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 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->getNamespace() == NS_SPECIAL || $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 = MediaWiki::articleFromTitle( $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                 } else if ( 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 a given user
 366          *
 367          * @param $user User
 368          * @return Array contextlines, contextchars
 369          */
 370         public static function userHighlightPrefs( &$user ) {
 371                 // $contextlines = $user->getOption( 'contextlines',  5 );
 372                 // $contextchars = $user->getOption( 'contextchars', 50 );
 373                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 374                 $contextchars = 75; // same as above.... :P
 375                 return array( $contextlines, $contextchars );
 376         }
 377
 378         /**
 379          * An array of namespaces indexes to be searched by default
 380          *
 381          * @return Array
 382          */
 383         public static function defaultNamespaces() {
 384                 global $wgNamespacesToBeSearchedDefault;
 385
 386                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 387         }
 388
 389         /**
 390          * Get a list of namespace names useful for showing in tooltips
 391          * and preferences
 392          *
 393          * @param $namespaces Array
 394          */
 395         public static function namespacesAsText( $namespaces ) {
 396                 global $wgContLang;
 397
 398                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 399                 foreach ( $formatted as $key => $ns ) {
 400                         if ( empty( $ns ) )
 401                                 $formatted[$key] = wfMsg( 'blanknamespace' );
 402                 }
 403                 return $formatted;
 404         }
 405
 406         /**
 407          * Return the help namespaces to be shown on Special:Search
 408          *
 409          * @return Array
 410          */
 411         public static function helpNamespaces() {
 412                 global $wgNamespacesToBeSearchedHelp;
 413
 414                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 415         }
 416
 417         /**
 418          * Return a 'cleaned up' search string
 419          *
 420          * @param $text String
 421          * @return String
 422          */
 423         function filter( $text ) {
 424                 $lc = $this->legalSearchChars();
 425                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 426         }
 427         /**
 428          * Load up the appropriate search engine class for the currently
 429          * active database backend, and return a configured instance.
 430          *
 431          * @return SearchEngine
 432          */
 433         public static function create() {
 434                 global $wgSearchType;
 435                 $dbr = null;
 436                 if ( $wgSearchType ) {
 437                         $class = $wgSearchType;
 438                 } else {
 439                         $dbr = wfGetDB( DB_SLAVE );
 440                         $class = $dbr->getSearchEngine();
 441                 }
 442                 $search = new $class( $dbr );
 443                 $search->setLimitOffset( 0, 0 );
 444                 return $search;
 445         }
 446
 447         /**
 448          * Create or update the search index record for the given page.
 449          * Title and text should be pre-processed.
 450          * STUB
 451          *
 452          * @param $id Integer
 453          * @param $title String
 454          * @param $text String
 455          */
 456         function update( $id, $title, $text ) {
 457                 // no-op
 458         }
 459
 460         /**
 461          * Update a search index record's title only.
 462          * Title should be pre-processed.
 463          * STUB
 464          *
 465          * @param $id Integer
 466          * @param $title String
 467          */
 468         function updateTitle( $id, $title ) {
 469                 // no-op
 470         }
 471
 472         /**
 473          * Get OpenSearch suggestion template
 474          *
 475          * @return String
 476          */
 477         public static function getOpenSearchTemplate() {
 478                 global $wgOpenSearchTemplate, $wgServer;
 479                 if ( $wgOpenSearchTemplate )    {
 480                         return $wgOpenSearchTemplate;
 481                 } else {
 482                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 483                         if ( !$ns ) $ns = "0";
 484                         return $wgServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
 485                 }
 486         }
 487
 488         /**
 489          * Get internal MediaWiki Suggest template
 490          *
 491          * @return String
 492          */
 493         public static function getMWSuggestTemplate() {
 494                 global $wgMWSuggestTemplate, $wgServer;
 495                 if ( $wgMWSuggestTemplate )
 496                         return $wgMWSuggestTemplate;
 497                 else
 498                         return $wgServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
 499         }
 500 }
 501
 502 /**
 503  * @ingroup Search
 504  */
 505 class SearchResultSet {
 506         /**
 507          * Fetch an array of regular expression fragments for matching
 508          * the search terms as parsed by this engine in a text extract.
