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         /**
  26          * @var DatabaseBase
  27          */
  28         protected $db;
  29
  30         function __construct($db = null) {
  31                 if ( $db ) {
  32                         $this->db = $db;
  33                 } else {
  34                         $this->db = wfGetDB( DB_SLAVE );
  35                 }
  36         }
  37
  38         /**
  39          * Perform a full text search query and return a result set.
  40          * If title searches are not supported or disabled, return null.
  41          * STUB
  42          *
  43          * @param $term String: raw search term
  44          * @return SearchResultSet
  45          */
  46         function searchText( $term ) {
  47                 return null;
  48         }
  49
  50         /**
  51          * Perform a title-only search query and return a result set.
  52          * If title searches are not supported or disabled, return null.
  53          * STUB
  54          *
  55          * @param $term String: raw search term
  56          * @return SearchResultSet
  57          */
  58         function searchTitle( $term ) {
  59                 return null;
  60         }
  61
  62         /** If this search backend can list/unlist redirects */
  63         function acceptListRedirects() {
  64                 return true;
  65         }
  66
  67         /**
  68          * When overridden in derived class, performs database-specific conversions
  69          * on text to be used for searching or updating search index.
  70          * Default implementation does nothing (simply returns $string).
  71          *
  72          * @param $string string: String to process
  73          * @return string
  74          */
  75         public function normalizeText( $string ) {
  76                 global $wgContLang;
  77
  78                 // Some languages such as Chinese require word segmentation
  79                 return $wgContLang->segmentByWord( $string );
  80         }
  81
  82         /**
  83          * Transform search term in cases when parts of the query came as different GET params (when supported)
  84          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
  85          */
  86         function transformSearchTerm( $term ) {
  87                 return $term;
  88         }
  89
  90         /**
  91          * If an exact title match can be found, or a very slightly close match,
  92          * return the title. If no match, returns NULL.
  93          *
  94          * @param $searchterm String
  95          * @return Title
  96          */
  97         public static function getNearMatch( $searchterm ) {
  98                 $title = self::getNearMatchInternal( $searchterm );
  99
 100                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
 101                 return $title;
 102         }
 103
 104         /**
 105          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 106          * SearchResultSet.
 107          *
 108          * @param $searchterm string
 109          * @return SearchResultSet
 110          */
 111         public static function getNearMatchResultSet( $searchterm ) {
 112                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 113         }
 114
 115         /**
 116          * Really find the title match.
 117          */
 118         private static function getNearMatchInternal( $searchterm ) {
 119                 global $wgContLang;
 120
 121                 $allSearchTerms = array( $searchterm );
 122
 123                 if ( $wgContLang->hasVariants() ) {
 124                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 125                 }
 126
 127                 $titleResult = null;
 128                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 129                         return $titleResult;
 130                 }
 131
 132                 foreach ( $allSearchTerms as $term ) {
 133
 134                         # Exact match? No need to look further.
 135                         $title = Title::newFromText( $term );
 136                         if ( is_null( $title ) )
 137                                 return null;
 138
 139                         if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal() || $title->exists() ) {
 140                                 return $title;
 141                         }
 142
 143                         # See if it still otherwise has content is some sane sense
 144                         $article = MediaWiki::articleFromTitle( $title );
 145                         if ( $article->hasViewableContent() ) {
 146                                 return $title;
 147                         }
 148
 149                         # Now try all lower case (i.e. first letter capitalized)
 150                         #
 151                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 152                         if ( $title && $title->exists() ) {
 153                                 return $title;
 154                         }
 155
 156                         # Now try capitalized string
 157                         #
 158                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 159                         if ( $title && $title->exists() ) {
 160                                 return $title;
 161                         }
 162
 163                         # Now try all upper case
 164                         #
 165                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 166                         if ( $title && $title->exists() ) {
 167                                 return $title;
 168                         }
 169
 170                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 171                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 172                         if ( $title && $title->exists() ) {
 173                                 return $title;
 174                         }
 175
 176                         // Give hooks a chance at better match variants
 177                         $title = null;
 178                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 179                                 return $title;
 180                         }
 181                 }
 182
 183                 $title = Title::newFromText( $searchterm );
 184
 185                 # Entering an IP address goes to the contributions page
 186                 if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 187                         || User::isIP( trim( $searchterm ) ) ) {
 188                         return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 189                 }
 190
 191
 192                 # Entering a user goes to the user page whether it's there or not
 193                 if ( $title->getNamespace() == NS_USER ) {
 194                         return $title;
 195                 }
 196
 197                 # Go to images that exist even if there's no local page.
 198                 # There may have been a funny upload, or it may be on a shared
 199                 # file repository such as Wikimedia Commons.
