includes/search/SearchEngine.php

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