<?php
/**
- * Contain site class
- * See search.doc
- * @package MediaWiki
- */
-
-/**
+ * @defgroup Search Search
*
+ * @file
+ * @ingroup Search
*/
-define( 'MW_SEARCH_OK', true );
-define( 'MW_SEARCH_BAD_QUERY', false );
/**
- * @todo document
- * @package MediaWiki
+ * Contain a class for special pages
+ * @ingroup Search
*/
class SearchEngine {
- /* private */ var $rawText, $filteredText, $searchTerms;
- /* private */ var $titleCond, $textCond;
-
- var $doSearchRedirects = true;
- var $addToQuery = array();
- var $namespacesToSearch = array();
- var $alternateTitle;
- var $allTitles = false;
-
- function SearchEngine( $text ) {
- $this->rawText = trim( $text );
-
- # We display the query, so let's strip it for safety
- #
- global $wgDBmysql4;
- $lc = SearchEngine::legalSearchChars() . '()';
- if( $wgDBmysql4 ) {
- $lc .= "\"~<>*+-";
- }
- $this->filteredText = trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
- $this->searchTerms = array();
- $this->strictMatching = true; # Google-style, add '+' on all terms
-
- $this->db =& wfGetDB( DB_SLAVE );
- }
+ var $limit = 10;
+ var $offset = 0;
+ var $prefix = '';
+ var $searchTerms = array();
+ var $namespaces = array( NS_MAIN );
+ var $showRedirects = false;
/**
- * Return a partial WHERE clause to limit the search to the given namespaces
+ * Perform a full text search query and return a result set.
+ * If title searches are not supported or disabled, return null.
+ *
+ * @param string $term - Raw search term
+ * @return SearchResultSet
+ * @access public
+ * @abstract
*/
- function queryNamespaces() {
- $namespaces = implode( ',', $this->namespacesToSearch );
- if ($namespaces == '') {
- $namespaces = '0';
- }
- return "AND cur_namespace IN (" . $namespaces . ')';
+ function searchText( $term ) {
+ return null;
}
/**
- * Return a partial WHERE clause to include or exclude redirects from results
+ * Perform a title-only search query and return a result set.
+ * If title searches are not supported or disabled, return null.
+ *
+ * @param string $term - Raw search term
+ * @return SearchResultSet
+ * @access public
+ * @abstract
*/
- function searchRedirects() {
- if ( $this->doSearchRedirects ) {
- return '';
- } else {
- return 'AND cur_is_redirect=0 ';
- }
+ function searchTitle( $term ) {
+ return null;
}
-
+
+ /** If this search backend can list/unlist redirects */
+ function acceptListRedirects() {
+ return true;
+ }
+
/**
- * @access private
- */ function initNamespaceCheckbox( $i ) {
- global $wgUser, $wgNamespacesToBeSearchedDefault;
-
- if ($wgUser->getID()) {
- // User is logged in so we retrieve his default namespaces
- return $wgUser->getOption( 'searchNs'.$i );
- } else {
- // User is not logged in so we give him the global default namespaces
- return !empty($wgNamespacesToBeSearchedDefault[ $i ]);
- }
+ * Transform search term in cases when parts of the query came as different GET params (when supported)
+ * e.g. for prefix queries: search=test&prefix=Main_Page/Archive -> test prefix:Main Page/Archive
+ */
+ function transformSearchTerm( $term ) {
+ return $term;
}
-
+
/**
- * Display the "power search" footer. Does not actually perform the search,
- * that is done by showResults()
+ * If an exact title match can be find, or a very slightly close match,
+ * return the title. If no match, returns NULL.
+ *
+ * @param string $term
+ * @return Title
*/
- function powersearch() {
- global $wgUser, $wgOut, $wgContLang, $wgTitle, $wgRequest;
- $sk =& $wgUser->getSkin();
-
- $search = $this->rawText;
- $searchx = $wgRequest->getVal( 'searchx' );
- $listredirs = $wgRequest->getVal( 'redirs' );
-
- $ret = wfMsg('powersearchtext'); # Text to be returned
- $tempText = ''; # Temporary text, for substitution into $ret
+ public static function getNearMatch( $searchterm ) {
+ global $wgContLang;
- if( isset( $_REQUEST['searchx'] ) ) {
- $this->addToQuery['searchx'] = '1';
+ $allSearchTerms = array($searchterm);
+
+ if($wgContLang->hasVariants()){
+ $allSearchTerms = array_merge($allSearchTerms,$wgContLang->convertLinkToAllVariants($searchterm));
}
-
- # Do namespace checkboxes
- $namespaces = $wgContLang->getNamespaces();
- foreach ( $namespaces as $i => $namespace ) {
- # Skip virtual namespaces
- if ( $i < 0 ) {
- continue;
+
+ foreach($allSearchTerms as $term){
+
+ # Exact match? No need to look further.
+ $title = Title::newFromText( $term );
+ if (is_null($title))
+ return NULL;
+
+ if ( $title->getNamespace() == NS_SPECIAL || $title->isExternal() || $title->exists() ) {
+ return $title;
+ }
+
+ # See if it still otherwise has content is some sane sense
+ $article = MediaWiki::articleFromTitle( $title );
+ if( $article->hasViewableContent() ) {
+ return $title;
}
- $formVar = 'ns'.$i;
+ # Now try all lower case (i.e. first letter capitalized)
+ #
+ $title = Title::newFromText( $wgContLang->lc( $term ) );
+ if ( $title && $title->exists() ) {
+ return $title;
+ }
- # Initialise checkboxValues, either from defaults or from
- # a previous invocation
- if ( !isset( $searchx ) ) {
- $checkboxValue = $this->initNamespaceCheckbox( $i );
- } else {
- $checkboxValue = $wgRequest->getVal( $formVar );
+ # Now try capitalized string
+ #
+ $title = Title::newFromText( $wgContLang->ucwords( $term ) );
+ if ( $title && $title->exists() ) {
+ return $title;
}
- $checked = '';
- if ( $checkboxValue == 1 ) {
- $checked = ' checked="checked"';
- $this->addToQuery['ns'.$i] = 1;
- array_push( $this->namespacesToSearch, $i );
+ # Now try all upper case
+ #
+ $title = Title::newFromText( $wgContLang->uc( $term ) );
+ if ( $title && $title->exists() ) {
+ return $title;
}
- $name = str_replace( '_', ' ', $namespaces[$i] );
- if ( '' == $name ) {
- $name = wfMsg( 'blanknamespace' );
+
+ # Now try Word-Caps-Breaking-At-Word-Breaks, for hyphenated names etc
+ $title = Title::newFromText( $wgContLang->ucwordbreaks($term) );
+ if ( $title && $title->exists() ) {
+ return $title;
}
- if ( $tempText !== '' ) {
- $tempText .= ' ';
+ // Give hooks a chance at better match variants
+ $title = null;
+ if( !wfRunHooks( 'SearchGetNearMatch', array( $term, &$title ) ) ) {
+ return $title;
}
- $tempText .= "<input type='checkbox' value=\"1\" name=\"" .
