195                 if ( count( $snippets ) == 0 ) {
1196                         // couldn't find the target words, just show beginning of article
1197                         if ( array_key_exists( $first, $all ) ) {
1198                                 $targetchars = $contextchars * $contextlines;
1199                                 $snippets[$first] = '';
1200                                 $offsets[$first] = 0;
1201                         }
1202                 } else {
1203                         // if begin of the article contains the whole phrase, show only that !!
1204                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1205                                 && $offsets[$first] < $contextchars * 2 ) {
1206                                 $snippets = array( $first => $snippets[$first] );
1207                         }
1208
1209                         // calc by how much to extend existing snippets
1210                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1211                 }
1212
1213                 foreach ( $snippets as $index => $line ) {
1214                         $extended[$index] = $line;
1215                         $len = strlen( $line );
1216                         if ( $len < $targetchars - 20 ) {
1217                                 // complete this line
1218                                 if ( $len < strlen( $all[$index] ) ) {
1219                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1220                                         $len = strlen( $extended[$index] );
1221                                 }
1222
1223                                 // add more lines
1224                                 $add = $index + 1;
1225                                 while ( $len < $targetchars - 20
1226                                                 && array_key_exists( $add, $all )
1227                                                 && !array_key_exists( $add, $snippets ) ) {
1228                                         $offsets[$add] = 0;
1229                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1230                                         $extended[$add] = $tt;
1231                                         $len += strlen( $tt );
1232                                         $add++;
1233                                 }
1234                         }
1235                 }
1236
1237                 // $snippets = array_map( 'htmlspecialchars', $extended );
1238                 $snippets = $extended;
1239                 $last = - 1;
1240                 $extract = '';
1241                 foreach ( $snippets as $index => $line ) {
1242                         if ( $last == - 1 ) {
1243                                 $extract .= $line; // first line
1244                         } elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) ) {
1245                                 $extract .= " " . $line; // continous lines
1246                         } else {
1247                                 $extract .= '<b> ... </b>' . $line;
1248                         }
1249
1250                         $last = $index;
1251                 }
1252                 if ( $extract ) {
1253                         $extract .= '<b> ... </b>';
1254                 }
1255
1256                 $processed = array();
1257                 foreach ( $terms as $term ) {
1258                         if ( ! isset( $processed[$term] ) ) {
1259                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1260                                 $extract = preg_replace( $pat3,
1261                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1262                                 $processed[$term] = true;
1263                         }
1264                 }
1265
1266                 wfProfileOut( "$fname-extract" );
1267
1268                 return $extract;
1269         }
1270
1271         /**
1272          * Split text into lines and add it to extracts array
1273          *
1274          * @param array $extracts index -> $line
1275          * @param $count Integer
1276          * @param $text String
1277          */
1278         function splitAndAdd( &$extracts, &$count, $text ) {
1279                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1280                 foreach ( $split as $line ) {
1281                         $tt = trim( $line );
1282                         if ( $tt ) {
1283                                 $extracts[$count++] = $tt;
1284                         }
1285                 }
1286         }
1287
1288         /**
1289          * Do manual case conversion for non-ascii chars
1290          *
1291          * @param $matches Array
1292          * @return string
1293          */
1294         function caseCallback( $matches ) {
1295                 global $wgContLang;
1296                 if ( strlen( $matches[0] ) > 1 ) {
1297                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1298                 } else {
1299                         return $matches[0];
1300                 }
1301         }
1302
1303         /**
1304          * Extract part of the text from start to end, but by
1305          * not chopping up words
1306          * @param $text String
1307          * @param $start Integer
1308          * @param $end Integer
1309          * @param $posStart Integer: (out) actual start position
1310          * @param $posEnd Integer: (out) actual end position
1311          * @return String
1312          */
1313         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1314                 if ( $start != 0 ) {
1315                         $start = $this->position( $text, $start, 1 );
1316                 }
1317                 if ( $end >= strlen( $text ) ) {
1318                         $end = strlen( $text );
1319                 } else {
1320                         $end = $this->position( $text, $end );
1321                 }
1322
1323                 if ( !is_null( $posStart ) ) {
1324                         $posStart = $start;
1325                 }
1326                 if ( !is_null( $posEnd ) ) {
1327                         $posEnd = $end;
1328                 }
1329
1330                 if ( $end > $start ) {
1331                         return substr( $text, $start, $end - $start );
1332                 } else {
1333                         return '';
1334                 }
1335         }
1336
1337         /**
1338          * Find a nonletter near a point (index) in the text
1339          *
1340          * @param $text String
1341          * @param $point Integer
1342          * @param $offset Integer: offset to found index
1343          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1344          */
1345         function position( $text, $point, $offset = 0 ) {
1346                 $tolerance = 10;
1347                 $s = max( 0, $point - $tolerance );
1348                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1349                 $m = array();
1350                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1351                         return $m[0][1] + $s + $offset;
1352                 } else {
1353                         // check if point is on a valid first UTF8 char
1354                         $char = ord( $text[$point] );
1355                         while ( $char >= 0x80 && $char < 0xc0 ) {
1356                                 // skip trailing bytes
1357                                 $point++;
1358                                 if ( $point >= strlen( $text ) ) {
