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 $showRedirects = false;
  24
  25         function __construct($db = null) {
  26                 if ( $db ) {
  27                         $this->db = $db;
  28                 } else {
  29                         $this->db = wfGetDB( DB_SLAVE );
  30                 }
  31         }
  32
  33         /**
  34          * Perform a full text search query and return a result set.
  35          * If title searches are not supported or disabled, return null.
  36          * STUB
  37          *
  38          * @param $term String: raw search term
  39          * @return SearchResultSet
  40          */
  41         function searchText( $term ) {
  42                 return null;
  43         }
  44
  45         /**
  46          * Perform a title-only search query and return a result set.
  47          * If title searches are not supported or disabled, return null.
  48          * STUB
  49          *
  50          * @param $term String: raw search term
  51          * @return SearchResultSet
  52          */
  53         function searchTitle( $term ) {
  54                 return null;
  55         }
  56
  57         /** If this search backend can list/unlist redirects */
  58         function acceptListRedirects() {
  59                 return true;
  60         }
  61
  62         /**
  63          * When overridden in derived class, performs database-specific conversions
  64          * on text to be used for searching or updating search index.
  65          * Default implementation does nothing (simply returns $string).
  66          *
  67          * @param $string string: String to process
  68          * @return string
  69          */
  70         public function normalizeText( $string ) {
  71                 global $wgContLang;
  72
  73                 // Some languages such as Chinese require word segmentation
  74                 return $wgContLang->segmentByWord( $string );
  75         }
  76
  77         /**
  78          * Transform search term in cases when parts of the query came as different GET params (when supported)
  79          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
  80          */
  81         function transformSearchTerm( $term ) {
  82                 return $term;
  83         }
  84
  85         /**
  86          * If an exact title match can be found, or a very slightly close match,
  87          * return the title. If no match, returns NULL.
  88          *
  89          * @param $searchterm String
  90          * @return Title
  91          */
  92         public static function getNearMatch( $searchterm ) {
  93                 $title = self::getNearMatchInternal( $searchterm );
  94
  95                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
  96                 return $title;
  97         }
  98
  99         /**
 100          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 101          * SearchResultSet.
 102          *
 103          * @param $searchterm string
 104          * @return SearchResultSet
 105          */
 106         public static function getNearMatchResultSet( $searchterm ) {
 107                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 108         }
 109
 110         /**
 111          * Really find the title match.
 112          */
 113         private static function getNearMatchInternal( $searchterm ) {
 114                 global $wgContLang;
 115
 116                 $allSearchTerms = array( $searchterm );
 117
 118                 if ( $wgContLang->hasVariants() ) {
 119                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 120                 }
 121
 122                 $titleResult = null;
 123                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 124                         return $titleResult;
 125                 }
 126
 127                 foreach ( $allSearchTerms as $term ) {
 128
 129                         # Exact match? No need to look further.
 130                         $title = Title::newFromText( $term );
 131                         if ( is_null( $title ) )
 132                                 return null;
 133
 134                         if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal() || $title->exists() ) {
 135                                 return $title;
 136                         }
 137
 138                         # See if it still otherwise has content is some sane sense
 139                         $article = MediaWiki::articleFromTitle( $title );
 140                         if ( $article->hasViewableContent() ) {
 141                                 return $title;
 142                         }
 143
 144                         # Now try all lower case (i.e. first letter capitalized)
 145                         #
 146                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 147                         if ( $title && $title->exists() ) {
 148                                 return $title;
 149                         }
 150
 151                         # Now try capitalized string
 152                         #
 153                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 154                         if ( $title && $title->exists() ) {
 155                                 return $title;
 156                         }
 157
 158                         # Now try all upper case
 159                         #
 160                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 161                         if ( $title && $title->exists() ) {
 162                                 return $title;
 163                         }
 164
 165                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 166                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 167                         if ( $title && $title->exists() ) {
 168                                 return $title;
 169                         }
 170
 171                         // Give hooks a chance at better match variants
 172                         $title = null;
 173                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 174                                 return $title;
 175                         }
 176                 }
 177
 178                 $title = Title::newFromText( $searchterm );
 179
 180                 # Entering an IP address goes to the contributions page
 181                 if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 182                         || User::isIP( trim( $searchterm ) ) ) {
 183                         return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 184                 }
 185
 186
 187                 # Entering a user goes to the user page whether it's there or not
 188                 if ( $title->getNamespace() == NS_USER ) {
 189                         return $title;
 190                 }
 191
 192                 # Go to images that exist even if there's no local page.
 193                 # There may have been a funny upload, or it may be on a shared
 194                 # file repository such as Wikimedia Commons.
