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