includes/search/SearchEngine.php

   1 <?php
   2 /**
   3  * Basic search engine
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup Search
  22  */
  23
  24 /**
  25  * @defgroup Search Search
  26  */
  27
  28 /**
  29  * Contain a class for special pages
  30  * @ingroup Search
  31  */
  32 class SearchEngine {
  33         var $limit = 10;
  34         var $offset = 0;
  35         var $prefix = '';
  36         var $searchTerms = array();
  37         var $namespaces = array( NS_MAIN );
  38         var $showRedirects = false;
  39         protected $showSuggestion = true;
  40
  41         /// Feature values
  42         protected $features = array();
  43
  44         /**
  45          * Perform a full text search query and return a result set.
  46          * If title searches are not supported or disabled, return null.
  47          * STUB
  48          *
  49          * @param string $term raw search term
  50          * @return SearchResultSet|Status|null
  51          */
  52         function searchText( $term ) {
  53                 return null;
  54         }
  55
  56         /**
  57          * Perform a title-only search query and return a result set.
  58          * If title searches are not supported or disabled, return null.
  59          * STUB
  60          *
  61          * @param string $term raw search term
  62          * @return SearchResultSet|null
  63          */
  64         function searchTitle( $term ) {
  65                 return null;
  66         }
  67
  68         /**
  69          * @since 1.18
  70          * @param $feature String
  71          * @return Boolean
  72          */
  73         public function supports( $feature ) {
  74                 switch ( $feature ) {
  75                 case 'list-redirects':
  76                 case 'search-update':
  77                         return true;
  78                 case 'title-suffix-filter':
  79                 default:
  80                         return false;
  81                 }
  82         }
  83
  84         /**
  85          * Way to pass custom data for engines
  86          * @since 1.18
  87          * @param $feature String
  88          * @param $data Mixed
  89          * @return bool
  90          */
  91         public function setFeatureData( $feature, $data ) {
  92                 $this->features[$feature] = $data;
  93         }
  94
  95         /**
  96          * When overridden in derived class, performs database-specific conversions
  97          * on text to be used for searching or updating search index.
  98          * Default implementation does nothing (simply returns $string).
  99          *
 100          * @param string $string String to process
 101          * @return string
 102          */
 103         public function normalizeText( $string ) {
 104                 global $wgContLang;
 105
 106                 // Some languages such as Chinese require word segmentation
 107                 return $wgContLang->segmentByWord( $string );
 108         }
 109
 110         /**
 111          * Transform search term in cases when parts of the query came as different GET params (when supported)
 112          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
 113          */
 114         function transformSearchTerm( $term ) {
 115                 return $term;
 116         }
 117
 118         /**
 119          * If an exact title match can be found, or a very slightly close match,
 120          * return the title. If no match, returns NULL.
 121          *
 122          * @param $searchterm String
 123          * @return Title
 124          */
 125         public static function getNearMatch( $searchterm ) {
 126                 $title = self::getNearMatchInternal( $searchterm );
 127
 128                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
 129                 return $title;
 130         }
 131
 132         /**
 133          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 134          * SearchResultSet.
 135          *
 136          * @param $searchterm string
 137          * @return SearchResultSet
 138          */
 139         public static function getNearMatchResultSet( $searchterm ) {
 140                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 141         }
 142
 143         /**
 144          * Really find the title match.
 145          * @return null|Title
 146          */
 147         private static function getNearMatchInternal( $searchterm ) {
 148                 global $wgContLang, $wgEnableSearchContributorsByIP;
 149
 150                 $allSearchTerms = array( $searchterm );
 151
 152                 if ( $wgContLang->hasVariants() ) {
 153                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 154                 }
 155
 156                 $titleResult = null;
 157                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 158                         return $titleResult;
 159                 }
 160
 161                 foreach ( $allSearchTerms as $term ) {
 162
 163                         # Exact match? No need to look further.
 164                         $title = Title::newFromText( $term );
 165                         if ( is_null( $title ) ) {
 166                                 return null;
 167                         }
 168
 169                         # Try files if searching in the Media: namespace
 170                         if ( $title->getNamespace() == NS_MEDIA ) {
 171                                 $title = Title::makeTitle( NS_FILE, $title->getText() );
 172                         }
 173
 174                         if ( $title->isSpecialPage() || $title->isExternal() || $title->exists() ) {
 175                                 return $title;
 176                         }
 177
 178                         # See if it still otherwise has content is some sane sense
 179                         $page = WikiPage::factory( $title );
 180                         if ( $page->hasViewableContent() ) {
 181                                 return $title;
 182                         }
 183
 184                         if ( !wfRunHooks( 'SearchAfterNoDirectMatch', array( $term, &$title ) ) ) {
 185                                 return $title;
 186                         }
 187
 188                         # Now try all lower case (i.e. first letter capitalized)
 189                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 190                         if ( $title && $title->exists() ) {
 191                                 return $title;
 192                         }
 193
 194                         # Now try capitalized string
 195                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 196                         if ( $title && $title->exists() ) {
 197                                 return $title;
 198                         }
 199
 200                         # Now try all upper case
 201                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 202                         if ( $title && $title->exists() ) {
 203                                 return $title;
 204                         }
 205
 206                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 207                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 208                         if ( $title && $title->exists() ) {
 209                                 return $title;
 210                         }
 211
 212                         // Give hooks a chance at better match variants
 213                         $title = null;
 214                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 215                                 return $title;
 216                         }
 217                 }
 218
 219                 $title = Title::newFromText( $searchterm );
 220
 221                 # Entering an IP address goes to the contributions page
 222                 if ( $wgEnableSearchContributorsByIP ) {
 223                         if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 224                                 || User::isIP( trim( $searchterm ) ) ) {
 225                                 return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 226                         }
 227                 }
 228
 229                 # Entering a user goes to the user page whether it's there or not
 230                 if ( $title->getNamespace() == NS_USER ) {
 231                         return $title;
 232                 }
 233
 234                 # Go to images that exist even if there's no local page.
