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          * @return bool
  69          */
  70         function acceptListRedirects() {
  71                 wfDeprecated( __METHOD__, '1.18' );
  72                 return $this->supports( 'list-redirects' );
  73         }
  74
  75         /**
  76          * @since 1.18
  77          * @param $feature String
  78          * @return Boolean
  79          */
  80         public function supports( $feature ) {
  81                 switch( $feature ) {
  82                 case 'list-redirects':
  83                         return true;
  84                 case 'title-suffix-filter':
  85                 default:
  86                         return false;
  87                 }
  88         }
  89
  90         /**
  91          * Way to pass custom data for engines
  92          * @since 1.18
  93          * @param $feature String
  94          * @param $data Mixed
  95          * @return bool
  96          */
  97         public function setFeatureData( $feature, $data ) {
  98                 $this->features[$feature] = $data;
  99         }
 100
 101         /**
 102          * When overridden in derived class, performs database-specific conversions
 103          * on text to be used for searching or updating search index.
 104          * Default implementation does nothing (simply returns $string).
 105          *
 106          * @param $string string: String to process
 107          * @return string
 108          */
 109         public function normalizeText( $string ) {
 110                 global $wgContLang;
 111
 112                 // Some languages such as Chinese require word segmentation
 113                 return $wgContLang->segmentByWord( $string );
 114         }
 115
 116         /**
 117          * Transform search term in cases when parts of the query came as different GET params (when supported)
 118          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
 119          */
 120         function transformSearchTerm( $term ) {
 121                 return $term;
 122         }
 123
 124         /**
 125          * If an exact title match can be found, or a very slightly close match,
 126          * return the title. If no match, returns NULL.
 127          *
 128          * @param $searchterm String
 129          * @return Title
 130          */
 131         public static function getNearMatch( $searchterm ) {
 132                 $title = self::getNearMatchInternal( $searchterm );
 133
 134                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
 135                 return $title;
 136         }
 137
 138         /**
 139          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 140          * SearchResultSet.
 141          *
 142          * @param $searchterm string
 143          * @return SearchResultSet
 144          */
 145         public static function getNearMatchResultSet( $searchterm ) {
 146                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 147         }
 148
 149         /**
 150          * Really find the title match.
 151          * @return null|\Title
 152          */
 153         private static function getNearMatchInternal( $searchterm ) {
 154                 global $wgContLang, $wgEnableSearchContributorsByIP;
 155
 156                 $allSearchTerms = array( $searchterm );
 157
 158                 if ( $wgContLang->hasVariants() ) {
 159                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 160                 }
 161
 162                 $titleResult = null;
 163                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 164                         return $titleResult;
 165                 }
 166
 167                 foreach ( $allSearchTerms as $term ) {
 168
 169                         # Exact match? No need to look further.
 170                         $title = Title::newFromText( $term );
 171                         if ( is_null( $title ) ){
 172                                 return null;
 173                         }
 174
 175                         if ( $title->isSpecialPage() || $title->isExternal() || $title->exists() ) {
 176                                 return $title;
 177                         }
 178
 179                         # See if it still otherwise has content is some sane sense
 180                         $page = WikiPage::factory( $title );
 181                         if ( $page->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
 222                 # Entering an IP address goes to the contributions page
 223                 if ( $wgEnableSearchContributorsByIP ) {
 224                         if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 225                                 || User::isIP( trim( $searchterm ) ) ) {
 226                                 return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 227                         }
 228                 }
 229
 230
 231                 # Entering a user goes to the user page whether it's there or not
 232                 if ( $title->getNamespace() == NS_USER ) {
 233                         return $title;
 234                 }
 235
 236                 # Go to images that exist even if there's no local page.
 237                 # There may have been a funny upload, or it may be on a shared
 238                 # file repository such as Wikimedia Commons.
 239                 if ( $title->getNamespace() == NS_FILE ) {
 240                         $image = wfFindFile( $title );
 241                         if ( $image ) {
 242                                 return $title;
 243                         }
 244                 }
 245
 246                 # MediaWiki namespace? Page may be "implied" if not customized.
 247                 # Just return it, with caps forced as the message system likes it.
 248                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 249                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 250                 }
 251
 252                 # Quoted term? Try without the quotes...
 253                 $matches = array();
 254                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 255                         return SearchEngine::getNearMatch( $matches[1] );
 256                 }
 257
 258                 return null;
 259         }
 260
 261         public static function legalSearchChars() {
 262                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 263         }
 264
 265         /**
 266          * Set the maximum number of results to return
 267          * and how many to skip before returning the first.