 509          * STUB
 510          *
 511          * @return Array
 512          */
 513         function termMatches() {
 514                 return array();
 515         }
 516
 517         function numRows() {
 518                 return 0;
 519         }
 520
 521         /**
 522          * Return true if results are included in this result set.
 523          * STUB
 524          *
 525          * @return Boolean
 526          */
 527         function hasResults() {
 528                 return false;
 529         }
 530
 531         /**
 532          * Some search modes return a total hit count for the query
 533          * in the entire article database. This may include pages
 534          * in namespaces that would not be matched on the given
 535          * settings.
 536          *
 537          * Return null if no total hits number is supported.
 538          *
 539          * @return Integer
 540          */
 541         function getTotalHits() {
 542                 return null;
 543         }
 544
 545         /**
 546          * Some search modes return a suggested alternate term if there are
 547          * no exact hits. Returns true if there is one on this set.
 548          *
 549          * @return Boolean
 550          */
 551         function hasSuggestion() {
 552                 return false;
 553         }
 554
 555         /**
 556          * @return String: suggested query, null if none
 557          */
 558         function getSuggestionQuery() {
 559                 return null;
 560         }
 561
 562         /**
 563          * @return String: HTML highlighted suggested query, '' if none
 564          */
 565         function getSuggestionSnippet() {
 566                 return '';
 567         }
 568
 569         /**
 570          * Return information about how and from where the results were fetched,
 571          * should be useful for diagnostics and debugging
 572          *
 573          * @return String
 574          */
 575         function getInfo() {
 576                 return null;
 577         }
 578
 579         /**
 580          * Return a result set of hits on other (multiple) wikis associated with this one
 581          *
 582          * @return SearchResultSet
 583          */
 584         function getInterwikiResults() {
 585                 return null;
 586         }
 587
 588         /**
 589          * Check if there are results on other wikis
 590          *
 591          * @return Boolean
 592          */
 593         function hasInterwikiResults() {
 594                 return $this->getInterwikiResults() != null;
 595         }
 596
 597         /**
 598          * Fetches next search result, or false.
 599          * STUB
 600          *
 601          * @return SearchResult
 602          */
 603         function next() {
 604                 return false;
 605         }
 606
 607         /**
 608          * Frees the result set, if applicable.
 609          */
 610         function free() {
 611                 // ...
 612         }
 613 }
 614
 615 /**
 616  * This class is used for different SQL-based search engines shipped with MediaWiki
 617  */
 618 class SqlSearchResultSet extends SearchResultSet {
 619
 620         private $mResultSet;
 621
 622         function __construct( $resultSet, $terms ) {
 623                 $this->mResultSet = $resultSet;
 624                 $this->mTerms = $terms;
 625         }
 626
 627         function termMatches() {
 628                 return $this->mTerms;
 629         }
 630
 631         function numRows() {
 632                 if ( $this->mResultSet === false )
 633                         return false;
 634
 635                 return $this->mResultSet->numRows();
 636         }
 637
 638         function next() {
 639                 if ( $this->mResultSet === false )
 640                         return false;
 641
 642                 $row = $this->mResultSet->fetchObject();
 643                 if ( $row === false )
 644                         return false;
 645
 646                 return SearchResult::newFromRow( $row );
 647         }
 648
 649         function free() {
 650                 if ( $this->mResultSet === false )
 651                         return false;
 652
 653                 $this->mResultSet->free();
 654         }
 655 }
 656
 657 /**
 658  * @ingroup Search
 659  */
 660 class SearchResultTooMany {
 661         # # Some search engines may bail out if too many matches are found
 662 }
 663
 664
 665 /**
 666  * @todo Fixme: This class is horribly factored. It would probably be better to
 667  * have a useful base class to which you pass some standard information, then
 668  * let the fancy self-highlighters extend that.
 669  * @ingroup Search
 670  */
 671 class SearchResult {
 672
 673         /**
 674          * @var Revision
 675          */
 676         var $mRevision = null;
 677         var $mImage = null;
 678
 679         /**
 680          * @var Title
 681          */
 682         var $mTitle;
 683
 684         /**
 685          * @var String
 686          */
 687         var $mText;
 688
 689         /**
 690          * Return a new SearchResult and initializes it with a title.
