includes/search/SearchEngine.php

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