 200                 if ( $title->getNamespace() == NS_FILE ) {
 201                         $image = wfFindFile( $title );
 202                         if ( $image ) {
 203                                 return $title;
 204                         }
 205                 }
 206
 207                 # MediaWiki namespace? Page may be "implied" if not customized.
 208                 # Just return it, with caps forced as the message system likes it.
 209                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 210                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 211                 }
 212
 213                 # Quoted term? Try without the quotes...
 214                 $matches = array();
 215                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 216                         return SearchEngine::getNearMatch( $matches[1] );
 217                 }
 218
 219                 return null;
 220         }
 221
 222         public static function legalSearchChars() {
 223                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 224         }
 225
 226         /**
 227          * Set the maximum number of results to return
 228          * and how many to skip before returning the first.
 229          *
 230          * @param $limit Integer
 231          * @param $offset Integer
 232          */
 233         function setLimitOffset( $limit, $offset = 0 ) {
 234                 $this->limit = intval( $limit );
 235                 $this->offset = intval( $offset );
 236         }
 237
 238         /**
 239          * Set which namespaces the search should include.
 240          * Give an array of namespace index numbers.
 241          *
 242          * @param $namespaces Array
 243          */
 244         function setNamespaces( $namespaces ) {
 245                 $this->namespaces = $namespaces;
 246         }
 247
 248         /**
 249          * Parse some common prefixes: all (search everything)
 250          * or namespace names
 251          *
 252          * @param $query String
 253          */
 254         function replacePrefixes( $query ) {
 255                 global $wgContLang;
 256
 257                 $parsed = $query;
 258                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 259                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 260                         return $parsed;
 261                 }
 262
 263                 $allkeyword = wfMsgForContent( 'searchall' ) . ":";
 264                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 265                         $this->namespaces = null;
 266                         $parsed = substr( $query, strlen( $allkeyword ) );
 267                 } else if ( strpos( $query, ':' ) !== false ) {
 268                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 269                         $index = $wgContLang->getNsIndex( $prefix );
 270                         if ( $index !== false ) {
 271                                 $this->namespaces = array( $index );
 272                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 273                         }
 274                 }
 275                 if ( trim( $parsed ) == '' )
 276                         $parsed = $query; // prefix was the whole query
 277
 278                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 279
 280                 return $parsed;
 281         }
 282
 283         /**
 284          * Make a list of searchable namespaces and their canonical names.
 285          * @return Array
 286          */
 287         public static function searchableNamespaces() {
 288                 global $wgContLang;
 289                 $arr = array();
 290                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 291                         if ( $ns >= NS_MAIN ) {
 292                                 $arr[$ns] = $name;
 293                         }
 294                 }
 295
 296                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 297                 return $arr;
 298         }
 299
 300         /**
 301          * Extract default namespaces to search from the given user's
 302          * settings, returning a list of index numbers.
 303          *
 304          * @param $user User
 305          * @return Array
 306          */
 307         public static function userNamespaces( $user ) {
 308                 global $wgSearchEverythingOnlyLoggedIn;
 309
 310                 // get search everything preference, that can be set to be read for logged-in users
 311                 $searcheverything = false;
 312                 if ( ( $wgSearchEverythingOnlyLoggedIn && $user->isLoggedIn() )
 313                     || !$wgSearchEverythingOnlyLoggedIn )
 314                         $searcheverything = $user->getOption( 'searcheverything' );
 315
 316                 // searcheverything overrides other options
 317                 if ( $searcheverything )
 318                         return array_keys( SearchEngine::searchableNamespaces() );
 319
 320                 $arr = Preferences::loadOldSearchNs( $user );
 321                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 322
 323                 $arr = array_intersect( $arr, array_keys( $searchableNamespaces ) ); // Filter
 324
 325                 return $arr;
 326         }
 327
 328         /**
 329          * Find snippet highlight settings for a given user
 330          *
 331          * @param $user User
 332          * @return Array contextlines, contextchars
 333          */
 334         public static function userHighlightPrefs( &$user ) {
 335                 // $contextlines = $user->getOption( 'contextlines',  5 );
 336                 // $contextchars = $user->getOption( 'contextchars', 50 );
 337                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 338                 $contextchars = 75; // same as above.... :P
 339                 return array( $contextlines, $contextchars );
 340         }
 341
 342         /**
 343          * An array of namespaces indexes to be searched by default
 344          *
 345          * @return Array
 346          */
 347         public static function defaultNamespaces() {
 348                 global $wgNamespacesToBeSearchedDefault;
 349
 350                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 351         }
 352
 353         /**
 354          * Get a list of namespace names useful for showing in tooltips
 355          * and preferences
 356          *
 357          * @param $namespaces Array
 358          */
 359         public static function namespacesAsText( $namespaces ) {
 360                 global $wgContLang;
 361
 362                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 363                 foreach ( $formatted as $key => $ns ) {
 364                         if ( empty( $ns ) )
 365                                 $formatted[$key] = wfMsg( 'blanknamespace' );
 366                 }
 367                 return $formatted;
 368         }
 369
 370         /**
 371          * Return the help namespaces to be shown on Special:Search
 372          *
 373          * @return Array
 374          */
 375         public static function helpNamespaces() {
 376                 global $wgNamespacesToBeSearchedHelp;
 377
 378                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 379         }
 380
 381         /**
 382          * Return a 'cleaned up' search string
 383          *
 384          * @param $text String
 385          * @return String
 386          */
 387         function filter( $text ) {
 388                 $lc = $this->legalSearchChars();
 389                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 390         }
 391         /**
 392          * Load up the appropriate search engine class for the currently
 393          * active database backend, and return a configured instance.