 691          *
 692          * @param $title Title
 693          * @return SearchResult
 694          */
 695         public static function newFromTitle( $title ) {
 696                 $result = new self();
 697                 $result->initFromTitle( $title );
 698                 return $result;
 699         }
 700         /**
 701          * Return a new SearchResult and initializes it with a row.
 702          *
 703          * @param $row object
 704          * @return SearchResult
 705          */
 706         public static function newFromRow( $row ) {
 707                 $result = new self();
 708                 $result->initFromRow( $row );
 709                 return $result;
 710         }
 711
 712         public function __construct( $row = null ) {
 713                 if ( !is_null( $row ) ) {
 714                         // Backwards compatibility with pre-1.17 callers
 715                         $this->initFromRow( $row );
 716                 }
 717         }
 718
 719         /**
 720          * Initialize from a database row. Makes a Title and passes that to
 721          * initFromTitle.
 722          *
 723          * @param $row object
 724          */
 725         protected function initFromRow( $row ) {
 726                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 727         }
 728
 729         /**
 730          * Initialize from a Title and if possible initializes a corresponding
 731          * Revision and File.
 732          *
 733          * @param $title Title
 734          */
 735         protected function initFromTitle( $title ) {
 736                 $this->mTitle = $title;
 737                 if ( !is_null( $this->mTitle ) ) {
 738                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 739                         if ( $this->mTitle->getNamespace() === NS_FILE )
 740                                 $this->mImage = wfFindFile( $this->mTitle );
 741                 }
 742         }
 743
 744         /**
 745          * Check if this is result points to an invalid title
 746          *
 747          * @return Boolean
 748          */
 749         function isBrokenTitle() {
 750                 if ( is_null( $this->mTitle ) )
 751                         return true;
 752                 return false;
 753         }
 754
 755         /**
 756          * Check if target page is missing, happens when index is out of date
 757          *
 758          * @return Boolean
 759          */
 760         function isMissingRevision() {
 761                 return !$this->mRevision && !$this->mImage;
 762         }
 763
 764         /**
 765          * @return Title
 766          */
 767         function getTitle() {
 768                 return $this->mTitle;
 769         }
 770
 771         /**
 772          * @return Double or null if not supported
 773          */
 774         function getScore() {
 775                 return null;
 776         }
 777
 778         /**
 779          * Lazy initialization of article text from DB
 780          */
 781         protected function initText() {
 782                 if ( !isset( $this->mText ) ) {
 783                         if ( $this->mRevision != null )
 784                                 $this->mText = $this->mRevision->getText();
 785                         else // TODO: can we fetch raw wikitext for commons images?
 786                                 $this->mText = '';
 787
 788                 }
 789         }
 790
 791         /**
 792          * @param $terms Array: terms to highlight
 793          * @return String: highlighted text snippet, null (and not '') if not supported
 794          */
 795         function getTextSnippet( $terms ) {
 796                 global $wgUser, $wgAdvancedSearchHighlighting;
 797                 $this->initText();
 798                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
 799                 $h = new SearchHighlighter();
 800                 if ( $wgAdvancedSearchHighlighting )
 801                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 802                 else
 803                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 804         }
 805
 806         /**
 807          * @param $terms Array: terms to highlight
 808          * @return String: highlighted title, '' if not supported
 809          */
 810         function getTitleSnippet( $terms ) {
 811                 return '';
 812         }
 813
 814         /**
 815          * @param $terms Array: terms to highlight
 816          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 817          */
 818         function getRedirectSnippet( $terms ) {
 819                 return '';
 820         }
 821
 822         /**
 823          * @return Title object for the redirect to this page, null if none or not supported
 824          */
 825         function getRedirectTitle() {
 826                 return null;
 827         }
 828
 829         /**
 830          * @return string highlighted relevant section name, null if none or not supported
 831          */
 832         function getSectionSnippet() {
 833                 return '';
 834         }
 835
 836         /**
 837          * @return Title object (pagename+fragment) for the section, null if none or not supported
 838          */
 839         function getSectionTitle() {
 840                 return null;
 841         }
 842
 843         /**
 844          * @return String: timestamp
 845          */
 846         function getTimestamp() {
 847                 if ( $this->mRevision )
 848                         return $this->mRevision->getTimestamp();
 849                 else if ( $this->mImage )
 850                         return $this->mImage->getTimestamp();
 851                 return '';
 852         }
 853
 854         /**
 855          * @return Integer: number of words
 856          */
 857         function getWordCount() {
 858                 $this->initText();