 195                 if ( $title->getNamespace() == NS_FILE ) {
 196                         $image = wfFindFile( $title );
 197                         if ( $image ) {
 198                                 return $title;
 199                         }
 200                 }
 201
 202                 # MediaWiki namespace? Page may be "implied" if not customized.
 203                 # Just return it, with caps forced as the message system likes it.
 204                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 205                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 206                 }
 207
 208                 # Quoted term? Try without the quotes...
 209                 $matches = array();
 210                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 211                         return SearchEngine::getNearMatch( $matches[1] );
 212                 }
 213
 214                 return null;
 215         }
 216
 217         public static function legalSearchChars() {
 218                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 219         }
 220
 221         /**
 222          * Set the maximum number of results to return
 223          * and how many to skip before returning the first.
 224          *
 225          * @param $limit Integer
 226          * @param $offset Integer
 227          */
 228         function setLimitOffset( $limit, $offset = 0 ) {
 229                 $this->limit = intval( $limit );
 230                 $this->offset = intval( $offset );
 231         }
 232
 233         /**
 234          * Set which namespaces the search should include.
 235          * Give an array of namespace index numbers.
 236          *
 237          * @param $namespaces Array
 238          */
 239         function setNamespaces( $namespaces ) {
 240                 $this->namespaces = $namespaces;
 241         }
 242
 243         /**
 244          * Parse some common prefixes: all (search everything)
 245          * or namespace names
 246          *
 247          * @param $query String
 248          */
 249         function replacePrefixes( $query ) {
 250                 global $wgContLang;
 251
 252                 $parsed = $query;
 253                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 254                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 255                         return $parsed;
 256                 }
 257
 258                 $allkeyword = wfMsgForContent( 'searchall' ) . ":";
 259                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 260                         $this->namespaces = null;
 261                         $parsed = substr( $query, strlen( $allkeyword ) );
 262                 } else if ( strpos( $query, ':' ) !== false ) {
 263                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 264                         $index = $wgContLang->getNsIndex( $prefix );
 265                         if ( $index !== false ) {
 266                                 $this->namespaces = array( $index );
 267                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 268                         }
 269                 }
 270                 if ( trim( $parsed ) == '' )
 271                         $parsed = $query; // prefix was the whole query
 272
 273                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 274
 275                 return $parsed;
 276         }
 277
 278         /**
 279          * Make a list of searchable namespaces and their canonical names.
 280          * @return Array
 281          */
 282         public static function searchableNamespaces() {
 283                 global $wgContLang;
 284                 $arr = array();
 285                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 286                         if ( $ns >= NS_MAIN ) {
 287                                 $arr[$ns] = $name;
 288                         }
 289                 }
 290
 291                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 292                 return $arr;
 293         }
 294
 295         /**
 296          * Extract default namespaces to search from the given user's
 297          * settings, returning a list of index numbers.
 298          *
 299          * @param $user User
 300          * @return Array
 301          */
 302         public static function userNamespaces( $user ) {
 303                 global $wgSearchEverythingOnlyLoggedIn;
 304
 305                 // get search everything preference, that can be set to be read for logged-in users
 306                 $searcheverything = false;
 307                 if ( ( $wgSearchEverythingOnlyLoggedIn && $user->isLoggedIn() )
 308                     || !$wgSearchEverythingOnlyLoggedIn )
 309                         $searcheverything = $user->getOption( 'searcheverything' );
 310
 311                 // searcheverything overrides other options
 312                 if ( $searcheverything )
 313                         return array_keys( SearchEngine::searchableNamespaces() );
 314
 315                 $arr = Preferences::loadOldSearchNs( $user );
 316                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 317
 318                 $arr = array_intersect( $arr, array_keys( $searchableNamespaces ) ); // Filter
 319
 320                 return $arr;
 321         }
 322
 323         /**
 324          * Find snippet highlight settings for a given user
 325          *
 326          * @param $user User
 327          * @return Array contextlines, contextchars
 328          */
 329         public static function userHighlightPrefs( &$user ) {
 330                 // $contextlines = $user->getOption( 'contextlines',  5 );
 331                 // $contextchars = $user->getOption( 'contextchars', 50 );
 332                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 333                 $contextchars = 75; // same as above.... :P
 334                 return array( $contextlines, $contextchars );
 335         }
 336
 337         /**
 338          * An array of namespaces indexes to be searched by default
 339          *
 340          * @return Array
 341          */
 342         public static function defaultNamespaces() {
 343                 global $wgNamespacesToBeSearchedDefault;
 344
 345                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 346         }
 347
 348         /**
 349          * Get a list of namespace names useful for showing in tooltips
 350          * and preferences
 351          *
 352          * @param $namespaces Array
 353          */
 354         public static function namespacesAsText( $namespaces ) {
 355                 global $wgContLang;
 356
 357                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 358                 foreach ( $formatted as $key => $ns ) {
 359                         if ( empty( $ns ) )
 360                                 $formatted[$key] = wfMsg( 'blanknamespace' );
 361                 }
 362                 return $formatted;
 363         }
 364
 365         /**
 366          * Return the help namespaces to be shown on Special:Search
 367          *
 368          * @return Array
 369          */
 370         public static function helpNamespaces() {
 371                 global $wgNamespacesToBeSearchedHelp;
 372
 373                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 374         }
 375
 376         /**
 377          * Return a 'cleaned up' search string
 378          *
 379          * @param $text String
 380          * @return String
 381          */
 382         function filter( $text ) {
 383                 $lc = $this->legalSearchChars();
 384                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 385         }
 386         /**
 387          * Load up the appropriate search engine class for the currently
 388          * active database backend, and return a configured instance.
