includes/search/SearchEngine.php

   1 <?php
   2 /**
   3  * Basic search engine
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup Search
  22  */
  23
  24 /**
  25  * @defgroup Search Search
  26  */
  27
  28 /**
  29  * Contain a class for special pages
  30  * @ingroup Search
  31  */
  32 class SearchEngine {
  33         var $limit = 10;
  34         var $offset = 0;
  35         var $prefix = '';
  36         var $searchTerms = array();
  37         var $namespaces = array( NS_MAIN );
  38         var $showRedirects = false;
  39
  40         /// Feature values
  41         protected $features = array();
  42
  43         /**
  44          * @var DatabaseBase
  45          */
  46         protected $db;
  47
  48         function __construct( $db = null ) {
  49                 if ( $db ) {
  50                         $this->db = $db;
  51                 } else {
  52                         $this->db = wfGetDB( DB_SLAVE );
  53                 }
  54         }
  55
  56         /**
  57          * Perform a full text search query and return a result set.
  58          * If title searches are not supported or disabled, return null.
  59          * STUB
  60          *
  61          * @param string $term raw search term
  62          * @return SearchResultSet|Status|null
  63          */
  64         function searchText( $term ) {
  65                 return null;
  66         }
  67
  68         /**
  69          * Perform a title-only search query and return a result set.
  70          * If title searches are not supported or disabled, return null.
  71          * STUB
  72          *
  73          * @param string $term raw search term
  74          * @return SearchResultSet|null
  75          */
  76         function searchTitle( $term ) {
  77                 return null;
  78         }
  79
  80         /**
  81          * If this search backend can list/unlist redirects
  82          * @deprecated since 1.18 Call supports( 'list-redirects' );
  83          * @return bool
  84          */
  85         function acceptListRedirects() {
  86                 wfDeprecated( __METHOD__, '1.18' );
  87                 return $this->supports( 'list-redirects' );
  88         }
  89
  90         /**
  91          * @since 1.18
  92          * @param $feature String
  93          * @return Boolean
  94          */
  95         public function supports( $feature ) {
  96                 switch ( $feature ) {
  97                 case 'list-redirects':
  98                         return true;
  99                 case 'title-suffix-filter':
 100                 default:
 101                         return false;
 102                 }
 103         }
 104
 105         /**
 106          * Way to pass custom data for engines
 107          * @since 1.18
 108          * @param $feature String
 109          * @param $data Mixed
 110          * @return bool
 111          */
 112         public function setFeatureData( $feature, $data ) {
 113                 $this->features[$feature] = $data;
 114         }
 115
 116         /**
 117          * When overridden in derived class, performs database-specific conversions
 118          * on text to be used for searching or updating search index.
 119          * Default implementation does nothing (simply returns $string).
 120          *
 121          * @param string $string String to process
 122          * @return string
 123          */
 124         public function normalizeText( $string ) {
 125                 global $wgContLang;
 126
 127                 // Some languages such as Chinese require word segmentation
 128                 return $wgContLang->segmentByWord( $string );
 129         }
 130
 131         /**
 132          * Transform search term in cases when parts of the query came as different GET params (when supported)
 133          * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
 134          */
 135         function transformSearchTerm( $term ) {
 136                 return $term;
 137         }
 138
 139         /**
 140          * If an exact title match can be found, or a very slightly close match,
 141          * return the title. If no match, returns NULL.
 142          *
 143          * @param $searchterm String
 144          * @return Title
 145          */
 146         public static function getNearMatch( $searchterm ) {
 147                 $title = self::getNearMatchInternal( $searchterm );
 148
 149                 wfRunHooks( 'SearchGetNearMatchComplete', array( $searchterm, &$title ) );
 150                 return $title;
 151         }
 152
 153         /**
 154          * Do a near match (see SearchEngine::getNearMatch) and wrap it into a
 155          * SearchResultSet.
 156          *
 157          * @param $searchterm string
 158          * @return SearchResultSet
 159          */
 160         public static function getNearMatchResultSet( $searchterm ) {
 161                 return new SearchNearMatchResultSet( self::getNearMatch( $searchterm ) );
 162         }
 163
 164         /**
 165          * Really find the title match.
 166          * @return null|Title
 167          */
 168         private static function getNearMatchInternal( $searchterm ) {
 169                 global $wgContLang, $wgEnableSearchContributorsByIP;
 170
 171                 $allSearchTerms = array( $searchterm );
 172
 173                 if ( $wgContLang->hasVariants() ) {
 174                         $allSearchTerms = array_merge( $allSearchTerms, $wgContLang->autoConvertToAllVariants( $searchterm ) );
 175                 }
 176
 177                 $titleResult = null;
 178                 if ( !wfRunHooks( 'SearchGetNearMatchBefore', array( $allSearchTerms, &$titleResult ) ) ) {
 179                         return $titleResult;
 180                 }
 181
 182                 foreach ( $allSearchTerms as $term ) {
 183
 184                         # Exact match? No need to look further.