 394          *
 395          * @return SearchEngine
 396          */
 397         public static function create() {
 398                 global $wgSearchType;
 399                 $dbr = null;
 400                 if ( $wgSearchType ) {
 401                         $class = $wgSearchType;
 402                 } else {
 403                         $dbr = wfGetDB( DB_SLAVE );
 404                         $class = $dbr->getSearchEngine();
 405                 }
 406                 $search = new $class( $dbr );
 407                 $search->setLimitOffset( 0, 0 );
 408                 return $search;
 409         }
 410
 411         /**
 412          * Create or update the search index record for the given page.
 413          * Title and text should be pre-processed.
 414          * STUB
 415          *
 416          * @param $id Integer
 417          * @param $title String
 418          * @param $text String
 419          */
 420         function update( $id, $title, $text ) {
 421                 // no-op
 422         }
 423
 424         /**
 425          * Update a search index record's title only.
 426          * Title should be pre-processed.
 427          * STUB
 428          *
 429          * @param $id Integer
 430          * @param $title String
 431          */
 432         function updateTitle( $id, $title ) {
 433                 // no-op
 434         }
 435
 436         /**
 437          * Get OpenSearch suggestion template
 438          *
 439          * @return String
 440          */
 441         public static function getOpenSearchTemplate() {
 442                 global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
 443                 if ( $wgOpenSearchTemplate )    {
 444                         return $wgOpenSearchTemplate;
 445                 } else {
 446                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 447                         if ( !$ns ) $ns = "0";
 448                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace=' . $ns;
 449                 }
 450         }
 451
 452         /**
 453          * Get internal MediaWiki Suggest template
 454          *
 455          * @return String
 456          */
 457         public static function getMWSuggestTemplate() {
 458                 global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
 459                 if ( $wgMWSuggestTemplate )
 460                         return $wgMWSuggestTemplate;
 461                 else
 462                         return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
 463         }
 464 }
 465
 466 /**
 467  * @ingroup Search
 468  */
 469 class SearchResultSet {
 470         /**
 471          * Fetch an array of regular expression fragments for matching
 472          * the search terms as parsed by this engine in a text extract.
 473          * STUB
 474          *
 475          * @return Array
 476          */
 477         function termMatches() {
 478                 return array();
 479         }
 480
 481         function numRows() {
 482                 return 0;
 483         }
 484
 485         /**
 486          * Return true if results are included in this result set.
 487          * STUB
 488          *
 489          * @return Boolean
 490          */
 491         function hasResults() {
 492                 return false;
 493         }
 494
 495         /**
 496          * Some search modes return a total hit count for the query
 497          * in the entire article database. This may include pages
 498          * in namespaces that would not be matched on the given
 499          * settings.
 500          *
 501          * Return null if no total hits number is supported.
 502          *
 503          * @return Integer
 504          */
 505         function getTotalHits() {
 506                 return null;
 507         }
 508
 509         /**
 510          * Some search modes return a suggested alternate term if there are
 511          * no exact hits. Returns true if there is one on this set.
 512          *
 513          * @return Boolean
 514          */
 515         function hasSuggestion() {
 516                 return false;
 517         }
 518
 519         /**
 520          * @return String: suggested query, null if none
 521          */
 522         function getSuggestionQuery() {
 523                 return null;
 524         }
 525
 526         /**
 527          * @return String: HTML highlighted suggested query, '' if none
 528          */
 529         function getSuggestionSnippet() {
 530                 return '';
 531         }
 532
 533         /**
 534          * Return information about how and from where the results were fetched,
 535          * should be useful for diagnostics and debugging
 536          *
 537          * @return String
 538          */
 539         function getInfo() {
 540                 return null;
 541         }
 542
 543         /**
 544          * Return a result set of hits on other (multiple) wikis associated with this one
 545          *
 546          * @return SearchResultSet
 547          */
 548         function getInterwikiResults() {
 549                 return null;
 550         }
 551
 552         /**
 553          * Check if there are results on other wikis
 554          *
 555          * @return Boolean
 556          */
 557         function hasInterwikiResults() {
 558                 return $this->getInterwikiResults() != null;
 559         }
 560
 561         /**
 562          * Fetches next search result, or false.
