includes/search/SearchEngine.php

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