 235                 # There may have been a funny upload, or it may be on a shared
 236                 # file repository such as Wikimedia Commons.
 237                 if ( $title->getNamespace() == NS_FILE ) {
 238                         $image = wfFindFile( $title );
 239                         if ( $image ) {
 240                                 return $title;
 241                         }
 242                 }
 243
 244                 # MediaWiki namespace? Page may be "implied" if not customized.
 245                 # Just return it, with caps forced as the message system likes it.
 246                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 247                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 248                 }
 249
 250                 # Quoted term? Try without the quotes...
 251                 $matches = array();
 252                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 253                         return SearchEngine::getNearMatch( $matches[1] );
 254                 }
 255
 256                 return null;
 257         }
 258
 259         public static function legalSearchChars() {
 260                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 261         }
 262
 263         /**
 264          * Set the maximum number of results to return
 265          * and how many to skip before returning the first.
 266          *
 267          * @param $limit Integer
 268          * @param $offset Integer
 269          */
 270         function setLimitOffset( $limit, $offset = 0 ) {
 271                 $this->limit = intval( $limit );
 272                 $this->offset = intval( $offset );
 273         }
 274
 275         /**
 276          * Set which namespaces the search should include.
 277          * Give an array of namespace index numbers.
 278          *
 279          * @param $namespaces Array
 280          */
 281         function setNamespaces( $namespaces ) {
 282                 $this->namespaces = $namespaces;
 283         }
 284
 285         /**
 286          * Set whether the searcher should try to build a suggestion.  Note: some searchers
 287          * don't support building a suggestion in the first place and others don't respect
 288          * this flag.
 289          *
 290          * @param boolean $showSuggestion should the searcher try to build suggestions
 291          */
 292         function setShowSuggestion( $showSuggestion ) {
 293                 $this->showSuggestion = $showSuggestion;
 294         }
 295
 296         /**
 297          * Parse some common prefixes: all (search everything)
 298          * or namespace names
 299          *
 300          * @param $query String
 301          * @return string
 302          */
 303         function replacePrefixes( $query ) {
 304                 global $wgContLang;
 305
 306                 $parsed = $query;
 307                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 308                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 309                         return $parsed;
 310                 }
 311
 312                 $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
 313                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 314                         $this->namespaces = null;
 315                         $parsed = substr( $query, strlen( $allkeyword ) );
 316                 } elseif ( strpos( $query, ':' ) !== false ) {
 317                         $prefix = str_replace( ' ', '_', substr( $query, 0, strpos( $query, ':' ) ) );
 318                         $index = $wgContLang->getNsIndex( $prefix );
 319                         if ( $index !== false ) {
 320                                 $this->namespaces = array( $index );
 321                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 322                         }
 323                 }
 324                 if ( trim( $parsed ) == '' ) {
 325                         $parsed = $query; // prefix was the whole query
 326                 }
 327
 328                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 329
 330                 return $parsed;
 331         }
 332
 333         /**
 334          * Make a list of searchable namespaces and their canonical names.
 335          * @return Array
 336          */
 337         public static function searchableNamespaces() {
 338                 global $wgContLang;
 339                 $arr = array();
 340                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 341                         if ( $ns >= NS_MAIN ) {
 342                                 $arr[$ns] = $name;
 343                         }
 344                 }
 345
 346                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 347                 return $arr;
 348         }
 349
 350         /**
 351          * Extract default namespaces to search from the given user's
 352          * settings, returning a list of index numbers.
 353          *
 354          * @param $user User
 355          * @return Array
 356          */
 357         public static function userNamespaces( $user ) {
 358                 global $wgSearchEverythingOnlyLoggedIn;
 359
 360                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 361
 362                 // get search everything preference, that can be set to be read for logged-in users
 363                 // it overrides other options
 364                 if ( !$wgSearchEverythingOnlyLoggedIn || $user->isLoggedIn() ) {
 365                         if ( $user->getOption( 'searcheverything' ) ) {
 366                                 return array_keys( $searchableNamespaces );
 367                         }
 368                 }
 369
 370                 $arr = array();
 371                 foreach ( $searchableNamespaces as $ns => $name ) {
 372                         if ( $user->getOption( 'searchNs' . $ns ) ) {
 373                                 $arr[] = $ns;
 374                         }
 375                 }
 376
 377                 return $arr;
 378         }
 379
 380         /**
 381          * Find snippet highlight settings for all users
 382          *
 383          * @return Array contextlines, contextchars
 384          */
 385         public static function userHighlightPrefs() {
 386                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 387                 $contextchars = 75; // same as above.... :P
 388                 return array( $contextlines, $contextchars );
 389         }
 390
 391         /**
 392          * An array of namespaces indexes to be searched by default
 393          *
 394          * @return Array
 395          */
 396         public static function defaultNamespaces() {
 397                 global $wgNamespacesToBeSearchedDefault;
 398
 399                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 400         }
 401
 402         /**
 403          * Get a list of namespace names useful for showing in tooltips
 404          * and preferences
 405          *
 406          * @param $namespaces Array
 407          * @return array
 408          */
 409         public static function namespacesAsText( $namespaces ) {
 410                 global $wgContLang;
 411
 412                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 413                 foreach ( $formatted as $key => $ns ) {
 414                         if ( empty( $ns ) ) {
 415                                 $formatted[$key] = wfMessage( 'blanknamespace' )->text();
 416                         }
 417                 }
 418                 return $formatted;
 419         }
 420
 421         /**
 422          * Return the help namespaces to be shown on Special:Search
 423          *
 424          * @return Array
 425          */
 426         public static function helpNamespaces() {
 427                 global $wgNamespacesToBeSearchedHelp;
 428
 429                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 430         }
 431
 432         /**
 433          * Return a 'cleaned up' search string
 434          *
 435          * @param $text String
 436          * @return String
 437          */
 438         function filter( $text ) {
 439                 $lc = $this->legalSearchChars();
 440                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 441         }
 442         /**
 443          * Load up the appropriate search engine class for the currently
 444          * active database backend, and return a configured instance.
