includes/search/SearchEngine.php

   1 <?php
   2 /**
   3  * @defgroup Search Search
   4  *
   5  * @file
   6  * @ingroup Search
   7  */
   8
   9 /**
  10  * Contain a class for special pages
  11  * @ingroup Search
  12  */
  13 class SearchEngine {
  14         var $limit = 10;
  15         var $offset = 0;
  16         var $prefix = '';
  17         var $searchTerms = array();
  18         var $namespaces = array( NS_MAIN );
  19         var $showRedirects = false;
  20
  21         /**
  22          * Perform a full text search query and return a result set.
  23          * If title searches are not supported or disabled, return null.
  24          * STUB
  25          *
  26          * @param $term String: raw search term
  27          * @return SearchResultSet
  28          */
  29         function searchText( $term ) {
  30                 return null;
  31         }
  32
  33         /**
  34          * Perform a title-only 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 searchTitle( $term ) {
  42                 return null;
  43         }
  44
  45         /** If this search backend can list/unlist redirects */
  46         function acceptListRedirects() {
  47                 return true;
  48         }
  49
  50         /**
  51          * When overridden in derived class, performs database-specific conversions
  52          * on text to be used for searching or updating search index.
  53          * Default implementation does nothing (simply returns $string).
  54          *
  55          * @param $string string: String to process
  56          * @return string
  57          */
  58         public function normalizeText( $string ) {
  59                 global $wgContLang;
  60
  61                 // Some languages such as Chinese require word segmentation
  62                 return $wgContLang->segmentByWord( $string );
  63         }
  64
  65         /**
  66          * Transform search term in cases when parts of the query came as different GET params (when supported)
  67          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
  68          */
  69         function transformSearchTerm( $term ) {
  70                 return $term;
  71         }
  72
  73         /**
  74          * If an exact title match can be found, or a very slightly close match,
  75          * return the title. If no match, returns NULL.
  76          *
  77          * @param $searchterm String
  78          * @return Title
  79          */
  80         public static function getNearMatch( $searchterm ) {
  81                 $title = self::getNearMatchInternal( $searchterm );
  82
  83                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
  84                 return $title;
  85         }
  86
  87         /**
  88          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
  89          * SearchResultSet.
  90          *
  91          * @param $searchterm string
  92          * @return SearchResultSet
  93          */
  94         public static function getNearMatchResultSet( $searchterm ) {
  95                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
  96         }
  97
  98         /**
  99          * Really find the title match.
 100          */
 101         private static function getNearMatchInternal( $searchterm ) {
 102                 global $wgContLang;
 103
 104                 $allSearchTerms = array($searchterm);
 105
 106                 if ( $wgContLang->hasVariants() ) {
 107                         $allSearchTerms = array_merge($allSearchTerms,$wgContLang->convertLinkToAllVariants($searchterm));
 108                 }
 109
 110                 if( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 111                         return $titleResult;
 112                 }
 113
 114                 foreach($allSearchTerms as $term) {
 115
 116                         # Exact match? No need to look further.
 117                         $title = Title::newFromText( $term );
 118                         if (is_null($title))
 119                                 return null;
 120
 121                         if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal() || $title->exists() ) {
 122                                 return $title;
 123                         }
 124
 125                         # See if it still otherwise has content is some sane sense
 126                         $article = MediaWiki::articleFromTitle( $title );
 127                         if( $article->hasViewableContent() ) {
 128                                 return $title;
 129                         }
 130
 131                         # Now try all lower case (i.e. first letter capitalized)
 132                         #
 133                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 134                         if ( $title && $title->exists() ) {
 135                                 return $title;
 136                         }
 137
 138                         # Now try capitalized string
 139                         #
 140                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 141                         if ( $title && $title->exists() ) {
 142                                 return $title;
 143                         }
 144
 145                         # Now try all upper case
 146                         #
 147                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 148                         if ( $title && $title->exists() ) {
 149                                 return $title;
 150                         }
 151
 152                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 153                         $title = Title::newFromText( $wgContLang->ucwordbreaks($term) );
 154                         if ( $title && $title->exists() ) {
 155                                 return $title;
 156                         }
 157
 158                         // Give hooks a chance at better match variants
 159                         $title = null;
 160                         if( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 161                                 return $title;
 162                         }
 163                 }
 164
 165                 $title = Title::newFromText( $searchterm );
 166
 167                 # Entering an IP address goes to the contributions page
 168                 if ( ( $title->getNamespace() == NS_USER && User::isIP($title->getText() ) )
 169                         || User::isIP( trim( $searchterm ) ) ) {
 170                         return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 171                 }
 172
 173
 174                 # Entering a user goes to the user page whether it's there or not
 175                 if ( $title->getNamespace() == NS_USER ) {
 176                         return $title;
 177                 }
 178
 179                 # Go to images that exist even if there's no local page.
 180                 # There may have been a funny upload, or it may be on a shared
 181                 # file repository such as Wikimedia Commons.
 182                 if( $title->getNamespace() == NS_FILE ) {
 183                         $image = wfFindFile( $title );
 184                         if( $image ) {
 185                                 return $title;
 186                         }
 187                 }
 188
 189                 # MediaWiki namespace? Page may be "implied" if not customized.
