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 $term String: 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 $term String: 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                         # Now try all lower case (i.e. first letter capitalized)
 206                         #
 207                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 208                         if ( $title && $title->exists() ) {
 209                                 return $title;
 210                         }
 211
 212                         # Now try capitalized string
 213                         #
 214                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 215                         if ( $title && $title->exists() ) {
 216                                 return $title;
 217                         }
 218
 219                         # Now try all upper case
 220                         #
 221                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 222                         if ( $title && $title->exists() ) {
 223                                 return $title;
 224                         }
 225
 226                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 227                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 228                         if ( $title && $title->exists() ) {
 229                                 return $title;
 230                         }
 231
 232                         // Give hooks a chance at better match variants
 233                         $title = null;
 234                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 235                                 return $title;
 236                         }
 237                 }
 238
 239                 $title = Title::newFromText( $searchterm );
 240
 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
 251                 # Entering a user goes to the user page whether it's there or not
 252                 if ( $title->getNamespace() == NS_USER ) {
 253                         return $title;
 254                 }
 255
 256                 # Go to images that exist even if there's no local page.
 257                 # There may have been a funny upload, or it may be on a shared
 258                 # file repository such as Wikimedia Commons.
 259                 if ( $title->getNamespace() == NS_FILE ) {
 260                         $image = wfFindFile( $title );
 261                         if ( $image ) {
 262                                 return $title;
 263                         }
 264                 }
 265
 266                 # MediaWiki namespace? Page may be "implied" if not customized.
 267                 # Just return it, with caps forced as the message system likes it.
 268                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 269                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 270                 }
 271
 272                 # Quoted term? Try without the quotes...
 273                 $matches = array();
 274                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 275                         return SearchEngine::getNearMatch( $matches[1] );
 276                 }
 277
 278                 return null;
 279         }
 280
 281         public static function legalSearchChars() {
 282                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 283         }
 284
 285         /**
 286          * Set the maximum number of results to return
 287          * and how many to skip before returning the first.
 288          *
 289          * @param $limit Integer
 290          * @param $offset Integer
 291          */
 292         function setLimitOffset( $limit, $offset = 0 ) {
 293                 $this->limit = intval( $limit );
 294                 $this->offset = intval( $offset );
 295         }
 296
 297         /**
 298          * Set which namespaces the search should include.
 299          * Give an array of namespace index numbers.
 300          *
 301          * @param $namespaces Array
 302          */
 303         function setNamespaces( $namespaces ) {
 304                 $this->namespaces = $namespaces;
 305         }
 306
 307         /**
 308          * Parse some common prefixes: all (search everything)
 309          * or namespace names
 310          *
 311          * @param $query String
 312          * @return string
 313          */
 314         function replacePrefixes( $query ) {
 315                 global $wgContLang;
 316
 317                 $parsed = $query;
 318                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 319                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 320                         return $parsed;
 321                 }
 322
 323                 $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
 324                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 325                         $this->namespaces = null;
 326                         $parsed = substr( $query, strlen( $allkeyword ) );
 327                 } elseif ( strpos( $query, ':' ) !== false ) {
 328                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 329                         $index = $wgContLang->getNsIndex( $prefix );
 330                         if ( $index !== false ) {
 331                                 $this->namespaces = array( $index );
 332                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 333                         }
 334                 }
 335                 if ( trim( $parsed ) == '' )
 336                         $parsed = $query; // prefix was the whole query
 337
 338                 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                 return $formatted;
 428         }
 429
 430         /**
 431          * Return the help namespaces to be shown on Special:Search
 432          *
 433          * @return Array
 434          */
 435         public static function helpNamespaces() {
 436                 global $wgNamespacesToBeSearchedHelp;
 437
 438                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 439         }
 440
 441         /**
 442          * Return a 'cleaned up' search string
 443          *
 444          * @param $text String
 445          * @return String
 446          */
 447         function filter( $text ) {
 448                 $lc = $this->legalSearchChars();
 449                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 450         }
 451         /**
 452          * Load up the appropriate search engine class for the currently
 453          * active database backend, and return a configured instance.