 445          *
 446          * @param String $type Type of search backend, if not the default
 447          * @return SearchEngine
 448          */
 449         public static function create( $type = null ) {
 450                 global $wgSearchType;
 451                 $dbr = null;
 452
 453                 $alternatives = self::getSearchTypes();
 454
 455                 if ( $type && in_array( $type, $alternatives ) ) {
 456                         $class = $type;
 457                 } elseif ( $wgSearchType !== null ) {
 458                         $class = $wgSearchType;
 459                 } else {
 460                         $dbr = wfGetDB( DB_SLAVE );
 461                         $class = $dbr->getSearchEngine();
 462                 }
 463
 464                 $search = new $class( $dbr );
 465                 return $search;
 466         }
 467
 468         /**
 469          * Return the search engines we support. If only $wgSearchType
 470          * is set, it'll be an array of just that one item.
 471          *
 472          * @return array
 473          */
 474         public static function getSearchTypes() {
 475                 global $wgSearchType, $wgSearchTypeAlternatives;
 476
 477                 $alternatives = $wgSearchTypeAlternatives ?: array();
 478                 array_unshift( $alternatives, $wgSearchType );
 479
 480                 return $alternatives;
 481         }
 482
 483         /**
 484          * Create or update the search index record for the given page.
 485          * Title and text should be pre-processed.
 486          * STUB
 487          *
 488          * @param $id Integer
 489          * @param $title String
 490          * @param $text String
 491          */
 492         function update( $id, $title, $text ) {
 493                 // no-op
 494         }
 495
 496         /**
 497          * Update a search index record's title only.
 498          * Title should be pre-processed.
 499          * STUB
 500          *
 501          * @param $id Integer
 502          * @param $title String
 503          */
 504         function updateTitle( $id, $title ) {
 505                 // no-op
 506         }
 507
 508         /**
 509          * Delete an indexed page
 510          * Title should be pre-processed.
 511          * STUB
 512          *
 513          * @param Integer $id Page id that was deleted
 514          * @param String $title Title of page that was deleted
 515          */
 516         function delete( $id, $title ) {
 517                 // no-op
 518         }
 519
 520         /**
 521          * Get OpenSearch suggestion template
 522          *
 523          * @return String
 524          */
 525         public static function getOpenSearchTemplate() {
 526                 global $wgOpenSearchTemplate, $wgCanonicalServer;
 527                 if ( $wgOpenSearchTemplate ) {
 528                         return $wgOpenSearchTemplate;
 529                 } else {
 530                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 531                         if ( !$ns ) {
 532                                 $ns = "0";
 533                         }
 534                         return $wgCanonicalServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
 535                 }
 536         }
 537
 538         /**
 539          * Get the raw text for updating the index from a content object
 540          * Nicer search backends could possibly do something cooler than
 541          * just returning raw text
 542          *
 543          * @todo This isn't ideal, we'd really like to have content-specific handling here
 544          * @param Title $t Title we're indexing
 545          * @param Content $c Content of the page to index
 546          * @return string
 547          */
 548         public function getTextFromContent( Title $t, Content $c = null ) {
 549                 return $c ? $c->getTextForSearchIndex() : '';
 550         }
 551
 552         /**
 553          * If an implementation of SearchEngine handles all of its own text processing
 554          * in getTextFromContent() and doesn't require SearchUpdate::updateText()'s
 555          * rather silly handling, it should return true here instead.
 556          *
 557          * @return bool
 558          */
 559         public function textAlreadyUpdatedForIndex() {
 560                 return false;
 561         }
 562 }
 563
 564 /**
 565  * @ingroup Search
 566  */
 567 class SearchResultSet {
 568         /**
 569          * Fetch an array of regular expression fragments for matching
 570          * the search terms as parsed by this engine in a text extract.
 571          * STUB
 572          *
 573          * @return Array
 574          */
 575         function termMatches() {
 576                 return array();
 577         }
 578
 579         function numRows() {
 580                 return 0;
 581         }
 582
 583         /**
 584          * Return true if results are included in this result set.
 585          * STUB
 586          *
 587          * @return Boolean
 588          */
 589         function hasResults() {
 590                 return false;
 591         }
 592
 593         /**
 594          * Some search modes return a total hit count for the query
 595          * in the entire article database. This may include pages
 596          * in namespaces that would not be matched on the given
 597          * settings.