 389          *
 390          * @return SearchEngine
 391          */
 392         public static function create() {
 393                 global $wgSearchType;
 394                 $dbr = null;
 395                 if ( $wgSearchType ) {
 396                         $class = $wgSearchType;
 397                 } else {
 398                         $dbr = wfGetDB( DB_SLAVE );
 399                         $class = $dbr->getSearchEngine();
 400                 }
 401                 $search = new $class( $dbr );
 402                 $search->setLimitOffset( 0, 0 );
 403                 return $search;
 404         }
 405
 406         /**
 407          * Create or update the search index record for the given page.
 408          * Title and text should be pre-processed.
 409          * STUB
 410          *
 411          * @param $id Integer
 412          * @param $title String
 413          * @param $text String
 414          */
 415         function update( $id, $title, $text ) {
 416                 // no-op
 417         }
 418
 419         /**
 420          * Update a search index record's title only.
 421          * Title should be pre-processed.
 422          * STUB
 423          *
 424          * @param $id Integer
 425          * @param $title String
 426          */
 427         function updateTitle( $id, $title ) {
 428                 // no-op
 429         }
 430
 431         /**
 432          * Get OpenSearch suggestion template
 433          *
 434          * @return String
 435          */
 436         public static function getOpenSearchTemplate() {
 437                 global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
 438                 if ( $wgOpenSearchTemplate )    {
 439                         return $wgOpenSearchTemplate;
 440                 } else {
 441                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 442                         if ( !$ns ) $ns = "0";
 443                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace=' . $ns;
 444                 }
 445         }
 446
 447         /**
 448          * Get internal MediaWiki Suggest template
 449          *
 450          * @return String
 451          */
 452         public static function getMWSuggestTemplate() {
 453                 global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
 454                 if ( $wgMWSuggestTemplate )
 455                         return $wgMWSuggestTemplate;
 456                 else
 457                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
 458         }
 459 }
 460
 461 /**
 462  * @ingroup Search
 463  */
 464 class SearchResultSet {
 465         /**
 466          * Fetch an array of regular expression fragments for matching
 467          * the search terms as parsed by this engine in a text extract.
 468          * STUB
 469          *
 470          * @return Array
 471          */
 472         function termMatches() {
 473                 return array();
 474         }
 475
 476         function numRows() {
 477                 return 0;
 478         }
 479
 480         /**
 481          * Return true if results are included in this result set.
 482          * STUB
 483          *
 484          * @return Boolean
 485          */
 486         function hasResults() {
 487                 return false;
 488         }
 489
 490         /**
 491          * Some search modes return a total hit count for the query
 492          * in the entire article database. This may include pages
 493          * in namespaces that would not be matched on the given
 494          * settings.
 495          *
 496          * Return null if no total hits number is supported.
 497          *
 498          * @return Integer
 499          */
 500         function getTotalHits() {
 501                 return null;
 502         }
 503
 504         /**
 505          * Some search modes return a suggested alternate term if there are
 506          * no exact hits. Returns true if there is one on this set.
 507          *
 508          * @return Boolean
 509          */
 510         function hasSuggestion() {
 511                 return false;
 512         }
 513
 514         /**
 515          * @return String: suggested query, null if none
 516          */
 517         function getSuggestionQuery() {
 518                 return null;
 519         }
 520
 521         /**
 522          * @return String: HTML highlighted suggested query, '' if none
 523          */
 524         function getSuggestionSnippet() {
 525                 return '';
 526         }
 527
 528         /**
 529          * Return information about how and from where the results were fetched,
 530          * should be useful for diagnostics and debugging
 531          *
 532          * @return String
 533          */
 534         function getInfo() {
 535                 return null;
 536         }
 537
 538         /**
 539          * Return a result set of hits on other (multiple) wikis associated with this one
 540          *
 541          * @return SearchResultSet
 542          */
 543         function getInterwikiResults() {
 544                 return null;
 545         }
 546
 547         /**
 548          * Check if there are results on other wikis
 549          *
 550          * @return Boolean
 551          */
 552         function hasInterwikiResults() {
 553                 return $this->getInterwikiResults() != null;
 554         }
 555
 556         /**
 557          * Fetches next search result, or false.