- "ns{$i}\"{$checked} />{$name}\n";
}
- $ret = str_replace ( '$1', $tempText, $ret );
- # List redirects checkbox
+ $title = Title::newFromText( $searchterm );
- $checked = '';
- if ( $listredirs == 1 ) {
- $this->addToQuery['redirs'] = 1;
- $checked = ' checked="checked"';
+ # Entering an IP address goes to the contributions page
+ if ( ( $title->getNamespace() == NS_USER && User::isIP($title->getText() ) )
+ || User::isIP( trim( $searchterm ) ) ) {
+ return SpecialPage::getTitleFor( 'Contributions', $title->getDBkey() );
}
- $tempText = "<input type='checkbox' value='1' name=\"redirs\"{$checked} />\n";
- $ret = str_replace( '$2', $tempText, $ret );
- # Search field
- $tempText = "<input type='text' name=\"search\" value=\"" .
- htmlspecialchars( $search ) ."\" width=\"80\" />\n";
- $ret = str_replace( "$3", $tempText, $ret );
-
- # Searchx button
+ # Entering a user goes to the user page whether it's there or not
+ if ( $title->getNamespace() == NS_USER ) {
+ return $title;
+ }
- $tempText = '<input type="submit" name="searchx" value="' .
- wfMsg('powersearch') . "\" />\n";
- $ret = str_replace( '$9', $tempText, $ret );
+ # Go to images that exist even if there's no local page.
+ # There may have been a funny upload, or it may be on a shared
+ # file repository such as Wikimedia Commons.
+ if( $title->getNamespace() == NS_FILE ) {
+ $image = wfFindFile( $title );
+ if( $image ) {
+ return $title;
+ }
+ }
- $action = $sk->escapeSearchLink();
- $ret = "<br /><br />\n<form id=\"powersearch\" method=\"get\" " .
- "action=\"$action\">\n{$ret}\n</form>\n";
+ # MediaWiki namespace? Page may be "implied" if not customized.
+ # Just return it, with caps forced as the message system likes it.
+ if( $title->getNamespace() == NS_MEDIAWIKI ) {
+ return Title::makeTitle( NS_MEDIAWIKI, $wgContLang->ucfirst( $title->getText() ) );
+ }
- if ( isset ( $searchx ) ) {
- if ( ! $listredirs ) {
- $this->doSearchRedirects = false;
- }
+ # Quoted term? Try without the quotes...
+ $matches = array();
+ if( preg_match( '/^"([^"]+)"$/', $searchterm, $matches ) ) {
+ return SearchEngine::getNearMatch( $matches[1] );
}
- return $ret;
+
+ return NULL;
}
- function setupPage() {
- global $wgOut;
- $wgOut->setPageTitle( wfMsg( 'searchresults' ) );
- $wgOut->setSubtitle( wfMsg( 'searchquery', htmlspecialchars( $this->rawText ) ) );
- $wgOut->setArticleRelated( false );
- $wgOut->setRobotpolicy( 'noindex,nofollow' );
+ public static function legalSearchChars() {
+ return "A-Za-z_'.0-9\\x80-\\xFF\\-";
}
/**
- * Perform the search and construct the results page
+ * Set the maximum number of results to return
+ * and how many to skip before returning the first.
+ *
+ * @param int $limit
+ * @param int $offset
+ * @access public
*/
- function showResults() {
- global $wgUser, $wgTitle, $wgOut, $wgLang;
- global $wgDisableTextSearch, $wgInputEncoding;
- $fname = 'SearchEngine::showResults';
+ function setLimitOffset( $limit, $offset = 0 ) {
+ $this->limit = intval( $limit );
+ $this->offset = intval( $offset );
+ }
- $search = $this->rawText;
+ /**
+ * Set which namespaces the search should include.
+ * Give an array of namespace index numbers.
+ *
+ * @param array $namespaces
+ * @access public
+ */
+ function setNamespaces( $namespaces ) {
+ $this->namespaces = $namespaces;
+ }
- $powersearch = $this->powersearch(); /* Need side-effects here? */
+ /**
+ * Parse some common prefixes: all (search everything)
+ * or namespace names
+ *
+ * @param string $query
+ */
+ function replacePrefixes( $query ){
+ global $wgContLang;
- $this->setupPage();
+ if( strpos($query,':') === false )
+ return $query; // nothing to do
+
+ $parsed = $query;
+ $allkeyword = wfMsgForContent('searchall').":";
+ if( strncmp($query, $allkeyword, strlen($allkeyword)) == 0 ){
+ $this->namespaces = null;
+ $parsed = substr($query,strlen($allkeyword));
+ } else if( strpos($query,':') !== false ) {
+ $prefix = substr($query,0,strpos($query,':'));
+ $index = $wgContLang->getNsIndex($prefix);
+ if($index !== false){
+ $this->namespaces = array($index);
+ $parsed = substr($query,strlen($prefix)+1);
+ }
+ }
+ if(trim($parsed) == '')
+ return $query; // prefix was the whole query
- $sk = $wgUser->getSkin();
- $wgOut->addWikiText( wfMsg( 'searchresulttext' ) );
+ return $parsed;
+ }
- if ( !$this->parseQuery() ) {
- $wgOut->addWikiText(
- '==' . wfMsg( 'badquery' ) . "==\n" .
- wfMsg( 'badquerytext' ) );
- return;
+ /**
+ * Make a list of searchable namespaces and their canonical names.
+ * @return array
+ */
+ public static function searchableNamespaces() {
+ global $wgContLang;
+ $arr = array();
+ foreach( $wgContLang->getNamespaces() as $ns => $name ) {
+ if( $ns >= NS_MAIN ) {
+ $arr[$ns] = $name;
+ }
}
- list( $limit, $offset ) = wfCheckLimits( 20, 'searchlimit' );
+ return $arr;
+ }
+
+ /**
+ * Extract default namespaces to search from the given user's
+ * settings, returning a list of index numbers.