 268          *
 269          * @param $limit Integer
 270          * @param $offset Integer
 271          */
 272         function setLimitOffset( $limit, $offset = 0 ) {
 273                 $this->limit = intval( $limit );
 274                 $this->offset = intval( $offset );
 275         }
 276
 277         /**
 278          * Set which namespaces the search should include.
 279          * Give an array of namespace index numbers.
 280          *
 281          * @param $namespaces Array
 282          */
 283         function setNamespaces( $namespaces ) {
 284                 $this->namespaces = $namespaces;
 285         }
 286
 287         /**
 288          * Parse some common prefixes: all (search everything)
 289          * or namespace names
 290          *
 291          * @param $query String
 292          * @return string
 293          */
 294         function replacePrefixes( $query ) {
 295                 global $wgContLang;
 296
 297                 $parsed = $query;
 298                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 299                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 300                         return $parsed;
 301                 }
 302
 303                 $allkeyword = wfMsgForContent( 'searchall' ) . ":";
 304                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 305                         $this->namespaces = null;
 306                         $parsed = substr( $query, strlen( $allkeyword ) );
 307                 } elseif ( strpos( $query, ':' ) !== false ) {
 308                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 309                         $index = $wgContLang->getNsIndex( $prefix );
 310                         if ( $index !== false ) {
 311                                 $this->namespaces = array( $index );
 312                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 313                         }
 314                 }
 315                 if ( trim( $parsed ) == '' )
 316                         $parsed = $query; // prefix was the whole query
 317
 318                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 319
 320                 return $parsed;
 321         }
 322
 323         /**
 324          * Make a list of searchable namespaces and their canonical names.
 325          * @return Array
 326          */
 327         public static function searchableNamespaces() {
 328                 global $wgContLang;
 329                 $arr = array();
 330                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 331                         if ( $ns >= NS_MAIN ) {
 332                                 $arr[$ns] = $name;
 333                         }
 334                 }
 335
 336                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 337                 return $arr;
 338         }
 339
 340         /**
 341          * Extract default namespaces to search from the given user's
 342          * settings, returning a list of index numbers.
 343          *
 344          * @param $user User
 345          * @return Array
 346          */
 347         public static function userNamespaces( $user ) {
 348                 global $wgSearchEverythingOnlyLoggedIn;
 349
 350                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 351
 352                 // get search everything preference, that can be set to be read for logged-in users
 353                 // it overrides other options
 354                 if ( !$wgSearchEverythingOnlyLoggedIn || $user->isLoggedIn() ) {
 355                         if ( $user->getOption( 'searcheverything' ) ) {
 356                                 return array_keys( $searchableNamespaces );
 357                         }
 358                 }
 359
 360                 $arr = array();
 361                 foreach ( $searchableNamespaces as $ns => $name ) {
 362                         if ( $user->getOption( 'searchNs' . $ns ) ) {
 363                                 $arr[] = $ns;
 364                         }
 365                 }
 366
 367                 return $arr;
 368         }
 369
 370         /**
 371          * Find snippet highlight settings for all users
 372          *
 373          * @return Array contextlines, contextchars
 374          */
 375         public static function userHighlightPrefs() {
 376                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 377                 $contextchars = 75; // same as above.... :P
 378                 return array( $contextlines, $contextchars );
 379         }
 380
 381         /**
 382          * An array of namespaces indexes to be searched by default
 383          *
 384          * @return Array
 385          */
 386         public static function defaultNamespaces() {
 387                 global $wgNamespacesToBeSearchedDefault;
 388
 389                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 390         }
 391
 392         /**
 393          * Get a list of namespace names useful for showing in tooltips
 394          * and preferences
 395          *
 396          * @param $namespaces Array
 397          * @return array
 398          */
 399         public static function namespacesAsText( $namespaces ) {
 400                 global $wgContLang;
 401
 402                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 403                 foreach ( $formatted as $key => $ns ) {
 404                         if ( empty( $ns ) )
 405                                 $formatted[$key] = wfMsg( 'blanknamespace' );
 406                 }
 407                 return $formatted;
 408         }
 409
 410         /**
 411          * Return the help namespaces to be shown on Special:Search
 412          *
 413          * @return Array
 414          */
 415         public static function helpNamespaces() {
 416                 global $wgNamespacesToBeSearchedHelp;
 417
 418                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 419         }
 420
 421         /**
 422          * Return a 'cleaned up' search string
 423          *
 424          * @param $text String
 425          * @return String
 426          */
 427         function filter( $text ) {
 428                 $lc = $this->legalSearchChars();
 429                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 430         }
 431         /**
 432          * Load up the appropriate search engine class for the currently
 433          * active database backend, and return a configured instance.