 558          * STUB
 559          *
 560          * @return SearchResult
 561          */
 562         function next() {
 563                 return false;
 564         }
 565
 566         /**
 567          * Frees the result set, if applicable.
 568          */
 569         function free() {
 570                 // ...
 571         }
 572 }
 573
 574 /**
 575  * This class is used for different SQL-based search engines shipped with MediaWiki
 576  */
 577 class SqlSearchResultSet extends SearchResultSet {
 578         function __construct( $resultSet, $terms ) {
 579                 $this->mResultSet = $resultSet;
 580                 $this->mTerms = $terms;
 581         }
 582
 583         function termMatches() {
 584                 return $this->mTerms;
 585         }
 586
 587         function numRows() {
 588                 if ( $this->mResultSet === false )
 589                         return false;
 590
 591                 return $this->mResultSet->numRows();
 592         }
 593
 594         function next() {
 595                 if ( $this->mResultSet === false )
 596                         return false;
 597
 598                 $row = $this->mResultSet->fetchObject();
 599                 if ( $row === false )
 600                         return false;
 601
 602                 return SearchResult::newFromRow( $row );
 603         }
 604
 605         function free() {
 606                 if ( $this->mResultSet === false )
 607                         return false;
 608
 609                 $this->mResultSet->free();
 610         }
 611 }
 612
 613 /**
 614  * @ingroup Search
 615  */
 616 class SearchResultTooMany {
 617         # # Some search engines may bail out if too many matches are found
 618 }
 619
 620
 621 /**
 622  * @todo Fixme: This class is horribly factored. It would probably be better to
 623  * have a useful base class to which you pass some standard information, then
 624  * let the fancy self-highlighters extend that.
 625  * @ingroup Search
 626  */
 627 class SearchResult {
 628         var $mRevision = null;
 629         var $mImage = null;
 630
 631         /**
 632          * Return a new SearchResult and initializes it with a title.
 633          *
 634          * @param $title Title
 635          * @return SearchResult
 636          */
 637         public static function newFromTitle( $title ) {
 638                 $result = new self();
 639                 $result->initFromTitle( $title );
 640                 return $result;
 641         }
 642         /**
 643          * Return a new SearchResult and initializes it with a row.
 644          *
 645          * @param $row object
 646          * @return SearchResult
 647          */
 648         public static function newFromRow( $row ) {
 649                 $result = new self();
 650                 $result->initFromRow( $row );
 651                 return $result;
 652         }
 653
 654         public function __construct( $row = null ) {
 655                 if ( !is_null( $row ) ) {
 656                         // Backwards compatibility with pre-1.17 callers
 657                         $this->initFromRow( $row );
 658                 }
 659         }
 660
 661         /**
 662          * Initialize from a database row. Makes a Title and passes that to
 663          * initFromTitle.
 664          *
 665          * @param $row object
 666          */
 667         protected function initFromRow( $row ) {
 668                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 669         }
 670
 671         /**
 672          * Initialize from a Title and if possible initializes a corresponding
 673          * Revision and File.
 674          *
 675          * @param $title Title
 676          */
 677         protected function initFromTitle( $title ) {
 678                 $this->mTitle = $title;
 679                 if ( !is_null( $this->mTitle ) ) {
 680                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 681                         if ( $this->mTitle->getNamespace() === NS_FILE )
 682                                 $this->mImage = wfFindFile( $this->mTitle );
 683                 }
 684         }
 685
 686         /**
 687          * Check if this is result points to an invalid title
 688          *
 689          * @return Boolean
 690          */
 691         function isBrokenTitle() {
 692                 if ( is_null( $this->mTitle ) )
 693                         return true;
 694                 return false;
 695         }
 696
 697         /**
 698          * Check if target page is missing, happens when index is out of date
 699          *
 700          * @return Boolean
 701          */
 702         function isMissingRevision() {
 703                 return !$this->mRevision && !$this->mImage;
 704         }
 705
 706         /**
 707          * @return Title
 708          */
 709         function getTitle() {
 710                 return $this->mTitle;
 711         }
 712
 713         /**
 714          * @return Double or null if not supported
 715          */
 716         function getScore() {
 717                 return null;
 718         }
 719
 720         /**
 721          * Lazy initialization of article text from DB
 722          */
 723         protected function initText() {
 724                 if ( !isset( $this->mText ) ) {
 725                         if ( $this->mRevision != null )
 726                                 $this->mText = $this->mRevision->getText();
 727                         else // TODO: can we fetch raw wikitext for commons images?