 454          *
 455          * @return SearchEngine
 456          */
 457         public static function create() {
 458                 global $wgSearchType;
 459                 $dbr = null;
 460                 if ( $wgSearchType ) {
 461                         $class = $wgSearchType;
 462                 } else {
 463                         $dbr = wfGetDB( DB_SLAVE );
 464                         $class = $dbr->getSearchEngine();
 465                 }
 466                 $search = new $class( $dbr );
 467                 $search->setLimitOffset( 0, 0 );
 468                 return $search;
 469         }
 470
 471         /**
 472          * Create or update the search index record for the given page.
 473          * Title and text should be pre-processed.
 474          * STUB
 475          *
 476          * @param $id Integer
 477          * @param $title String
 478          * @param $text String
 479          */
 480         function update( $id, $title, $text ) {
 481                 // no-op
 482         }
 483
 484         /**
 485          * Update a search index record's title only.
 486          * Title should be pre-processed.
 487          * STUB
 488          *
 489          * @param $id Integer
 490          * @param $title String
 491          */
 492         function updateTitle( $id, $title ) {
 493                 // no-op
 494         }
 495
 496         /**
 497          * Get OpenSearch suggestion template
 498          *
 499          * @return String
 500          */
 501         public static function getOpenSearchTemplate() {
 502                 global $wgOpenSearchTemplate, $wgCanonicalServer;
 503                 if ( $wgOpenSearchTemplate ) {
 504                         return $wgOpenSearchTemplate;
 505                 } else {
 506                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 507                         if ( !$ns ) {
 508                                 $ns = "0";
 509                         }
 510                         return $wgCanonicalServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
 511                 }
 512         }
 513 }
 514
 515 /**
 516  * @ingroup Search
 517  */
 518 class SearchResultSet {
 519         /**
 520          * Fetch an array of regular expression fragments for matching
 521          * the search terms as parsed by this engine in a text extract.
 522          * STUB
 523          *
 524          * @return Array
 525          */
 526         function termMatches() {
 527                 return array();
 528         }
 529
 530         function numRows() {
 531                 return 0;
 532         }
 533
 534         /**
 535          * Return true if results are included in this result set.
 536          * STUB
 537          *
 538          * @return Boolean
 539          */
 540         function hasResults() {
 541                 return false;
 542         }
 543
 544         /**
 545          * Some search modes return a total hit count for the query
 546          * in the entire article database. This may include pages
 547          * in namespaces that would not be matched on the given
 548          * settings.
 549          *
 550          * Return null if no total hits number is supported.
 551          *
 552          * @return Integer
 553          */
 554         function getTotalHits() {
 555                 return null;
 556         }
 557
 558         /**
 559          * Some search modes return a suggested alternate term if there are
 560          * no exact hits. Returns true if there is one on this set.
 561          *
 562          * @return Boolean
 563          */
 564         function hasSuggestion() {
 565                 return false;
 566         }
 567
 568         /**
 569          * @return String: suggested query, null if none
 570          */
 571         function getSuggestionQuery() {
 572                 return null;
 573         }
 574
 575         /**
 576          * @return String: HTML highlighted suggested query, '' if none
 577          */
 578         function getSuggestionSnippet() {
 579                 return '';
 580         }
 581
 582         /**
 583          * Return information about how and from where the results were fetched,
 584          * should be useful for diagnostics and debugging
 585          *
 586          * @return String
 587          */
 588         function getInfo() {
 589                 return null;
 590         }
 591
 592         /**
 593          * Return a result set of hits on other (multiple) wikis associated with this one
 594          *
 595          * @return SearchResultSet
 596          */
 597         function getInterwikiResults() {
 598                 return null;
 599         }
 600
 601         /**
 602          * Check if there are results on other wikis
 603          *
 604          * @return Boolean
 605          */
 606         function hasInterwikiResults() {
 607                 return $this->getInterwikiResults() != null;
 608         }
 609
 610         /**
 611          * Fetches next search result, or false.