+ *
+ * @param User $user
+ * @return array
+ * @static
+ */
+ public static function userNamespaces( &$user ) {
+ $arr = array();
+ foreach( SearchEngine::searchableNamespaces() as $ns => $name ) {
+ if( $user->getOption( 'searchNs' . $ns ) ) {
+ $arr[] = $ns;
+ }
+ }
+ return $arr;
+ }
+
+ /**
+ * Find snippet highlight settings for a given user
+ *
+ * @param User $user
+ * @return array contextlines, contextchars
+ * @static
+ */
+ public static function userHighlightPrefs( &$user ){
+ //$contextlines = $user->getOption( 'contextlines', 5 );
+ //$contextchars = $user->getOption( 'contextchars', 50 );
+ $contextlines = 2; // Hardcode this. Old defaults sucked. :)
+ $contextchars = 75; // same as above.... :P
+ return array($contextlines, $contextchars);
+ }
+
+ /**
+ * An array of namespaces indexes to be searched by default
+ *
+ * @return array
+ * @static
+ */
+ public static function defaultNamespaces(){
+ global $wgNamespacesToBeSearchedDefault;
+
+ return array_keys($wgNamespacesToBeSearchedDefault, true);
+ }
+
+ /**
+ * Get a list of namespace names useful for showing in tooltips
+ * and preferences
+ *
+ * @param unknown_type $namespaces
+ */
+ public static function namespacesAsText( $namespaces ){
+ global $wgContLang;
- if ( $wgDisableTextSearch ) {
- $wgOut->addHTML( wfMsg( 'searchdisabled' ) );
- $wgOut->addHTML( wfMsg( 'googlesearch',
- htmlspecialchars( $this->rawText ),
- htmlspecialchars( $wgInputEncoding ) ) );
- return;
+ $formatted = array_map( array($wgContLang,'getFormattedNsText'), $namespaces );
+ foreach( $formatted as $key => $ns ){
+ if ( empty($ns) )
+ $formatted[$key] = wfMsg( 'blanknamespace' );
}
-
- $titleMatches = $this->getMatches( $this->titleCond, $limit, $offset );
- $textMatches = $this->getMatches( $this->textCond, $limit, $offset );
-
- $sk = $wgUser->getSkin();
+ return $formatted;
+ }
+
+ /**
+ * An array of "project" namespaces indexes typically searched
+ * by logged-in users
+ *
+ * @return array
+ * @static
+ */
+ public static function projectNamespaces() {
+ global $wgNamespacesToBeSearchedDefault, $wgNamespacesToBeSearchedProject;
+
+ return array_keys( $wgNamespacesToBeSearchedProject, true );
+ }
+
+ /**
+ * An array of "project" namespaces indexes typically searched
+ * by logged-in users in addition to the default namespaces
+ *
+ * @return array
+ * @static
+ */
+ public static function defaultAndProjectNamespaces() {
+ global $wgNamespacesToBeSearchedDefault, $wgNamespacesToBeSearchedProject;
- $num = count( $titleMatches ) + count( $textMatches );
- if ( $num >= $limit ) {
- $top = wfShowingResults( $offset, $limit );
+ return array_keys( $wgNamespacesToBeSearchedDefault +
+ $wgNamespacesToBeSearchedProject, true);
+ }
+
+ /**
+ * Return a 'cleaned up' search string
+ *
+ * @return string
+ * @access public
+ */
+ function filter( $text ) {
+ $lc = $this->legalSearchChars();
+ return trim( preg_replace( "/[^{$lc}]/", " ", $text ) );
+ }
+ /**
+ * Load up the appropriate search engine class for the currently
+ * active database backend, and return a configured instance.
+ *
+ * @return SearchEngine
+ */
+ public static function create() {
+ global $wgSearchType;
+ $dbr = wfGetDB( DB_SLAVE );
+ if( $wgSearchType ) {
+ $class = $wgSearchType;
} else {
- $top = wfShowingResultsNum( $offset, $limit, $num );
- }
- $wgOut->addHTML( "<p>{$top}</p>\n" );
-
- # For powersearch
- $a2l = '';
- $akk = array_keys( $this->addToQuery );
- foreach ( $akk AS $ak ) {
- $a2l .= "&{$ak}={$this->addToQuery[$ak]}" ;
+ $class = $dbr->getSearchEngine();
}
+ $search = new $class( $dbr );
+ $search->setLimitOffset(0,0);
+ return $search;
+ }
- $prevnext = wfViewPrevNext( $offset, $limit, '',
- 'search=' . wfUrlencode( $this->filteredText ) . $a2l );
- $wgOut->addHTML( "<br />{$prevnext}\n" );
+ /**
+ * Create or update the search index record for the given page.
+ * Title and text should be pre-processed.
+ *
+ * @param int $id
+ * @param string $title
+ * @param string $text
+ * @abstract
+ */
+ function update( $id, $title, $text ) {
+ // no-op
+ }
- $foundsome = $this->showMatches( $titleMatches, $offset, 'notitlematches', 'titlematches' )
- || $this->showMatches( $textMatches, $offset, 'notextmatches', 'textmatches' );
-
- if ( !$foundsome ) {
- $wgOut->addWikiText( wfMsg( 'nonefound' ) );
+ /**
+ * Update a search index record's title only.
+ * Title should be pre-processed.
+ *
+ * @param int $id
+ * @param string $title
+ * @abstract
+ */
+ function updateTitle( $id, $title ) {
+ // no-op
+ }
+
+ /**
+ * Get OpenSearch suggestion template
+ *
+ * @return string
+ * @static
+ */
+ public static function getOpenSearchTemplate() {
+ global $wgOpenSearchTemplate, $wgServer, $wgScriptPath;
+ if( $wgOpenSearchTemplate ) {
+ return $wgOpenSearchTemplate;
+ } else {
+ $ns = implode( '|', SearchEngine::defaultNamespaces() );
+ if( !$ns ) $ns = "0";
+ return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace='.$ns;
}
- $wgOut->addHTML( "<p>{$prevnext}</p>\n" );
- $wgOut->addHTML( $powersearch );
}
+
+ /**
+ * Get internal MediaWiki Suggest template
+ *
+ * @return string
+ * @static
+ */
+ public static function getMWSuggestTemplate() {
+ global $wgMWSuggestTemplate, $wgServer, $wgScriptPath;
+ if($wgMWSuggestTemplate)
+ return $wgMWSuggestTemplate;
+ else
+ return $wgServer . $wgScriptPath . '/api.php?action=opensearch&search={searchTerms}&namespace={namespaces}&suggest';
+ }
+}
- function legalSearchChars() {
- $lc = "A-Za-z_'0-9\\x80-\\xFF\\-";
- return $lc;
+/**
+ * @ingroup Search
+ */
+class SearchResultSet {
+ /**
+ * Fetch an array of regular expression fragments for matching
+ * the search terms as parsed by this engine in a text extract.
+ *
+ * @return array
+ * @access public
+ * @abstract
+ */
+ function termMatches() {
+ return array();
}
- function parseQuery() {
- global $wgDBmysql4;
- if( $wgDBmysql4 ) {
- # Use cleaner boolean search if available
- return $this->parseQuery4();
- } else {
- # Fall back to ugly hack with multiple search clauses
- return $this->parseQuery3();
- }
+ function numRows() {
+ return 0;
}
-
- function parseQuery3() {
- global $wgDBminWordLen, $wgContLang;
- # on non mysql4 database: get list of words we don't want to search for
- require_once( 'FulltextStoplist.php' );
+ /**
+ * Return true if results are included in this result set.
+ * @return bool
+ * @abstract
+ */
+ function hasResults() {
+ return false;
+ }
- $lc = SearchEngine::legalSearchChars() . '()';
- $q = preg_replace( "/([()])/", " \\1 ", $this->filteredText );
- $q = preg_replace( "/\\s+/", " ", $q );
- $w = explode( ' ', trim( $q ) );
+ /**
+ * Some search modes return a total hit count for the query
+ * in the entire article database. This may include pages
+ * in namespaces that would not be matched on the given
+ * settings.