 185                         $title = Title::newFromText( $term );
 186                         if ( is_null( $title ) ) {
 187                                 return null;
 188                         }
 189
 190                         # Try files if searching in the Media: namespace
 191                         if ( $title->getNamespace() == NS_MEDIA ) {
 192                                 $title = Title::makeTitle( NS_FILE, $title->getText() );
 193                         }
 194
 195                         if ( $title->isSpecialPage() || $title->isExternal() || $title->exists() ) {
 196                                 return $title;
 197                         }
 198
 199                         # See if it still otherwise has content is some sane sense
 200                         $page = WikiPage::factory( $title );
 201                         if ( $page->hasViewableContent() ) {
 202                                 return $title;
 203                         }
 204
 205                         if ( !wfRunHooks( 'SearchAfterNoDirectMatch', array( $term, &$title ) ) ) {
 206                                 return $title;
 207                         }
 208
 209                         # Now try all lower case (i.e. first letter capitalized)
 210                         $title = Title::newFromText( $wgContLang->lc( $term ) );
 211                         if ( $title && $title->exists() ) {
 212                                 return $title;
 213                         }
 214
 215                         # Now try capitalized string
 216                         $title = Title::newFromText( $wgContLang->ucwords( $term ) );
 217                         if ( $title && $title->exists() ) {
 218                                 return $title;
 219                         }
 220
 221                         # Now try all upper case
 222                         $title = Title::newFromText( $wgContLang->uc( $term ) );
 223                         if ( $title && $title->exists() ) {
 224                                 return $title;
 225                         }
 226
 227                         # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
 228                         $title = Title::newFromText( $wgContLang->ucwordbreaks( $term ) );
 229                         if ( $title && $title->exists() ) {
 230                                 return $title;
 231                         }
 232
 233                         // Give hooks a chance at better match variants
 234                         $title = null;
 235                         if ( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
 236                                 return $title;
 237                         }
 238                 }
 239
 240                 $title = Title::newFromText( $searchterm );
 241
 242                 # Entering an IP address goes to the contributions page
 243                 if ( $wgEnableSearchContributorsByIP ) {
 244                         if ( ( $title->getNamespace() == NS_USER && User::isIP( $title->getText() ) )
 245                                 || User::isIP( trim( $searchterm ) ) ) {
 246                                 return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
 247                         }
 248                 }
 249
 250                 # Entering a user goes to the user page whether it's there or not
 251                 if ( $title->getNamespace() == NS_USER ) {
 252                         return $title;
 253                 }
 254
 255                 # Go to images that exist even if there's no local page.
 256                 # There may have been a funny upload, or it may be on a shared
 257                 # file repository such as Wikimedia Commons.
 258                 if ( $title->getNamespace() == NS_FILE ) {
 259                         $image = wfFindFile( $title );
 260                         if ( $image ) {
 261                                 return $title;
 262                         }
 263                 }
 264
 265                 # MediaWiki namespace? Page may be "implied" if not customized.
 266                 # Just return it, with caps forced as the message system likes it.
 267                 if ( $title->getNamespace() == NS_MEDIAWIKI ) {
 268                         return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
 269                 }
 270
 271                 # Quoted term? Try without the quotes...
 272                 $matches = array();
 273                 if ( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
 274                         return SearchEngine::getNearMatch( $matches[1] );
 275                 }
 276
 277                 return null;
 278         }
 279
 280         public static function legalSearchChars() {
 281                 return "A-Za-z_'.0-9\\x80-\\xFF\\-";
 282         }
 283
 284         /**
 285          * Set the maximum number of results to return
 286          * and how many to skip before returning the first.
 287          *
 288          * @param $limit Integer
 289          * @param $offset Integer
 290          */
 291         function setLimitOffset( $limit, $offset = 0 ) {
 292                 $this->limit = intval( $limit );
 293                 $this->offset = intval( $offset );
 294         }
 295
 296         /**
 297          * Set which namespaces the search should include.
 298          * Give an array of namespace index numbers.
 299          *
 300          * @param $namespaces Array
 301          */
 302         function setNamespaces( $namespaces ) {
 303                 $this->namespaces = $namespaces;
 304         }
 305
 306         /**
 307          * Parse some common prefixes: all (search everything)
 308          * or namespace names
 309          *
 310          * @param $query String
 311          * @return string
 312          */
 313         function replacePrefixes( $query ) {
 314                 global $wgContLang;
 315
 316                 $parsed = $query;
 317                 if ( strpos( $query, ':' ) === false ) { // nothing to do
 318                         wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 319                         return $parsed;
 320                 }
 321
 322                 $allkeyword = wfMessage( 'searchall' )->inContentLanguage()->text() . ":";
 323                 if ( strncmp( $query, $allkeyword, strlen( $allkeyword ) ) == 0 ) {
 324                         $this->namespaces = null;
 325                         $parsed = substr( $query, strlen( $allkeyword ) );
 326                 } elseif ( strpos( $query, ':' ) !== false ) {
 327                         $prefix = substr( $query, 0, strpos( $query, ':' ) );
 328                         $index = $wgContLang->getNsIndex( $prefix );
 329                         if ( $index !== false ) {
 330                                 $this->namespaces = array( $index );
 331                                 $parsed = substr( $query, strlen( $prefix ) + 1 );
 332                         }
 333                 }
 334                 if ( trim( $parsed ) == '' ) {
 335                         $parsed = $query; // prefix was the whole query
 336                 }
 337
 338                 wfRunHooks( 'SearchEngineReplacePrefixesComplete', array( $this, $query, &$parsed ) );
 339
 340                 return $parsed;
 341         }
 342
 343         /**
 344          * Make a list of searchable namespaces and their canonical names.
 345          * @return Array
 346          */
 347         public static function searchableNamespaces() {
 348                 global $wgContLang;
 349                 $arr = array();
 350                 foreach ( $wgContLang->getNamespaces() as $ns => $name ) {
 351                         if ( $ns >= NS_MAIN ) {
 352                                 $arr[$ns] = $name;
 353                         }
 354                 }
 355
 356                 wfRunHooks( 'SearchableNamespaces', array( &$arr ) );
 357                 return $arr;
 358         }
 359
 360         /**
 361          * Extract default namespaces to search from the given user's
 362          * settings, returning a list of index numbers.