 598          *
 599          * Return null if no total hits number is supported.
 600          *
 601          * @return Integer
 602          */
 603         function getTotalHits() {
 604                 return null;
 605         }
 606
 607         /**
 608          * Some search modes return a suggested alternate term if there are
 609          * no exact hits. Returns true if there is one on this set.
 610          *
 611          * @return Boolean
 612          */
 613         function hasSuggestion() {
 614                 return false;
 615         }
 616
 617         /**
 618          * @return String: suggested query, null if none
 619          */
 620         function getSuggestionQuery() {
 621                 return null;
 622         }
 623
 624         /**
 625          * @return String: HTML highlighted suggested query, '' if none
 626          */
 627         function getSuggestionSnippet() {
 628                 return '';
 629         }
 630
 631         /**
 632          * Return information about how and from where the results were fetched,
 633          * should be useful for diagnostics and debugging
 634          *
 635          * @return String
 636          */
 637         function getInfo() {
 638                 return null;
 639         }
 640
 641         /**
 642          * Return a result set of hits on other (multiple) wikis associated with this one
 643          *
 644          * @return SearchResultSet
 645          */
 646         function getInterwikiResults() {
 647                 return null;
 648         }
 649
 650         /**
 651          * Check if there are results on other wikis
 652          *
 653          * @return Boolean
 654          */
 655         function hasInterwikiResults() {
 656                 return $this->getInterwikiResults() != null;
 657         }
 658
 659         /**
 660          * Fetches next search result, or false.
 661          * STUB
 662          *
 663          * @return SearchResult
 664          */
 665         function next() {
 666                 return false;
 667         }
 668
 669         /**
 670          * Frees the result set, if applicable.
 671          */
 672         function free() {
 673                 // ...
 674         }
 675
 676         /**
 677          * Did the search contain search syntax?  If so, Special:Search won't offer
 678          * the user a link to a create a page named by the search string because the
 679          * name would contain the search syntax.
 680          */
 681         public function searchContainedSyntax() {
 682                 return false;
 683         }
 684 }
 685
 686 /**
 687  * This class is used for different SQL-based search engines shipped with MediaWiki
 688  */
 689 class SqlSearchResultSet extends SearchResultSet {
 690
 691         protected $mResultSet;
 692
 693         function __construct( $resultSet, $terms ) {
 694                 $this->mResultSet = $resultSet;
 695                 $this->mTerms = $terms;
 696         }
 697
 698         function termMatches() {
 699                 return $this->mTerms;
 700         }
 701
 702         function numRows() {
 703                 if ( $this->mResultSet === false ) {
 704                         return false;
 705                 }
 706
 707                 return $this->mResultSet->numRows();
 708         }
 709
 710         function next() {
 711                 if ( $this->mResultSet === false ) {
 712                         return false;
 713                 }
 714
 715                 $row = $this->mResultSet->fetchObject();
 716                 if ( $row === false ) {
 717                         return false;
 718                 }
 719
 720                 return SearchResult::newFromRow( $row );
 721         }
 722
 723         function free() {
 724                 if ( $this->mResultSet === false ) {
 725                         return false;
 726                 }
 727
 728                 $this->mResultSet->free();
 729         }
 730 }
 731
 732 /**
 733  * @ingroup Search
 734  */
 735 class SearchResultTooMany {
 736         # # Some search engines may bail out if too many matches are found
 737 }
 738
 739 /**
 740  * @todo FIXME: This class is horribly factored. It would probably be better to
 741  * have a useful base class to which you pass some standard information, then
 742  * let the fancy self-highlighters extend that.
 743  * @ingroup Search
 744  */
 745 class SearchResult {
 746
 747         /**
 748          * @var Revision
 749          */
 750         protected $mRevision = null;
 751
 752         /**
 753          * @var File
 754          */
 755         protected $mImage = null;
 756
 757         /**
 758          * @var Title
 759          */
 760         protected $mTitle;
 761
 762         /**
 763          * @var String
 764          */
 765         protected $mText;
 766
 767         /**
 768          * Return a new SearchResult and initializes it with a title.
 769          *
 770          * @param $title Title
 771          * @return SearchResult
 772          */
 773         public static function newFromTitle( $title ) {
 774                 $result = new self();
 775                 $result->initFromTitle( $title );
 776                 return $result;
 777         }
 778         /**
 779          * Return a new SearchResult and initializes it with a row.
 780          *
 781          * @param $row object
 782          * @return SearchResult
 783          */
 784         public static function newFromRow( $row ) {
 785                 $result = new self();
 786                 $result->initFromRow( $row );
 787                 return $result;
 788         }
 789
 790         public function __construct( $row = null ) {
 791                 if ( !is_null( $row ) ) {
 792                         // Backwards compatibility with pre-1.17 callers
 793                         $this->initFromRow( $row );
 794                 }
 795         }
 796
 797         /**
 798          * Initialize from a database row. Makes a Title and passes that to
 799          * initFromTitle.
 800          *
 801          * @param $row object
 802          */
 803         protected function initFromRow( $row ) {
 804                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 805         }
 806
 807         /**
 808          * Initialize from a Title and if possible initializes a corresponding
 809          * Revision and File.