 563          * STUB
 564          *
 565          * @return SearchResult
 566          */
 567         function next() {
 568                 return false;
 569         }
 570
 571         /**
 572          * Frees the result set, if applicable.
 573          */
 574         function free() {
 575                 // ...
 576         }
 577 }
 578
 579 /**
 580  * This class is used for different SQL-based search engines shipped with MediaWiki
 581  */
 582 class SqlSearchResultSet extends SearchResultSet {
 583         function __construct( $resultSet, $terms ) {
 584                 $this->mResultSet = $resultSet;
 585                 $this->mTerms = $terms;
 586         }
 587
 588         function termMatches() {
 589                 return $this->mTerms;
 590         }
 591
 592         function numRows() {
 593                 if ( $this->mResultSet === false )
 594                         return false;
 595
 596                 return $this->mResultSet->numRows();
 597         }
 598
 599         function next() {
 600                 if ( $this->mResultSet === false )
 601                         return false;
 602
 603                 $row = $this->mResultSet->fetchObject();
 604                 if ( $row === false )
 605                         return false;
 606
 607                 return SearchResult::newFromRow( $row );
 608         }
 609
 610         function free() {
 611                 if ( $this->mResultSet === false )
 612                         return false;
 613
 614                 $this->mResultSet->free();
 615         }
 616 }
 617
 618 /**
 619  * @ingroup Search
 620  */
 621 class SearchResultTooMany {
 622         # # Some search engines may bail out if too many matches are found
 623 }
 624
 625
 626 /**
 627  * @todo Fixme: This class is horribly factored. It would probably be better to
 628  * have a useful base class to which you pass some standard information, then
 629  * let the fancy self-highlighters extend that.
 630  * @ingroup Search
 631  */
 632 class SearchResult {
 633         var $mRevision = null;
 634         var $mImage = null;
 635
 636         /**
 637          * Return a new SearchResult and initializes it with a title.
 638          *
 639          * @param $title Title
 640          * @return SearchResult
 641          */
 642         public static function newFromTitle( $title ) {
 643                 $result = new self();
 644                 $result->initFromTitle( $title );
 645                 return $result;
 646         }
 647         /**
 648          * Return a new SearchResult and initializes it with a row.
 649          *
 650          * @param $row object
 651          * @return SearchResult
 652          */
 653         public static function newFromRow( $row ) {
 654                 $result = new self();
 655                 $result->initFromRow( $row );
 656                 return $result;
 657         }
 658
 659         public function __construct( $row = null ) {
 660                 if ( !is_null( $row ) ) {
 661                         // Backwards compatibility with pre-1.17 callers
 662                         $this->initFromRow( $row );
 663                 }
 664         }
 665
 666         /**
 667          * Initialize from a database row. Makes a Title and passes that to
 668          * initFromTitle.
 669          *
 670          * @param $row object
 671          */
 672         protected function initFromRow( $row ) {
 673                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 674         }
 675
 676         /**
 677          * Initialize from a Title and if possible initializes a corresponding
 678          * Revision and File.
 679          *
 680          * @param $title Title
 681          */
 682         protected function initFromTitle( $title ) {
 683                 $this->mTitle = $title;
 684                 if ( !is_null( $this->mTitle ) ) {
 685                         $this->mRevision = Revision::newFromTitle( $this->mTitle );
 686                         if ( $this->mTitle->getNamespace() === NS_FILE )
 687                                 $this->mImage = wfFindFile( $this->mTitle );
 688                 }
 689         }
 690
 691         /**
 692          * Check if this is result points to an invalid title
 693          *
 694          * @return Boolean
 695          */
 696         function isBrokenTitle() {
 697                 if ( is_null( $this->mTitle ) )
 698                         return true;
 699                 return false;
 700         }
 701
 702         /**
 703          * Check if target page is missing, happens when index is out of date
 704          *
 705          * @return Boolean
 706          */
 707         function isMissingRevision() {
 708                 return !$this->mRevision && !$this->mImage;
 709         }
 710
 711         /**
 712          * @return Title
 713          */
 714         function getTitle() {
 715                 return $this->mTitle;
 716         }
 717
 718         /**
 719          * @return Double or null if not supported
 720          */
 721         function getScore() {
 722                 return null;
 723         }
 724
 725         /**
 726          * Lazy initialization of article text from DB
 727          */
 728         protected function initText() {
 729                 if ( !isset( $this->mText ) ) {
 730                         if ( $this->mRevision != null )
 731                                 $this->mText = $this->mRevision->getText();
 732                         else // TODO: can we fetch raw wikitext for commons images?