 363          *
 364          * @param $user User
 365          * @return Array
 366          */
 367         public static function userNamespaces( $user ) {
 368                 global $wgSearchEverythingOnlyLoggedIn;
 369
 370                 $searchableNamespaces = SearchEngine::searchableNamespaces();
 371
 372                 // get search everything preference, that can be set to be read for logged-in users
 373                 // it overrides other options
 374                 if ( !$wgSearchEverythingOnlyLoggedIn || $user->isLoggedIn() ) {
 375                         if ( $user->getOption( 'searcheverything' ) ) {
 376                                 return array_keys( $searchableNamespaces );
 377                         }
 378                 }
 379
 380                 $arr = array();
 381                 foreach ( $searchableNamespaces as $ns => $name ) {
 382                         if ( $user->getOption( 'searchNs' . $ns ) ) {
 383                                 $arr[] = $ns;
 384                         }
 385                 }
 386
 387                 return $arr;
 388         }
 389
 390         /**
 391          * Find snippet highlight settings for all users
 392          *
 393          * @return Array contextlines, contextchars
 394          */
 395         public static function userHighlightPrefs() {
 396                 $contextlines = 2; // Hardcode this. Old defaults sucked. :)
 397                 $contextchars = 75; // same as above.... :P
 398                 return array( $contextlines, $contextchars );
 399         }
 400
 401         /**
 402          * An array of namespaces indexes to be searched by default
 403          *
 404          * @return Array
 405          */
 406         public static function defaultNamespaces() {
 407                 global $wgNamespacesToBeSearchedDefault;
 408
 409                 return array_keys( $wgNamespacesToBeSearchedDefault, true );
 410         }
 411
 412         /**
 413          * Get a list of namespace names useful for showing in tooltips
 414          * and preferences
 415          *
 416          * @param $namespaces Array
 417          * @return array
 418          */
 419         public static function namespacesAsText( $namespaces ) {
 420                 global $wgContLang;
 421
 422                 $formatted = array_map( array( $wgContLang, 'getFormattedNsText' ), $namespaces );
 423                 foreach ( $formatted as $key => $ns ) {
 424                         if ( empty( $ns ) ) {
 425                                 $formatted[$key] = wfMessage( 'blanknamespace' )->text();
 426                         }
 427                 }
 428                 return $formatted;
 429         }
 430
 431         /**
 432          * Return the help namespaces to be shown on Special:Search
 433          *
 434          * @return Array
 435          */
 436         public static function helpNamespaces() {
 437                 global $wgNamespacesToBeSearchedHelp;
 438
 439                 return array_keys( $wgNamespacesToBeSearchedHelp, true );
 440         }
 441
 442         /**
 443          * Return a 'cleaned up' search string
 444          *
 445          * @param $text String
 446          * @return String
 447          */
 448         function filter( $text ) {
 449                 $lc = $this->legalSearchChars();
 450                 return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
 451         }
 452         /**
 453          * Load up the appropriate search engine class for the currently
 454          * active database backend, and return a configured instance.
 455          *
 456          * @return SearchEngine
 457          */
 458         public static function create() {
 459                 global $wgSearchType;
 460                 $dbr = null;
 461                 if ( $wgSearchType ) {
 462                         $class = $wgSearchType;
 463                 } else {
 464                         $dbr = wfGetDB( DB_SLAVE );
 465                         $class = $dbr->getSearchEngine();
 466                 }
 467                 $search = new $class( $dbr );
 468                 $search->setLimitOffset( 0, 0 );
 469                 return $search;
 470         }
 471
 472         /**
 473          * Create or update the search index record for the given page.
 474          * Title and text should be pre-processed.
 475          * STUB
 476          *
 477          * @param $id Integer
 478          * @param $title String
 479          * @param $text String
 480          */
 481         function update( $id, $title, $text ) {
 482                 // no-op
 483         }
 484
 485         /**
 486          * Update a search index record's title only.
 487          * Title should be pre-processed.
 488          * STUB
 489          *
 490          * @param $id Integer
 491          * @param $title String
 492          */
 493         function updateTitle( $id, $title ) {
 494                 // no-op
 495         }
 496
 497         /**
 498          * Delete an indexed page
 499          * Title should be pre-processed.
 500          * STUB
 501          *
 502          * @param Integer $id Page id that was deleted
 503          * @param String $title Title of page that was deleted
 504          */
 505         function delete( $id, $title ) {
 506                 // no-op
 507         }
 508
 509         /**
 510          * Get OpenSearch suggestion template
 511          *
 512          * @return String
 513          */
 514         public static function getOpenSearchTemplate() {
 515                 global $wgOpenSearchTemplate, $wgCanonicalServer;
 516                 if ( $wgOpenSearchTemplate ) {
 517                         return $wgOpenSearchTemplate;
 518                 } else {
 519                         $ns = implode( '|', SearchEngine::defaultNamespaces() );
 520                         if ( !$ns ) {
 521                                 $ns = "0";
 522                         }
 523                         return $wgCanonicalServer . wfScript( 'api' ) . '?action=opensearch&search={searchTerms}&namespace=' . $ns;
 524                 }
 525         }
 526 }
 527
 528 /**
 529  * @ingroup Search
 530  */
 531 class SearchResultSet {
 532         /**
 533          * Fetch an array of regular expression fragments for matching
 534          * the search terms as parsed by this engine in a text extract.