 190                 # Just return it, with caps forced as the message system likes it.
 191                 if( $title->getNamespace() == NS_MEDIAWIKI ) {
 192                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 193                 }
 194
 195                 # Quoted term? Try without the quotes...
 196                 $matches = array();
 197                 if( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 198                         return SearchEngine::getNearMatch( $matches[1] );
 199                 }
 200
 201                 return null;
 202         }
 203
 204         public static function legalSearchChars() {
 205                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 206         }
 207
 208         /**
 209          * Set the maximum number of results to return
 210          * and how many to skip before returning the first.
 211          *
 212          * @param $limit Integer
 213          * @param $offset Integer
 214          */
 215         function setLimitOffset( $limit, $offset = 0 ) {
 216                 $this->limit = intval( $limit );
 217                 $this->offset = intval( $offset );
 218         }
 219
 220         /**
 221          * Set which namespaces the search should include.
 222          * Give an array of namespace index numbers.
 223          *
 224          * @param $namespaces Array
 225          */
 226         function setNamespaces( $namespaces ) {
 227                 $this->namespaces = $namespaces;
 228         }
 229
 230         /**
 231          * Parse some common prefixes: all (search everything)
 232          * or namespace names
 233          *
 234          * @param $query String
 235          */
 236         function replacePrefixes( $query ){
 237                 global $wgContLang;
 238
 239                 $parsed = $query;
 240                 if( strpos($query,':') === false ) { // nothing to do
 241                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 242                         return $parsed;
 243                 }
 244
 245                 $allkeyword = wfMsgForContent('searchall').":";
 246                 if( strncmp($query, $allkeyword, strlen($allkeyword)) == 0 ){
 247                         $this->namespaces = null;
 248                         $parsed = substr($query,strlen($allkeyword));
 249                 } else if( strpos($query,':') !== false ) {
 250                         $prefix = substr($query,0,strpos($query,':'));
 251                         $index = $wgContLang->getNsIndex($prefix);
 252                         if($index !== false){
 253                                 $this->namespaces = array($index);
 254                                 $parsed = substr($query,strlen($prefix)+1);
 255                         }
 256                 }
 257                 if(trim($parsed) == '')
 258                         $parsed = $query; // prefix was the whole query
 259
 260                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 261
 262                 return $parsed;
 263         }
 264
 265         /**
 266          * Make a list of searchable namespaces and their canonical names.
 267          * @return Array
 268          */
 269         public static function searchableNamespaces() {
 270                 global $wgContLang;
 271                 $arr = array();
 272                 foreach( $wgContLang->getNamespaces() as $ns => $name ) {
 273                         if( $ns >= NS_MAIN ) {
 274                                 $arr[$ns] = $name;
 275                         }
 276                 }
 277
 278                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 279                 return $arr;
 280         }
 281
 282         /**
 283          * Extract default namespaces to search from the given user's
 284          * settings, returning a list of index numbers.
 285          *
 286          * @param $user User
 287          * @return Array
 288          */
 289         public static function userNamespaces( $user ) {
 290                 global $wgSearchEverythingOnlyLoggedIn;
 291
 292                 // get search everything preference, that can be set to be read for logged-in users
 293                 $searcheverything = false;
 294                 if( ( $wgSearchEverythingOnlyLoggedIn && $user->isLoggedIn() )
 295                     || !$wgSearchEverythingOnlyLoggedIn )
 296                         $searcheverything = $user->getOption('searcheverything');
 297
 298                 // searcheverything overrides other options
 299                 if( $searcheverything )
 300                         return array_keys(SearchEngine::searchableNamespaces());
 301
 302                 $arr = Preferences::loadOldSearchNs( $user );
 303                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 304
 305                 $arr = array_intersect( $arr, array_keys($searchableNamespaces) ); // Filter
 306
 307                 return $arr;
 308         }
 309
 310         /**
 311          * Find snippet highlight settings for a given user
 312          *
 313          * @param $user User
 314          * @return Array contextlines, contextchars
 315          */
 316         public static function userHighlightPrefs( &$user ){
 317                 //$contextlines = $user->getOption( 'contextlines',  5 );
 318                 //$contextchars = $user->getOption( 'contextchars', 50 );
 319                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 320                 $contextchars = 75; // same as above.... :P
 321                 return array($contextlines, $contextchars);
 322         }
 323
 324         /**
 325          * An array of namespaces indexes to be searched by default
 326          *
 327          * @return Array
 328          */
 329         public static function defaultNamespaces(){
 330                 global $wgNamespacesToBeSearchedDefault;
 331
 332                 return array_keys($wgNamespacesToBeSearchedDefault, true);
 333         }
 334
 335         /**
 336          * Get a list of namespace names useful for showing in tooltips
 337          * and preferences
 338          *
 339          * @param $namespaces Array
 340          */
 341         public static function namespacesAsText( $namespaces ){
 342                 global $wgContLang;
 343
 344                 $formatted = array_map( array($wgContLang,'getFormattedNsText'), $namespaces );
 345                 foreach( $formatted as $key => $ns ){
 346                         if ( empty($ns) )
 347                                 $formatted[$key] = wfMsg( 'blanknamespace' );
 348                 }
 349                 return $formatted;
 350         }
 351
 352         /**
 353          * Return the help namespaces to be shown on Special:Search
 354          *
 355          * @return Array
 356          */
 357         public static function helpNamespaces() {
 358                 global $wgNamespacesToBeSearchedHelp;
 359
 360                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 361         }
 362
 363         /**
 364          * Return a 'cleaned up' search string
 365          *
 366          * @param $text String
 367          * @return String
 368          */
 369         function filter( $text ) {
 370                 $lc = $this->legalSearchChars();
 371                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 372         }
 373         /**
 374          * Load up the appropriate search engine class for the currently
 375          * active database backend, and return a configured instance.