 733                                 $this->mText = '';
 734
 735                 }
 736         }
 737
 738         /**
 739          * @param $terms Array: terms to highlight
 740          * @return String: highlighted text snippet, null (and not '') if not supported
 741          */
 742         function getTextSnippet( $terms ) {
 743                 global $wgUser, $wgAdvancedSearchHighlighting;
 744                 $this->initText();
 745                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
 746                 $h = new SearchHighlighter();
 747                 if ( $wgAdvancedSearchHighlighting )
 748                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 749                 else
 750                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 751         }
 752
 753         /**
 754          * @param $terms Array: terms to highlight
 755          * @return String: highlighted title, '' if not supported
 756          */
 757         function getTitleSnippet( $terms ) {
 758                 return '';
 759         }
 760
 761         /**
 762          * @param $terms Array: terms to highlight
 763          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 764          */
 765         function getRedirectSnippet( $terms ) {
 766                 return '';
 767         }
 768
 769         /**
 770          * @return Title object for the redirect to this page, null if none or not supported
 771          */
 772         function getRedirectTitle() {
 773                 return null;
 774         }
 775
 776         /**
 777          * @return string highlighted relevant section name, null if none or not supported
 778          */
 779         function getSectionSnippet() {
 780                 return '';
 781         }
 782
 783         /**
 784          * @return Title object (pagename+fragment) for the section, null if none or not supported
 785          */
 786         function getSectionTitle() {
 787                 return null;
 788         }
 789
 790         /**
 791          * @return String: timestamp
 792          */
 793         function getTimestamp() {
 794                 if ( $this->mRevision )
 795                         return $this->mRevision->getTimestamp();
 796                 else if ( $this->mImage )
 797                         return $this->mImage->getTimestamp();
 798                 return '';
 799         }
 800
 801         /**
 802          * @return Integer: number of words
 803          */
 804         function getWordCount() {
 805                 $this->initText();
 806                 return str_word_count( $this->mText );
 807         }
 808
 809         /**
 810          * @return Integer: size in bytes
 811          */
 812         function getByteSize() {
 813                 $this->initText();
 814                 return strlen( $this->mText );
 815         }
 816
 817         /**
 818          * @return Boolean if hit has related articles
 819          */
 820         function hasRelated() {
 821                 return false;
 822         }
 823
 824         /**
 825          * @return String: interwiki prefix of the title (return iw even if title is broken)
 826          */
 827         function getInterwikiPrefix() {
 828                 return '';
 829         }
 830 }
 831 /**
 832  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 833  */
 834 class SearchNearMatchResultSet extends SearchResultSet {
 835         private $fetched = false;
 836         /**
 837          * @param $match mixed Title if matched, else null
 838          */
 839         public function __construct( $match ) {
 840                 $this->result = $match;
 841         }
 842         public function hasResult() {
 843                 return (bool)$this->result;
 844         }
 845         public function numRows() {
 846                 return $this->hasResults() ? 1 : 0;
 847         }
 848         public function next() {
 849                 if ( $this->fetched || !$this->result ) {
 850                         return false;
 851                 }
 852                 $this->fetched = true;
 853                 return SearchResult::newFromTitle( $this->result );
 854         }
 855 }
 856
 857 /**
 858  * Highlight bits of wikitext
 859  *
 860  * @ingroup Search
 861  */
 862 class SearchHighlighter {
 863         var $mCleanWikitext = true;
 864
 865         function __construct( $cleanupWikitext = true ) {
 866                 $this->mCleanWikitext = $cleanupWikitext;
 867         }
 868
 869         /**
 870          * Default implementation of wikitext highlighting
 871          *
 872          * @param $text String
 873          * @param $terms Array: terms to highlight (unescaped)
 874          * @param $contextlines Integer
 875          * @param $contextchars Integer
 876          * @return String
 877          */
 878         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 879                 global $wgContLang;
 880                 global $wgSearchHighlightBoundaries;
 881                 $fname = __METHOD__;
 882
 883                 if ( $text == '' )
 884                         return '';
 885
 886                 // spli text into text + templates/links/tables
 887                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 888                 // first capture group is for detecting nested templates/links/tables/references
 889                 $endPatterns = array(
 890                         1 => '/(\{\{)|(\}\})/', // template
 891                         2 => '/(\[\[)|(\]\])/', // image
 892                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
 893
 894                 // FIXME: this should prolly be a hook or something
 895                 if ( function_exists( 'wfCite' ) ) {
 896                         $spat .= '|(<ref>)'; // references via cite extension
 897                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 898                 }
 899                 $spat .= '/';
 900                 $textExt = array(); // text extracts
 901                 $otherExt = array();  // other extracts
 902                 wfProfileIn( "$fname-split" );
 903                 $start = 0;
 904                 $textLen = strlen( $text );
 905                 $count = 0; // sequence number to maintain ordering
 906                 while ( $start < $textLen ) {
 907                         // find start of template/image/table
 908                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
 909                                 $epat = '';
 910                                 foreach ( $matches as $key => $val ) {
 911                                         if ( $key > 0 && $val[1] != - 1 ) {
 912                                                 if ( $key == 2 ) {
 913                                                         // see if this is an image link
 914                                                         $ns = substr( $val[0], 2, - 1 );
 915                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE )
 916                                                                 break;
 917
 918                                                 }
 919                                                 $epat = $endPatterns[$key];
