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