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