 920                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
 921                                                 $start = $val[1];
 922                                                 break;
 923                                         }
 924                                 }
 925                                 if ( $epat ) {
 926                                         // find end (and detect any nested elements)
 927                                         $level = 0;
 928                                         $offset = $start + 1;
 929                                         $found = false;
 930                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
 931                                                 if ( array_key_exists( 2, $endMatches ) ) {
 932                                                         // found end
 933                                                         if ( $level == 0 ) {
 934                                                                 $len = strlen( $endMatches[2][0] );
 935                                                                 $off = $endMatches[2][1];
 936                                                                 $this->splitAndAdd( $otherExt, $count,
 937                                                                         substr( $text, $start, $off + $len  - $start ) );
 938                                                                 $start = $off + $len;
 939                                                                 $found = true;
 940                                                                 break;
 941                                                         } else {
 942                                                                 // end of nested element
 943                                                                 $level -= 1;
 944                                                         }
 945                                                 } else {
 946                                                         // nested
 947                                                         $level += 1;
 948                                                 }
 949                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
 950                                         }
 951                                         if ( ! $found ) {
 952                                                 // couldn't find appropriate closing tag, skip
 953                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
 954                                                 $start += strlen( $matches[0][0] );
 955                                         }
 956                                         continue;
 957                                 }
 958                         }
 959                         // else: add as text extract
 960                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
 961                         break;
 962                 }
 963
 964                 $all = $textExt + $otherExt; // these have disjunct key sets
 965
 966                 wfProfileOut( "$fname-split" );
 967
 968                 // prepare regexps
 969                 foreach ( $terms as $index => $term ) {
 970                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 971                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
 972                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
 973                         } else {
 974                                 $terms[$index] = $term;
 975                         }
 976                 }
 977                 $anyterm = implode( '|', $terms );
 978                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
 979
 980                 // FIXME: a hack to scale contextchars, a correct solution
 981                 // would be to have contextchars actually be char and not byte
 982                 // length, and do proper utf-8 substrings and lengths everywhere,
 983                 // but PHP is making that very hard and unclean to implement :(
 984                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
 985                 $contextchars = intval( $contextchars * $scale );
 986
 987                 $patPre = "(^|$wgSearchHighlightBoundaries)";
 988                 $patPost = "($wgSearchHighlightBoundaries|$)";
 989
 990                 $pat1 = "/(" . $phrase . ")/ui";
 991                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
 992
 993                 wfProfileIn( "$fname-extract" );
 994
 995                 $left = $contextlines;
 996
 997                 $snippets = array();
 998                 $offsets = array();
 999
1000                 // show beginning only if it contains all words
1001                 $first = 0;
1002                 $firstText = '';
1003                 foreach ( $textExt as $index => $line ) {
1004                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1005                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1006                                 $first = $index;
1007                                 break;
1008                         }
1009                 }
1010                 if ( $firstText ) {
1011                         $succ = true;
1012                         // check if first text contains all terms
1013                         foreach ( $terms as $term ) {
1014                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1015                                         $succ = false;
1016                                         break;
1017                                 }
1018                         }
1019                         if ( $succ ) {
1020                                 $snippets[$first] = $firstText;
1021                                 $offsets[$first] = 0;
1022                         }
1023                 }
1024                 if ( ! $snippets ) {
1025                         // match whole query on text
1026                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1027                         // match whole query on templates/tables/images
1028                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1029                         // match any words on text
1030                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1031                         // match any words on templates/tables/images
1032                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1033
1034                         ksort( $snippets );
1035                 }
1036
1037                 // add extra chars to each snippet to make snippets constant size
1038                 $extended = array();
1039                 if ( count( $snippets ) == 0 ) {
1040                         // couldn't find the target words, just show beginning of article
1041                         if ( array_key_exists( $first, $all ) ) {
1042                                 $targetchars = $contextchars * $contextlines;
1043                                 $snippets[$first] = '';
1044                                 $offsets[$first] = 0;
1045                         }
1046                 } else {
1047                         // if begin of the article contains the whole phrase, show only that !!