1359                                         return strlen( $text );
1360                                 }
1361                                 $char = ord( $text[$point] );
1362                         }
1363                         return $point;
1364
1365                 }
1366         }
1367
1368         /**
1369          * Search extracts for a pattern, and return snippets
1370          *
1371          * @param string $pattern regexp for matching lines
1372          * @param array $extracts extracts to search
1373          * @param $linesleft Integer: number of extracts to make
1374          * @param $contextchars Integer: length of snippet
1375          * @param array $out map for highlighted snippets
1376          * @param array $offsets map of starting points of snippets
1377          * @protected
1378          */
1379         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1380                 if ( $linesleft == 0 ) {
1381                         return; // nothing to do
1382                 }
1383                 foreach ( $extracts as $index => $line ) {
1384                         if ( array_key_exists( $index, $out ) ) {
1385                                 continue; // this line already highlighted
1386                         }
1387
1388                         $m = array();
1389                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) ) {
1390                                 continue;
1391                         }
1392
1393                         $offset = $m[0][1];
1394                         $len = strlen( $m[0][0] );
1395                         if ( $offset + $len < $contextchars ) {
1396                                 $begin = 0;
1397                         } elseif ( $len > $contextchars ) {
1398                                 $begin = $offset;
1399                         } else {
1400                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1401                         }
1402
1403                         $end = $begin + $contextchars;
1404
1405                         $posBegin = $begin;
1406                         // basic snippet from this line
1407                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1408                         $offsets[$index] = $posBegin;
1409                         $linesleft--;
1410                         if ( $linesleft == 0 ) {
1411                                 return;
1412                         }
1413                 }
1414         }
1415
1416         /**
1417          * Basic wikitext removal
1418          * @protected
1419          * @return mixed
1420          */
1421         function removeWiki( $text ) {
1422                 $fname = __METHOD__;
1423                 wfProfileIn( $fname );
1424
1425                 // $text = preg_replace( "/'{2,5}/", "", $text );
1426                 // $text = preg_replace( "/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text );
1427                 // $text = preg_replace( "/\[\[([^]|]+)\]\]/", "\\1", $text );
1428                 // $text = preg_replace( "/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text );
1429                 // $text = preg_replace( "/\\{\\|(.*?)\\|\\}/", "", $text );
1430                 // $text = preg_replace( "/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text );
1431                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1432                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1433                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1434                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1435                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1436                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1437                 $text = preg_replace( "/'''''/", "", $text );
1438                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1439                 $text = preg_replace( "/''/", "", $text );
1440
1441                 wfProfileOut( $fname );
1442                 return $text;
1443         }
1444
1445         /**
1446          * callback to replace [[target|caption]] kind of links, if
1447          * the target is category or image, leave it
1448          *
1449          * @param $matches Array
1450          */
1451         function linkReplace( $matches ) {
1452                 $colon = strpos( $matches[1], ':' );
1453                 if ( $colon === false ) {
1454                         return $matches[2]; // replace with caption
1455                 }
1456                 global $wgContLang;
1457                 $ns = substr( $matches[1], 0, $colon );
1458                 $index = $wgContLang->getNsIndex( $ns );
1459                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) ) {
1460                         return $matches[0]; // return the whole thing
1461                 } else {
1462                         return $matches[2];
1463                 }
1464         }
1465
1466         /**
1467          * Simple & fast snippet extraction, but gives completely unrelevant
1468          * snippets
1469          *
1470          * @param $text String
1471          * @param $terms Array
1472          * @param $contextlines Integer
1473          * @param $contextchars Integer
1474          * @return String
1475          */
1476         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1477                 global $wgContLang;
1478                 $fname = __METHOD__;
1479
1480                 $lines = explode( "\n", $text );
1481
1482                 $terms = implode( '|', $terms );
1483                 $max = intval( $contextchars ) + 1;
1484                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
1485
1486                 $lineno = 0;
1487
1488                 $extract = "";
1489                 wfProfileIn( "$fname-extract" );
1490                 foreach ( $lines as $line ) {
1491                         if ( 0 == $contextlines ) {
1492                                 break;
1493                         }
1494                         ++$lineno;
1495                         $m = array();
1496                         if ( ! preg_match( $pat1, $line, $m ) ) {
1497                                 continue;
1498                         }
1499                         --$contextlines;
1500                         // truncate function changes ... to relevant i18n message.
1501                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1502
1503                         if ( count( $m ) < 3 ) {
1504                                 $post = '';
1505                         } else {
1506                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1507                         }
1508
1509                         $found = $m[2];
1510
1511                         $line = htmlspecialchars( $pre . $found . $post );
1512                         $pat2 = '/(' . $terms . ")/i";
1513                         $line = preg_replace( $pat2, "<span class='searchmatch'>\\1</span>", $line );
1514
1515                         $extract .= "${line}\n";
1516                 }
1517                 wfProfileOut( "$fname-extract" );
1518
1519                 return $extract;
1520         }
1521
1522 }
1523
1524 /**
1525  * Dummy class to be used when non-supported Database engine is present.
1526  * @todo FIXME: Dummy class should probably try something at least mildly useful,
1527  * such as a LIKE search through titles.
1528  * @ingroup Search
1529  */
1530 class SearchEngineDummy extends SearchEngine {
1531         // no-op
1532 }