 434          *
 435          * @return SearchEngine
 436          */
 437         public static function create() {
 438                 global $wgSearchType;
 439                 $dbr = null;
 440                 if ( $wgSearchType ) {
 441                         $class = $wgSearchType;
 442                 } else {
 443                         $dbr = wfGetDB( DB_SLAVE );
 444                         $class = $dbr->getSearchEngine();
 445                 }
 446                 $search = new $class( $dbr );
 447                 $search->setLimitOffset( 0, 0 );
 448                 return $search;
 449         }
 450
 451         /**
 452          * Create or update the search index record for the given page.
 453          * Title and text should be pre-processed.
 454          * STUB
 455          *
 456          * @param $id Integer
 457          * @param $title String
 458          * @param $text String
 459          */
 460         function update( $id, $title, $text ) {
 461                 // no-op
 462         }
 463
 464         /**
 465          * Update a search index record's title only.
 466          * Title should be pre-processed.
 467          * STUB
 468          *
 469          * @param $id Integer
 470          * @param $title String
 471          */
 472         function updateTitle( $id, $title ) {
 473                 // no-op
 474         }
 475
 476         /**
 477          * Get OpenSearch suggestion template
 478          *
 479          * @return String
 480          */
 481         public static function getOpenSearchTemplate() {
 482                 global $wgOpenSearchTemplate, $wgCanonicalServer;
 483                 if ( $wgOpenSearchTemplate ) {
 484                         return $wgOpenSearchTemplate;
 485                 } else {
 486                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 487                         if ( !$ns ) {
 488                                 $ns = "0";
 489                         }
 490                         return $wgCanonicalServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
 491                 }
 492         }
 493
 494         /**
 495          * Get internal MediaWiki Suggest template
 496          *
 497          * @return String
 498          */
 499         public static function getMWSuggestTemplate() {
 500                 global $wgMWSuggestTemplate, $wgServer;
 501                 if ( $wgMWSuggestTemplate )
 502                         return $wgMWSuggestTemplate;
 503                 else
 504                         return $wgServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
 505         }
 506 }
 507
 508 /**
 509  * @ingroup Search
 510  */
 511 class SearchResultSet {
 512         /**
 513          * Fetch an array of regular expression fragments for matching
 514          * the search terms as parsed by this engine in a text extract.
 515          * STUB
 516          *
 517          * @return Array
 518          */
 519         function termMatches() {
 520                 return array();
 521         }
 522
 523         function numRows() {
 524                 return 0;
 525         }
 526
 527         /**
 528          * Return true if results are included in this result set.
 529          * STUB
 530          *
 531          * @return Boolean
 532          */
 533         function hasResults() {
 534                 return false;
 535         }
 536
 537         /**
 538          * Some search modes return a total hit count for the query
 539          * in the entire article database. This may include pages
 540          * in namespaces that would not be matched on the given
 541          * settings.
 542          *
 543          * Return null if no total hits number is supported.
 544          *
 545          * @return Integer
 546          */
 547         function getTotalHits() {
 548                 return null;
 549         }
 550
 551         /**
 552          * Some search modes return a suggested alternate term if there are
 553          * no exact hits. Returns true if there is one on this set.
 554          *
 555          * @return Boolean
 556          */
 557         function hasSuggestion() {
 558                 return false;
 559         }
 560
 561         /**
 562          * @return String: suggested query, null if none
 563          */
 564         function getSuggestionQuery() {
 565                 return null;
 566         }
 567
 568         /**
 569          * @return String: HTML highlighted suggested query, '' if none
 570          */
 571         function getSuggestionSnippet() {
 572                 return '';
 573         }
 574
 575         /**
 576          * Return information about how and from where the results were fetched,
 577          * should be useful for diagnostics and debugging
 578          *
 579          * @return String
 580          */
 581         function getInfo() {
 582                 return null;
 583         }
 584
 585         /**
 586          * Return a result set of hits on other (multiple) wikis associated with this one
 587          *
 588          * @return SearchResultSet
 589          */
 590         function getInterwikiResults() {
 591                 return null;
 592         }
 593
 594         /**
 595          * Check if there are results on other wikis
 596          *
 597          * @return Boolean
 598          */
 599         function hasInterwikiResults() {
 600                 return $this->getInterwikiResults() != null;
 601         }
 602
 603         /**
 604          * Fetches next search result, or false.
 605          * STUB
 606          *
 607          * @return SearchResult
 608          */
 609         function next() {
 610                 return false;
 611         }
 612
 613         /**
 614          * Frees the result set, if applicable.
 615          */
 616         function free() {
 617                 // ...