1048                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1049                             && $offsets[$first] < $contextchars * 2 ) {
1050                                 $snippets = array ( $first => $snippets[$first] );
1051                         }
1052
1053                         // calc by how much to extend existing snippets
1054                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1055                 }
1056
1057                 foreach ( $snippets as $index => $line ) {
1058                         $extended[$index] = $line;
1059                         $len = strlen( $line );
1060                         if ( $len < $targetchars - 20 ) {
1061                                 // complete this line
1062                                 if ( $len < strlen( $all[$index] ) ) {
1063                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1064                                         $len = strlen( $extended[$index] );
1065                                 }
1066
1067                                 // add more lines
1068                                 $add = $index + 1;
1069                                 while ( $len < $targetchars - 20
1070                                        && array_key_exists( $add, $all )
1071                                        && !array_key_exists( $add, $snippets ) ) {
1072                                     $offsets[$add] = 0;
1073                                     $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1074                                         $extended[$add] = $tt;
1075                                         $len += strlen( $tt );
1076                                         $add++;
1077                                 }
1078                         }
1079                 }
1080
1081                 // $snippets = array_map('htmlspecialchars', $extended);
1082                 $snippets = $extended;
1083                 $last = - 1;
1084                 $extract = '';
1085                 foreach ( $snippets as $index => $line ) {
1086                         if ( $last == - 1 )
1087                                 $extract .= $line; // first line
1088                         elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) )
1089                                 $extract .= " " . $line; // continous lines
1090                         else
1091                                 $extract .= '<b> ... </b>' . $line;
1092
1093                         $last = $index;
1094                 }
1095                 if ( $extract )
1096                         $extract .= '<b> ... </b>';
1097
1098                 $processed = array();
1099                 foreach ( $terms as $term ) {
1100                         if ( ! isset( $processed[$term] ) ) {
1101                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1102                                 $extract = preg_replace( $pat3,
1103                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1104                                 $processed[$term] = true;
1105                         }
1106                 }
1107
1108                 wfProfileOut( "$fname-extract" );
1109
1110                 return $extract;
1111         }
1112
1113         /**
1114          * Split text into lines and add it to extracts array
1115          *
1116          * @param $extracts Array: index -> $line
1117          * @param $count Integer
1118          * @param $text String
1119          */
1120         function splitAndAdd( &$extracts, &$count, $text ) {
1121                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1122                 foreach ( $split as $line ) {
1123                         $tt = trim( $line );
1124                         if ( $tt )
1125                                 $extracts[$count++] = $tt;
1126                 }
1127         }
1128
1129         /**
1130          * Do manual case conversion for non-ascii chars
1131          *
1132          * @param $matches Array
1133          */
1134         function caseCallback( $matches ) {
1135                 global $wgContLang;
1136                 if ( strlen( $matches[0] ) > 1 ) {
1137                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1138                 } else
1139                         return $matches[0];
1140         }
1141
1142         /**
1143          * Extract part of the text from start to end, but by
1144          * not chopping up words
1145          * @param $text String
1146          * @param $start Integer
1147          * @param $end Integer
1148          * @param $posStart Integer: (out) actual start position
1149          * @param $posEnd Integer: (out) actual end position
1150          * @return String
1151          */
1152         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1153                 if ( $start != 0 )
1154                         $start = $this->position( $text, $start, 1 );
1155                 if ( $end >= strlen( $text ) )
1156                         $end = strlen( $text );
1157                 else
1158                         $end = $this->position( $text, $end );
1159
1160                 if ( !is_null( $posStart ) )
1161                         $posStart = $start;
1162                 if ( !is_null( $posEnd ) )
1163                         $posEnd = $end;
1164
1165                 if ( $end > $start )
1166                         return substr( $text, $start, $end - $start );
1167                 else
1168                         return '';
1169         }
1170
1171         /**
1172          * Find a nonletter near a point (index) in the text
1173          *
1174          * @param $text String
1175          * @param $point Integer
1176          * @param $offset Integer: offset to found index
1177          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1178          */
1179         function position( $text, $point, $offset = 0 ) {
1180                 $tolerance = 10;
1181                 $s = max( 0, $point - $tolerance );
1182                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1183                 $m = array();
1184                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1185                         return $m[0][1] + $s + $offset;
1186                 } else {
1187                         // check if point is on a valid first UTF8 char
1188                         $char = ord( $text[$point] );
1189                         while ( $char >= 0x80 && $char < 0xc0 ) {
1190                                 // skip trailing bytes
1191                                 $point++;
1192                                 if ( $point >= strlen( $text ) )
1193                                         return strlen( $text );