+ *
+ * Return null if no total hits number is supported.
+ *
+ * @return int
+ * @access public
+ */
+ function getTotalHits() {
+ return null;
+ }
- $last = $cond = '';
- foreach ( $w as $word ) {
- $word = $wgContLang->stripForSearch( $word );
- if ( 'and' == $word || 'or' == $word || 'not' == $word
- || '(' == $word || ')' == $word ) {
- $cond .= ' ' . strtoupper( $word );
- $last = '';
- } else if ( strlen( $word ) < $wgDBminWordLen ) {
- continue;
- } else if ( FulltextStoplist::inList( $word ) ) {
- continue;
- } else {
- if ( '' != $last ) { $cond .= ' AND'; }
- $cond .= " (MATCH (##field##) AGAINST ('" .
- $this->db->strencode( $word ). "'))";
- $last = $word;
- array_push( $this->searchTerms, "\\b" . $word . "\\b" );
- }
- }
- if ( 0 == count( $this->searchTerms ) ) {
- return MW_SEARCH_BAD_QUERY;
- }
+ /**
+ * Some search modes return a suggested alternate term if there are
+ * no exact hits. Returns true if there is one on this set.
+ *
+ * @return bool
+ * @access public
+ */
+ function hasSuggestion() {
+ return false;
+ }
- $this->titleCond = '(' . str_replace( '##field##',
- 'si_title', $cond ) . ' )';
+ /**
+ * @return string suggested query, null if none
+ */
+ function getSuggestionQuery(){
+ return null;
+ }
- $this->textCond = '(' . str_replace( '##field##',
- 'si_text', $cond ) . ' AND (cur_is_redirect=0) )';
-
- return MW_SEARCH_OK;
+ /**
+ * @return string HTML highlighted suggested query, '' if none
+ */
+ function getSuggestionSnippet(){
+ return '';
}
- function parseQuery4() {
- global $wgContLang;
- $lc = SearchEngine::legalSearchChars();
- $searchon = '';
- $this->searchTerms = array();
-
- # FIXME: This doesn't handle parenthetical expressions.
- if( preg_match_all( '/([-+<>~]?)(([' . $lc . ']+)(\*?)|"[^"]*")/',
- $this->filteredText, $m, PREG_SET_ORDER ) ) {
- foreach( $m as $terms ) {
- if( $searchon !== '' ) $searchon .= ' ';
- if( $this->strictMatching && ($terms[1] == '') ) {
- $terms[1] = '+';
- }
- $searchon .= $terms[1] . $wgContLang->stripForSearch( $terms[2] );
- if( !empty( $terms[3] ) ) {
- $regexp = preg_quote( $terms[3] );
- if( $terms[4] ) $regexp .= "[0-9A-Za-z_]+";
- } else {
- $regexp = preg_quote( str_replace( '"', '', $terms[2] ) );
- }
- $this->searchTerms[] = $regexp;
- }
- wfDebug( "Would search with '$searchon'\n" );
- wfDebug( "Match with /\b" . implode( '\b|\b', $this->searchTerms ) . "\b/\n" );
- } else {
- wfDebug( "Can't understand search query '{$this->filteredText}'\n" );
- }
-
- $searchon = $this->db->strencode( $searchon );
- $this->titleCond = " MATCH(si_title) AGAINST('$searchon' IN BOOLEAN MODE)";
- $this->textCond = " (MATCH(si_text) AGAINST('$searchon' IN BOOLEAN MODE) AND cur_is_redirect=0)";
- return MW_SEARCH_OK;
+ /**
+ * Return information about how and from where the results were fetched,
+ * should be useful for diagnostics and debugging
+ *
+ * @return string
+ */
+ function getInfo() {
+ return null;
}
-
- function &getMatches( $cond, $limit, $offset = 0 ) {
- $searchindex = $this->db->tableName( 'searchindex' );
- $cur = $this->db->tableName( 'cur' );
- $searchnamespaces = $this->queryNamespaces();
- $redircond = $this->searchRedirects();
-
- $sql = "SELECT cur_id,cur_namespace,cur_title," .
- "cur_text FROM $cur,$searchindex " .
- "WHERE cur_id=si_page AND {$cond} " .
- "{$searchnamespaces} {$redircond} " .
- $this->db->limitResult( $limit, $offset );
-
- $res = $this->db->query( $sql, 'SearchEngine::getMatches' );
- $matches = array();
- while ( $row = $this->db->fetchObject( $res ) ) {
- $matches[] = $row;
- }
- $this->db->freeResult( $res );
-
- return $matches;
+
+ /**
+ * Return a result set of hits on other (multiple) wikis associated with this one
+ *
+ * @return SearchResultSet
+ */
+ function getInterwikiResults() {
+ return null;
}
+
+ /**
+ * Check if there are results on other wikis
+ *
+ * @return boolean
+ */
+ function hasInterwikiResults() {
+ return $this->getInterwikiResults() != null;
+ }
+
- function showMatches( &$matches, $offset, $msgEmpty, $msgFound ) {
- global $wgOut;
- if ( 0 == count( $matches ) ) {
- $wgOut->addHTML( "<h2>" . wfMsg( $msgEmpty ) .
- "</h2>\n" );
- return false;
- } else {
- $off = $offset + 1;
- $wgOut->addHTML( "<h2>" . wfMsg( $msgFound ) .
- "</h2>\n<ol start='{$off}'>" );
+ /**
+ * Fetches next search result, or false.
+ * @return SearchResult
+ * @access public
+ * @abstract
+ */
+ function next() {
+ return false;
+ }
- foreach( $matches as $row ) {
- $this->showHit( $row );
- }
- $wgOut->addHTML( "</ol>\n" );
- return true;
- }
+ /**
+ * Frees the result set, if applicable.
+ * @ access public
+ */
+ function free() {
+ // ...
}
+}
- function showHit( $row ) {
- global $wgUser, $wgOut, $wgContLang;
- $t = Title::makeName( $row->cur_namespace, $row->cur_title );
- if( is_null( $t ) ) {
- $wgOut->addHTML( "<!-- Broken link in search result -->\n" );
- return;
- }
- $sk = $wgUser->getSkin();
+/**
+ * @ingroup Search
+ */
+class SearchResultTooMany {
+ ## Some search engines may bail out if too many matches are found
+}
- $contextlines = $wgUser->getOption( 'contextlines' );
- if ( '' == $contextlines ) { $contextlines = 5; }
- $contextchars = $wgUser->getOption( 'contextchars' );
- if ( '' == $contextchars ) { $contextchars = 50; }
- $link = $sk->makeKnownLink( $t, '' );
- $size = wfMsg( 'nbytes', strlen( $row->cur_text ) );
- $wgOut->addHTML( "<li>{$link} ({$size})" );
+/**
+ * @fixme This class is horribly factored. It would probably be better to have
+ * a useful base class to which you pass some standard information, then let
+ * the fancy self-highlighters extend that.