 810          *
 811          * @param $title Title
 812          */
 813         protected function initFromTitle( $title ) {
 814                 $this->mTitle = $title;
 815                 if ( !is_null( $this->mTitle ) ) {
 816                         $id = false;
 817                         wfRunHooks( 'SearchResultInitFromTitle', array( $title, &$id ) );
 818                         $this->mRevision = Revision::newFromTitle(
 819                                 $this->mTitle, $id, Revision::READ_NORMAL );
 820                         if ( $this->mTitle->getNamespace() === NS_FILE ) {
 821                                 $this->mImage = wfFindFile( $this->mTitle );
 822                         }
 823                 }
 824         }
 825
 826         /**
 827          * Check if this is result points to an invalid title
 828          *
 829          * @return Boolean
 830          */
 831         function isBrokenTitle() {
 832                 return is_null( $this->mTitle );
 833         }
 834
 835         /**
 836          * Check if target page is missing, happens when index is out of date
 837          *
 838          * @return Boolean
 839          */
 840         function isMissingRevision() {
 841                 return !$this->mRevision && !$this->mImage;
 842         }
 843
 844         /**
 845          * @return Title
 846          */
 847         function getTitle() {
 848                 return $this->mTitle;
 849         }
 850
 851         /**
 852          * Get the file for this page, if one exists
 853          * @return File|null
 854          */
 855         function getFile() {
 856                 return $this->mImage;
 857         }
 858
 859         /**
 860          * @return float|null if not supported
 861          */
 862         function getScore() {
 863                 return null;
 864         }
 865
 866         /**
 867          * Lazy initialization of article text from DB
 868          */
 869         protected function initText() {
 870                 if ( !isset( $this->mText ) ) {
 871                         if ( $this->mRevision != null ) {
 872                                 $this->mText = SearchEngine::create()
 873                                         ->getTextFromContent( $this->mTitle, $this->mRevision->getContent() );
 874                         } else { // TODO: can we fetch raw wikitext for commons images?
 875                                 $this->mText = '';
 876                         }
 877                 }
 878         }
 879
 880         /**
 881          * @param array $terms terms to highlight
 882          * @return String: highlighted text snippet, null (and not '') if not supported
 883          */
 884         function getTextSnippet( $terms ) {
 885                 global $wgAdvancedSearchHighlighting;
 886                 $this->initText();
 887
 888                 // TODO: make highliter take a content object. Make ContentHandler a factory for SearchHighliter.
 889                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs();
 890                 $h = new SearchHighlighter();
 891                 if ( $wgAdvancedSearchHighlighting ) {
 892                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 893                 } else {
 894                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 895                 }
 896         }
 897
 898         /**
 899          * @return String: highlighted title, '' if not supported
 900          */
 901         function getTitleSnippet() {
 902                 return '';
 903         }
 904
 905         /**
 906          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 907          */
 908         function getRedirectSnippet() {
 909                 return '';
 910         }
 911
 912         /**
 913          * @return Title object for the redirect to this page, null if none or not supported
 914          */
 915         function getRedirectTitle() {
 916                 return null;
 917         }
 918
 919         /**
 920          * @return string highlighted relevant section name, null if none or not supported
 921          */
 922         function getSectionSnippet() {
 923                 return '';
 924         }
 925
 926         /**
 927          * @return Title object (pagename+fragment) for the section, null if none or not supported
 928          */
 929         function getSectionTitle() {
 930                 return null;
 931         }
 932
 933         /**
 934          * @return String: timestamp
 935          */
 936         function getTimestamp() {
 937                 if ( $this->mRevision ) {
 938                         return $this->mRevision->getTimestamp();
 939                 } elseif ( $this->mImage ) {
 940                         return $this->mImage->getTimestamp();
 941                 }
 942                 return '';
 943         }
 944
 945         /**
 946          * @return Integer: number of words
 947          */
 948         function getWordCount() {
 949                 $this->initText();
 950                 return str_word_count( $this->mText );
 951         }
 952
 953         /**
 954          * @return Integer: size in bytes
 955          */
 956         function getByteSize() {
 957                 $this->initText();
 958                 return strlen( $this->mText );
 959         }
 960
 961         /**
 962          * @return Boolean if hit has related articles
 963          */
 964         function hasRelated() {
 965                 return false;
 966         }
 967
 968         /**
 969          * @return String: interwiki prefix of the title (return iw even if title is broken)
 970          */
 971         function getInterwikiPrefix() {
 972                 return '';
 973         }
 974
 975         /**
 976          * @return string interwiki namespace of the title (since we likely can't resolve it locally)
 977          */
 978         function getInterwikiNamespaceText() {
 979                 return '';
 980         }
 981
 982         /**
 983          * Did this match file contents (eg: PDF/DJVU)?
