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