 618         }
 619 }
 620
 621 /**
 622  * This class is used for different SQL-based search engines shipped with MediaWiki
 623  */
 624 class SqlSearchResultSet extends SearchResultSet {
 625
 626         protected $mResultSet;
 627
 628         function __construct( $resultSet, $terms ) {
 629                 $this->mResultSet = $resultSet;
 630                 $this->mTerms = $terms;
 631         }
 632
 633         function termMatches() {
 634                 return $this->mTerms;
 635         }
 636
 637         function numRows() {
 638                 if ( $this->mResultSet === false )
 639                         return false;
 640
 641                 return $this->mResultSet->numRows();
 642         }
 643
 644         function next() {
 645                 if ( $this->mResultSet === false )
 646                         return false;
 647
 648                 $row = $this->mResultSet->fetchObject();
 649                 if ( $row === false )
 650                         return false;
 651
 652                 return SearchResult::newFromRow( $row );
 653         }
 654
 655         function free() {
 656                 if ( $this->mResultSet === false )
 657                         return false;
 658
 659                 $this->mResultSet->free();
 660         }
 661 }
 662
 663 /**
 664  * @ingroup Search
 665  */
 666 class SearchResultTooMany {
 667         # # Some search engines may bail out if too many matches are found
 668 }
 669
 670
 671 /**
 672  * @todo FIXME: This class is horribly factored. It would probably be better to
 673  * have a useful base class to which you pass some standard information, then
 674  * let the fancy self-highlighters extend that.
 675  * @ingroup Search
 676  */
 677 class SearchResult {
 678
 679         /**
 680          * @var Revision
 681          */
 682         var $mRevision = null;
 683         var $mImage = null;
 684
 685         /**
 686          * @var Title
 687          */
 688         var $mTitle;
 689
 690         /**
 691          * @var String
 692          */
 693         var $mText;
 694
 695         /**
 696          * Return a new SearchResult and initializes it with a title.
 697          *
 698          * @param $title Title
 699          * @return SearchResult
 700          */
 701         public static function newFromTitle( $title ) {
 702                 $result = new self();
 703                 $result->initFromTitle( $title );
 704                 return $result;
 705         }
 706         /**
 707          * Return a new SearchResult and initializes it with a row.
 708          *
 709          * @param $row object
 710          * @return SearchResult
 711          */
 712         public static function newFromRow( $row ) {
 713                 $result = new self();
 714                 $result->initFromRow( $row );
 715                 return $result;
 716         }
 717
 718         public function __construct( $row = null ) {
 719                 if ( !is_null( $row ) ) {
 720                         // Backwards compatibility with pre-1.17 callers
 721                         $this->initFromRow( $row );
 722                 }
 723         }
 724
 725         /**
 726          * Initialize from a database row. Makes a Title and passes that to
 727          * initFromTitle.
 728          *
 729          * @param $row object
 730          */
 731         protected function initFromRow( $row ) {
 732                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 733         }
 734
 735         /**
 736          * Initialize from a Title and if possible initializes a corresponding
 737          * Revision and File.
 738          *
 739          * @param $title Title
 740          */
 741         protected function initFromTitle( $title ) {
 742                 $this->mTitle = $title;
 743                 if ( !is_null( $this->mTitle ) ) {
 744                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 745                         if ( $this->mTitle->getNamespace() === NS_FILE )
 746                                 $this->mImage = wfFindFile( $this->mTitle );
 747                 }
 748         }
 749
 750         /**
 751          * Check if this is result points to an invalid title
 752          *
 753          * @return Boolean
 754          */
 755         function isBrokenTitle() {
 756                 if ( is_null( $this->mTitle ) )
 757                         return true;
 758                 return false;
 759         }
 760
 761         /**
 762          * Check if target page is missing, happens when index is out of date
 763          *
 764          * @return Boolean
 765          */
 766         function isMissingRevision() {
 767                 return !$this->mRevision && !$this->mImage;
 768         }
 769
 770         /**
 771          * @return Title
 772          */
 773         function getTitle() {
 774                 return $this->mTitle;
 775         }
 776
 777         /**
 778          * @return float|null if not supported
 779          */
 780         function getScore() {
 781                 return null;
 782         }
 783
 784         /**
 785          * Lazy initialization of article text from DB
 786          */
 787         protected function initText() {
 788                 if ( !isset( $this->mText ) ) {
 789                         if ( $this->mRevision != null )
 790                                 $this->mText = $this->mRevision->getText();
 791                         else // TODO: can we fetch raw wikitext for commons images?