+ * @ingroup Search
+ */
+class SearchResult {
+ var $mRevision = null;
+ var $mImage = null;
+
+ function __construct( $row ) {
+ $this->mTitle = Title::makeTitle( $row->page_namespace, $row->page_title );
+ if( !is_null($this->mTitle) ){
+ $this->mRevision = Revision::newFromTitle( $this->mTitle );
+ if( $this->mTitle->getNamespace() === NS_FILE )
+ $this->mImage = wfFindFile( $this->mTitle );
+ }
+ }
+
+ /**
+ * Check if this is result points to an invalid title
+ *
+ * @return boolean
+ * @access public
+ */
+ function isBrokenTitle(){
+ if( is_null($this->mTitle) )
+ return true;
+ return false;
+ }
+
+ /**
+ * Check if target page is missing, happens when index is out of date
+ *
+ * @return boolean
+ * @access public
+ */
+ function isMissingRevision(){
+ return !$this->mRevision && !$this->mImage;
+ }
- $lines = explode( "\n", $row->cur_text );
- $pat1 = "/(.*)(" . implode( "|", $this->searchTerms ) . ")(.*)/i";
- $lineno = 0;
+ /**
+ * @return Title
+ * @access public
+ */
+ function getTitle() {
+ return $this->mTitle;
+ }
- foreach ( $lines as $line ) {
- if ( 0 == $contextlines ) {
- break;
- }
- --$contextlines;
- ++$lineno;
- if ( ! preg_match( $pat1, $line, $m ) ) {
- continue;
- }
+ /**
+ * @return double or null if not supported
+ */
+ function getScore() {
+ return null;
+ }
- $pre = $wgContLang->truncate( $m[1], -$contextchars, '...' );
+ /**
+ * Lazy initialization of article text from DB
+ */
+ protected function initText(){
+ if( !isset($this->mText) ){
+ if($this->mRevision != null)
+ $this->mText = $this->mRevision->getText();
+ else // TODO: can we fetch raw wikitext for commons images?
+ $this->mText = '';
+
+ }
+ }
+
+ /**
+ * @param array $terms terms to highlight
+ * @return string highlighted text snippet, null (and not '') if not supported
+ */
+ function getTextSnippet($terms){
+ global $wgUser, $wgAdvancedSearchHighlighting;
+ $this->initText();
+ list($contextlines,$contextchars) = SearchEngine::userHighlightPrefs($wgUser);
+ $h = new SearchHighlighter();
+ if( $wgAdvancedSearchHighlighting )
+ return $h->highlightText( $this->mText, $terms, $contextlines, $contextchars );
+ else
+ return $h->highlightSimple( $this->mText, $terms, $contextlines, $contextchars );
+ }
+
+ /**
+ * @param array $terms terms to highlight
+ * @return string highlighted title, '' if not supported
+ */
+ function getTitleSnippet($terms){
+ return '';
+ }
- if ( count( $m ) < 3 ) {
- $post = '';
- } else {
- $post = $wgContLang->truncate( $m[3], $contextchars, '...' );
- }
+ /**
+ * @param array $terms terms to highlight
+ * @return string highlighted redirect name (redirect to this page), '' if none or not supported
+ */
+ function getRedirectSnippet($terms){
+ return '';
+ }
- $found = $m[2];
+ /**
+ * @return Title object for the redirect to this page, null if none or not supported
+ */
+ function getRedirectTitle(){
+ return null;
+ }
- $line = htmlspecialchars( $pre . $found . $post );
- $pat2 = '/(' . implode( '|', $this->searchTerms ) . ")/i";
- $line = preg_replace( $pat2,
- "<span class='searchmatch'>\\1</span>", $line );
+ /**
+ * @return string highlighted relevant section name, null if none or not supported
+ */
+ function getSectionSnippet(){
+ return '';
+ }
- $wgOut->addHTML( "<br /><small>{$lineno}: {$line}</small>\n" );
- }
- $wgOut->addHTML( "</li>\n" );
+ /**
+ * @return Title object (pagename+fragment) for the section, null if none or not supported
+ */
+ function getSectionTitle(){
+ return null;
}
- function getNearMatch() {
- # Exact match? No need to look further.
- $title = Title::newFromText( $this->rawText );
- if ( $title->getNamespace() == NS_SPECIAL || 0 != $title->getArticleID() ) {
- return $title;
- }
+ /**
+ * @return string timestamp
+ */
+ function getTimestamp(){
+ if( $this->mRevision )
+ return $this->mRevision->getTimestamp();
+ else if( $this->mImage )
+ return $this->mImage->getTimestamp();
+ return '';
+ }
- # Now try all lower case (i.e. first letter capitalized)
- #
- $title = Title::newFromText( strtolower( $this->rawText ) );
- if ( 0 != $title->getArticleID() ) {
- return $title;
- }
+ /**
+ * @return int number of words
+ */
+ function getWordCount(){
+ $this->initText();
+ return str_word_count( $this->mText );
+ }
- # Now try capitalized string
- #
- $title = Title::newFromText( ucwords( strtolower( $this->rawText ) ) );
- if ( 0 != $title->getArticleID() ) {
- return $title;
- }
+ /**
+ * @return int size in bytes
+ */
+ function getByteSize(){
+ $this->initText();
+ return strlen( $this->mText );
+ }
+
+ /**
+ * @return boolean if hit has related articles
+ */
+ function hasRelated(){
+ return false;
+ }
+
+ /**
+ * @return interwiki prefix of the title (return iw even if title is broken)
+ */
+ function getInterwikiPrefix(){
+ return '';
+ }
+}
- # Now try all upper case
- #
- $title = Title::newFromText( strtoupper( $this->rawText ) );
- if ( 0 != $title->getArticleID() ) {
- return $title;
+/**
+ * Highlight bits of wikitext
+ *
+ * @ingroup Search
+ */
+class SearchHighlighter {
+ var $mCleanWikitext = true;
+
+ function SearchHighlighter($cleanupWikitext = true){
+ $this->mCleanWikitext = $cleanupWikitext;
+ }
+
+ /**
+ * Default implementation of wikitext highlighting
+ *
+ * @param string $text
+ * @param array $terms Terms to highlight (unescaped)
+ * @param int $contextlines
+ * @param int $contextchars
+ * @return string
+ */
+ public function highlightText( $text, $terms, $contextlines, $contextchars ) {
+ global $wgLang, $wgContLang;
+ global $wgSearchHighlightBoundaries;
+ $fname = __METHOD__;
+
+ if($text == '')
+ return '';
+
+ // spli text into text + templates/links/tables
+ $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
+ // first capture group is for detecting nested templates/links/tables/references
+ $endPatterns = array(
+ 1 => '/(\{\{)|(\}\})/', // template
+ 2 => '/(\[\[)|(\]\])/', // image
+ 3 => "/(\n\\{\\|)|(\n\\|\\})/"); // table
+
+ // FIXME: this should prolly be a hook or something
+ if(function_exists('wfCite')){
+ $spat .= '|(<ref>)'; // references via cite extension
+ $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
}
-
- # Entering an IP address goes to the contributions page