 984          */
 985         function isFileMatch() {
 986                 return false;
 987         }
 988 }
 989 /**
 990  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 991  */
 992 class SearchNearMatchResultSet extends SearchResultSet {
 993         private $fetched = false;
 994         /**
 995          * @param $match mixed Title if matched, else null
 996          */
 997         public function __construct( $match ) {
 998                 $this->result = $match;
 999         }
1000         public function hasResult() {
1001                 return (bool)$this->result;
1002         }
1003         public function numRows() {
1004                 return $this->hasResults() ? 1 : 0;
1005         }
1006         public function next() {
1007                 if ( $this->fetched || !$this->result ) {
1008                         return false;
1009                 }
1010                 $this->fetched = true;
1011                 return SearchResult::newFromTitle( $this->result );
1012         }
1013 }
1014
1015 /**
1016  * Highlight bits of wikitext
1017  *
1018  * @ingroup Search
1019  */
1020 class SearchHighlighter {
1021         var $mCleanWikitext = true;
1022
1023         function __construct( $cleanupWikitext = true ) {
1024                 $this->mCleanWikitext = $cleanupWikitext;
1025         }
1026
1027         /**
1028          * Default implementation of wikitext highlighting
1029          *
1030          * @param $text String
1031          * @param array $terms terms to highlight (unescaped)
1032          * @param $contextlines Integer
1033          * @param $contextchars Integer
1034          * @return String
1035          */
1036         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
1037                 global $wgContLang;
1038                 global $wgSearchHighlightBoundaries;
1039                 $fname = __METHOD__;
1040
1041                 if ( $text == '' ) {
1042                         return '';
1043                 }
1044
1045                 // spli text into text + templates/links/tables
1046                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
1047                 // first capture group is for detecting nested templates/links/tables/references
1048                 $endPatterns = array(
1049                         1 => '/(\{\{)|(\}\})/', // template
1050                         2 => '/(\[\[)|(\]\])/', // image
1051                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
1052
1053                 // @todo FIXME: This should prolly be a hook or something
1054                 if ( function_exists( 'wfCite' ) ) {
1055                         $spat .= '|(<ref>)'; // references via cite extension
1056                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
1057                 }
1058                 $spat .= '/';
1059                 $textExt = array(); // text extracts
1060                 $otherExt = array(); // other extracts
1061                 wfProfileIn( "$fname-split" );
1062                 $start = 0;
1063                 $textLen = strlen( $text );
1064                 $count = 0; // sequence number to maintain ordering
1065                 while ( $start < $textLen ) {
1066                         // find start of template/image/table
1067                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
1068                                 $epat = '';
1069                                 foreach ( $matches as $key => $val ) {
1070                                         if ( $key > 0 && $val[1] != - 1 ) {
1071                                                 if ( $key == 2 ) {
1072                                                         // see if this is an image link
1073                                                         $ns = substr( $val[0], 2, - 1 );
1074                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE ) {
1075                                                                 break;
1076                                                         }
1077
1078                                                 }
1079                                                 $epat = $endPatterns[$key];
1080                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
1081                                                 $start = $val[1];
1082                                                 break;
1083                                         }
1084                                 }
1085                                 if ( $epat ) {
1086                                         // find end (and detect any nested elements)
1087                                         $level = 0;
1088                                         $offset = $start + 1;
1089                                         $found = false;
1090                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
1091                                                 if ( array_key_exists( 2, $endMatches ) ) {
1092                                                         // found end
1093                                                         if ( $level == 0 ) {
1094                                                                 $len = strlen( $endMatches[2][0] );
1095                                                                 $off = $endMatches[2][1];
1096                                                                 $this->splitAndAdd( $otherExt, $count,
1097                                                                         substr( $text, $start, $off + $len - $start ) );
1098                                                                 $start = $off + $len;
1099                                                                 $found = true;
1100                                                                 break;
1101                                                         } else {
1102                                                                 // end of nested element
1103                                                                 $level -= 1;
1104                                                         }
1105                                                 } else {
1106                                                         // nested
1107                                                         $level += 1;
1108                                                 }
1109                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1110                                         }
1111                                         if ( ! $found ) {
1112                                                 // couldn't find appropriate closing tag, skip
1113                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1114                                                 $start += strlen( $matches[0][0] );
1115                                         }
1116                                         continue;
1117                                 }
1118                         }
1119                         // else: add as text extract
1120                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1121                         break;
1122                 }
1123
1124                 $all = $textExt + $otherExt; // these have disjunct key sets
1125
1126                 wfProfileOut( "$fname-split" );
1127
1128                 // prepare regexps
1129                 foreach ( $terms as $index => $term ) {
1130                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1131                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1132                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1133                         } else {
1134                                 $terms[$index] = $term;
1135                         }
1136                 }
1137                 $anyterm = implode( '|', $terms );
1138                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1139
1140                 // @todo FIXME: A hack to scale contextchars, a correct solution
1141                 // would be to have contextchars actually be char and not byte
1142                 // length, and do proper utf-8 substrings and lengths everywhere,
1143                 // but PHP is making that very hard and unclean to implement :(
1144                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1145                 $contextchars = intval( $contextchars * $scale );
1146
1147                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1148                 $patPost = "($wgSearchHighlightBoundaries|$)";
1149
1150                 $pat1 = "/(" . $phrase . ")/ui";
1151                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1152
1153                 wfProfileIn( "$fname-extract" );
1154
1155                 $left = $contextlines;
1156
1157                 $snippets = array();
1158                 $offsets = array();
1159
1160                 // show beginning only if it contains all words
1161                 $first = 0;
1162                 $firstText = '';
1163                 foreach ( $textExt as $index => $line ) {
1164                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1165                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1166                                 $first = $index;
1167                                 break;
1168                         }
1169                 }
1170                 if ( $firstText ) {
1171                         $succ = true;
1172                         // check if first text contains all terms
1173                         foreach ( $terms as $term ) {
1174                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1175                                         $succ = false;
1176                                         break;
1177                                 }
1178                         }
1179                         if ( $succ ) {
1180                                 $snippets[$first] = $firstText;
1181                                 $offsets[$first] = 0;
1182                         }
1183                 }
1184                 if ( ! $snippets ) {
1185                         // match whole query on text
1186                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1187                         // match whole query on templates/tables/images
1188                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1189                         // match any words on text
1190                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1191                         // match any words on templates/tables/images
1192                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1193
1194                         ksort( $snippets );
1195                 }
1196
1197                 // add extra chars to each snippet to make snippets constant size
1198                 $extended = array();
1199                 if ( count( $snippets ) == 0 ) {
1200                         // couldn't find the target words, just show beginning of article
1201                         if ( array_key_exists( $first, $all ) ) {
1202                                 $targetchars = $contextchars * $contextlines;
1203                                 $snippets[$first] = '';
1204                                 $offsets[$first] = 0;
1205                         }
1206                 } else {
1207                         // if begin of the article contains the whole phrase, show only that !!