 376          *
 377          * @return SearchEngine
 378          */
 379         public static function create() {
 380                 global $wgSearchType;
 381                 $dbr = wfGetDB( DB_SLAVE );
 382                 if( $wgSearchType ) {
 383                         $class = $wgSearchType;
 384                 } else {
 385                         $class = $dbr->getSearchEngine();
 386                 }
 387                 $search = new $class( $dbr );
 388                 $search->setLimitOffset(0,0);
 389                 return $search;
 390         }
 391
 392         /**
 393          * Create or update the search index record for the given page.
 394          * Title and text should be pre-processed.
 395          * STUB
 396          *
 397          * @param $id Integer
 398          * @param $title String
 399          * @param $text String
 400          */
 401         function update( $id, $title, $text ) {
 402                 // no-op
 403         }
 404
 405         /**
 406          * Update a search index record's title only.
 407          * Title should be pre-processed.
 408          * STUB
 409          *
 410          * @param $id Integer
 411          * @param $title String
 412          */
 413         function updateTitle( $id, $title ) {
 414                 // no-op
 415         }
 416
 417         /**
 418          * Get OpenSearch suggestion template
 419          *
 420          * @return String
 421          */
 422         public static function getOpenSearchTemplate() {
 423                 global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
 424                 if( $wgOpenSearchTemplate )     {
 425                         return $wgOpenSearchTemplate;
 426                 } else {
 427                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 428                         if( !$ns ) $ns = "0";
 429                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace='.$ns;
 430                 }
 431         }
 432
 433         /**
 434          * Get internal MediaWiki Suggest template
 435          *
 436          * @return String
 437          */
 438         public static function getMWSuggestTemplate() {
 439                 global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
 440                 if($wgMWSuggestTemplate)
 441                         return $wgMWSuggestTemplate;
 442                 else
 443                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
 444         }
 445 }
 446
 447 /**
 448  * @ingroup Search
 449  */
 450 class SearchResultSet {
 451         /**
 452          * Fetch an array of regular expression fragments for matching
 453          * the search terms as parsed by this engine in a text extract.
 454          * STUB
 455          *
 456          * @return Array
 457          */
 458         function termMatches() {
 459                 return array();
 460         }
 461
 462         function numRows() {
 463                 return 0;
 464         }
 465
 466         /**
 467          * Return true if results are included in this result set.
 468          * STUB
 469          *
 470          * @return Boolean
 471          */
 472         function hasResults() {
 473                 return false;
 474         }
 475
 476         /**
 477          * Some search modes return a total hit count for the query
 478          * in the entire article database. This may include pages
 479          * in namespaces that would not be matched on the given
 480          * settings.
 481          *
 482          * Return null if no total hits number is supported.
 483          *
 484          * @return Integer
 485          */
 486         function getTotalHits() {
 487                 return null;
 488         }
 489
 490         /**
 491          * Some search modes return a suggested alternate term if there are
 492          * no exact hits. Returns true if there is one on this set.
 493          *
 494          * @return Boolean
 495          */
 496         function hasSuggestion() {
 497                 return false;
 498         }
 499
 500         /**
 501          * @return String: suggested query, null if none
 502          */
 503         function getSuggestionQuery(){
 504                 return null;
 505         }
 506
 507         /**
 508          * @return String: HTML highlighted suggested query, '' if none
 509          */
 510         function getSuggestionSnippet(){
 511                 return '';
 512         }
 513
 514         /**
 515          * Return information about how and from where the results were fetched,
 516          * should be useful for diagnostics and debugging
 517          *
 518          * @return String
 519          */
 520         function getInfo() {
 521                 return null;
 522         }
 523
 524         /**
 525          * Return a result set of hits on other (multiple) wikis associated with this one
 526          *
 527          * @return SearchResultSet
 528          */
 529         function getInterwikiResults() {
 530                 return null;
 531         }
 532
 533         /**
 534          * Check if there are results on other wikis
 535          *
 536          * @return Boolean
 537          */
 538         function hasInterwikiResults() {
 539                 return $this->getInterwikiResults() != null;
 540         }
 541
 542         /**
 543          * Fetches next search result, or false.
 544          * STUB
 545          *
 546          * @return SearchResult
 547          */
 548         function next() {
 549                 return false;
 550         }
 551
 552         /**
 553          * Frees the result set, if applicable.