 612          * STUB
 613          *
 614          * @return SearchResult
 615          */
 616         function next() {
 617                 return false;
 618         }
 619
 620         /**
 621          * Frees the result set, if applicable.
 622          */
 623         function free() {
 624                 // ...
 625         }
 626 }
 627
 628 /**
 629  * This class is used for different SQL-based search engines shipped with MediaWiki
 630  */
 631 class SqlSearchResultSet extends SearchResultSet {
 632
 633         protected $mResultSet;
 634
 635         function __construct( $resultSet, $terms ) {
 636                 $this->mResultSet = $resultSet;
 637                 $this->mTerms = $terms;
 638         }
 639
 640         function termMatches() {
 641                 return $this->mTerms;
 642         }
 643
 644         function numRows() {
 645                 if ( $this->mResultSet === false )
 646                         return false;
 647
 648                 return $this->mResultSet->numRows();
 649         }
 650
 651         function next() {
 652                 if ( $this->mResultSet === false )
 653                         return false;
 654
 655                 $row = $this->mResultSet->fetchObject();
 656                 if ( $row === false )
 657                         return false;
 658
 659                 return SearchResult::newFromRow( $row );
 660         }
 661
 662         function free() {
 663                 if ( $this->mResultSet === false )
 664                         return false;
 665
 666                 $this->mResultSet->free();
 667         }
 668 }
 669
 670 /**
 671  * @ingroup Search
 672  */
 673 class SearchResultTooMany {
 674         # # Some search engines may bail out if too many matches are found
 675 }
 676
 677
 678 /**
 679  * @todo FIXME: This class is horribly factored. It would probably be better to
 680  * have a useful base class to which you pass some standard information, then
 681  * let the fancy self-highlighters extend that.
 682  * @ingroup Search
 683  */
 684 class SearchResult {
 685
 686         /**
 687          * @var Revision
 688          */
 689         var $mRevision = null;
 690         var $mImage = null;
 691
 692         /**
 693          * @var Title
 694          */
 695         var $mTitle;
 696
 697         /**
 698          * @var String
 699          */
 700         var $mText;
 701
 702         /**
 703          * Return a new SearchResult and initializes it with a title.
 704          *
 705          * @param $title Title
 706          * @return SearchResult
 707          */
 708         public static function newFromTitle( $title ) {
 709                 $result = new self();
 710                 $result->initFromTitle( $title );
 711                 return $result;
 712         }
 713         /**
 714          * Return a new SearchResult and initializes it with a row.
 715          *
 716          * @param $row object
 717          * @return SearchResult
 718          */
 719         public static function newFromRow( $row ) {
 720                 $result = new self();
 721                 $result->initFromRow( $row );
 722                 return $result;
 723         }
 724
 725         public function __construct( $row = null ) {
 726                 if ( !is_null( $row ) ) {
 727                         // Backwards compatibility with pre-1.17 callers
 728                         $this->initFromRow( $row );
 729                 }
 730         }
 731
 732         /**
 733          * Initialize from a database row. Makes a Title and passes that to
 734          * initFromTitle.
 735          *
 736          * @param $row object
 737          */
 738         protected function initFromRow( $row ) {
 739                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 740         }
 741
 742         /**
 743          * Initialize from a Title and if possible initializes a corresponding
 744          * Revision and File.
