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