 792                                 $this->mText = '';
 793
 794                 }
 795         }
 796
 797         /**
 798          * @param $terms Array: terms to highlight
 799          * @return String: highlighted text snippet, null (and not '') if not supported
 800          */
 801         function getTextSnippet( $terms ) {
 802                 global $wgUser, $wgAdvancedSearchHighlighting;
 803                 $this->initText();
 804                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
 805                 $h = new SearchHighlighter();
 806                 if ( $wgAdvancedSearchHighlighting )
 807                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 808                 else
 809                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 810         }
 811
 812         /**
 813          * @param $terms Array: terms to highlight
 814          * @return String: highlighted title, '' if not supported
 815          */
 816         function getTitleSnippet( $terms ) {
 817                 return '';
 818         }
 819
 820         /**
 821          * @param $terms Array: terms to highlight
 822          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 823          */
 824         function getRedirectSnippet( $terms ) {
 825                 return '';
 826         }
 827
 828         /**
 829          * @return Title object for the redirect to this page, null if none or not supported
 830          */
 831         function getRedirectTitle() {
 832                 return null;
 833         }
 834
 835         /**
 836          * @return string highlighted relevant section name, null if none or not supported
 837          */
 838         function getSectionSnippet() {
 839                 return '';
 840         }
 841
 842         /**
 843          * @return Title object (pagename+fragment) for the section, null if none or not supported
 844          */
 845         function getSectionTitle() {
 846                 return null;
 847         }
 848
 849         /**
 850          * @return String: timestamp
 851          */
 852         function getTimestamp() {
 853                 if ( $this->mRevision )
 854                         return $this->mRevision->getTimestamp();
 855                 elseif ( $this->mImage )
 856                         return $this->mImage->getTimestamp();
 857                 return '';
 858         }
 859
 860         /**
 861          * @return Integer: number of words
 862          */
 863         function getWordCount() {
 864                 $this->initText();
 865                 return str_word_count( $this->mText );
 866         }
 867
 868         /**
 869          * @return Integer: size in bytes
 870          */
 871         function getByteSize() {
 872                 $this->initText();
 873                 return strlen( $this->mText );
 874         }
 875
 876         /**
 877          * @return Boolean if hit has related articles
 878          */
 879         function hasRelated() {
 880                 return false;
 881         }
 882
 883         /**
 884          * @return String: interwiki prefix of the title (return iw even if title is broken)
 885          */
 886         function getInterwikiPrefix() {
 887                 return '';
 888         }
 889 }
 890 /**
 891  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 892  */
 893 class SearchNearMatchResultSet extends SearchResultSet {
 894         private $fetched = false;
 895         /**
 896          * @param $match mixed Title if matched, else null
 897          */
 898         public function __construct( $match ) {
 899                 $this->result = $match;
 900         }
 901         public function hasResult() {
 902                 return (bool)$this->result;
 903         }
 904         public function numRows() {
 905                 return $this->hasResults() ? 1 : 0;
 906         }
 907         public function next() {
 908                 if ( $this->fetched || !$this->result ) {
 909                         return false;
 910                 }
 911                 $this->fetched = true;
 912                 return SearchResult::newFromTitle( $this->result );
 913         }
 914 }
 915
 916 /**
 917  * Highlight bits of wikitext
 918  *
 919  * @ingroup Search
 920  */
 921 class SearchHighlighter {
 922         var $mCleanWikitext = true;
 923
 924         function __construct( $cleanupWikitext = true ) {
 925                 $this->mCleanWikitext = $cleanupWikitext;
 926         }
 927
 928         /**
 929          * Default implementation of wikitext highlighting
 930          *
 931          * @param $text String
 932          * @param $terms Array: terms to highlight (unescaped)
 933          * @param $contextlines Integer
 934          * @param $contextchars Integer
 935          * @return String
 936          */
 937         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 938                 global $wgContLang;
 939                 global $wgSearchHighlightBoundaries;
 940                 $fname = __METHOD__;
 941
 942                 if ( $text == '' )
 943                         return '';
 944
 945                 // spli text into text + templates/links/tables
 946                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 947                 // first capture group is for detecting nested templates/links/tables/references
 948                 $endPatterns = array(
 949                         1 => '/(\{\{)|(\}\})/', // template
 950                         2 => '/(\[\[)|(\]\])/', // image
 951                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
 952
 953                 // @todo FIXME: This should prolly be a hook or something
 954                 if ( function_exists( 'wfCite' ) ) {
 955                         $spat .= '|(<ref>)'; // references via cite extension
 956                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 957                 }
 958                 $spat .= '/';
 959                 $textExt = array(); // text extracts
 960                 $otherExt = array();  // other extracts
 961                 wfProfileIn( "$fname-split" );
 962                 $start = 0;
 963                 $textLen = strlen( $text );
 964                 $count = 0; // sequence number to maintain ordering
 965                 while ( $start < $textLen ) {
 966                         // find start of template/image/table
 967                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
 968                                 $epat = '';
 969                                 foreach ( $matches as $key => $val ) {
 970                                         if ( $key > 0 && $val[1] != - 1 ) {
 971                                                 if ( $key == 2 ) {
 972                                                         // see if this is an image link
 973                                                         $ns = substr( $val[0], 2, - 1 );
 974                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE )
 975                                                                 break;
 976
 977                                                 }
 978                                                 $epat = $endPatterns[$key];