 859                 return str_word_count( $this->mText );
 860         }
 861
 862         /**
 863          * @return Integer: size in bytes
 864          */
 865         function getByteSize() {
 866                 $this->initText();
 867                 return strlen( $this->mText );
 868         }
 869
 870         /**
 871          * @return Boolean if hit has related articles
 872          */
 873         function hasRelated() {
 874                 return false;
 875         }
 876
 877         /**
 878          * @return String: interwiki prefix of the title (return iw even if title is broken)
 879          */
 880         function getInterwikiPrefix() {
 881                 return '';
 882         }
 883 }
 884 /**
 885  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 886  */
 887 class SearchNearMatchResultSet extends SearchResultSet {
 888         private $fetched = false;
 889         /**
 890          * @param $match mixed Title if matched, else null
 891          */
 892         public function __construct( $match ) {
 893                 $this->result = $match;
 894         }
 895         public function hasResult() {
 896                 return (bool)$this->result;
 897         }
 898         public function numRows() {
 899                 return $this->hasResults() ? 1 : 0;
 900         }
 901         public function next() {
 902                 if ( $this->fetched || !$this->result ) {
 903                         return false;
 904                 }
 905                 $this->fetched = true;
 906                 return SearchResult::newFromTitle( $this->result );
 907         }
 908 }
 909
 910 /**
 911  * Highlight bits of wikitext
 912  *
 913  * @ingroup Search
 914  */
 915 class SearchHighlighter {
 916         var $mCleanWikitext = true;
 917
 918         function __construct( $cleanupWikitext = true ) {
 919                 $this->mCleanWikitext = $cleanupWikitext;
 920         }
 921
 922         /**
 923          * Default implementation of wikitext highlighting
 924          *
 925          * @param $text String
 926          * @param $terms Array: terms to highlight (unescaped)
 927          * @param $contextlines Integer
 928          * @param $contextchars Integer
 929          * @return String
 930          */
 931         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 932                 global $wgContLang;
 933                 global $wgSearchHighlightBoundaries;
 934                 $fname = __METHOD__;
 935
 936                 if ( $text == '' )
 937                         return '';
 938
 939                 // spli text into text + templates/links/tables
 940                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 941                 // first capture group is for detecting nested templates/links/tables/references
 942                 $endPatterns = array(
 943                         1 => '/(\{\{)|(\}\})/', // template
 944                         2 => '/(\[\[)|(\]\])/', // image
 945                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
 946
 947                 // FIXME: this should prolly be a hook or something
 948                 if ( function_exists( 'wfCite' ) ) {
 949                         $spat .= '|(<ref>)'; // references via cite extension
 950                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 951                 }
 952                 $spat .= '/';
 953                 $textExt = array(); // text extracts
 954                 $otherExt = array();  // other extracts
 955                 wfProfileIn( "$fname-split" );
 956                 $start = 0;
 957                 $textLen = strlen( $text );
 958                 $count = 0; // sequence number to maintain ordering
 959                 while ( $start < $textLen ) {
 960                         // find start of template/image/table
 961                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
 962                                 $epat = '';
 963                                 foreach ( $matches as $key => $val ) {
 964                                         if ( $key > 0 && $val[1] != - 1 ) {
 965                                                 if ( $key == 2 ) {
 966                                                         // see if this is an image link
 967                                                         $ns = substr( $val[0], 2, - 1 );
 968                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE )
 969                                                                 break;
 970
 971                                                 }
 972                                                 $epat = $endPatterns[$key];
 973                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 974                                                 $start = $val[1];
 975                                                 break;
 976                                         }
 977                                 }
 978                                 if ( $epat ) {
 979                                         // find end (and detect any nested elements)
 980                                         $level = 0;
 981                                         $offset = $start + 1;
 982                                         $found = false;
 983                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
 984                                                 if ( array_key_exists( 2, $endMatches ) ) {
 985                                                         // found end
 986                                                         if ( $level == 0 ) {
 987                                                                 $len = strlen( $endMatches[2][0] );
 988                                                                 $off = $endMatches[2][1];
 989                                                                 $this->splitAndAdd( $otherExt, $count,
 990                                                                         substr( $text, $start, $off + $len  - $start ) );
