Note incorrect docs in DefaultSettings.php
[lhc/web/wiklou.git] / includes / SearchPostgres.php
index b5ee174..fa9d842 100644 (file)
 # http://www.gnu.org/copyleft/gpl.html
 
 /**
- * Search engine hook base class for Postgres
- * @addtogroup Search
+ * @file
+ * @ingroup Search
  */
 
+/**
+ * Search engine hook base class for Postgres
+ * @ingroup Search
+ */
 class SearchPostgres extends SearchEngine {
 
-       function SearchPostgres( $db ) {
+       function __construct( $db ) {
                $this->db = $db;
        }
 
        /**
         * Perform a full text search query via tsearch2 and return a result set.
-        * Currently searches a page's current title (page.page_title) and 
+        * Currently searches a page's current title (page.page_title) and
         * latest revision article text (pagecontent.old_text)
         *
         * @param string $term - Raw search term
@@ -38,21 +42,35 @@ class SearchPostgres extends SearchEngine {
         * @access public
         */
        function searchTitle( $term ) {
-               $resultSet = $this->db->resultObject( $this->db->query( $this->searchQuery( $term , 'titlevector', 'page_title' )));
+               $q = $this->searchQuery( $term , 'titlevector', 'page_title' );
+               $olderror = error_reporting(E_ERROR);
+               $resultSet = $this->db->resultObject( $this->db->query( $q, 'SearchPostgres', true ) );
+               error_reporting($olderror);
+               if (!$resultSet) {
+                       // Needed for "Query requires full scan, GIN doesn't support it"
+                       return new SearchResultTooMany();
+               }
                return new PostgresSearchResultSet( $resultSet, $this->searchTerms );
        }
        function searchText( $term ) {
-               $resultSet = $this->db->resultObject( $this->db->query( $this->searchQuery( $term, 'textvector', 'old_text' )));
+               $q = $this->searchQuery( $term, 'textvector', 'old_text' );
+               $olderror = error_reporting(E_ERROR);
+               $resultSet = $this->db->resultObject( $this->db->query( $q, 'SearchPostgres', true ) );
+               error_reporting($olderror);
+               if (!$resultSet) {
+                       return new SearchResultTooMany();
+               }
                return new PostgresSearchResultSet( $resultSet, $this->searchTerms );
        }
 
 
        /*
         * Transform the user's search string into a better form for tsearch2
+        * Returns an SQL fragment consisting of quoted text to search for.
        */
        function parseQuery( $term ) {
 
-               wfDebug( "parseQuery received: $term" );
+               wfDebug( "parseQuery received: $term \n" );
 
                ## No backslashes allowed
                $term = preg_replace('/\\\/', '', $term);
@@ -64,9 +82,8 @@ class SearchPostgres extends SearchEngine {
                ## Treat colons as word separators:
                $term = preg_replace('/:/', ' ', $term);
 
-               $this->searchTerms = array();
-               $m = array();
                $searchstring = '';
+               $m = array();
                if( preg_match_all('/([-!]?)(\S+)\s*/', $term, $m, PREG_SET_ORDER ) ) {
                        foreach( $m as $terms ) {
                                if (strlen($terms[1])) {
@@ -83,8 +100,6 @@ class SearchPostgres extends SearchEngine {
                                }
                                else {
                                        $searchstring .= " & $terms[2]";
-                                       $safeterm = preg_replace('/\W+/', '', $terms[2]);
-                                       $this->searchTerms[$safeterm] = $safeterm;
                                }
                        }
                }
@@ -98,16 +113,16 @@ class SearchPostgres extends SearchEngine {
                ## Remove any non-spaced operators (e.g. "Zounds!")
                $searchstring = preg_replace('/([^ ])[\!\&\|]/', "$1", $searchstring);
 
-               ## Remove any trailing whitespace
-               $searchstring = preg_replace('/\s+$/', '', $searchstring);
+               ## Remove any trailing whitespace or operators
+               $searchstring = preg_replace('/[\s\!\&\|]+$/', '', $searchstring);
 
-               ## Remove any trailing operators
-               $searchstring = preg_replace('/(?: [\!\&\|])*$/', '', $searchstring);
+               ## Remove unnecessary quotes around everything
+               $searchstring = preg_replace('/^[\'"](.*)[\'"]$/', "$1", $searchstring);
 
                ## Quote the whole thing
                $searchstring = $this->db->addQuotes($searchstring);
 
-               wfDebug( "parseQuery returned: $searchstring" );
+               wfDebug( "parseQuery returned: $searchstring \n" );
 
                return $searchstring;
 
@@ -120,11 +135,19 @@ class SearchPostgres extends SearchEngine {
         * @private
         */
        function searchQuery( $term, $fulltext, $colname ) {
+               global $wgDBversion;
 
