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