 554          */
 555         function free() {
 556                 // ...
 557         }
 558 }
 559
 560 /**
 561  * This class is used for different SQL-based search engines shipped with MediaWiki
 562  */
 563 class SqlSearchResultSet extends SearchResultSet {
 564         function __construct( $resultSet, $terms ) {
 565                 $this->mResultSet = $resultSet;
 566                 $this->mTerms = $terms;
 567         }
 568
 569         function termMatches() {
 570                 return $this->mTerms;
 571         }
 572
 573         function numRows() {
 574                 if ($this->mResultSet === false )
 575                         return false;
 576
 577                 return $this->mResultSet->numRows();
 578         }
 579
 580         function next() {
 581                 if ($this->mResultSet === false )
 582                         return false;
 583
 584                 $row = $this->mResultSet->fetchObject();
 585                 if ($row === false)
 586                         return false;
 587
 588                 return SearchResult::newFromRow( $row );
 589         }
 590
 591         function free() {
 592                 if ($this->mResultSet === false )
 593                         return false;
 594
 595                 $this->mResultSet->free();
 596         }
 597 }
 598
 599 /**
 600  * @ingroup Search
 601  */
 602 class SearchResultTooMany {
 603         ## Some search engines may bail out if too many matches are found
 604 }
 605
 606
 607 /**
 608  * @todo Fixme: This class is horribly factored. It would probably be better to
 609  * have a useful base class to which you pass some standard information, then
 610  * let the fancy self-highlighters extend that.
 611  * @ingroup Search
 612  */
 613 class SearchResult {
 614         var $mRevision = null;
 615         var $mImage = null;
 616
 617         /**
 618          * Return a new SearchResult and initializes it with a title.
 619          *
 620          * @param $title Title
 621          * @return SearchResult
 622          */
 623         public static function newFromTitle( $title ) {
 624                 $result = new self();
 625                 $result->initFromTitle( $title );
 626                 return $result;
 627         }
 628         /**
 629          * Return a new SearchResult and initializes it with a row.
 630          *
 631          * @param $row object
 632          * @return SearchResult
 633          */
 634         public static function newFromRow( $row ) {
 635                 $result = new self();
 636                 $result->initFromRow( $row );
 637                 return $result;
 638         }
 639
 640         public function __construct( $row = null ) {
 641                 if ( !is_null( $row ) ) {
 642                         // Backwards compatibility with pre-1.17 callers
 643                         $this->initFromRow( $row );
 644                 }
 645         }
 646
 647         /**
 648          * Initialize from a database row. Makes a Title and passes that to
 649          * initFromTitle.
 650          *
 651          * @param $row object
 652          */
 653         protected function initFromRow( $row ) {
 654                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 655         }
 656
 657         /**
 658          * Initialize from a Title and if possible initializes a corresponding
 659          * Revision and File.
 660          *
 661          * @param $title Title
 662          */
 663         protected function initFromTitle( $title ) {
 664                 $this->mTitle = $title;
 665                 if( !is_null( $this->mTitle ) ){
 666                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 667                         if( $this->mTitle->getNamespace() === NS_FILE )
 668                                 $this->mImage = wfFindFile( $this->mTitle );
 669                 }
 670         }
 671
 672         /**
 673          * Check if this is result points to an invalid title
 674          *
 675          * @return Boolean
 676          */
 677         function isBrokenTitle(){
 678                 if( is_null($this->mTitle) )
 679                         return true;
 680                 return false;
 681         }
 682
 683         /**
 684          * Check if target page is missing, happens when index is out of date
 685          *
 686          * @return Boolean
 687          */
 688         function isMissingRevision(){
 689                 return !$this->mRevision && !$this->mImage;
 690         }
 691
 692         /**
 693          * @return Title
 694          */
 695         function getTitle() {
 696                 return $this->mTitle;
 697         }
 698
 699         /**
 700          * @return Double or null if not supported
 701          */
 702         function getScore() {
 703                 return null;
 704         }
 705
 706         /**
 707          * Lazy initialization of article text from DB
 708          */
 709         protected function initText(){
 710                 if( !isset($this->mText) ){
 711                         if($this->mRevision != null)
 712                                 $this->mText = $this->mRevision->getText();
 713                         else // TODO: can we fetch raw wikitext for commons images?