 991                                                                 $start = $off + $len;
 992                                                                 $found = true;
 993                                                                 break;
 994                                                         } else {
 995                                                                 // end of nested element
 996                                                                 $level -= 1;
 997                                                         }
 998                                                 } else {
 999                                                         // nested
1000                                                         $level += 1;
1001                                                 }
1002                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1003                                         }
1004                                         if ( ! $found ) {
1005                                                 // couldn't find appropriate closing tag, skip
1006                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1007                                                 $start += strlen( $matches[0][0] );
1008                                         }
1009                                         continue;
1010                                 }
1011                         }
1012                         // else: add as text extract
1013                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1014                         break;
1015                 }
1016
1017                 $all = $textExt + $otherExt; // these have disjunct key sets
1018
1019                 wfProfileOut( "$fname-split" );
1020
1021                 // prepare regexps
1022                 foreach ( $terms as $index => $term ) {
1023                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1024                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1025                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1026                         } else {
1027                                 $terms[$index] = $term;
1028                         }
1029                 }
1030                 $anyterm = implode( '|', $terms );
1031                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1032
1033                 // FIXME: a hack to scale contextchars, a correct solution
1034                 // would be to have contextchars actually be char and not byte
1035                 // length, and do proper utf-8 substrings and lengths everywhere,
1036                 // but PHP is making that very hard and unclean to implement :(
1037                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1038                 $contextchars = intval( $contextchars * $scale );
1039
1040                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1041                 $patPost = "($wgSearchHighlightBoundaries|$)";
1042
1043                 $pat1 = "/(" . $phrase . ")/ui";
1044                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1045
1046                 wfProfileIn( "$fname-extract" );
1047
1048                 $left = $contextlines;
1049
1050                 $snippets = array();
1051                 $offsets = array();
1052
1053                 // show beginning only if it contains all words
1054                 $first = 0;
1055                 $firstText = '';
1056                 foreach ( $textExt as $index => $line ) {
1057                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1058                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1059                                 $first = $index;
1060                                 break;
1061                         }
1062                 }
1063                 if ( $firstText ) {
1064                         $succ = true;
1065                         // check if first text contains all terms
1066                         foreach ( $terms as $term ) {
1067                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1068                                         $succ = false;
1069                                         break;
1070                                 }
1071                         }
1072                         if ( $succ ) {
1073                                 $snippets[$first] = $firstText;
1074                                 $offsets[$first] = 0;
1075                         }
1076                 }
1077                 if ( ! $snippets ) {
1078                         // match whole query on text
1079                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1080                         // match whole query on templates/tables/images
1081                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1082                         // match any words on text
1083                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1084                         // match any words on templates/tables/images
1085                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1086
1087                         ksort( $snippets );
1088                 }
1089
1090                 // add extra chars to each snippet to make snippets constant size
1091                 $extended = array();
1092                 if ( count( $snippets ) == 0 ) {
1093                         // couldn't find the target words, just show beginning of article
1094                         if ( array_key_exists( $first, $all ) ) {
1095                                 $targetchars = $contextchars * $contextlines;
1096                                 $snippets[$first] = '';
1097                                 $offsets[$first] = 0;
1098                         }
1099                 } else {
1100                         // if begin of the article contains the whole phrase, show only that !!