 728                                 $this->mText = '';
 729
 730                 }
 731         }
 732
 733         /**
 734          * @param $terms Array: terms to highlight
 735          * @return String: highlighted text snippet, null (and not '') if not supported
 736          */
 737         function getTextSnippet( $terms ) {
 738                 global $wgUser, $wgAdvancedSearchHighlighting;
 739                 $this->initText();
 740                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
 741                 $h = new SearchHighlighter();
 742                 if ( $wgAdvancedSearchHighlighting )
 743                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 744                 else
 745                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 746         }
 747
 748         /**
 749          * @param $terms Array: terms to highlight
 750          * @return String: highlighted title, '' if not supported
 751          */
 752         function getTitleSnippet( $terms ) {
 753                 return '';
 754         }
 755
 756         /**
 757          * @param $terms Array: terms to highlight
 758          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 759          */
 760         function getRedirectSnippet( $terms ) {
 761                 return '';
 762         }
 763
 764         /**
 765          * @return Title object for the redirect to this page, null if none or not supported
 766          */
 767         function getRedirectTitle() {
 768                 return null;
 769         }
 770
 771         /**
 772          * @return string highlighted relevant section name, null if none or not supported
 773          */
 774         function getSectionSnippet() {
 775                 return '';
 776         }
 777
 778         /**
 779          * @return Title object (pagename+fragment) for the section, null if none or not supported
 780          */
 781         function getSectionTitle() {
 782                 return null;
 783         }
 784
 785         /**
 786          * @return String: timestamp
 787          */
 788         function getTimestamp() {
 789                 if ( $this->mRevision )
 790                         return $this->mRevision->getTimestamp();
 791                 else if ( $this->mImage )
 792                         return $this->mImage->getTimestamp();
 793                 return '';
 794         }
 795
 796         /**
 797          * @return Integer: number of words
 798          */
 799         function getWordCount() {
 800                 $this->initText();
 801                 return str_word_count( $this->mText );
 802         }
 803
 804         /**
 805          * @return Integer: size in bytes
 806          */
 807         function getByteSize() {
 808                 $this->initText();
 809                 return strlen( $this->mText );
 810         }
 811
 812         /**
 813          * @return Boolean if hit has related articles
 814          */
 815         function hasRelated() {
 816                 return false;
 817         }
 818
 819         /**
 820          * @return String: interwiki prefix of the title (return iw even if title is broken)
 821          */
 822         function getInterwikiPrefix() {
 823                 return '';
 824         }
 825 }
 826 /**
 827  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 828  */
 829 class SearchNearMatchResultSet extends SearchResultSet {
 830         private $fetched = false;
 831         /**
 832          * @param $match mixed Title if matched, else null
 833          */
 834         public function __construct( $match ) {
 835                 $this->result = $match;
 836         }
 837         public function hasResult() {
 838                 return (bool)$this->result;
 839         }
 840         public function numRows() {
 841                 return $this->hasResults() ? 1 : 0;
 842         }
 843         public function next() {
 844                 if ( $this->fetched || !$this->result ) {
 845                         return false;
 846                 }
 847                 $this->fetched = true;
 848                 return SearchResult::newFromTitle( $this->result );
 849         }
 850 }
 851
 852 /**
 853  * Highlight bits of wikitext
 854  *
 855  * @ingroup Search
 856  */
 857 class SearchHighlighter {
 858         var $mCleanWikitext = true;
 859
 860         function __construct( $cleanupWikitext = true ) {
 861                 $this->mCleanWikitext = $cleanupWikitext;
 862         }
 863
 864         /**
 865          * Default implementation of wikitext highlighting
 866          *
 867          * @param $text String
 868          * @param $terms Array: terms to highlight (unescaped)
 869          * @param $contextlines Integer
 870          * @param $contextchars Integer
 871          * @return String
 872          */
 873         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 874                 global $wgContLang;
 875                 global $wgSearchHighlightBoundaries;
 876                 $fname = __METHOD__;
 877
 878                 if ( $text == '' )
 879                         return '';
 880
 881                 // spli text into text + templates/links/tables
 882                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 883                 // first capture group is for detecting nested templates/links/tables/references
 884                 $endPatterns = array(
 885                         1 => '/(\{\{)|(\}\})/', // template
 886                         2 => '/(\[\[)|(\]\])/', // image
 887                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
 888
 889                 // FIXME: this should prolly be a hook or something
 890                 if ( function_exists( 'wfCite' ) ) {
 891                         $spat .= '|(<ref>)'; // references via cite extension
 892                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 893                 }
 894                 $spat .= '/';
 895                 $textExt = array(); // text extracts
 896                 $otherExt = array();  // other extracts
 897                 wfProfileIn( "$fname-split" );
 898                 $start = 0;
 899                 $textLen = strlen( $text );
 900                 $count = 0; // sequence number to maintain ordering
 901                 while ( $start < $textLen ) {
 902                         // find start of template/image/table
 903                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
 904                                 $epat = '';
 905                                 foreach ( $matches as $key => $val ) {
 906                                         if ( $key > 0 && $val[1] != - 1 ) {
 907                                                 if ( $key == 2 ) {
 908                                                         // see if this is an image link
 909                                                         $ns = substr( $val[0], 2, - 1 );
 910                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE )
 911                                                                 break;
 912
 913                                                 }
 914                                                 $epat = $endPatterns[$key];