 745          *
 746          * @param $title Title
 747          */
 748         protected function initFromTitle( $title ) {
 749                 $this->mTitle = $title;
 750                 if ( !is_null( $this->mTitle ) ) {
 751                         $id = false;
 752                         wfRunHooks( 'SearchResultInitFromTitle', array( $title, &$id ) );
 753                         $this->mRevision = Revision::newFromTitle(
 754                                 $this->mTitle, $id, Revision::READ_NORMAL );
 755                         if ( $this->mTitle->getNamespace() === NS_FILE )
 756                                 $this->mImage = wfFindFile( $this->mTitle );
 757                 }
 758         }
 759
 760         /**
 761          * Check if this is result points to an invalid title
 762          *
 763          * @return Boolean
 764          */
 765         function isBrokenTitle() {
 766                 if ( is_null( $this->mTitle ) )
 767                         return true;
 768                 return false;
 769         }
 770
 771         /**
 772          * Check if target page is missing, happens when index is out of date
 773          *
 774          * @return Boolean
 775          */
 776         function isMissingRevision() {
 777                 return !$this->mRevision && !$this->mImage;
 778         }
 779
 780         /**
 781          * @return Title
 782          */
 783         function getTitle() {
 784                 return $this->mTitle;
 785         }
 786
 787         /**
 788          * @return float|null if not supported
 789          */
 790         function getScore() {
 791                 return null;
 792         }
 793
 794         /**
 795          * Lazy initialization of article text from DB
 796          */
 797         protected function initText() {
 798                 if ( !isset( $this->mText ) ) {
 799                         if ( $this->mRevision != null ) {
 800                                 //TODO: if we could plug in some code that knows about special content models *and* about
 801                                 //      special features of the search engine, the search could benefit.
 802                                 $content = $this->mRevision->getContent();
 803                                 $this->mText = $content ? $content->getTextForSearchIndex() : '';
 804                         } else { // TODO: can we fetch raw wikitext for commons images?
 805                                 $this->mText = '';
 806                         }
 807                 }
 808         }
 809
 810         /**
 811          * @param $terms Array: terms to highlight
 812          * @return String: highlighted text snippet, null (and not '') if not supported
 813          */
 814         function getTextSnippet( $terms ) {
 815                 global $wgUser, $wgAdvancedSearchHighlighting;
 816                 $this->initText();
 817
 818                 // TODO: make highliter take a content object. Make ContentHandler a factory for SearchHighliter.
 819                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
 820                 $h = new SearchHighlighter();
 821                 if ( $wgAdvancedSearchHighlighting )
 822                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 823                 else
 824                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 825         }
 826
 827         /**
 828          * @param $terms Array: terms to highlight
 829          * @return String: highlighted title, '' if not supported
 830          */
 831         function getTitleSnippet( $terms ) {
 832                 return '';
 833         }
 834
 835         /**
 836          * @param $terms Array: terms to highlight
 837          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 838          */
 839         function getRedirectSnippet( $terms ) {
 840                 return '';
 841         }
 842
 843         /**
 844          * @return Title object for the redirect to this page, null if none or not supported
 845          */
 846         function getRedirectTitle() {
 847                 return null;
 848         }
 849
 850         /**
 851          * @return string highlighted relevant section name, null if none or not supported
 852          */
 853         function getSectionSnippet() {
 854                 return '';
 855         }
 856
 857         /**
 858          * @return Title object (pagename+fragment) for the section, null if none or not supported
 859          */
 860         function getSectionTitle() {
 861                 return null;
 862         }
 863
 864         /**
 865          * @return String: timestamp
 866          */
 867         function getTimestamp() {
 868                 if ( $this->mRevision )
 869                         return $this->mRevision->getTimestamp();
 870                 elseif ( $this->mImage )
 871                         return $this->mImage->getTimestamp();
 872                 return '';
 873         }
 874
 875         /**
 876          * @return Integer: number of words
 877          */
 878         function getWordCount() {
 879                 $this->initText();
 880                 return str_word_count( $this->mText );
 881         }
 882
 883         /**
 884          * @return Integer: size in bytes
 885          */
 886         function getByteSize() {
 887                 $this->initText();
 888                 return strlen( $this->mText );
 889         }
 890
 891         /**
 892          * @return Boolean if hit has related articles
 893          */
 894         function hasRelated() {
 895                 return false;
 896         }
 897
 898         /**
 899          * @return String: interwiki prefix of the title (return iw even if title is broken)
 900          */
 901         function getInterwikiPrefix() {
 902                 return '';
 903         }
 904 }
 905 /**
 906  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 907  */
 908 class SearchNearMatchResultSet extends SearchResultSet {
 909         private $fetched = false;
 910         /**
 911          * @param $match mixed Title if matched, else null
 912          */
 913         public function __construct( $match ) {