- if ( preg_match( '/^\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}$/', $this->rawText ) ) {
- $title = Title::makeTitle( NS_SPECIAL, "Contributions/" . $this->rawText );
- return $title;
+ $spat .= '/';
+ $textExt = array(); // text extracts
+ $otherExt = array(); // other extracts
+ wfProfileIn( "$fname-split" );
+ $start = 0;
+ $textLen = strlen($text);
+ $count = 0; // sequence number to maintain ordering
+ while( $start < $textLen ){
+ // find start of template/image/table
+ if( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ){
+ $epat = '';
+ foreach($matches as $key => $val){
+ if($key > 0 && $val[1] != -1){
+ if($key == 2){
+ // see if this is an image link
+ $ns = substr($val[0],2,-1);
+ if( $wgContLang->getNsIndex($ns) != NS_FILE )
+ break;
+
+ }
+ $epat = $endPatterns[$key];
+ $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
+ $start = $val[1];
+ break;
+ }
+ }
+ if( $epat ){
+ // find end (and detect any nested elements)
+ $level = 0;
+ $offset = $start + 1;
+ $found = false;
+ while( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ){
+ if( array_key_exists(2,$endMatches) ){
+ // found end
+ if($level == 0){
+ $len = strlen($endMatches[2][0]);
+ $off = $endMatches[2][1];
+ $this->splitAndAdd( $otherExt, $count,
+ substr( $text, $start, $off + $len - $start ) );
+ $start = $off + $len;
+ $found = true;
+ break;
+ } else{
+ // end of nested element
+ $level -= 1;
+ }
+ } else{
+ // nested
+ $level += 1;
+ }
+ $offset = $endMatches[0][1] + strlen($endMatches[0][0]);
+ }
+ if( ! $found ){
+ // couldn't find appropriate closing tag, skip
+ $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen($matches[0][0]) ) );
+ $start += strlen($matches[0][0]);
+ }
+ continue;
+ }
+ }
+ // else: add as text extract
+ $this->splitAndAdd( $textExt, $count, substr($text,$start) );
+ break;
}
- return NULL;
- }
-
- function goResult() {
- global $wgOut, $wgGoToEdit;
- global $wgDisableTextSearch;
- $fname = 'SearchEngine::goResult';
+ $all = $textExt + $otherExt; // these have disjunct key sets
+
+ wfProfileOut( "$fname-split" );
- # Try to go to page as entered.
- #
- $t = Title::newFromText( $this->rawText );
-
- # If the string cannot be used to create a title
- if( is_null( $t ) ){
- $this->showResults();
- return;
+ // prepare regexps
+ foreach( $terms as $index => $term ) {
+ // manually do upper/lowercase stuff for utf-8 since PHP won't do it
+ if(preg_match('/[\x80-\xff]/', $term) ){
+ $terms[$index] = preg_replace_callback('/./us',array($this,'caseCallback'),$terms[$index]);
+ } else {
+ $terms[$index] = $term;
+ }
}
+ $anyterm = implode( '|', $terms );
+ $phrase = implode("$wgSearchHighlightBoundaries+", $terms );
+
+ // FIXME: a hack to scale contextchars, a correct solution
+ // would be to have contextchars actually be char and not byte
+ // length, and do proper utf-8 substrings and lengths everywhere,
+ // but PHP is making that very hard and unclean to implement :(
+ $scale = strlen($anyterm) / mb_strlen($anyterm);
+ $contextchars = intval( $contextchars * $scale );
+
+ $patPre = "(^|$wgSearchHighlightBoundaries)";
+ $patPost = "($wgSearchHighlightBoundaries|$)";
+
+ $pat1 = "/(".$phrase.")/ui";
+ $pat2 = "/$patPre(".$anyterm.")$patPost/ui";
+
+ wfProfileIn( "$fname-extract" );
+
+ $left = $contextlines;
- # If there's an exact or very near match, jump right there.
- $t = $this->getNearMatch();
- if( !is_null( $t ) ) {
- $wgOut->redirect( $t->getFullURL() );
- return;
- }
+ $snippets = array();
+ $offsets = array();
- # No match, generate an edit URL
- $t = Title::newFromText( $this->rawText );
+ // show beginning only if it contains all words
+ $first = 0;
+ $firstText = '';
+ foreach($textExt as $index => $line){
+ if(strlen($line)>0 && $line[0] != ';' && $line[0] != ':'){
+ $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
+ $first = $index;
+ break;
+ }
+ }
+ if( $firstText ){
+ $succ = true;
+ // check if first text contains all terms
+ foreach($terms as $term){
+ if( ! preg_match("/$patPre".$term."$patPost/ui", $firstText) ){
+ $succ = false;
+ break;
+ }
+ }
+ if( $succ ){
+ $snippets[$first] = $firstText;
+ $offsets[$first] = 0;
+ }
+ }
+ if( ! $snippets ) {
+ // match whole query on text
+ $this->process($pat1, $textExt, $left, $contextchars, $snippets, $offsets);
+ // match whole query on templates/tables/images
+ $this->process($pat1, $otherExt, $left, $contextchars, $snippets, $offsets);
+ // match any words on text
+ $this->process($pat2, $textExt, $left, $contextchars, $snippets, $offsets);
+ // match any words on templates/tables/images
+ $this->process($pat2, $otherExt, $left, $contextchars, $snippets, $offsets);
+
+ ksort($snippets);
+ }
- # If the feature is enabled, go straight to the edit page
- if ( $wgGoToEdit ) {
- $wgOut->redirect( $t->getFullURL( 'action=edit' ) );
- return;
+ // add extra chars to each snippet to make snippets constant size
+ $extended = array();
+ if( count( $snippets ) == 0){
+ // couldn't find the target words, just show beginning of article
+ $targetchars = $contextchars * $contextlines;
+ $snippets[$first] = '';
+ $offsets[$first] = 0;
+ } else{
+ // if begin of the article contains the whole phrase, show only that !!
+ if( array_key_exists($first,$snippets) && preg_match($pat1,$snippets[$first])
+ && $offsets[$first] < $contextchars * 2 ){
+ $snippets = array ($first => $snippets[$first]);
+ }
+
+ // calc by how much to extend existing snippets
+ $targetchars = intval( ($contextchars * $contextlines) / count ( $snippets ) );
+ }
+
+ foreach($snippets as $index => $line){
+ $extended[$index] = $line;
+ $len = strlen($line);
+ if( $len < $targetchars - 20 ){
+ // complete this line
+ if($len < strlen( $all[$index] )){
+ $extended[$index] = $this->extract( $all[$index], $offsets[$index], $offsets[$index]+$targetchars, $offsets[$index]);
+ $len = strlen( $extended[$index] );
+ }
+
+ // add more lines
+ $add = $index + 1;
+ while( $len < $targetchars - 20
+ && array_key_exists($add,$all)
+ && !array_key_exists($add,$snippets) ){
+ $offsets[$add] = 0;
+ $tt = "\n".$this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
+ $extended[$add] = $tt;
+ $len += strlen( $tt );
+ $add++;
+ }
+ }
}
- if( $t ) {
- $editurl = $t->escapeLocalURL( 'action=edit' );
- } else {
- $editurl = ''; # ??