 714                                 $this->mText = '';
 715
 716                 }
 717         }
 718
 719         /**
 720          * @param $terms Array: terms to highlight
 721          * @return String: highlighted text snippet, null (and not '') if not supported
 722          */
 723         function getTextSnippet($terms){
 724                 global $wgUser, $wgAdvancedSearchHighlighting;
 725                 $this->initText();
 726                 list($contextlines,$contextchars) = SearchEngine::userHighlightPrefs($wgUser);
 727                 $h = new SearchHighlighter();
 728                 if( $wgAdvancedSearchHighlighting )
 729                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 730                 else
 731                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 732         }
 733
 734         /**
 735          * @param $terms Array: terms to highlight
 736          * @return String: highlighted title, '' if not supported
 737          */
 738         function getTitleSnippet($terms){
 739                 return '';
 740         }
 741
 742         /**
 743          * @param $terms Array: terms to highlight
 744          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 745          */
 746         function getRedirectSnippet($terms){
 747                 return '';
 748         }
 749
 750         /**
 751          * @return Title object for the redirect to this page, null if none or not supported
 752          */
 753         function getRedirectTitle(){
 754                 return null;
 755         }
 756
 757         /**
 758          * @return string highlighted relevant section name, null if none or not supported
 759          */
 760         function getSectionSnippet(){
 761                 return '';
 762         }
 763
 764         /**
 765          * @return Title object (pagename+fragment) for the section, null if none or not supported
 766          */
 767         function getSectionTitle(){
 768                 return null;
 769         }
 770
 771         /**
 772          * @return String: timestamp
 773          */
 774         function getTimestamp(){
 775                 if( $this->mRevision )
 776                         return $this->mRevision->getTimestamp();
 777                 else if( $this->mImage )
 778                         return $this->mImage->getTimestamp();
 779                 return '';
 780         }
 781
 782         /**
 783          * @return Integer: number of words
 784          */
 785         function getWordCount(){
 786                 $this->initText();
 787                 return str_word_count( $this->mText );
 788         }
 789
 790         /**
 791          * @return Integer: size in bytes
 792          */
 793         function getByteSize(){
 794                 $this->initText();
 795                 return strlen( $this->mText );
 796         }
 797
 798         /**
 799          * @return Boolean if hit has related articles
 800          */
 801         function hasRelated(){
 802                 return false;
 803         }
 804
 805         /**
 806          * @return String: interwiki prefix of the title (return iw even if title is broken)
 807          */
 808         function getInterwikiPrefix(){
 809                 return '';
 810         }
 811 }
 812 /**
 813  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 814  */
 815 class SearchNearMatchResultSet extends SearchResultSet {
 816         private $fetched = false;
 817         /**
 818          * @param $match mixed Title if matched, else null
 819          */
 820         public function __construct( $match ) {
 821                 $this->result = $match;
 822         }
 823         public function hasResult() {
 824                 return (bool)$this->result;
 825         }
 826         public function numRows() {
 827                 return $this->hasResults() ? 1 : 0;
 828         }
 829         public function next() {
 830                 if ( $this->fetched || !$this->result ) {
 831                         return false;
 832                 }
 833                 $this->fetched = true;
 834                 return SearchResult::newFromTitle( $this->result );
 835         }
 836 }
 837
 838 /**
 839  * Highlight bits of wikitext
 840  *
 841  * @ingroup Search
 842  */
 843 class SearchHighlighter {
 844         var $mCleanWikitext = true;
 845
 846         function SearchHighlighter($cleanupWikitext = true){
 847                 $this->mCleanWikitext = $cleanupWikitext;
 848         }
 849
 850         /**
 851          * Default implementation of wikitext highlighting
 852          *
 853          * @param $text String
 854          * @param $terms Array: terms to highlight (unescaped)
 855          * @param $contextlines Integer
 856          * @param $contextchars Integer
 857          * @return String
 858          */
 859         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 860                 global $wgLang, $wgContLang;
 861                 global $wgSearchHighlightBoundaries;
 862                 $fname = __METHOD__;
 863
 864                 if($text == '')
 865                         return '';
 866
 867                 // spli text into text + templates/links/tables
 868                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 869                 // first capture group is for detecting nested templates/links/tables/references
 870                 $endPatterns = array(
 871                         1 => '/(\{\{)|(\}\})/', // template
 872                         2 => '/(\[\[)|(\]\])/', // image
 873                         3 => "/(\n\\{\\|)|(\n\\|\\})/"); // table
 874
 875                 // FIXME: this should prolly be a hook or something
 876                 if(function_exists('wfCite')){
 877                         $spat .= '|(<ref>)'; // references via cite extension
 878                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 879                 }
 880                 $spat .= '/';
 881                 $textExt = array(); // text extracts
 882                 $otherExt = array();  // other extracts
 883                 wfProfileIn( "$fname-split" );
 884                 $start = 0;
 885                 $textLen = strlen($text);
 886                 $count = 0; // sequence number to maintain ordering
 887                 while( $start < $textLen ){
 888                         // find start of template/image/table
 889                         if( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ){
 890                                 $epat = '';
 891                                 foreach($matches as $key => $val){
 892                                         if($key > 0 && $val[1] != -1){
 893                                                 if($key == 2){
 894                                                         // see if this is an image link
 895                                                         $ns = substr($val[0],2,-1);
 896                                                         if( $wgContLang->getNsIndex($ns) != NS_FILE )
 897                                                                 break;
 898
 899                                                 }
 900                                                 $epat = $endPatterns[$key];
 901                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 902                                                 $start = $val[1];