 535          * STUB
 536          *
 537          * @return Array
 538          */
 539         function termMatches() {
 540                 return array();
 541         }
 542
 543         function numRows() {
 544                 return 0;
 545         }
 546
 547         /**
 548          * Return true if results are included in this result set.
 549          * STUB
 550          *
 551          * @return Boolean
 552          */
 553         function hasResults() {
 554                 return false;
 555         }
 556
 557         /**
 558          * Some search modes return a total hit count for the query
 559          * in the entire article database. This may include pages
 560          * in namespaces that would not be matched on the given
 561          * settings.
 562          *
 563          * Return null if no total hits number is supported.
 564          *
 565          * @return Integer
 566          */
 567         function getTotalHits() {
 568                 return null;
 569         }
 570
 571         /**
 572          * Some search modes return a suggested alternate term if there are
 573          * no exact hits. Returns true if there is one on this set.
 574          *
 575          * @return Boolean
 576          */
 577         function hasSuggestion() {
 578                 return false;
 579         }
 580
 581         /**
 582          * @return String: suggested query, null if none
 583          */
 584         function getSuggestionQuery() {
 585                 return null;
 586         }
 587
 588         /**
 589          * @return String: HTML highlighted suggested query, '' if none
 590          */
 591         function getSuggestionSnippet() {
 592                 return '';
 593         }
 594
 595         /**
 596          * Return information about how and from where the results were fetched,
 597          * should be useful for diagnostics and debugging
 598          *
 599          * @return String
 600          */
 601         function getInfo() {
 602                 return null;
 603         }
 604
 605         /**
 606          * Return a result set of hits on other (multiple) wikis associated with this one
 607          *
 608          * @return SearchResultSet
 609          */
 610         function getInterwikiResults() {
 611                 return null;
 612         }
 613
 614         /**
 615          * Check if there are results on other wikis
 616          *
 617          * @return Boolean
 618          */
 619         function hasInterwikiResults() {
 620                 return $this->getInterwikiResults() != null;
 621         }
 622
 623         /**
 624          * Fetches next search result, or false.
 625          * STUB
 626          *
 627          * @return SearchResult
 628          */
 629         function next() {
 630                 return false;
 631         }
 632
 633         /**
 634          * Frees the result set, if applicable.
 635          */
 636         function free() {
 637                 // ...
 638         }
 639 }
 640
 641 /**
 642  * This class is used for different SQL-based search engines shipped with MediaWiki
 643  */
 644 class SqlSearchResultSet extends SearchResultSet {
 645
 646         protected $mResultSet;
 647
 648         function __construct( $resultSet, $terms ) {
 649                 $this->mResultSet = $resultSet;
 650                 $this->mTerms = $terms;
 651         }
 652
 653         function termMatches() {
 654                 return $this->mTerms;
 655         }
 656
 657         function numRows() {
 658                 if ( $this->mResultSet === false ) {
 659                         return false;
 660                 }
 661
 662                 return $this->mResultSet->numRows();
 663         }
 664
 665         function next() {
 666                 if ( $this->mResultSet === false ) {
 667                         return false;
 668                 }
 669
 670                 $row = $this->mResultSet->fetchObject();
 671                 if ( $row === false ) {
 672                         return false;
 673                 }
 674
 675                 return SearchResult::newFromRow( $row );
 676         }
 677
 678         function free() {
 679                 if ( $this->mResultSet === false ) {
 680                         return false;
 681                 }
 682
 683                 $this->mResultSet->free();
 684         }
 685 }
 686
 687 /**
 688  * @ingroup Search
 689  */
 690 class SearchResultTooMany {
 691         # # Some search engines may bail out if too many matches are found
 692 }
 693
 694 /**
 695  * @todo FIXME: This class is horribly factored. It would probably be better to
 696  * have a useful base class to which you pass some standard information, then
 697  * let the fancy self-highlighters extend that.
 698  * @ingroup Search
 699  */
 700 class SearchResult {
 701
 702         /**
 703          * @var Revision
 704          */
 705         var $mRevision = null;
 706         var $mImage = null;
 707
 708         /**
 709          * @var Title
 710          */
 711         var $mTitle;
 712
 713         /**
 714          * @var String
 715          */
 716         var $mText;
 717
 718         /**
 719          * Return a new SearchResult and initializes it with a title.
 720          *
 721          * @param $title Title
 722          * @return SearchResult
 723          */
 724         public static function newFromTitle( $title ) {
 725                 $result = new self();
 726                 $result->initFromTitle( $title );
 727                 return $result;
 728         }
 729         /**
 730          * Return a new SearchResult and initializes it with a row.
 731          *
 732          * @param $row object
 733          * @return SearchResult
 734          */
 735         public static function newFromRow( $row ) {
 736                 $result = new self();
 737                 $result->initFromRow( $row );
 738                 return $result;
 739         }
 740
 741         public function __construct( $row = null ) {
 742                 if ( !is_null( $row ) ) {
 743                         // Backwards compatibility with pre-1.17 callers
 744                         $this->initFromRow( $row );
 745                 }
 746         }
 747
 748         /**
 749          * Initialize from a database row. Makes a Title and passes that to
 750          * initFromTitle.
 751          *
 752          * @param $row object
 753          */
 754         protected function initFromRow( $row ) {
 755                 $this->initFromTitle( Title::makeTitle( $row->page_namespace, $row->page_title ) );
 756         }
 757
 758         /**
 759          * Initialize from a Title and if possible initializes a corresponding
 760          * Revision and File.