1101                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1102                             && $offsets[$first] < $contextchars * 2 ) {
1103                                 $snippets = array ( $first => $snippets[$first] );
1104                         }
1105
1106                         // calc by how much to extend existing snippets
1107                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1108                 }
1109
1110                 foreach ( $snippets as $index => $line ) {
1111                         $extended[$index] = $line;
1112                         $len = strlen( $line );
1113                         if ( $len < $targetchars - 20 ) {
1114                                 // complete this line
1115                                 if ( $len < strlen( $all[$index] ) ) {
1116                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1117                                         $len = strlen( $extended[$index] );
1118                                 }
1119
1120                                 // add more lines
1121                                 $add = $index + 1;
1122                                 while ( $len < $targetchars - 20
1123                                        && array_key_exists( $add, $all )
1124                                        && !array_key_exists( $add, $snippets ) ) {
1125                                     $offsets[$add] = 0;
1126                                     $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1127                                         $extended[$add] = $tt;
1128                                         $len += strlen( $tt );
1129                                         $add++;
1130                                 }
1131                         }
1132                 }
1133
1134                 // $snippets = array_map('htmlspecialchars', $extended);
1135                 $snippets = $extended;
1136                 $last = - 1;
1137                 $extract = '';
1138                 foreach ( $snippets as $index => $line ) {
1139                         if ( $last == - 1 )
1140                                 $extract .= $line; // first line
1141                         elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) )
1142                                 $extract .= " " . $line; // continous lines
1143                         else
1144                                 $extract .= '<b> ... </b>' . $line;
1145
1146                         $last = $index;
1147                 }
1148                 if ( $extract )
1149                         $extract .= '<b> ... </b>';
1150
1151                 $processed = array();
1152                 foreach ( $terms as $term ) {
1153                         if ( ! isset( $processed[$term] ) ) {
1154                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1155                                 $extract = preg_replace( $pat3,
1156                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1157                                 $processed[$term] = true;
1158                         }
1159                 }
1160
1161                 wfProfileOut( "$fname-extract" );
1162
1163                 return $extract;
1164         }
1165
1166         /**
1167          * Split text into lines and add it to extracts array
1168          *
1169          * @param $extracts Array: index -> $line
1170          * @param $count Integer
1171          * @param $text String
1172          */
1173         function splitAndAdd( &$extracts, &$count, $text ) {
1174                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1175                 foreach ( $split as $line ) {
1176                         $tt = trim( $line );
1177                         if ( $tt )
1178                                 $extracts[$count++] = $tt;
1179                 }
1180         }
1181
1182         /**
1183          * Do manual case conversion for non-ascii chars
1184          *
1185          * @param $matches Array
1186          */
1187         function caseCallback( $matches ) {
1188                 global $wgContLang;
1189                 if ( strlen( $matches[0] ) > 1 ) {
1190                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1191                 } else
1192                         return $matches[0];
1193         }
1194
1195         /**
1196          * Extract part of the text from start to end, but by
1197          * not chopping up words
1198          * @param $text String
1199          * @param $start Integer
1200          * @param $end Integer
1201          * @param $posStart Integer: (out) actual start position
1202          * @param $posEnd Integer: (out) actual end position
1203          * @return String
1204          */
1205         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1206                 if ( $start != 0 )
1207                         $start = $this->position( $text, $start, 1 );
1208                 if ( $end >= strlen( $text ) )
1209                         $end = strlen( $text );
1210                 else
1211                         $end = $this->position( $text, $end );
1212
1213                 if ( !is_null( $posStart ) )
1214                         $posStart = $start;
1215                 if ( !is_null( $posEnd ) )
1216                         $posEnd = $end;
1217
1218                 if ( $end > $start )
1219                         return substr( $text, $start, $end - $start );
1220                 else
1221                         return '';
1222         }
1223
1224         /**
1225          * Find a nonletter near a point (index) in the text
1226          *
1227          * @param $text String
1228          * @param $point Integer
1229          * @param $offset Integer: offset to found index
1230          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1231          */
1232         function position( $text, $point, $offset = 0 ) {
1233                 $tolerance = 10;
1234                 $s = max( 0, $point - $tolerance );
1235                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1236                 $m = array();
1237                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1238                         return $m[0][1] + $s + $offset;
1239                 } else {
1240                         // check if point is on a valid first UTF8 char
1241                         $char = ord( $text[$point] );
1242                         while ( $char >= 0x80 && $char < 0xc0 ) {
1243                                 // skip trailing bytes
1244                                 $point++;
1245                                 if ( $point >= strlen( $text ) )
1246                                         return strlen( $text );