 979                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 980                                                 $start = $val[1];
 981                                                 break;
 982                                         }
 983                                 }
 984                                 if ( $epat ) {
 985                                         // find end (and detect any nested elements)
 986                                         $level = 0;
 987                                         $offset = $start + 1;
 988                                         $found = false;
 989                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
 990                                                 if ( array_key_exists( 2, $endMatches ) ) {
 991                                                         // found end
 992                                                         if ( $level == 0 ) {
 993                                                                 $len = strlen( $endMatches[2][0] );
 994                                                                 $off = $endMatches[2][1];
 995                                                                 $this->splitAndAdd( $otherExt, $count,
 996                                                                         substr( $text, $start, $off + $len  - $start ) );
 997                                                                 $start = $off + $len;
 998                                                                 $found = true;
 999                                                                 break;
1000                                                         } else {
1001                                                                 // end of nested element
1002                                                                 $level -= 1;
1003                                                         }
1004                                                 } else {
1005                                                         // nested
1006                                                         $level += 1;
1007                                                 }
1008                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1009                                         }
1010                                         if ( ! $found ) {
1011                                                 // couldn't find appropriate closing tag, skip
1012                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1013                                                 $start += strlen( $matches[0][0] );
1014                                         }
1015                                         continue;
1016                                 }
1017                         }
1018                         // else: add as text extract
1019                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1020                         break;
1021                 }
1022
1023                 $all = $textExt + $otherExt; // these have disjunct key sets
1024
1025                 wfProfileOut( "$fname-split" );
1026
1027                 // prepare regexps
1028                 foreach ( $terms as $index => $term ) {
1029                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1030                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1031                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1032                         } else {
1033                                 $terms[$index] = $term;
1034                         }
1035                 }
1036                 $anyterm = implode( '|', $terms );
1037                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1038
1039                 // @todo FIXME: A hack to scale contextchars, a correct solution
1040                 // would be to have contextchars actually be char and not byte
1041                 // length, and do proper utf-8 substrings and lengths everywhere,
1042                 // but PHP is making that very hard and unclean to implement :(
1043                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1044                 $contextchars = intval( $contextchars * $scale );
1045
1046                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1047                 $patPost = "($wgSearchHighlightBoundaries|$)";
1048
1049                 $pat1 = "/(" . $phrase . ")/ui";
1050                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1051
1052                 wfProfileIn( "$fname-extract" );
1053
1054                 $left = $contextlines;
1055
1056                 $snippets = array();
1057                 $offsets = array();
1058
1059                 // show beginning only if it contains all words
1060                 $first = 0;
1061                 $firstText = '';
1062                 foreach ( $textExt as $index => $line ) {
1063                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1064                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1065                                 $first = $index;
1066                                 break;
1067                         }
1068                 }
1069                 if ( $firstText ) {
1070                         $succ = true;
1071                         // check if first text contains all terms
1072                         foreach ( $terms as $term ) {
1073                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1074                                         $succ = false;
1075                                         break;
1076                                 }
1077                         }
1078                         if ( $succ ) {
1079                                 $snippets[$first] = $firstText;
1080                                 $offsets[$first] = 0;
1081                         }
1082                 }
1083                 if ( ! $snippets ) {
1084                         // match whole query on text
1085                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1086                         // match whole query on templates/tables/images
1087                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1088                         // match any words on text
1089                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1090                         // match any words on templates/tables/images
1091                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1092
1093                         ksort( $snippets );
1094                 }
1095
1096                 // add extra chars to each snippet to make snippets constant size
1097                 $extended = array();
1098                 if ( count( $snippets ) == 0 ) {
1099                         // couldn't find the target words, just show beginning of article
1100                         if ( array_key_exists( $first, $all ) ) {
1101                                 $targetchars = $contextchars * $contextlines;
1102                                 $snippets[$first] = '';
1103                                 $offsets[$first] = 0;
1104                         }
1105                 } else {
1106                         // if begin of the article contains the whole phrase, show only that !!