 915                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 916                                                 $start = $val[1];
 917                                                 break;
 918                                         }
 919                                 }
 920                                 if ( $epat ) {
 921                                         // find end (and detect any nested elements)
 922                                         $level = 0;
 923                                         $offset = $start + 1;
 924                                         $found = false;
 925                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
 926                                                 if ( array_key_exists( 2, $endMatches ) ) {
 927                                                         // found end
 928                                                         if ( $level == 0 ) {
 929                                                                 $len = strlen( $endMatches[2][0] );
 930                                                                 $off = $endMatches[2][1];
 931                                                                 $this->splitAndAdd( $otherExt, $count,
 932                                                                         substr( $text, $start, $off + $len  - $start ) );
 933                                                                 $start = $off + $len;
 934                                                                 $found = true;
 935                                                                 break;
 936                                                         } else {
 937                                                                 // end of nested element
 938                                                                 $level -= 1;
 939                                                         }
 940                                                 } else {
 941                                                         // nested
 942                                                         $level += 1;
 943                                                 }
 944                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
 945                                         }
 946                                         if ( ! $found ) {
 947                                                 // couldn't find appropriate closing tag, skip
 948                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
 949                                                 $start += strlen( $matches[0][0] );
 950                                         }
 951                                         continue;
 952                                 }
 953                         }
 954                         // else: add as text extract
 955                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
 956                         break;
 957                 }
 958
 959                 $all = $textExt + $otherExt; // these have disjunct key sets
 960
 961                 wfProfileOut( "$fname-split" );
 962
 963                 // prepare regexps
 964                 foreach ( $terms as $index => $term ) {
 965                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 966                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
 967                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
 968                         } else {
 969                                 $terms[$index] = $term;
 970                         }
 971                 }
 972                 $anyterm = implode( '|', $terms );
 973                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
 974
 975                 // FIXME: a hack to scale contextchars, a correct solution
 976                 // would be to have contextchars actually be char and not byte
 977                 // length, and do proper utf-8 substrings and lengths everywhere,
 978                 // but PHP is making that very hard and unclean to implement :(
 979                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
 980                 $contextchars = intval( $contextchars * $scale );
 981
 982                 $patPre = "(^|$wgSearchHighlightBoundaries)";
 983                 $patPost = "($wgSearchHighlightBoundaries|$)";
 984
 985                 $pat1 = "/(" . $phrase . ")/ui";
 986                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
 987
 988                 wfProfileIn( "$fname-extract" );
 989
 990                 $left = $contextlines;
 991
 992                 $snippets = array();
 993                 $offsets = array();
 994
 995                 // show beginning only if it contains all words
 996                 $first = 0;
 997                 $firstText = '';
 998                 foreach ( $textExt as $index => $line ) {
 999                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1000                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1001                                 $first = $index;
1002                                 break;
1003                         }
1004                 }
1005                 if ( $firstText ) {
1006                         $succ = true;
1007                         // check if first text contains all terms
1008                         foreach ( $terms as $term ) {
1009                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1010                                         $succ = false;
1011                                         break;
1012                                 }
1013                         }
1014                         if ( $succ ) {
1015                                 $snippets[$first] = $firstText;
1016                                 $offsets[$first] = 0;
1017                         }
1018                 }
1019                 if ( ! $snippets ) {
1020                         // match whole query on text
1021                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1022                         // match whole query on templates/tables/images
1023                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1024                         // match any words on text
1025                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1026                         // match any words on templates/tables/images
1027                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1028
1029                         ksort( $snippets );
1030                 }
1031
1032                 // add extra chars to each snippet to make snippets constant size
1033                 $extended = array();
1034                 if ( count( $snippets ) == 0 ) {
1035                         // couldn't find the target words, just show beginning of article
1036                         if ( array_key_exists( $first, $all ) ) {
1037                                 $targetchars = $contextchars * $contextlines;
1038                                 $snippets[$first] = '';
1039                                 $offsets[$first] = 0;
1040                         }
1041                 } else {
1042                         // if begin of the article contains the whole phrase, show only that !!