1208                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1209                                 && $offsets[$first] < $contextchars * 2 ) {
1210                                 $snippets = array( $first => $snippets[$first] );
1211                         }
1212
1213                         // calc by how much to extend existing snippets
1214                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1215                 }
1216
1217                 foreach ( $snippets as $index => $line ) {
1218                         $extended[$index] = $line;
1219                         $len = strlen( $line );
1220                         if ( $len < $targetchars - 20 ) {
1221                                 // complete this line
1222                                 if ( $len < strlen( $all[$index] ) ) {
1223                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1224                                         $len = strlen( $extended[$index] );
1225                                 }
1226
1227                                 // add more lines
1228                                 $add = $index + 1;
1229                                 while ( $len < $targetchars - 20
1230                                                 && array_key_exists( $add, $all )
1231                                                 && !array_key_exists( $add, $snippets ) ) {
1232                                         $offsets[$add] = 0;
1233                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1234                                         $extended[$add] = $tt;
1235                                         $len += strlen( $tt );
1236                                         $add++;
1237                                 }
1238                         }
1239                 }
1240
1241                 // $snippets = array_map( 'htmlspecialchars', $extended );
1242                 $snippets = $extended;
1243                 $last = - 1;
1244                 $extract = '';
1245                 foreach ( $snippets as $index => $line ) {
1246                         if ( $last == - 1 ) {
1247                                 $extract .= $line; // first line
1248                         } elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) ) {
1249                                 $extract .= " " . $line; // continous lines
1250                         } else {
1251                                 $extract .= '<b> ... </b>' . $line;
1252                         }
1253
1254                         $last = $index;
1255                 }
1256                 if ( $extract ) {
1257                         $extract .= '<b> ... </b>';
1258                 }
1259
1260                 $processed = array();
1261                 foreach ( $terms as $term ) {
1262                         if ( ! isset( $processed[$term] ) ) {
1263                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1264                                 $extract = preg_replace( $pat3,
1265                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1266                                 $processed[$term] = true;
1267                         }
1268                 }
1269
1270                 wfProfileOut( "$fname-extract" );
1271
1272                 return $extract;
1273         }
1274
1275         /**
1276          * Split text into lines and add it to extracts array
1277          *
1278          * @param array $extracts index -> $line
1279          * @param $count Integer
1280          * @param $text String
1281          */
1282         function splitAndAdd( &$extracts, &$count, $text ) {
1283                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1284                 foreach ( $split as $line ) {
1285                         $tt = trim( $line );
1286                         if ( $tt ) {
1287                                 $extracts[$count++] = $tt;
1288                         }
1289                 }
1290         }
1291
1292         /**
1293          * Do manual case conversion for non-ascii chars
1294          *
1295          * @param $matches Array
1296          * @return string
1297          */
1298         function caseCallback( $matches ) {
1299                 global $wgContLang;
1300                 if ( strlen( $matches[0] ) > 1 ) {
1301                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1302                 } else {
1303                         return $matches[0];
1304                 }
1305         }
1306
1307         /**
1308          * Extract part of the text from start to end, but by
1309          * not chopping up words
1310          * @param $text String
1311          * @param $start Integer
1312          * @param $end Integer
1313          * @param $posStart Integer: (out) actual start position
1314          * @param $posEnd Integer: (out) actual end position
1315          * @return String
1316          */
1317         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1318                 if ( $start != 0 ) {
1319                         $start = $this->position( $text, $start, 1 );
1320                 }
1321                 if ( $end >= strlen( $text ) ) {
1322                         $end = strlen( $text );
1323                 } else {
1324                         $end = $this->position( $text, $end );
1325                 }
1326
1327                 if ( !is_null( $posStart ) ) {
1328                         $posStart = $start;
1329                 }
1330                 if ( !is_null( $posEnd ) ) {
1331                         $posEnd = $end;
1332                 }
1333
1334                 if ( $end > $start ) {
1335                         return substr( $text, $start, $end - $start );
1336                 } else {
1337                         return '';
1338                 }
1339         }
1340
1341         /**
1342          * Find a nonletter near a point (index) in the text
1343          *
1344          * @param $text String
1345          * @param $point Integer
1346          * @param $offset Integer: offset to found index
1347          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1348          */
1349         function position( $text, $point, $offset = 0 ) {
1350                 $tolerance = 10;
1351                 $s = max( 0, $point - $tolerance );
1352                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1353                 $m = array();
1354                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1355                         return $m[0][1] + $s + $offset;
1356                 } else {
1357                         // check if point is on a valid first UTF8 char
1358                         $char = ord( $text[$point] );
1359                         while ( $char >= 0x80 && $char < 0xc0 ) {
1360                                 // skip trailing bytes
1361                                 $point++;
1362                                 if ( $point >= strlen( $text ) ) {
1363                                         return strlen( $text );
1364                                 }
1365                                 $char = ord( $text[$point] );
1366                         }
1367                         return $point;
1368
1369                 }
1370         }
1371