 914                 $this->result = $match;
 915         }
 916         public function hasResult() {
 917                 return (bool)$this->result;
 918         }
 919         public function numRows() {
 920                 return $this->hasResults() ? 1 : 0;
 921         }
 922         public function next() {
 923                 if ( $this->fetched || !$this->result ) {
 924                         return false;
 925                 }
 926                 $this->fetched = true;
 927                 return SearchResult::newFromTitle( $this->result );
 928         }
 929 }
 930
 931 /**
 932  * Highlight bits of wikitext
 933  *
 934  * @ingroup Search
 935  */
 936 class SearchHighlighter {
 937         var $mCleanWikitext = true;
 938
 939         function __construct( $cleanupWikitext = true ) {
 940                 $this->mCleanWikitext = $cleanupWikitext;
 941         }
 942
 943         /**
 944          * Default implementation of wikitext highlighting
 945          *
 946          * @param $text String
 947          * @param $terms Array: terms to highlight (unescaped)
 948          * @param $contextlines Integer
 949          * @param $contextchars Integer
 950          * @return String
 951          */
 952         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 953                 global $wgContLang;
 954                 global $wgSearchHighlightBoundaries;
 955                 $fname = __METHOD__;
 956
 957                 if ( $text == '' )
 958                         return '';
 959
 960                 // spli text into text + templates/links/tables
 961                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 962                 // first capture group is for detecting nested templates/links/tables/references
 963                 $endPatterns = array(
 964                         1 => '/(\{\{)|(\}\})/', // template
 965                         2 => '/(\[\[)|(\]\])/', // image
 966                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
 967
 968                 // @todo FIXME: This should prolly be a hook or something
 969                 if ( function_exists( 'wfCite' ) ) {
 970                         $spat .= '|(<ref>)'; // references via cite extension
 971                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 972                 }
 973                 $spat .= '/';
 974                 $textExt = array(); // text extracts
 975                 $otherExt = array();  // other extracts
 976                 wfProfileIn( "$fname-split" );
 977                 $start = 0;
 978                 $textLen = strlen( $text );
 979                 $count = 0; // sequence number to maintain ordering
 980                 while ( $start < $textLen ) {
 981                         // find start of template/image/table
 982                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
 983                                 $epat = '';
 984                                 foreach ( $matches as $key => $val ) {
 985                                         if ( $key > 0 && $val[1] != - 1 ) {
 986                                                 if ( $key == 2 ) {
 987                                                         // see if this is an image link
 988                                                         $ns = substr( $val[0], 2, - 1 );
 989                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE )
 990                                                                 break;
 991
 992                                                 }
 993                                                 $epat = $endPatterns[$key];
 994                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 995                                                 $start = $val[1];
 996                                                 break;
 997                                         }
 998                                 }
 999                                 if ( $epat ) {
1000                                         // find end (and detect any nested elements)
1001                                         $level = 0;
1002                                         $offset = $start + 1;
1003                                         $found = false;
1004                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
1005                                                 if ( array_key_exists( 2, $endMatches ) ) {
1006                                                         // found end
1007                                                         if ( $level == 0 ) {
1008                                                                 $len = strlen( $endMatches[2][0] );
1009                                                                 $off = $endMatches[2][1];
1010                                                                 $this->splitAndAdd( $otherExt, $count,
1011                                                                         substr( $text, $start, $off + $len  - $start ) );
1012                                                                 $start = $off + $len;
1013                                                                 $found = true;
1014                                                                 break;
1015                                                         } else {
1016                                                                 // end of nested element
1017                                                                 $level -= 1;
1018                                                         }
1019                                                 } else {
1020                                                         // nested
1021                                                         $level += 1;
1022                                                 }
1023                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1024                                         }
1025                                         if ( ! $found ) {
1026                                                 // couldn't find appropriate closing tag, skip
1027                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1028                                                 $start += strlen( $matches[0][0] );