1043                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1044                             && $offsets[$first] < $contextchars * 2 ) {
1045                                 $snippets = array ( $first => $snippets[$first] );
1046                         }
1047
1048                         // calc by how much to extend existing snippets
1049                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1050                 }
1051
1052                 foreach ( $snippets as $index => $line ) {
1053                         $extended[$index] = $line;
1054                         $len = strlen( $line );
1055                         if ( $len < $targetchars - 20 ) {
1056                                 // complete this line
1057                                 if ( $len < strlen( $all[$index] ) ) {
1058                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1059                                         $len = strlen( $extended[$index] );
1060                                 }
1061
1062                                 // add more lines
1063                                 $add = $index + 1;
1064                                 while ( $len < $targetchars - 20
1065                                        && array_key_exists( $add, $all )
1066                                        && !array_key_exists( $add, $snippets ) ) {
1067                                     $offsets[$add] = 0;
1068                                     $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1069                                         $extended[$add] = $tt;
1070                                         $len += strlen( $tt );
1071                                         $add++;
1072                                 }
1073                         }
1074                 }
1075
1076                 // $snippets = array_map('htmlspecialchars', $extended);
1077                 $snippets = $extended;
1078                 $last = - 1;
1079                 $extract = '';
1080                 foreach ( $snippets as $index => $line ) {
1081                         if ( $last == - 1 )
1082                                 $extract .= $line; // first line
1083                         elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) )
1084                                 $extract .= " " . $line; // continous lines
1085                         else
1086                                 $extract .= '<b> ... </b>' . $line;
1087
1088                         $last = $index;
1089                 }
1090                 if ( $extract )
1091                         $extract .= '<b> ... </b>';
1092
1093                 $processed = array();
1094                 foreach ( $terms as $term ) {
1095                         if ( ! isset( $processed[$term] ) ) {
1096                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1097                                 $extract = preg_replace( $pat3,
1098                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1099                                 $processed[$term] = true;
1100                         }
1101                 }
1102
1103                 wfProfileOut( "$fname-extract" );
1104
1105                 return $extract;
1106         }
1107
1108         /**
1109          * Split text into lines and add it to extracts array
1110          *
1111          * @param $extracts Array: index -> $line
1112          * @param $count Integer
1113          * @param $text String
1114          */
1115         function splitAndAdd( &$extracts, &$count, $text ) {
1116                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1117                 foreach ( $split as $line ) {
1118                         $tt = trim( $line );
1119                         if ( $tt )
1120                                 $extracts[$count++] = $tt;
1121                 }
1122         }
1123
1124         /**
1125          * Do manual case conversion for non-ascii chars
1126          *
1127          * @param $matches Array
1128          */
1129         function caseCallback( $matches ) {
1130                 global $wgContLang;
1131                 if ( strlen( $matches[0] ) > 1 ) {
1132                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1133                 } else
1134                         return $matches[0];
1135         }
1136
1137         /**
1138          * Extract part of the text from start to end, but by
1139          * not chopping up words
1140          * @param $text String
1141          * @param $start Integer
1142          * @param $end Integer
1143          * @param $posStart Integer: (out) actual start position
1144          * @param $posEnd Integer: (out) actual end position
1145          * @return String
1146          */
1147         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1148                 if ( $start != 0 )
1149                         $start = $this->position( $text, $start, 1 );
1150                 if ( $end >= strlen( $text ) )
1151                         $end = strlen( $text );
1152                 else
1153                         $end = $this->position( $text, $end );
1154
1155                 if ( !is_null( $posStart ) )
1156                         $posStart = $start;
1157                 if ( !is_null( $posEnd ) )
1158                         $posEnd = $end;
1159
1160                 if ( $end > $start )
1161                         return substr( $text, $start, $end - $start );
1162                 else
1163                         return '';
1164         }
1165
1166         /**
1167          * Find a nonletter near a point (index) in the text
1168          *
1169          * @param $text String
1170          * @param $point Integer
1171          * @param $offset Integer: offset to found index
1172          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1173          */
1174         function position( $text, $point, $offset = 0 ) {
1175                 $tolerance = 10;
1176                 $s = max( 0, $point - $tolerance );
1177                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1178                 $m = array();
1179                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1180                         return $m[0][1] + $s + $offset;
1181                 } else {
1182                         // check if point is on a valid first UTF8 char
1183                         $char = ord( $text[$point] );
1184                         while ( $char >= 0x80 && $char < 0xc0 ) {
1185                                 // skip trailing bytes
1186                                 $point++;
1187                                 if ( $point >= strlen( $text ) )
1188                                         return strlen( $text );