 761          *
 762          * @param $title Title
 763          */
 764         protected function initFromTitle( $title ) {
 765                 $this->mTitle = $title;
 766                 if ( !is_null( $this->mTitle ) ) {
 767                         $id = false;
 768                         wfRunHooks( 'SearchResultInitFromTitle', array( $title, &$id ) );
 769                         $this->mRevision = Revision::newFromTitle(
 770                                 $this->mTitle, $id, Revision::READ_NORMAL );
 771                         if ( $this->mTitle->getNamespace() === NS_FILE ) {
 772                                 $this->mImage = wfFindFile( $this->mTitle );
 773                         }
 774                 }
 775         }
 776
 777         /**
 778          * Check if this is result points to an invalid title
 779          *
 780          * @return Boolean
 781          */
 782         function isBrokenTitle() {
 783                 if ( is_null( $this->mTitle ) ) {
 784                         return true;
 785                 }
 786                 return false;
 787         }
 788
 789         /**
 790          * Check if target page is missing, happens when index is out of date
 791          *
 792          * @return Boolean
 793          */
 794         function isMissingRevision() {
 795                 return !$this->mRevision && !$this->mImage;
 796         }
 797
 798         /**
 799          * @return Title
 800          */
 801         function getTitle() {
 802                 return $this->mTitle;
 803         }
 804
 805         /**
 806          * @return float|null if not supported
 807          */
 808         function getScore() {
 809                 return null;
 810         }
 811
 812         /**
 813          * Lazy initialization of article text from DB
 814          */
 815         protected function initText() {
 816                 if ( !isset( $this->mText ) ) {
 817                         if ( $this->mRevision != null ) {
 818                                 //TODO: if we could plug in some code that knows about special content models *and* about
 819                                 //      special features of the search engine, the search could benefit.
 820                                 $content = $this->mRevision->getContent();
 821                                 $this->mText = $content ? $content->getTextForSearchIndex() : '';
 822                         } else { // TODO: can we fetch raw wikitext for commons images?
 823                                 $this->mText = '';
 824                         }
 825                 }
 826         }
 827
 828         /**
 829          * @param array $terms terms to highlight
 830          * @return String: highlighted text snippet, null (and not '') if not supported
 831          */
 832         function getTextSnippet( $terms ) {
 833                 global $wgUser, $wgAdvancedSearchHighlighting;
 834                 $this->initText();
 835
 836                 // TODO: make highliter take a content object. Make ContentHandler a factory for SearchHighliter.
 837                 list( $contextlines, $contextchars ) = SearchEngine::userHighlightPrefs( $wgUser );
 838                 $h = new SearchHighlighter();
 839                 if ( $wgAdvancedSearchHighlighting ) {
 840                         return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
 841                 } else {
 842                         return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
 843                 }
 844         }
 845
 846         /**
 847          * @param array $terms terms to highlight
 848          * @return String: highlighted title, '' if not supported
 849          */
 850         function getTitleSnippet( $terms ) {
 851                 return '';
 852         }
 853
 854         /**
 855          * @param array $terms terms to highlight
 856          * @return String: highlighted redirect name (redirect to this page), '' if none or not supported
 857          */
 858         function getRedirectSnippet( $terms ) {
 859                 return '';
 860         }
 861
 862         /**
 863          * @return Title object for the redirect to this page, null if none or not supported
 864          */
 865         function getRedirectTitle() {
 866                 return null;
 867         }
 868
 869         /**
 870          * @return string highlighted relevant section name, null if none or not supported
 871          */
 872         function getSectionSnippet() {
 873                 return '';
 874         }
 875
 876         /**
 877          * @return Title object (pagename+fragment) for the section, null if none or not supported
 878          */
 879         function getSectionTitle() {
 880                 return null;
 881         }
 882
 883         /**
 884          * @return String: timestamp
 885          */
 886         function getTimestamp() {
 887                 if ( $this->mRevision ) {
 888                         return $this->mRevision->getTimestamp();
 889                 } elseif ( $this->mImage ) {
 890                         return $this->mImage->getTimestamp();
 891                 }
 892                 return '';
 893         }
 894
 895         /**
 896          * @return Integer: number of words
 897          */
 898         function getWordCount() {
 899                 $this->initText();
 900                 return str_word_count( $this->mText );
 901         }
 902
 903         /**
 904          * @return Integer: size in bytes
 905          */
 906         function getByteSize() {
 907                 $this->initText();
 908                 return strlen( $this->mText );
 909         }
 910
 911         /**
 912          * @return Boolean if hit has related articles
 913          */
 914         function hasRelated() {
 915                 return false;
 916         }
 917
 918         /**
 919          * @return String: interwiki prefix of the title (return iw even if title is broken)
 920          */
 921         function getInterwikiPrefix() {
 922                 return '';
 923         }
 924 }
 925 /**
 926  * A SearchResultSet wrapper for SearchEngine::getNearMatch
 927  */
 928 class SearchNearMatchResultSet extends SearchResultSet {
 929         private $fetched = false;
 930         /**
 931          * @param $match mixed Title if matched, else null
 932          */
 933         public function __construct( $match ) {
 934                 $this->result = $match;
 935         }
 936         public function hasResult() {
 937                 return (bool)$this->result;
 938         }
 939         public function numRows() {
 940                 return $this->hasResults() ? 1 : 0;
 941         }
 942         public function next() {
 943                 if ( $this->fetched || !$this->result ) {
 944                         return false;
 945                 }
 946                 $this->fetched = true;
 947                 return SearchResult::newFromTitle( $this->result );
 948         }
 949 }
 950
 951 /**
 952  * Highlight bits of wikitext
 953  *
 954  * @ingroup Search
 955  */
 956 class SearchHighlighter {
 957         var $mCleanWikitext = true;
 958
 959         function __construct( $cleanupWikitext = true ) {
 960                 $this->mCleanWikitext = $cleanupWikitext;
 961         }
 962
 963         /**
 964          * Default implementation of wikitext highlighting
 965          *
 966          * @param $text String
 967          * @param array $terms terms to highlight (unescaped)
 968          * @param $contextlines Integer
 969          * @param $contextchars Integer
 970          * @return String
 971          */
 972         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
 973                 global $wgContLang;
 974                 global $wgSearchHighlightBoundaries;
 975                 $fname = __METHOD__;
 976
 977                 if ( $text == '' ) {
 978                         return '';
 979                 }
 980
 981                 // spli text into text + templates/links/tables
 982                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
 983                 // first capture group is for detecting nested templates/links/tables/references
 984                 $endPatterns = array(
 985                         1 => '/(\{\{)|(\}\})/', // template
 986                         2 => '/(\[\[)|(\]\])/', // image
 987                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
 988
 989                 // @todo FIXME: This should prolly be a hook or something
 990                 if ( function_exists( 'wfCite' ) ) {
 991                         $spat .= '|(<ref>)'; // references via cite extension
 992                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