1029                                         }
1030                                         continue;
1031                                 }
1032                         }
1033                         // else: add as text extract
1034                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1035                         break;
1036                 }
1037
1038                 $all = $textExt + $otherExt; // these have disjunct key sets
1039
1040                 wfProfileOut( "$fname-split" );
1041
1042                 // prepare regexps
1043                 foreach ( $terms as $index => $term ) {
1044                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1045                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1046                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1047                         } else {
1048                                 $terms[$index] = $term;
1049                         }
1050                 }
1051                 $anyterm = implode( '|', $terms );
1052                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1053
1054                 // @todo FIXME: A hack to scale contextchars, a correct solution
1055                 // would be to have contextchars actually be char and not byte
1056                 // length, and do proper utf-8 substrings and lengths everywhere,
1057                 // but PHP is making that very hard and unclean to implement :(
1058                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1059                 $contextchars = intval( $contextchars * $scale );
1060
1061                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1062                 $patPost = "($wgSearchHighlightBoundaries|$)";
1063
1064                 $pat1 = "/(" . $phrase . ")/ui";
1065                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1066
1067                 wfProfileIn( "$fname-extract" );
1068
1069                 $left = $contextlines;
1070
1071                 $snippets = array();
1072                 $offsets = array();
1073
1074                 // show beginning only if it contains all words
1075                 $first = 0;
1076                 $firstText = '';
1077                 foreach ( $textExt as $index => $line ) {
1078                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1079                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1080                                 $first = $index;
1081                                 break;
1082                         }
1083                 }
1084                 if ( $firstText ) {
1085                         $succ = true;
1086                         // check if first text contains all terms
1087                         foreach ( $terms as $term ) {
1088                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1089                                         $succ = false;
1090                                         break;
1091                                 }
1092                         }
1093                         if ( $succ ) {
1094                                 $snippets[$first] = $firstText;
1095                                 $offsets[$first] = 0;
1096                         }
1097                 }
1098                 if ( ! $snippets ) {
1099                         // match whole query on text
1100                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1101                         // match whole query on templates/tables/images
1102                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1103                         // match any words on text
1104                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1105                         // match any words on templates/tables/images
1106                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1107
1108                         ksort( $snippets );
1109                 }
1110
1111                 // add extra chars to each snippet to make snippets constant size
1112                 $extended = array();
1113                 if ( count( $snippets ) == 0 ) {
1114                         // couldn't find the target words, just show beginning of article
1115                         if ( array_key_exists( $first, $all ) ) {
1116                                 $targetchars = $contextchars * $contextlines;
1117                                 $snippets[$first] = '';
1118                                 $offsets[$first] = 0;
1119                         }
1120                 } else {
1121                         // if begin of the article contains the whole phrase, show only that !!
1122                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1123                                 && $offsets[$first] < $contextchars * 2 ) {
1124                                 $snippets = array ( $first => $snippets[$first] );
1125                         }
1126
1127                         // calc by how much to extend existing snippets
1128                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1129                 }
1130
1131                 foreach ( $snippets as $index => $line ) {
1132                         $extended[$index] = $line;
1133                         $len = strlen( $line );
1134                         if ( $len < $targetchars - 20 ) {
1135                                 // complete this line
1136                                 if ( $len < strlen( $all[$index] ) ) {
1137                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1138                                         $len = strlen( $extended[$index] );
1139                                 }
1140
1141                                 // add more lines
1142                                 $add = $index + 1;
1143                                 while ( $len < $targetchars - 20
1144                                            && array_key_exists( $add, $all )
1145                                            && !array_key_exists( $add, $snippets ) ) {
1146                                         $offsets[$add] = 0;
1147                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1148                                         $extended[$add] = $tt;