+ //$snippets = array_map('htmlspecialchars', $extended);
+ $snippets = $extended;
+ $last = -1;
+ $extract = '';
+ foreach($snippets as $index => $line){
+ if($last == -1)
+ $extract .= $line; // first line
+ elseif($last+1 == $index && $offsets[$last]+strlen($snippets[$last]) >= strlen($all[$last]))
+ $extract .= " ".$line; // continous lines
+ else
+ $extract .= '<b> ... </b>' . $line;
+
+ $last = $index;
}
- $wgOut->addHTML( '<p>' . wfMsg('nogomatch', $editurl ) . "</p>\n" );
-
- # Try a fuzzy title search
- $anyhit = false;
- global $wgDisableFuzzySearch;
- if(! $wgDisableFuzzySearch ){
- foreach( array(NS_MAIN, NS_PROJECT, NS_USER, NS_IMAGE, NS_MEDIAWIKI) as $namespace){
- $anyhit |= SearchEngine::doFuzzyTitleSearch( $this->rawText, $namespace );
+ if( $extract )
+ $extract .= '<b> ... </b>';
+
+ $processed = array();
+ foreach($terms as $term){
+ if( ! isset($processed[$term]) ){
+ $pat3 = "/$patPre(".$term.")$patPost/ui"; // highlight word
+ $extract = preg_replace( $pat3,
+ "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
+ $processed[$term] = true;
}
}
- if( ! $anyhit ){
- return $this->showResults();
- }
+ wfProfileOut( "$fname-extract" );
+
+ return $extract;
}
-
+
/**
- * @static
+ * Split text into lines and add it to extracts array
+ *
+ * @param array $extracts index -> $line
+ * @param int $count
+ * @param string $text
*/
- function doFuzzyTitleSearch( $search, $namespace ){
- global $wgContLang, $wgOut;
-
- $this->setupPage();
-
- $sstr = ucfirst($search);
- $sstr = str_replace(' ', '_', $sstr);
- $fuzzymatches = SearchEngine::fuzzyTitles( $sstr, $namespace );
- $fuzzymatches = array_slice($fuzzymatches, 0, 10);
- $slen = strlen( $search );
- $wikitext = '';
- foreach($fuzzymatches as $res){
- $t = str_replace('_', ' ', $res[1]);
- $tfull = $wgContLang->getNsText( $namespace ) . ":$t|$t";
- if( $namespace == NS_MAIN )
- $tfull = $t;
- $distance = $res[0];
- $closeness = (strlen( $search ) - $distance) / strlen( $search );
- $percent = intval( $closeness * 100 ) . '%';
- $stars = str_repeat('*', ceil(5 * $closeness) );
- $wikitext .= "* [[$tfull]] $percent ($stars)\n";
- }
- if( $wikitext ){
- if( $namespace != NS_MAIN )
- $wikitext = '=== ' . $wgContLang->getNsText( $namespace ) . " ===\n" . $wikitext;
- $wgOut->addWikiText( $wikitext );
- return true;
+ function splitAndAdd(&$extracts, &$count, $text){
+ $split = explode( "\n", $this->mCleanWikitext? $this->removeWiki($text) : $text );
+ foreach($split as $line){
+ $tt = trim($line);
+ if( $tt )
+ $extracts[$count++] = $tt;
}
- return false;
}
-
+
/**
- * @static
+ * Do manual case conversion for non-ascii chars
+ *
+ * @param unknown_type $matches
*/
- function fuzzyTitles( $sstr, $namespace = NS_MAIN ){
- $span = 0.10; // weed on title length before doing levenshtein.
- $tolerance = 0.35; // allowed percentage of erronous characters
- $slen = strlen($sstr);
- $tolerance_count = ceil($tolerance * $slen);
- $spanabs = ceil($slen * (1 + $span)) - $slen;
- # print "Word: $sstr, len = $slen, range = [$min, $max], tolerance_count = $tolerance_count<BR>\n";
- $result = array();
- $cnt = 0;
- for( $i=0; $i <= $spanabs; $i++ ){
- $titles = SearchEngine::getTitlesByLength( $slen + $i, $namespace );
- if( $i != 0) {
- $titles = array_merge($titles, SearchEngine::getTitlesByLength( $slen - $i, $namespace ) );
- }
- foreach($titles as $t){
- $d = levenshtein($sstr, $t);
- if($d < $tolerance_count)
- $result[] = array($d, $t);
- $cnt++;
- }
- }
- usort($result, 'SearchEngine_pcmp');
- return $result;
+ function caseCallback($matches){
+ global $wgContLang;
+ if( strlen($matches[0]) > 1 ){
+ return '['.$wgContLang->lc($matches[0]).$wgContLang->uc($matches[0]).']';
+ } else
+ return $matches[0];
}
-
+
/**
- * static
+ * Extract part of the text from start to end, but by
+ * not chopping up words
+ * @param string $text
+ * @param int $start
+ * @param int $end
+ * @param int $posStart (out) actual start position
+ * @param int $posEnd (out) actual end position
+ * @return string
*/
- function getTitlesByLength($aLength, $aNamespace = 0){
- global $wgMemc, $wgDBname;
- $fname = 'SearchEngin::getTitlesByLength';
+ function extract($text, $start, $end, &$posStart = null, &$posEnd = null ){
+ global $wgContLang;
- // to avoid multiple costly SELECTs in case of no memcached
- if( $this->allTitles ){
- if( isset( $this->allTitles[$aLength][$aNamespace] ) ){
- return $this->allTitles[$aLength][$aNamespace];
- } else {
- return array();
+ if( $start != 0)
+ $start = $this->position( $text, $start, 1 );
+ if( $end >= strlen($text) )
+ $end = strlen($text);
+ else
+ $end = $this->position( $text, $end );
+
+ if(!is_null($posStart))
+ $posStart = $start;
+ if(!is_null($posEnd))
+ $posEnd = $end;
+
+ if($end > $start)
+ return substr($text, $start, $end-$start);
+ else
+ return '';
+ }
+
+ /**
+ * Find a nonletter near a point (index) in the text
+ *
+ * @param string $text
+ * @param int $point
+ * @param int $offset to found index
+ * @return int nearest nonletter index, or beginning of utf8 char if none
+ */
+ function position($text, $point, $offset=0 ){
+ $tolerance = 10;
+ $s = max( 0, $point - $tolerance );
+ $l = min( strlen($text), $point + $tolerance ) - $s;
+ $m = array();
+ if( preg_match('/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/', substr($text,$s,$l), $m, PREG_OFFSET_CAPTURE ) ){
+ return $m[0][1] + $s + $offset;
+ } else{
+ // check if point is on a valid first UTF8 char
+ $char = ord( $text[$point] );
+ while( $char >= 0x80 && $char < 0xc0 ) {
+ // skip trailing bytes
+ $point++;
+ if($point >= strlen($text))
+ return strlen($text);
+ $char = ord( $text[$point] );
}
+ return $point;
+
}
-
- $mkey = "$wgDBname:titlesbylength:$aLength:$aNamespace";
- $mkeyts = "$wgDBname:titlesbylength:createtime";
- $ts = $wgMemc->get( $mkeyts );
- $result = $wgMemc->get( $mkey );
-
- if( time() - $ts < 3600 ){
- // note: in case of insufficient memcached space, we return
- // an empty list instead of starting to hit the DB.