 993                 }
 994                 $spat .= '/';
 995                 $textExt = array(); // text extracts
 996                 $otherExt = array(); // other extracts
 997                 wfProfileIn( "$fname-split" );
 998                 $start = 0;
 999                 $textLen = strlen( $text );
1000                 $count = 0; // sequence number to maintain ordering
1001                 while ( $start < $textLen ) {
1002                         // find start of template/image/table
1003                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
1004                                 $epat = '';
1005                                 foreach ( $matches as $key => $val ) {
1006                                         if ( $key > 0 && $val[1] != - 1 ) {
1007                                                 if ( $key == 2 ) {
1008                                                         // see if this is an image link
1009                                                         $ns = substr( $val[0], 2, - 1 );
1010                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE ) {
1011                                                                 break;
1012                                                         }
1013
1014                                                 }
1015                                                 $epat = $endPatterns[$key];
1016                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
1017                                                 $start = $val[1];
1018                                                 break;
1019                                         }
1020                                 }
1021                                 if ( $epat ) {
1022                                         // find end (and detect any nested elements)
1023                                         $level = 0;
1024                                         $offset = $start + 1;
1025                                         $found = false;
1026                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
1027                                                 if ( array_key_exists( 2, $endMatches ) ) {
1028                                                         // found end
1029                                                         if ( $level == 0 ) {
1030                                                                 $len = strlen( $endMatches[2][0] );
1031                                                                 $off = $endMatches[2][1];
1032                                                                 $this->splitAndAdd( $otherExt, $count,
1033                                                                         substr( $text, $start, $off + $len - $start ) );
1034                                                                 $start = $off + $len;
1035                                                                 $found = true;
1036                                                                 break;
1037                                                         } else {
1038                                                                 // end of nested element
1039                                                                 $level -= 1;
1040                                                         }
1041                                                 } else {
1042                                                         // nested
1043                                                         $level += 1;
1044                                                 }
1045                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
1046                                         }
1047                                         if ( ! $found ) {
1048                                                 // couldn't find appropriate closing tag, skip
1049                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
1050                                                 $start += strlen( $matches[0][0] );
1051                                         }
1052                                         continue;
1053                                 }
1054                         }
1055                         // else: add as text extract
1056                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
1057                         break;
1058                 }
1059
1060                 $all = $textExt + $otherExt; // these have disjunct key sets
1061
1062                 wfProfileOut( "$fname-split" );
1063
1064                 // prepare regexps
1065                 foreach ( $terms as $index => $term ) {
1066                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
1067                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
1068                                 $terms[$index] = preg_replace_callback( '/./us', array( $this, 'caseCallback' ), $terms[$index] );
1069                         } else {
1070                                 $terms[$index] = $term;
1071                         }
1072                 }
1073                 $anyterm = implode( '|', $terms );
1074                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
1075
1076                 // @todo FIXME: A hack to scale contextchars, a correct solution
1077                 // would be to have contextchars actually be char and not byte
1078                 // length, and do proper utf-8 substrings and lengths everywhere,
1079                 // but PHP is making that very hard and unclean to implement :(
1080                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
1081                 $contextchars = intval( $contextchars * $scale );
1082
1083                 $patPre = "(^|$wgSearchHighlightBoundaries)";
1084                 $patPost = "($wgSearchHighlightBoundaries|$)";
1085
1086                 $pat1 = "/(" . $phrase . ")/ui";
1087                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
1088
1089                 wfProfileIn( "$fname-extract" );
1090
1091                 $left = $contextlines;
1092
1093                 $snippets = array();
1094                 $offsets = array();
1095
1096                 // show beginning only if it contains all words
1097                 $first = 0;
1098                 $firstText = '';
1099                 foreach ( $textExt as $index => $line ) {
1100                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
1101                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
1102                                 $first = $index;
1103                                 break;
1104                         }
1105                 }
1106                 if ( $firstText ) {
1107                         $succ = true;
1108                         // check if first text contains all terms
1109                         foreach ( $terms as $term ) {
1110                                 if ( ! preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
1111                                         $succ = false;
1112                                         break;
1113                                 }
1114                         }
1115                         if ( $succ ) {
1116                                 $snippets[$first] = $firstText;
1117                                 $offsets[$first] = 0;
1118                         }
1119                 }
1120                 if ( ! $snippets ) {
1121                         // match whole query on text
1122                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
1123                         // match whole query on templates/tables/images
1124                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
1125                         // match any words on text
1126                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
1127                         // match any words on templates/tables/images
1128                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
1129
1130                         ksort( $snippets );
1131                 }
1132
1133                 // add extra chars to each snippet to make snippets constant size
1134                 $extended = array();
1135                 if ( count( $snippets ) == 0 ) {
1136                         // couldn't find the target words, just show beginning of article
1137                         if ( array_key_exists( $first, $all ) ) {
1138                                 $targetchars = $contextchars * $contextlines;
1139                                 $snippets[$first] = '';
1140                                 $offsets[$first] = 0;
1141                         }
1142                 } else {
1143                         // if begin of the article contains the whole phrase, show only that !!