1149                                         $len += strlen( $tt );
1150                                         $add++;
1151                                 }
1152                         }
1153                 }
1154
1155                 // $snippets = array_map('htmlspecialchars', $extended);
1156                 $snippets = $extended;
1157                 $last = - 1;
1158                 $extract = '';
1159                 foreach ( $snippets as $index => $line ) {
1160                         if ( $last == - 1 )
1161                                 $extract .= $line; // first line
1162                         elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) )
1163                                 $extract .= " " . $line; // continous lines
1164                         else
1165                                 $extract .= '<b> ... </b>' . $line;
1166
1167                         $last = $index;
1168                 }
1169                 if ( $extract )
1170                         $extract .= '<b> ... </b>';
1171
1172                 $processed = array();
1173                 foreach ( $terms as $term ) {
1174                         if ( ! isset( $processed[$term] ) ) {
1175                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1176                                 $extract = preg_replace( $pat3,
1177                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1178                                 $processed[$term] = true;
1179                         }
1180                 }
1181
1182                 wfProfileOut( "$fname-extract" );
1183
1184                 return $extract;
1185         }
1186
1187         /**
1188          * Split text into lines and add it to extracts array
1189          *
1190          * @param $extracts Array: index -> $line
1191          * @param $count Integer
1192          * @param $text String
1193          */
1194         function splitAndAdd( &$extracts, &$count, $text ) {
1195                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1196                 foreach ( $split as $line ) {
1197                         $tt = trim( $line );
1198                         if ( $tt )
1199                                 $extracts[$count++] = $tt;
1200                 }
1201         }
1202
1203         /**
1204          * Do manual case conversion for non-ascii chars
1205          *
1206          * @param $matches Array
1207          * @return string
1208          */
1209         function caseCallback( $matches ) {
1210                 global $wgContLang;
1211                 if ( strlen( $matches[0] ) > 1 ) {
1212                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1213                 } else {
1214                         return $matches[0];
1215                 }
1216         }
1217
1218         /**
1219          * Extract part of the text from start to end, but by
1220          * not chopping up words
1221          * @param $text String
1222          * @param $start Integer
1223          * @param $end Integer
1224          * @param $posStart Integer: (out) actual start position
1225          * @param $posEnd Integer: (out) actual end position
1226          * @return String
1227          */
1228         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1229                 if ( $start != 0 ) {
1230                         $start = $this->position( $text, $start, 1 );
1231                 }
1232                 if ( $end >= strlen( $text ) ) {
1233                         $end = strlen( $text );
1234                 } else {
1235                         $end = $this->position( $text, $end );
1236                 }
1237
1238                 if ( !is_null( $posStart ) ) {
1239                         $posStart = $start;
1240                 }
1241                 if ( !is_null( $posEnd ) ) {
1242                         $posEnd = $end;
1243                 }
1244
1245                 if ( $end > $start )  {
1246                         return substr( $text, $start, $end - $start );
1247                 } else {
1248                         return '';
1249                 }
1250         }
1251
1252         /**
1253          * Find a nonletter near a point (index) in the text
1254          *
1255          * @param $text String
1256          * @param $point Integer
1257          * @param $offset Integer: offset to found index
1258          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1259          */
1260         function position( $text, $point, $offset = 0 ) {
1261                 $tolerance = 10;
1262                 $s = max( 0, $point - $tolerance );
1263                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1264                 $m = array();
1265                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1266                         return $m[0][1] + $s + $offset;
1267                 } else {
1268                         // check if point is on a valid first UTF8 char
1269                         $char = ord( $text[$point] );
1270                         while ( $char >= 0x80 && $char < 0xc0 ) {
1271                                 // skip trailing bytes
1272                                 $point++;
1273                                 if ( $point >= strlen( $text ) )
1274                                         return strlen( $text );
1275                                 $char = ord( $text[$point] );
1276                         }
1277                         return $point;
1278
1279                 }
1280         }
1281
1282         /**
1283          * Search extracts for a pattern, and return snippets
1284          *
1285          * @param $pattern String: regexp for matching lines
1286          * @param $extracts Array: extracts to search
1287          * @param $linesleft Integer: number of extracts to make
1288          * @param $contextchars Integer: length of snippet
1289          * @param $out Array: map for highlighted snippets
1290          * @param $offsets Array: map of starting points of snippets
1291          * @protected
1292          */
1293         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1294                 if ( $linesleft == 0 )
1295                         return; // nothing to do
1296                 foreach ( $extracts as $index => $line ) {