- return is_array( $result ) ? $result : array();
+ }
+
+ /**
+ * Search extracts for a pattern, and return snippets
+ *
+ * @param string $pattern regexp for matching lines
+ * @param array $extracts extracts to search
+ * @param int $linesleft number of extracts to make
+ * @param int $contextchars length of snippet
+ * @param array $out map for highlighted snippets
+ * @param array $offsets map of starting points of snippets
+ * @protected
+ */
+ function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ){
+ if($linesleft == 0)
+ return; // nothing to do
+ foreach($extracts as $index => $line){
+ if( array_key_exists($index,$out) )
+ continue; // this line already highlighted
+
+ $m = array();
+ if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) )
+ continue;
+
+ $offset = $m[0][1];
+ $len = strlen($m[0][0]);
+ if($offset + $len < $contextchars)
+ $begin = 0;
+ elseif( $len > $contextchars)
+ $begin = $offset;
+ else
+ $begin = $offset + intval( ($len - $contextchars) / 2 );
+
+ $end = $begin + $contextchars;
+
+ $posBegin = $begin;
+ // basic snippet from this line
+ $out[$index] = $this->extract($line,$begin,$end,$posBegin);
+ $offsets[$index] = $posBegin;
+ $linesleft--;
+ if($linesleft == 0)
+ return;
}
-
- $wgMemc->set( $mkeyts, time() );
+ }
+
+ /**
+ * Basic wikitext removal
+ * @protected
+ */
+ function removeWiki($text) {
+ $fname = __METHOD__;
+ wfProfileIn( $fname );
- $res = $this->db->select( 'cur', array( 'cur_title', 'cur_namespace' ), false, $fname );
- $titles = array(); // length, ns, [titles]
- while( $obj = $this->db->fetchObject( $res ) ){
- $title = $obj->cur_title;
- $ns = $obj->cur_namespace;
- $len = strlen( $title );
- $titles[$len][$ns][] = $title;
- }
- foreach($titles as $length => $length_arr){
- foreach($length_arr as $ns => $title_arr){
- $mkey = "$wgDBname:titlesbylength:$length:$ns";
- $wgMemc->set( $mkey, $title_arr, 3600 * 24 );
- }
- }
- $this->allTitles = $titles;
- if( isset( $titles[$aLength][$aNamespace] ) )
- return $titles[$aLength][$aNamespace];
+ //$text = preg_replace("/'{2,5}/", "", $text);
+ //$text = preg_replace("/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text);
+ //$text = preg_replace("/\[\[([^]|]+)\]\]/", "\\1", $text);
+ //$text = preg_replace("/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text);
+ //$text = preg_replace("/\\{\\|(.*?)\\|\\}/", "", $text);
+ //$text = preg_replace("/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text);
+ $text = preg_replace("/\\{\\{([^|]+?)\\}\\}/", "", $text);
+ $text = preg_replace("/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text);
+ $text = preg_replace("/\\[\\[([^|]+?)\\]\\]/", "\\1", $text);
+ $text = preg_replace_callback("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", array($this,'linkReplace'), $text);
+ //$text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
+ $text = preg_replace("/<\/?[^>]+>/", "", $text);
+ $text = preg_replace("/'''''/", "", $text);
+ $text = preg_replace("/('''|<\/?[iIuUbB]>)/", "", $text);
+ $text = preg_replace("/''/", "", $text);
+
+ wfProfileOut( $fname );
+ return $text;
+ }
+
+ /**
+ * callback to replace [[target|caption]] kind of links, if
+ * the target is category or image, leave it
+ *
+ * @param array $matches
+ */
+ function linkReplace($matches){
+ $colon = strpos( $matches[1], ':' );
+ if( $colon === false )
+ return $matches[2]; // replace with caption
+ global $wgContLang;
+ $ns = substr( $matches[1], 0, $colon );
+ $index = $wgContLang->getNsIndex($ns);
+ if( $index !== false && ($index == NS_FILE || $index == NS_CATEGORY) )
+ return $matches[0]; // return the whole thing
else
- return array();
+ return $matches[2];
+
}
+
+ /**
+ * Simple & fast snippet extraction, but gives completely unrelevant
+ * snippets
+ *
+ * @param string $text
+ * @param array $terms
+ * @param int $contextlines
+ * @param int $contextchars
+ * @return string
+ */
+ public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
+ global $wgLang, $wgContLang;
+ $fname = __METHOD__;
+
+ $lines = explode( "\n", $text );
+
+ $terms = implode( '|', $terms );
+ $max = intval( $contextchars ) + 1;
+ $pat1 = "/(.*)($terms)(.{0,$max})/i";
+
+ $lineno = 0;
+
+ $extract = "";
+ wfProfileIn( "$fname-extract" );
+ foreach ( $lines as $line ) {
+ if ( 0 == $contextlines ) {
+ break;
+ }
+ ++$lineno;
+ $m = array();
+ if ( ! preg_match( $pat1, $line, $m ) ) {
+ continue;
+ }
+ --$contextlines;
+ $pre = $wgContLang->truncate( $m[1], -$contextchars );
+
+ if ( count( $m ) < 3 ) {
+ $post = '';
+ } else {
+ $post = $wgContLang->truncate( $m[3], $contextchars );
+ }
+
+ $found = $m[2];
+
+ $line = htmlspecialchars( $pre . $found . $post );
+ $pat2 = '/(' . $terms . ")/i";
+ $line = preg_replace( $pat2,
+ "<span class='searchmatch'>\\1</span>", $line );
+
+ $extract .= "${line}\n";
+ }
+ wfProfileOut( "$fname-extract" );
+
+ return $extract;
+ }
+
}
/**
- * @access private
- * @static
+ * Dummy class to be used when non-supported Database engine is present.
+ * @fixme Dummy class should probably try something at least mildly useful,
+ * such as a LIKE search through titles.
+ * @ingroup Search
*/
-function SearchEngine_pcmp($a, $b){ return $a[0] - $b[0]; }
-
-?>
+class SearchEngineDummy extends SearchEngine {
+ // no-op
+}
dépôts / lhc / web / wiklou.git / blobdiff