 903                                                 break;
 904                                         }
 905                                 }
 906                                 if( $epat ){
 907                                         // find end (and detect any nested elements)
 908                                         $level = 0;
 909                                         $offset = $start + 1;
 910                                         $found = false;
 911                                         while( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ){
 912                                                 if( array_key_exists(2,$endMatches) ){
 913                                                         // found end
 914                                                         if($level == 0){
 915                                                                 $len = strlen($endMatches[2][0]);
 916                                                                 $off = $endMatches[2][1];
 917                                                                 $this->splitAndAdd( $otherExt, $count,
 918                                                                         substr( $text, $start, $off + $len  - $start ) );
 919                                                                 $start = $off + $len;
 920                                                                 $found = true;
 921                                                                 break;
 922                                                         } else{
 923                                                                 // end of nested element
 924                                                                 $level -= 1;
 925                                                         }
 926                                                 } else{
 927                                                         // nested
 928                                                         $level += 1;
 929                                                 }
 930                                                 $offset = $endMatches[0][1] + strlen($endMatches[0][0]);
 931                                         }
 932                                         if( ! $found ){
 933                                                 // couldn't find appropriate closing tag, skip
 934                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen($matches[0][0]) ) );
 935                                                 $start += strlen($matches[0][0]);
 936                                         }
 937                                         continue;
 938                                 }
 939                         }
 940                         // else: add as text extract
 941                         $this->splitAndAdd( $textExt, $count, substr($text,$start) );
 942                         break;
 943                 }
 944
 945                 $all = $textExt + $otherExt; // these have disjunct key sets
 946
 947                 wfProfileOut( "$fname-split" );
 948
 949                 // prepare regexps
 950                 foreach( $terms as $index => $term ) {
 951                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 952                         if(preg_match('/[\x80-\xff]/', $term) ){
 953                                 $terms[$index] = preg_replace_callback('/./us',array($this,'caseCallback'),$terms[$index]);
 954                         } else {
 955                                 $terms[$index] = $term;
 956                         }
 957                 }
 958                 $anyterm = implode( '|', $terms );
 959                 $phrase = implode("$wgSearchHighlightBoundaries+", $terms );
 960
 961                 // FIXME: a hack to scale contextchars, a correct solution
 962                 // would be to have contextchars actually be char and not byte
 963                 // length, and do proper utf-8 substrings and lengths everywhere,
 964                 // but PHP is making that very hard and unclean to implement :(
 965                 $scale = strlen($anyterm) / mb_strlen($anyterm);
 966                 $contextchars = intval( $contextchars * $scale );
 967
 968                 $patPre = "(^|$wgSearchHighlightBoundaries)";
 969                 $patPost = "($wgSearchHighlightBoundaries|$)";
 970
 971                 $pat1 = "/(".$phrase.")/ui";
 972                 $pat2 = "/$patPre(".$anyterm.")$patPost/ui";
 973
 974                 wfProfileIn( "$fname-extract" );
 975
 976                 $left = $contextlines;
 977
 978                 $snippets = array();
 979                 $offsets = array();
 980
 981                 // show beginning only if it contains all words
 982                 $first = 0;
 983                 $firstText = '';
 984                 foreach($textExt as $index => $line){
 985                         if(strlen($line)>0 && $line[0] != ';' && $line[0] != ':'){
 986                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
 987                                 $first = $index;
 988                                 break;
 989                         }
 990                 }
 991                 if( $firstText ){
 992                         $succ = true;
 993                         // check if first text contains all terms
 994                         foreach($terms as $term){
 995                                 if( ! preg_match("/$patPre".$term."$patPost/ui", $firstText) ){
 996                                         $succ = false;
 997                                         break;
 998                                 }
 999                         }
1000                         if( $succ ){
1001                                 $snippets[$first] = $firstText;
1002                                 $offsets[$first] = 0;
1003                         }
1004                 }
1005                 if( ! $snippets ) {
1006                         // match whole query on text
1007                         $this->process($pat1, $textExt, $left, $contextchars, $snippets, $offsets);
1008                         // match whole query on templates/tables/images
1009                         $this->process($pat1, $otherExt, $left, $contextchars, $snippets, $offsets);
1010                         // match any words on text
1011                         $this->process($pat2, $textExt, $left, $contextchars, $snippets, $offsets);
1012                         // match any words on templates/tables/images
1013                         $this->process($pat2, $otherExt, $left, $contextchars, $snippets, $offsets);
1014
1015                         ksort($snippets);
1016                 }
1017
1018                 // add extra chars to each snippet to make snippets constant size
1019                 $extended = array();
1020                 if( count( $snippets ) == 0 ){
1021                         // couldn't find the target words, just show beginning of article
1022                         if ( array_key_exists( $first, $all ) ) {
1023                                 $targetchars = $contextchars * $contextlines;
1024                                 $snippets[$first] = '';
1025                                 $offsets[$first] = 0;
1026                         }
1027                 } else {
1028                         // if begin of the article contains the whole phrase, show only that !!