1297                         if ( array_key_exists( $index, $out ) )
1298                                 continue; // this line already highlighted
1299
1300                         $m = array();
1301                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1302                                 continue;
1303
1304                         $offset = $m[0][1];
1305                         $len = strlen( $m[0][0] );
1306                         if ( $offset + $len < $contextchars )
1307                                 $begin = 0;
1308                         elseif ( $len > $contextchars )
1309                                 $begin = $offset;
1310                         else
1311                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1312
1313                         $end = $begin + $contextchars;
1314
1315                         $posBegin = $begin;
1316                         // basic snippet from this line
1317                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1318                         $offsets[$index] = $posBegin;
1319                         $linesleft--;
1320                         if ( $linesleft == 0 )
1321                                 return;
1322                 }
1323         }
1324
1325         /**
1326          * Basic wikitext removal
1327          * @protected
1328          * @return mixed
1329          */
1330         function removeWiki( $text ) {
1331                 $fname = __METHOD__;
1332                 wfProfileIn( $fname );
1333
1334                 // $text = preg_replace("/'{2,5}/", "", $text);
1335                 // $text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1336                 // $text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1337                 // $text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1338                 // $text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1339                 // $text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1340                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1341                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1342                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1343                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1344                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1345                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1346                 $text = preg_replace( "/'''''/", "", $text );
1347                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1348                 $text = preg_replace( "/''/", "", $text );
1349
1350                 wfProfileOut( $fname );
1351                 return $text;
1352         }
1353
1354         /**
1355          * callback to replace [[target|caption]] kind of links, if
1356          * the target is category or image, leave it
1357          *
1358          * @param $matches Array
1359          */
1360         function linkReplace( $matches ) {
1361                 $colon = strpos( $matches[1], ':' );
1362                 if ( $colon === false )
1363                         return $matches[2]; // replace with caption
1364                 global $wgContLang;
1365                 $ns = substr( $matches[1], 0, $colon );
1366                 $index = $wgContLang->getNsIndex( $ns );
1367                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) )
1368                         return $matches[0]; // return the whole thing
1369                 else
1370                         return $matches[2];
1371
1372         }
1373
1374         /**
1375          * Simple & fast snippet extraction, but gives completely unrelevant
1376          * snippets
1377          *
1378          * @param $text String
1379          * @param $terms Array
1380          * @param $contextlines Integer
1381          * @param $contextchars Integer
1382          * @return String
1383          */
1384         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1385                 global $wgContLang;
1386                 $fname = __METHOD__;
1387
1388                 $lines = explode( "\n", $text );
1389
1390                 $terms = implode( '|', $terms );
1391                 $max = intval( $contextchars ) + 1;
1392                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
1393
1394                 $lineno = 0;
1395
1396                 $extract = "";
1397                 wfProfileIn( "$fname-extract" );
1398                 foreach ( $lines as $line ) {
1399                         if ( 0 == $contextlines ) {
1400                                 break;
1401                         }
1402                         ++$lineno;
1403                         $m = array();
1404                         if ( ! preg_match( $pat1, $line, $m ) ) {
1405                                 continue;
1406                         }
1407                         --$contextlines;
1408                         // truncate function changes ... to relevant i18n message.
1409                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1410
1411                         if ( count( $m ) < 3 ) {
1412                                 $post = '';
1413                         } else {
1414                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1415                         }
1416
1417                         $found = $m[2];
1418
1419                         $line = htmlspecialchars( $pre . $found . $post );
1420                         $pat2 = '/(' . $terms . ")/i";
1421                         $line = preg_replace( $pat2,
1422                           "<span class='searchmatch'>\\1</span>", $line );
1423
1424                         $extract .= "${line}\n";
1425                 }
1426                 wfProfileOut( "$fname-extract" );
1427
1428                 return $extract;
1429         }
1430
1431 }
1432
1433 /**
1434  * Dummy class to be used when non-supported Database engine is present.
1435  * @todo FIXME: Dummy class should probably try something at least mildly useful,
1436  * such as a LIKE search through titles.
1437  * @ingroup Search
1438  */
1439 class SearchEngineDummy extends SearchEngine {
1440         // no-op
1441 }