1194                                 $char = ord( $text[$point] );
1195                         }
1196                         return $point;
1197
1198                 }
1199         }
1200
1201         /**
1202          * Search extracts for a pattern, and return snippets
1203          *
1204          * @param $pattern String: regexp for matching lines
1205          * @param $extracts Array: extracts to search
1206          * @param $linesleft Integer: number of extracts to make
1207          * @param $contextchars Integer: length of snippet
1208          * @param $out Array: map for highlighted snippets
1209          * @param $offsets Array: map of starting points of snippets
1210          * @protected
1211          */
1212         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1213                 if ( $linesleft == 0 )
1214                         return; // nothing to do
1215                 foreach ( $extracts as $index => $line ) {
1216                         if ( array_key_exists( $index, $out ) )
1217                                 continue; // this line already highlighted
1218
1219                         $m = array();
1220                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
1221                                 continue;
1222
1223                         $offset = $m[0][1];
1224                         $len = strlen( $m[0][0] );
1225                         if ( $offset + $len < $contextchars )
1226                                 $begin = 0;
1227                         elseif ( $len > $contextchars )
1228                                 $begin = $offset;
1229                         else
1230                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1231
1232                         $end = $begin + $contextchars;
1233
1234                         $posBegin = $begin;
1235                         // basic snippet from this line
1236                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1237                         $offsets[$index] = $posBegin;
1238                         $linesleft--;
1239                         if ( $linesleft == 0 )
1240                                 return;
1241                 }
1242         }
1243
1244         /**
1245          * Basic wikitext removal
1246          * @protected
1247          */
1248         function removeWiki( $text ) {
1249                 $fname = __METHOD__;
1250                 wfProfileIn( $fname );
1251
1252                 // $text = preg_replace("/'{2,5}/", "", $text);
1253                 // $text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
1254                 // $text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
1255                 // $text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
1256                 // $text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
1257                 // $text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
1258                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1259                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1260                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1261                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1262                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1263                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1264                 $text = preg_replace( "/'''''/", "", $text );
1265                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1266                 $text = preg_replace( "/''/", "", $text );
1267
1268                 wfProfileOut( $fname );
1269                 return $text;
1270         }
1271
1272         /**
1273          * callback to replace [[target|caption]] kind of links, if
1274          * the target is category or image, leave it
1275          *
1276          * @param $matches Array
1277          */
1278         function linkReplace( $matches ) {
1279                 $colon = strpos( $matches[1], ':' );
1280                 if ( $colon === false )
1281                         return $matches[2]; // replace with caption
1282                 global $wgContLang;
1283                 $ns = substr( $matches[1], 0, $colon );
1284                 $index = $wgContLang->getNsIndex( $ns );
1285                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) )
1286                         return $matches[0]; // return the whole thing
1287                 else
1288                         return $matches[2];
1289
1290         }
1291
1292         /**
1293      * Simple & fast snippet extraction, but gives completely unrelevant
1294      * snippets
1295      *
1296      * @param $text String
1297      * @param $terms Array
1298      * @param $contextlines Integer
1299      * @param $contextchars Integer
1300      * @return String
1301      */
1302     public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1303         global $wgContLang;
1304         $fname = __METHOD__;
1305
1306         $lines = explode( "\n", $text );
1307
1308         $terms = implode( '|', $terms );
1309         $max = intval( $contextchars ) + 1;
1310         $pat1 = "/(.*)($terms)(.{0,$max})/i";
1311
1312         $lineno = 0;
1313
1314         $extract = "";
1315         wfProfileIn( "$fname-extract" );
1316         foreach ( $lines as $line ) {
1317             if ( 0 == $contextlines ) {
1318                 break;
1319             }
1320             ++$lineno;
1321             $m = array();
1322             if ( ! preg_match( $pat1, $line, $m ) ) {
1323                 continue;
1324             }
1325             --$contextlines;
1326             // truncate function changes ... to relevant i18n message.
1327             $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1328
1329             if ( count( $m ) < 3 ) {
1330                 $post = '';
1331             } else {
1332                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1333             }
1334
1335             $found = $m[2];
1336
1337             $line = htmlspecialchars( $pre . $found . $post );
1338             $pat2 = '/(' . $terms . ")/i";
1339             $line = preg_replace( $pat2,
1340               "<span class='searchmatch'>\\1</span>", $line );
1341
1342             $extract .= "${line}\n";
1343         }
1344         wfProfileOut( "$fname-extract" );
1345
1346         return $extract;
1347     }
1348
1349 }
1350
1351 /**
1352  * Dummy class to be used when non-supported Database engine is present.
1353  * @todo Fixme: dummy class should probably try something at least mildly useful,
1354  * such as a LIKE search through titles.
1355  * @ingroup Search
1356  */
1357 class SearchEngineDummy extends SearchEngine {
1358         // no-op
1359 }