1372         /**
1373          * Search extracts for a pattern, and return snippets
1374          *
1375          * @param string $pattern regexp for matching lines
1376          * @param array $extracts extracts to search
1377          * @param $linesleft Integer: number of extracts to make
1378          * @param $contextchars Integer: length of snippet
1379          * @param array $out map for highlighted snippets
1380          * @param array $offsets map of starting points of snippets
1381          * @protected
1382          */
1383         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1384                 if ( $linesleft == 0 ) {
1385                         return; // nothing to do
1386                 }
1387                 foreach ( $extracts as $index => $line ) {
1388                         if ( array_key_exists( $index, $out ) ) {
1389                                 continue; // this line already highlighted
1390                         }
1391
1392                         $m = array();
1393                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) ) {
1394                                 continue;
1395                         }
1396
1397                         $offset = $m[0][1];
1398                         $len = strlen( $m[0][0] );
1399                         if ( $offset + $len < $contextchars ) {
1400                                 $begin = 0;
1401                         } elseif ( $len > $contextchars ) {
1402                                 $begin = $offset;
1403                         } else {
1404                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1405                         }
1406
1407                         $end = $begin + $contextchars;
1408
1409                         $posBegin = $begin;
1410                         // basic snippet from this line
1411                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1412                         $offsets[$index] = $posBegin;
1413                         $linesleft--;
1414                         if ( $linesleft == 0 ) {
1415                                 return;
1416                         }
1417                 }
1418         }
1419
1420         /**
1421          * Basic wikitext removal
1422          * @protected
1423          * @return mixed
1424          */
1425         function removeWiki( $text ) {
1426                 $fname = __METHOD__;
1427                 wfProfileIn( $fname );
1428
1429                 // $text = preg_replace( "/'{2,5}/", "", $text );
1430                 // $text = preg_replace( "/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text );
1431                 // $text = preg_replace( "/\[\[([^]|]+)\]\]/", "\\1", $text );
1432                 // $text = preg_replace( "/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text );
1433                 // $text = preg_replace( "/\\{\\|(.*?)\\|\\}/", "", $text );
1434                 // $text = preg_replace( "/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text );
1435                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1436                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1437                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1438                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1439                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1440                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1441                 $text = preg_replace( "/'''''/", "", $text );
1442                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1443                 $text = preg_replace( "/''/", "", $text );
1444
1445                 wfProfileOut( $fname );
1446                 return $text;
1447         }
1448
1449         /**
1450          * callback to replace [[target|caption]] kind of links, if
1451          * the target is category or image, leave it
1452          *
1453          * @param $matches Array
1454          */
1455         function linkReplace( $matches ) {
1456                 $colon = strpos( $matches[1], ':' );
1457                 if ( $colon === false ) {
1458                         return $matches[2]; // replace with caption
1459                 }
1460                 global $wgContLang;
1461                 $ns = substr( $matches[1], 0, $colon );
1462                 $index = $wgContLang->getNsIndex( $ns );
1463                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) ) {
1464                         return $matches[0]; // return the whole thing
1465                 } else {
1466                         return $matches[2];
1467                 }
1468         }
1469
1470         /**
1471          * Simple & fast snippet extraction, but gives completely unrelevant
1472          * snippets
1473          *
1474          * @param $text String
1475          * @param $terms Array
1476          * @param $contextlines Integer
1477          * @param $contextchars Integer
1478          * @return String
1479          */
1480         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1481                 global $wgContLang;
1482                 $fname = __METHOD__;
1483
1484                 $lines = explode( "\n", $text );
1485
1486                 $terms = implode( '|', $terms );
1487                 $max = intval( $contextchars ) + 1;
1488                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
1489
1490                 $lineno = 0;
1491
1492                 $extract = "";
1493                 wfProfileIn( "$fname-extract" );
1494                 foreach ( $lines as $line ) {
1495                         if ( 0 == $contextlines ) {
1496                                 break;
1497                         }
1498                         ++$lineno;
1499                         $m = array();
1500                         if ( ! preg_match( $pat1, $line, $m ) ) {
1501                                 continue;
1502                         }
1503                         --$contextlines;
1504                         // truncate function changes ... to relevant i18n message.
1505                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1506
1507                         if ( count( $m ) < 3 ) {
1508                                 $post = '';
1509                         } else {
1510                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1511                         }
1512
1513                         $found = $m[2];
1514
1515                         $line = htmlspecialchars( $pre . $found . $post );
1516                         $pat2 = '/(' . $terms . ")/i";
1517                         $line = preg_replace( $pat2, "<span class='searchmatch'>\\1</span>", $line );
1518
1519                         $extract .= "${line}\n";
1520                 }
1521                 wfProfileOut( "$fname-extract" );
1522
1523                 return $extract;
1524         }
1525
1526 }
1527
1528 /**
1529  * Dummy class to be used when non-supported Database engine is present.
1530  * @todo FIXME: Dummy class should probably try something at least mildly useful,
1531  * such as a LIKE search through titles.
1532  * @ingroup Search
1533  */
1534 class SearchEngineDummy extends SearchEngine {
1535         // no-op
1536 }