1107                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1108                                 && $offsets[$first] < $contextchars * 2 ) {
1109                                 $snippets = array ( $first => $snippets[$first] );
1110                         }
1111
1112                         // calc by how much to extend existing snippets
1113                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1114                 }
1115
1116                 foreach ( $snippets as $index => $line ) {
1117                         $extended[$index] = $line;
1118                         $len = strlen( $line );
1119                         if ( $len < $targetchars - 20 ) {
1120                                 // complete this line
1121                                 if ( $len < strlen( $all[$index] ) ) {
1122                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1123                                         $len = strlen( $extended[$index] );
1124                                 }
1125
1126                                 // add more lines
1127                                 $add = $index + 1;
1128                                 while ( $len < $targetchars - 20
1129                                            && array_key_exists( $add, $all )
1130                                            && !array_key_exists( $add, $snippets ) ) {
1131                                         $offsets[$add] = 0;
1132                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1133                                         $extended[$add] = $tt;
1134                                         $len += strlen( $tt );
1135                                         $add++;
1136                                 }
1137                         }
1138                 }
1139
1140                 // $snippets = array_map('htmlspecialchars', $extended);
1141                 $snippets = $extended;
1142                 $last = - 1;
1143                 $extract = '';
1144                 foreach ( $snippets as $index => $line ) {
1145                         if ( $last == - 1 )
1146                                 $extract .= $line; // first line
1147                         elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) )
1148                                 $extract .= " " . $line; // continous lines
1149                         else
1150                                 $extract .= '<b> ... </b>' . $line;
1151
1152                         $last = $index;
1153                 }
1154                 if ( $extract )
1155                         $extract .= '<b> ... </b>';
1156
1157                 $processed = array();
1158                 foreach ( $terms as $term ) {
1159                         if ( ! isset( $processed[$term] ) ) {
1160                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1161                                 $extract = preg_replace( $pat3,
1162                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1163                                 $processed[$term] = true;
1164                         }
1165                 }
1166
1167                 wfProfileOut( "$fname-extract" );
1168
1169                 return $extract;
1170         }
1171
1172         /**
1173          * Split text into lines and add it to extracts array
1174          *
1175          * @param $extracts Array: index -> $line
1176          * @param $count Integer
1177          * @param $text String
1178          */
1179         function splitAndAdd( &$extracts, &$count, $text ) {
1180                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1181                 foreach ( $split as $line ) {
1182                         $tt = trim( $line );
1183                         if ( $tt )
1184                                 $extracts[$count++] = $tt;
1185                 }
1186         }
1187
1188         /**
1189          * Do manual case conversion for non-ascii chars
1190          *
1191          * @param $matches Array
1192          * @return string
1193          */
1194         function caseCallback( $matches ) {
1195                 global $wgContLang;
1196                 if ( strlen( $matches[0] ) > 1 ) {
1197                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1198                 } else {
1199                         return $matches[0];
1200                 }
1201         }
1202
1203         /**
1204          * Extract part of the text from start to end, but by
1205          * not chopping up words
1206          * @param $text String
1207          * @param $start Integer
1208          * @param $end Integer
1209          * @param $posStart Integer: (out) actual start position
1210          * @param $posEnd Integer: (out) actual end position
1211          * @return String
1212          */
1213         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1214                 if ( $start != 0 ) {
1215                         $start = $this->position( $text, $start, 1 );
1216                 }
1217                 if ( $end >= strlen( $text ) ) {
1218                         $end = strlen( $text );
1219                 } else {
1220                         $end = $this->position( $text, $end );
1221                 }
1222
1223                 if ( !is_null( $posStart ) ) {
1224                         $posStart = $start;
1225                 }
1226                 if ( !is_null( $posEnd ) ) {
1227                         $posEnd = $end;
1228                 }
1229
1230                 if ( $end > $start )  {
1231                         return substr( $text, $start, $end - $start );
1232                 } else {
1233                         return '';
1234                 }
1235         }
1236
1237         /**
1238          * Find a nonletter near a point (index) in the text
1239          *
1240          * @param $text String
1241          * @param $point Integer
1242          * @param $offset Integer: offset to found index
1243          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1244          */
1245         function position( $text, $point, $offset = 0 ) {
1246                 $tolerance = 10;
1247                 $s = max( 0, $point - $tolerance );
1248                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1249                 $m = array();
1250                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1251                         return $m[0][1] + $s + $offset;
1252                 } else {
1253                         // check if point is on a valid first UTF8 char
1254                         $char = ord( $text[$point] );
1255                         while ( $char >= 0x80 && $char < 0xc0 ) {
1256                                 // skip trailing bytes
1257                                 $point++;
1258                                 if ( $point >= strlen( $text ) )
1259                                         return strlen( $text );
1260                                 $char = ord( $text[$point] );
1261                         }
1262                         return $point;