+               if ( !isset( $wgDBversion ) ) {
+                       $this->db->getServerVersion();
+                       $wgDBversion = $this->db->numeric_version;
+               }
+               $prefix = $wgDBversion < 8.3 ? "'default'," : '';
+
+               # Get the SQL fragment for the given term
                $searchstring = $this->parseQuery( $term );
 
                ## We need a separate query here so gin does not complain about empty searches
-               $SQL = "SELECT to_tsquery('default',$searchstring)";
+               $SQL = "SELECT to_tsquery($prefix $searchstring)";
                $res = $this->db->doQuery($SQL);
                if (!$res) {
                        ## TODO: Better output (example to catch: one 'two)
@@ -138,29 +161,40 @@ class SearchPostgres extends SearchEngine {
                                "AND r.rev_text_id = c.old_id AND 1=0";
                }
                else {
+                       $m = array();
+                       if( preg_match_all("/'([^']+)'/", $top, $m, PREG_SET_ORDER ) ) {
+                               foreach( $m as $terms ) {
+                                       $this->searchTerms[$terms[1]] = $terms[1];
+                               }
+                       }
+
+                       $rankscore = $wgDBversion > 8.2 ? 5 : 1;
+                       $rank = $wgDBversion < 8.3 ? 'rank' : 'ts_rank';
                        $query = "SELECT page_id, page_namespace, page_title, ".
-                       "rank($fulltext, to_tsquery('default',$searchstring),5) AS score ".
+                       "$rank($fulltext, to_tsquery($prefix $searchstring), $rankscore) AS score ".
                        "FROM page p, revision r, pagecontent c WHERE p.page_latest = r.rev_id " .
-                       "AND r.rev_text_id = c.old_id AND $fulltext @@ to_tsquery('default',$searchstring)";
+                       "AND r.rev_text_id = c.old_id AND $fulltext @@ to_tsquery($prefix $searchstring)";
                }
 
                ## Redirects
                if (! $this->showRedirects)
-                       $query .= ' AND page_is_redirect = 0'; ## IS FALSE
+                       $query .= ' AND page_is_redirect = 0';
 
                ## Namespaces - defaults to 0
-               if ( count($this->namespaces) < 1)
-                       $query .= ' AND page_namespace = 0';
-               else {
-                       $namespaces = implode( ',', $this->namespaces );
-                       $query .= " AND page_namespace IN ($namespaces)";
+               if( !is_null($this->namespaces) ){ // null -> search all
+                       if ( count($this->namespaces) < 1)
+                               $query .= ' AND page_namespace = 0';
+                       else {
+                               $namespaces = $this->db->makeList( $this->namespaces );
+                               $query .= " AND page_namespace IN ($namespaces)";
+                       }
                }
 
                $query .= " ORDER BY score DESC, page_id DESC";
 
                $query .= $this->db->limitResult( '', $this->limit, $this->offset );
 
-               wfDebug( "searchQuery returned: $query" );
+               wfDebug( "searchQuery returned: $query \n" );
 
                return $query;
        }
@@ -169,9 +203,9 @@ class SearchPostgres extends SearchEngine {
 
        function update( $pageid, $title, $text ) {
                ## We don't want to index older revisions
-               $SQL = "UPDATE pagecontent SET textvector = NULL WHERE old_id = ".
-                               "(SELECT rev_text_id FROM revision WHERE rev_page = $pageid ".
-                               "ORDER BY rev_text_id DESC LIMIT 1 OFFSET 1)";
+               $SQL = "UPDATE pagecontent SET textvector = NULL WHERE old_id IN ".
+                               "(SELECT rev_text_id FROM revision WHERE rev_page = " . intval( $pageid ) . 
+                               " ORDER BY rev_text_id DESC OFFSET 1)";
                $this->db->doQuery($SQL);
                return true;
        }
@@ -182,10 +216,12 @@ class SearchPostgres extends SearchEngine {
 
 } ## end of the SearchPostgres class
 
-
+/**
+ * @ingroup Search
+ */
 class PostgresSearchResult extends SearchResult {
-       function PostgresSearchResult( $row ) {
-               $this->mTitle = Title::makeTitle( $row->page_namespace, $row->page_title );
+       function __construct( $row ) {
+               parent::__construct($row);
                $this->score = $row->score;
        }
        function getScore() {
@@ -193,8 +229,11 @@ class PostgresSearchResult extends SearchResult {
        }
 }
 
+/**
+ * @ingroup Search
+ */
 class PostgresSearchResultSet extends SearchResultSet {
-       function PostgresSearchResultSet( $resultSet, $terms ) {
+       function __construct( $resultSet, $terms ) {
                $this->mResultSet = $resultSet;
                $this->mTerms = $terms;
        }
@@ -216,6 +255,3 @@ class PostgresSearchResultSet extends SearchResultSet {
                }
        }
 }
-
-
-?>