1144                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
1145                                 && $offsets[$first] < $contextchars * 2 ) {
1146                                 $snippets = array( $first => $snippets[$first] );
1147                         }
1148
1149                         // calc by how much to extend existing snippets
1150                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
1151                 }
1152
1153                 foreach ( $snippets as $index => $line ) {
1154                         $extended[$index] = $line;
1155                         $len = strlen( $line );
1156                         if ( $len < $targetchars - 20 ) {
1157                                 // complete this line
1158                                 if ( $len < strlen( $all[$index] ) ) {
1159                                         $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index] + $targetchars, $offsets[$index] );
1160                                         $len = strlen( $extended[$index] );
1161                                 }
1162
1163                                 // add more lines
1164                                 $add = $index + 1;
1165                                 while ( $len < $targetchars - 20
1166                                                 && array_key_exists( $add, $all )
1167                                                 && !array_key_exists( $add, $snippets ) ) {
1168                                         $offsets[$add] = 0;
1169                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
1170                                         $extended[$add] = $tt;
1171                                         $len += strlen( $tt );
1172                                         $add++;
1173                                 }
1174                         }
1175                 }
1176
1177                 // $snippets = array_map( 'htmlspecialchars', $extended );
1178                 $snippets = $extended;
1179                 $last = - 1;
1180                 $extract = '';
1181                 foreach ( $snippets as $index => $line ) {
1182                         if ( $last == - 1 ) {
1183                                 $extract .= $line; // first line
1184                         } elseif ( $last + 1 == $index && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] ) ) {
1185                                 $extract .= " " . $line; // continous lines
1186                         } else {
1187                                 $extract .= '<b> ... </b>' . $line;
1188                         }
1189
1190                         $last = $index;
1191                 }
1192                 if ( $extract ) {
1193                         $extract .= '<b> ... </b>';
1194                 }
1195
1196                 $processed = array();
1197                 foreach ( $terms as $term ) {
1198                         if ( ! isset( $processed[$term] ) ) {
1199                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
1200                                 $extract = preg_replace( $pat3,
1201                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
1202                                 $processed[$term] = true;
1203                         }
1204                 }
1205
1206                 wfProfileOut( "$fname-extract" );
1207
1208                 return $extract;
1209         }
1210
1211         /**
1212          * Split text into lines and add it to extracts array
1213          *
1214          * @param array $extracts index -> $line
1215          * @param $count Integer
1216          * @param $text String
1217          */
1218         function splitAndAdd( &$extracts, &$count, $text ) {
1219                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
1220                 foreach ( $split as $line ) {
1221                         $tt = trim( $line );
1222                         if ( $tt ) {
1223                                 $extracts[$count++] = $tt;
1224                         }
1225                 }
1226         }
1227
1228         /**
1229          * Do manual case conversion for non-ascii chars
1230          *
1231          * @param $matches Array
1232          * @return string
1233          */
1234         function caseCallback( $matches ) {
1235                 global $wgContLang;
1236                 if ( strlen( $matches[0] ) > 1 ) {
1237                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
1238                 } else {
1239                         return $matches[0];
1240                 }
1241         }
1242
1243         /**
1244          * Extract part of the text from start to end, but by
1245          * not chopping up words
1246          * @param $text String
1247          * @param $start Integer
1248          * @param $end Integer
1249          * @param $posStart Integer: (out) actual start position
1250          * @param $posEnd Integer: (out) actual end position
1251          * @return String
1252          */
1253         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
1254                 if ( $start != 0 ) {
1255                         $start = $this->position( $text, $start, 1 );
1256                 }
1257                 if ( $end >= strlen( $text ) ) {
1258                         $end = strlen( $text );
1259                 } else {
1260                         $end = $this->position( $text, $end );
1261                 }
1262
1263                 if ( !is_null( $posStart ) ) {
1264                         $posStart = $start;
1265                 }
1266                 if ( !is_null( $posEnd ) ) {
1267                         $posEnd = $end;
1268                 }
1269
1270                 if ( $end > $start ) {
1271                         return substr( $text, $start, $end - $start );
1272                 } else {
1273                         return '';
1274                 }
1275         }
1276
1277         /**
1278          * Find a nonletter near a point (index) in the text
1279          *
1280          * @param $text String
1281          * @param $point Integer
1282          * @param $offset Integer: offset to found index
1283          * @return Integer: nearest nonletter index, or beginning of utf8 char if none
1284          */
1285         function position( $text, $point, $offset = 0 ) {
1286                 $tolerance = 10;
1287                 $s = max( 0, $point - $tolerance );
1288                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
1289                 $m = array();
1290                 if ( preg_match( '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr( $text, $s, $l ), $m, PREG_OFFSET_CAPTURE ) ) {
1291                         return $m[0][1] + $s + $offset;
1292                 } else {
1293                         // check if point is on a valid first UTF8 char
1294                         $char = ord( $text[$point] );
1295                         while ( $char >= 0x80 && $char < 0xc0 ) {
1296                                 // skip trailing bytes
1297                                 $point++;
1298                                 if ( $point >= strlen( $text ) ) {
1299                                         return strlen( $text );
1300                                 }
1301                                 $char = ord( $text[$point] );
1302                         }
1303                         return $point;