1247                                 $char = ord( $text[$point] );
1248                         }
1249                         return $point;
1250
1251                 }
1252         }
1253
1254         /**
1255          * Search extracts for a pattern, and return snippets
1256          *
1257          * @param $pattern String: regexp for matching lines
1258          * @param $extracts Array: extracts to search
1259          * @param $linesleft Integer: number of extracts to make
1260          * @param $contextchars Integer: length of snippet
1261          * @param $out Array: map for highlighted snippets
1262          * @param $offsets Array: map of starting points of snippets
1263          * @protected
1264          */
1265         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1266                 if ( $linesleft == 0 )
1267                         return; // nothing to do
1268                 foreach ( $extracts as $index => $line ) {
1269                         if ( array_key_exists( $index, $out ) )
1270                                 continue; // this line already highlighted
1271
1272                         $m = array();
1273                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1274                                 continue;
1275
1276                         $offset = $m[0][1];
1277                         $len = strlen( $m[0][0] );
1278                         if ( $offset + $len < $contextchars )
1279                                 $begin = 0;
1280                         elseif ( $len > $contextchars )
1281                                 $begin = $offset;
1282                         else
1283                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1284
1285                         $end = $begin + $contextchars;
1286
1287                         $posBegin = $begin;
1288                         // basic snippet from this line
1289                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1290                         $offsets[$index] = $posBegin;
1291                         $linesleft--;
1292                         if ( $linesleft == 0 )
1293                                 return;
1294                 }
1295         }
1296
1297         /**
1298          * Basic wikitext removal
1299          * @protected
1300          */
1301         function removeWiki( $text ) {
1302                 $fname = __METHOD__;
1303                 wfProfileIn( $fname );
1304
1305                 // $text = preg_replace("/'{2,5}/", "", $text);
1306                 // $text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1307                 // $text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1308                 // $text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1309                 // $text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1310                 // $text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1311                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1312                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1313                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1314                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1315                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1316                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1317                 $text = preg_replace( "/'''''/", "", $text );
1318                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1319                 $text = preg_replace( "/''/", "", $text );
1320
1321                 wfProfileOut( $fname );
1322                 return $text;
1323         }
1324
1325         /**
1326          * callback to replace [[target|caption]] kind of links, if
1327          * the target is category or image, leave it
1328          *
1329          * @param $matches Array
1330          */
1331         function linkReplace( $matches ) {
1332                 $colon = strpos( $matches[1], ':' );
1333                 if ( $colon === false )
1334                         return $matches[2]; // replace with caption
1335                 global $wgContLang;
1336                 $ns = substr( $matches[1], 0, $colon );
1337                 $index = $wgContLang->getNsIndex( $ns );
1338                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) )
1339                         return $matches[0]; // return the whole thing
1340                 else
1341                         return $matches[2];
1342
1343         }
1344
1345         /**
1346      * Simple & fast snippet extraction, but gives completely unrelevant
1347      * snippets
1348      *
1349      * @param $text String
1350      * @param $terms Array
1351      * @param $contextlines Integer
1352      * @param $contextchars Integer
1353      * @return String
1354      */
1355     public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1356         global $wgContLang;
1357         $fname = __METHOD__;
1358
1359         $lines = explode( "\n", $text );
1360
1361         $terms = implode( '|', $terms );
1362         $max = intval( $contextchars ) + 1;
1363         $pat1 = "/(.*)($terms)(.{0,$max})/i";
1364
1365         $lineno = 0;
1366
1367         $extract = "";
1368         wfProfileIn( "$fname-extract" );
1369         foreach ( $lines as $line ) {
1370             if ( 0 == $contextlines ) {
1371                 break;
1372             }
1373             ++$lineno;
1374             $m = array();
1375             if ( ! preg_match( $pat1, $line, $m ) ) {
1376                 continue;
1377             }
1378             --$contextlines;
1379             // truncate function changes ... to relevant i18n message.
1380             $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1381
1382             if ( count( $m ) < 3 ) {
1383                 $post = '';
1384             } else {
1385                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1386             }
1387
1388             $found = $m[2];
1389
1390             $line = htmlspecialchars( $pre . $found . $post );
1391             $pat2 = '/(' . $terms . ")/i";
1392             $line = preg_replace( $pat2,
1393               "<span class='searchmatch'>\\1</span>", $line );
1394
1395             $extract .= "${line}\n";
1396         }
1397         wfProfileOut( "$fname-extract" );
1398
1399         return $extract;
1400     }
1401
1402 }
1403
1404 /**
1405  * Dummy class to be used when non-supported Database engine is present.
1406  * @todo Fixme: dummy class should probably try something at least mildly useful,
1407  * such as a LIKE search through titles.
1408  * @ingroup Search
1409  */
1410 class SearchEngineDummy extends SearchEngine {
1411         // no-op
1412 }