1029                         if( array_key_exists($first,$snippets) && preg_match($pat1,$snippets[$first])
1030                             && $offsets[$first] < $contextchars * 2 ){
1031                                 $snippets = array ($first => $snippets[$first]);
1032                         }
1033
1034                         // calc by how much to extend existing snippets
1035                         $targetchars = intval( ($contextchars * $contextlines) / count ( $snippets ) );
1036                 }
1037
1038                 foreach($snippets as $index => $line){
1039                         $extended[$index] = $line;
1040                         $len = strlen($line);
1041                         if( $len < $targetchars - 20 ){
1042                                 // complete this line
1043                                 if($len < strlen( $all[$index] )){
1044                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index]+$targetchars, $offsets[$index]);
1045                                         $len = strlen( $extended[$index] );
1046                                 }
1047
1048                                 // add more lines
1049                                 $add = $index + 1;
1050                                 while( $len < $targetchars - 20
1051                                        && array_key_exists($add,$all)
1052                                        && !array_key_exists($add,$snippets) ){
1053                                     $offsets[$add] = 0;
1054                                     $tt = "\n".$this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1055                                         $extended[$add] = $tt;
1056                                         $len += strlen( $tt );
1057                                         $add++;
1058                                 }
1059                         }
1060                 }
1061
1062                 //$snippets = array_map('htmlspecialchars', $extended);
1063                 $snippets = $extended;
1064                 $last = -1;
1065                 $extract = '';
1066                 foreach($snippets as $index => $line){
1067                         if($last == -1)
1068                                 $extract .= $line; // first line
1069                         elseif($last+1 == $index && $offsets[$last]+strlen($snippets[$last]) >= strlen($all[$last]))
1070                                 $extract .= " ".$line; // continous lines
1071                         else
1072                                 $extract .= '<b> ... </b>' . $line;
1073
1074                         $last = $index;
1075                 }
1076                 if( $extract )
1077                         $extract .= '<b> ... </b>';
1078
1079                 $processed = array();
1080                 foreach($terms as $term){
1081                         if( ! isset($processed[$term]) ){
1082                                 $pat3 = "/$patPre(".$term.")$patPost/ui"; // highlight word
1083                                 $extract = preg_replace( $pat3,
1084                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1085                                 $processed[$term] = true;
1086                         }
1087                 }
1088
1089                 wfProfileOut( "$fname-extract" );
1090
1091                 return $extract;
1092         }
1093
1094         /**
1095          * Split text into lines and add it to extracts array
1096          *
1097          * @param $extracts Array: index -> $line
1098          * @param $count Integer
1099          * @param $text String
1100          */
1101         function splitAndAdd(&$extracts, &$count, $text){
1102                 $split = explode( "\n", $this->mCleanWikitext? $this->removeWiki($text) : $text );
1103                 foreach($split as $line){
1104                         $tt = trim($line);
1105                         if( $tt )
1106                                 $extracts[$count++] = $tt;
1107                 }
1108         }
1109
1110         /**
1111          * Do manual case conversion for non-ascii chars
1112          *
1113          * @param $matches Array
1114          */
1115         function caseCallback($matches){
1116                 global $wgContLang;
1117                 if( strlen($matches[0]) > 1 ){
1118                         return '['.$wgContLang->lc($matches[0]).$wgContLang->uc($matches[0]).']';
1119                 } else
1120                         return $matches[0];
1121         }
1122
1123         /**
1124          * Extract part of the text from start to end, but by
1125          * not chopping up words
1126          * @param $text String
1127          * @param $start Integer
1128          * @param $end Integer
1129          * @param $posStart Integer: (out) actual start position
1130          * @param $posEnd Integer: (out) actual end position
1131          * @return String
1132          */
1133         function extract($text, $start, $end, &$posStart = null, &$posEnd = null ){
1134                 global $wgContLang;
1135
1136                 if( $start != 0)
1137                         $start = $this->position( $text, $start, 1 );
1138                 if( $end >= strlen($text) )
1139                         $end = strlen($text);
1140                 else
1141                         $end = $this->position( $text, $end );
1142
1143                 if(!is_null($posStart))
1144                         $posStart = $start;
1145                 if(!is_null($posEnd))
1146                         $posEnd = $end;
1147
1148                 if($end > $start)
1149                         return substr($text, $start, $end-$start);
1150                 else
1151                         return '';
1152         }
1153
1154         /**
1155          * Find a nonletter near a point (index) in the text
1156          *
1157          * @param $text String
1158          * @param $point Integer
1159          * @param $offset Integer: offset to found index
1160          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1161          */
1162         function position($text, $point, $offset=0 ){
1163                 $tolerance = 10;
1164                 $s = max( 0, $point - $tolerance );
1165                 $l = min( strlen($text), $point + $tolerance ) - $s;
1166                 $m = array();
1167                 if( preg_match('/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr($text,$s,$l), $m, PREG_OFFSET_CAPTURE ) ){
1168                         return $m[0][1] + $s + $offset;
1169                 } else{
1170                         // check if point is on a valid first UTF8 char
1171                         $char = ord( $text[$point] );
1172                         while( $char >= 0x80 && $char < 0xc0 ) {
1173                                 // skip trailing bytes
1174                                 $point++;
1175                                 if($point >= strlen($text))