1189                                 $char = ord( $text[$point] );
1190                         }
1191                         return $point;
1192
1193                 }
1194         }
1195
1196         /**
1197          * Search extracts for a pattern, and return snippets
1198          *
1199          * @param $pattern String: regexp for matching lines
1200          * @param $extracts Array: extracts to search
1201          * @param $linesleft Integer: number of extracts to make
1202          * @param $contextchars Integer: length of snippet
1203          * @param $out Array: map for highlighted snippets
1204          * @param $offsets Array: map of starting points of snippets
1205          * @protected
1206          */
1207         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1208                 if ( $linesleft == 0 )
1209                         return; // nothing to do
1210                 foreach ( $extracts as $index => $line ) {
1211                         if ( array_key_exists( $index, $out ) )
1212                                 continue; // this line already highlighted
1213
1214                         $m = array();
1215                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1216                                 continue;
1217
1218                         $offset = $m[0][1];
1219                         $len = strlen( $m[0][0] );
1220                         if ( $offset + $len < $contextchars )
1221                                 $begin = 0;
1222                         elseif ( $len > $contextchars )
1223                                 $begin = $offset;
1224                         else
1225                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1226
1227                         $end = $begin + $contextchars;
1228
1229                         $posBegin = $begin;
1230                         // basic snippet from this line
1231                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1232                         $offsets[$index] = $posBegin;
1233                         $linesleft--;
1234                         if ( $linesleft == 0 )
1235                                 return;
1236                 }
1237         }
1238
1239         /**
1240          * Basic wikitext removal
1241          * @protected
1242          */
1243         function removeWiki( $text ) {
1244                 $fname = __METHOD__;
1245                 wfProfileIn( $fname );
1246
1247                 // $text = preg_replace("/'{2,5}/", "", $text);
1248                 // $text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1249                 // $text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1250                 // $text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1251                 // $text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1252                 // $text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1253                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1254                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1255                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1256                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1257                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1258                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1259                 $text = preg_replace( "/'''''/", "", $text );
1260                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1261                 $text = preg_replace( "/''/", "", $text );
1262
1263                 wfProfileOut( $fname );
1264                 return $text;
1265         }
1266
1267         /**
1268          * callback to replace [[target|caption]] kind of links, if
1269          * the target is category or image, leave it
1270          *
1271          * @param $matches Array
1272          */
1273         function linkReplace( $matches ) {
1274                 $colon = strpos( $matches[1], ':' );
1275                 if ( $colon === false )
1276                         return $matches[2]; // replace with caption
1277                 global $wgContLang;
1278                 $ns = substr( $matches[1], 0, $colon );
1279                 $index = $wgContLang->getNsIndex( $ns );
1280                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) )
1281                         return $matches[0]; // return the whole thing
1282                 else
1283                         return $matches[2];
1284
1285         }
1286
1287         /**
1288      * Simple & fast snippet extraction, but gives completely unrelevant
1289      * snippets
1290      *
1291      * @param $text String
1292      * @param $terms Array
1293      * @param $contextlines Integer
1294      * @param $contextchars Integer
1295      * @return String
1296      */
1297     public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1298         global $wgContLang;
1299         $fname = __METHOD__;
1300
1301         $lines = explode( "\n", $text );
1302
1303         $terms = implode( '|', $terms );
1304         $max = intval( $contextchars ) + 1;
1305         $pat1 = "/(.*)($terms)(.{0,$max})/i";
1306
1307         $lineno = 0;
1308
1309         $extract = "";
1310         wfProfileIn( "$fname-extract" );
1311         foreach ( $lines as $line ) {
1312             if ( 0 == $contextlines ) {
1313                 break;
1314             }
1315             ++$lineno;
1316             $m = array();
1317             if ( ! preg_match( $pat1, $line, $m ) ) {
1318                 continue;
1319             }
1320             --$contextlines;
1321             $pre = $wgContLang->truncate( $m[1], - $contextchars );
1322
1323             if ( count( $m ) < 3 ) {
1324                 $post = '';
1325             } else {
1326                 $post = $wgContLang->truncate( $m[3], $contextchars );
1327             }
1328
1329             $found = $m[2];
1330
1331             $line = htmlspecialchars( $pre . $found . $post );
1332             $pat2 = '/(' . $terms . ")/i";
1333             $line = preg_replace( $pat2,
1334               "<span class='searchmatch'>\\1</span>", $line );
1335
1336             $extract .= "${line}\n";
1337         }
1338         wfProfileOut( "$fname-extract" );
1339
1340         return $extract;
1341     }
1342
1343 }
1344
1345 /**
1346  * Dummy class to be used when non-supported Database engine is present.
1347  * @todo Fixme: dummy class should probably try something at least mildly useful,
1348  * such as a LIKE search through titles.
1349  * @ingroup Search
1350  */
1351 class SearchEngineDummy extends SearchEngine {
1352         // no-op
1353 }