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