1176                                         return strlen($text);
1177                                 $char = ord( $text[$point] );
1178                         }
1179                         return $point;
1180
1181                 }
1182         }
1183
1184         /**
1185          * Search extracts for a pattern, and return snippets
1186          *
1187          * @param $pattern String: regexp for matching lines
1188          * @param $extracts Array: extracts to search
1189          * @param $linesleft Integer: number of extracts to make
1190          * @param $contextchars Integer: length of snippet
1191          * @param $out Array: map for highlighted snippets
1192          * @param $offsets Array: map of starting points of snippets
1193          * @protected
1194          */
1195         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ){
1196                 if($linesleft == 0)
1197                         return; // nothing to do
1198                 foreach($extracts as $index => $line){
1199                         if( array_key_exists($index,$out) )
1200                                 continue; // this line already highlighted
1201
1202                         $m = array();
1203                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1204                                 continue;
1205
1206                         $offset = $m[0][1];
1207                         $len = strlen($m[0][0]);
1208                         if($offset + $len < $contextchars)
1209                                 $begin = 0;
1210                         elseif( $len > $contextchars)
1211                                 $begin = $offset;
1212                         else
1213                                 $begin = $offset + intval( ($len - $contextchars) / 2 );
1214
1215                         $end = $begin + $contextchars;
1216
1217                         $posBegin = $begin;
1218                         // basic snippet from this line
1219                         $out[$index] = $this->extract($line,$begin,$end,$posBegin);
1220                         $offsets[$index] = $posBegin;
1221                         $linesleft--;
1222                         if($linesleft == 0)
1223                                 return;
1224                 }
1225         }
1226
1227         /**
1228          * Basic wikitext removal
1229          * @protected
1230          */
1231         function removeWiki($text) {
1232                 $fname = __METHOD__;
1233                 wfProfileIn( $fname );
1234
1235                 //$text = preg_replace("/'{2,5}/", "", $text);
1236                 //$text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1237                 //$text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1238                 //$text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1239                 //$text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1240                 //$text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1241                 $text = preg_replace("/\\{\\{([^|]+?)\\}\\}/", "", $text);
1242                 $text = preg_replace("/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text);
1243                 $text = preg_replace("/\\[\\[([^|]+?)\\]\\]/", "\\1", $text);
1244                 $text = preg_replace_callback("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array($this,'linkReplace'), $text);
1245                 //$text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1246                 $text = preg_replace("/<\/?[^>]+>/", "", $text);
1247                 $text = preg_replace("/'''''/", "", $text);
1248                 $text = preg_replace("/('''|<\/?[iIuUbB]>)/", "", $text);
1249                 $text = preg_replace("/''/", "", $text);
1250
1251                 wfProfileOut( $fname );
1252                 return $text;
1253         }
1254
1255         /**
1256          * callback to replace [[target|caption]] kind of links, if
1257          * the target is category or image, leave it
1258          *
1259          * @param $matches Array
1260          */
1261         function linkReplace($matches){
1262                 $colon = strpos( $matches[1], ':' );
1263                 if( $colon === false )
1264                         return $matches[2]; // replace with caption
1265                 global $wgContLang;
1266                 $ns = substr( $matches[1], 0, $colon );
1267                 $index = $wgContLang->getNsIndex($ns);
1268                 if( $index !== false && ($index == NS_FILE || $index == NS_CATEGORY) )
1269                         return $matches[0]; // return the whole thing
1270                 else
1271                         return $matches[2];
1272
1273         }
1274
1275         /**
1276      * Simple & fast snippet extraction, but gives completely unrelevant
1277      * snippets
1278      *
1279      * @param $text String
1280      * @param $terms Array
1281      * @param $contextlines Integer
1282      * @param $contextchars Integer
1283      * @return String
1284      */
1285     public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1286         global $wgLang, $wgContLang;
1287         $fname = __METHOD__;
1288
1289         $lines = explode( "\n", $text );
1290
1291         $terms = implode( '|', $terms );
1292         $max = intval( $contextchars ) + 1;
1293         $pat1 = "/(.*)($terms)(.{0,$max})/i";
1294
1295         $lineno = 0;
1296
1297         $extract = "";
1298         wfProfileIn( "$fname-extract" );
1299         foreach ( $lines as $line ) {
1300             if ( 0 == $contextlines ) {
1301                 break;
1302             }
1303             ++$lineno;
1304             $m = array();
1305             if ( ! preg_match( $pat1, $line, $m ) ) {
1306                 continue;
1307             }
1308             --$contextlines;
1309             $pre = $wgContLang->truncate( $m[1], -$contextchars );
1310
1311             if ( count( $m ) < 3 ) {
1312                 $post = '';
1313             } else {
1314                 $post = $wgContLang->truncate( $m[3], $contextchars );
1315             }
1316
1317             $found = $m[2];
1318
1319             $line = htmlspecialchars( $pre . $found . $post );
1320             $pat2 = '/(' . $terms . ")/i";
1321             $line = preg_replace( $pat2,
1322               "<span class='searchmatch'>\\1</span>", $line );
1323
1324             $extract .= "${line}\n";
1325         }
1326         wfProfileOut( "$fname-extract" );
1327
1328         return $extract;
1329     }
1330
1331 }
1332
1333 /**
1334  * Dummy class to be used when non-supported Database engine is present.
1335  * @todo Fixme: dummy class should probably try something at least mildly useful,
1336  * such as a LIKE search through titles.
1337  * @ingroup Search
1338  */
1339 class SearchEngineDummy extends SearchEngine {
1340         // no-op
1341 }