1304
1305                 }
1306         }
1307
1308         /**
1309          * Search extracts for a pattern, and return snippets
1310          *
1311          * @param string $pattern regexp for matching lines
1312          * @param array $extracts extracts to search
1313          * @param $linesleft Integer: number of extracts to make
1314          * @param $contextchars Integer: length of snippet
1315          * @param array $out map for highlighted snippets
1316          * @param array $offsets map of starting points of snippets
1317          * @protected
1318          */
1319         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
1320                 if ( $linesleft == 0 ) {
1321                         return; // nothing to do
1322                 }
1323                 foreach ( $extracts as $index => $line ) {
1324                         if ( array_key_exists( $index, $out ) ) {
1325                                 continue; // this line already highlighted
1326                         }
1327
1328                         $m = array();
1329                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) ) {
1330                                 continue;
1331                         }
1332
1333                         $offset = $m[0][1];
1334                         $len = strlen( $m[0][0] );
1335                         if ( $offset + $len < $contextchars ) {
1336                                 $begin = 0;
1337                         } elseif ( $len > $contextchars ) {
1338                                 $begin = $offset;
1339                         } else {
1340                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
1341                         }
1342
1343                         $end = $begin + $contextchars;
1344
1345                         $posBegin = $begin;
1346                         // basic snippet from this line
1347                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
1348                         $offsets[$index] = $posBegin;
1349                         $linesleft--;
1350                         if ( $linesleft == 0 ) {
1351                                 return;
1352                         }
1353                 }
1354         }
1355
1356         /**
1357          * Basic wikitext removal
1358          * @protected
1359          * @return mixed
1360          */
1361         function removeWiki( $text ) {
1362                 $fname = __METHOD__;
1363                 wfProfileIn( $fname );
1364
1365                 // $text = preg_replace( "/'{2,5}/", "", $text );
1366                 // $text = preg_replace( "/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text );
1367                 // $text = preg_replace( "/\[\[([^]|]+)\]\]/", "\\1", $text );
1368                 // $text = preg_replace( "/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text );
1369                 // $text = preg_replace( "/\\{\\|(.*?)\\|\\}/", "", $text );
1370                 // $text = preg_replace( "/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text );
1371                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
1372                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
1373                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
1374                 $text = preg_replace_callback( "/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array( $this, 'linkReplace' ), $text );
1375                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
1376                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
1377                 $text = preg_replace( "/'''''/", "", $text );
1378                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
1379                 $text = preg_replace( "/''/", "", $text );
1380
1381                 wfProfileOut( $fname );
1382                 return $text;
1383         }
1384
1385         /**
1386          * callback to replace [[target|caption]] kind of links, if
1387          * the target is category or image, leave it
1388          *
1389          * @param $matches Array
1390          */
1391         function linkReplace( $matches ) {
1392                 $colon = strpos( $matches[1], ':' );
1393                 if ( $colon === false ) {
1394                         return $matches[2]; // replace with caption
1395                 }
1396                 global $wgContLang;
1397                 $ns = substr( $matches[1], 0, $colon );
1398                 $index = $wgContLang->getNsIndex( $ns );
1399                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) ) {
1400                         return $matches[0]; // return the whole thing
1401                 } else {
1402                         return $matches[2];
1403                 }
1404         }
1405
1406         /**
1407          * Simple & fast snippet extraction, but gives completely unrelevant
1408          * snippets
1409          *
1410          * @param $text String
1411          * @param $terms Array
1412          * @param $contextlines Integer
1413          * @param $contextchars Integer
1414          * @return String
1415          */
1416         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
1417                 global $wgContLang;
1418                 $fname = __METHOD__;
1419
1420                 $lines = explode( "\n", $text );
1421
1422                 $terms = implode( '|', $terms );
1423                 $max = intval( $contextchars ) + 1;
1424                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
1425
1426                 $lineno = 0;
1427
1428                 $extract = "";
1429                 wfProfileIn( "$fname-extract" );
1430                 foreach ( $lines as $line ) {
1431                         if ( 0 == $contextlines ) {
1432                                 break;
1433                         }
1434                         ++$lineno;
1435                         $m = array();
1436                         if ( ! preg_match( $pat1, $line, $m ) ) {
1437                                 continue;
1438                         }
1439                         --$contextlines;
1440                         // truncate function changes ... to relevant i18n message.
1441                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
1442
1443                         if ( count( $m ) < 3 ) {
1444                                 $post = '';
1445                         } else {
1446                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
1447                         }
1448
1449                         $found = $m[2];
1450
1451                         $line = htmlspecialchars( $pre . $found . $post );
1452                         $pat2 = '/(' . $terms . ")/i";
1453                         $line = preg_replace( $pat2, "<span class='searchmatch'>\\1</span>", $line );
1454
1455                         $extract .= "${line}\n";
1456                 }
1457                 wfProfileOut( "$fname-extract" );
1458
1459                 return $extract;
1460         }
1461
1462 }
1463
1464 /**
1465  * Dummy class to be used when non-supported Database engine is present.
1466  * @todo FIXME: Dummy class should probably try something at least mildly useful,
1467  * such as a LIKE search through titles.
1468  * @ingroup Search
1469  */
1470 class SearchEngineDummy extends SearchEngine {
1471         // no-op
1472 }