1263
1264                 }
1265         }
1266
1267         /**
1268          * Search extracts for a pattern, and return snippets
1269          *
1270          * @param $pattern String: regexp for matching lines
1271          * @param $extracts Array: extracts to search
1272          * @param $linesleft Integer: number of extracts to make
1273          * @param $contextchars Integer: length of snippet
1274          * @param $out Array: map for highlighted snippets
1275          * @param $offsets Array: map of starting points of snippets
1276          * @protected
1277          */
1278         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1279                 if ( $linesleft == 0 )
1280                         return; // nothing to do
1281                 foreach ( $extracts as $index => $line ) {
1282                         if ( array_key_exists( $index, $out ) )
1283                                 continue; // this line already highlighted
1284
1285                         $m = array();
1286                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1287                                 continue;
1288
1289                         $offset = $m[0][1];
1290                         $len = strlen( $m[0][0] );
1291                         if ( $offset + $len < $contextchars )
1292                                 $begin = 0;
1293                         elseif ( $len > $contextchars )
1294                                 $begin = $offset;
1295                         else
1296                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1297
1298                         $end = $begin + $contextchars;
1299
1300                         $posBegin = $begin;
1301                         // basic snippet from this line
1302                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1303                         $offsets[$index] = $posBegin;
1304                         $linesleft--;
1305                         if ( $linesleft == 0 )
1306                                 return;
1307                 }
1308         }
1309
1310         /**
1311          * Basic wikitext removal
1312          * @protected
1313          * @return mixed
1314          */
1315         function removeWiki( $text ) {
1316                 $fname = __METHOD__;
1317                 wfProfileIn( $fname );
1318
1319                 // $text = preg_replace("/'{2,5}/", "", $text);
1320                 // $text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1321                 // $text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1322                 // $text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1323                 // $text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1324                 // $text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1325                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1326                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1327                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1328                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1329                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1330                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1331                 $text = preg_replace( "/'''''/", "", $text );
1332                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1333                 $text = preg_replace( "/''/", "", $text );
1334
1335                 wfProfileOut( $fname );
1336                 return $text;
1337         }
1338
1339         /**
1340          * callback to replace [[target|caption]] kind of links, if
1341          * the target is category or image, leave it
1342          *
1343          * @param $matches Array
1344          */
1345         function linkReplace( $matches ) {
1346                 $colon = strpos( $matches[1], ':' );
1347                 if ( $colon === false )
1348                         return $matches[2]; // replace with caption
1349                 global $wgContLang;
1350                 $ns = substr( $matches[1], 0, $colon );
1351                 $index = $wgContLang->getNsIndex( $ns );
1352                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) )
1353                         return $matches[0]; // return the whole thing
1354                 else
1355                         return $matches[2];
1356
1357         }
1358
1359         /**
1360          * Simple & fast snippet extraction, but gives completely unrelevant
1361          * snippets
1362          *
1363          * @param $text String
1364          * @param $terms Array
1365          * @param $contextlines Integer
1366          * @param $contextchars Integer
1367          * @return String
1368          */
1369         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1370                 global $wgContLang;
1371                 $fname = __METHOD__;
1372
1373                 $lines = explode( "\n", $text );
1374
1375                 $terms = implode( '|', $terms );
1376                 $max = intval( $contextchars ) + 1;
1377                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
1378
1379                 $lineno = 0;
1380
1381                 $extract = "";
1382                 wfProfileIn( "$fname-extract" );
1383                 foreach ( $lines as $line ) {
1384                         if ( 0 == $contextlines ) {
1385                                 break;
1386                         }
1387                         ++$lineno;
1388                         $m = array();
1389                         if ( ! preg_match( $pat1, $line, $m ) ) {
1390                                 continue;
1391                         }
1392                         --$contextlines;
1393                         // truncate function changes ... to relevant i18n message.
1394                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1395
1396                         if ( count( $m ) < 3 ) {
1397                                 $post = '';
1398                         } else {
1399                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1400                         }
1401
1402                         $found = $m[2];
1403
1404                         $line = htmlspecialchars( $pre . $found . $post );
1405                         $pat2 = '/(' . $terms . ")/i";
1406                         $line = preg_replace( $pat2,
1407                           "<span class='searchmatch'>\\1</span>", $line );
1408
1409                         $extract .= "${line}\n";
1410                 }
1411                 wfProfileOut( "$fname-extract" );
1412
1413                 return $extract;
1414         }
1415
1416 }
1417
1418 /**
1419  * Dummy class to be used when non-supported Database engine is present.
1420  * @todo FIXME: Dummy class should probably try something at least mildly useful,
1421  * such as a LIKE search through titles.
1422  * @ingroup Search
1423  */
1424 class SearchEngineDummy extends SearchEngine {
1425         // no-op
1426 }