includes/SearchUpdate.php

   1 <?php
   2 # $Id$
   3 /**
   4  * See deferred.doc
   5  * @package MediaWiki
   6  */
   7
   8 /**
   9  *
  10  * @package MediaWiki
  11  */
  12 class SearchUpdate {
  13
  14         /* private */ var $mId = 0, $mNamespace, $mTitle, $mText;
  15         /* private */ var $mTitleWords;
  16
  17         function SearchUpdate( $id, $title, $text = false ) {
  18                 $nt = Title::newFromText( $title );
  19                 if( $nt ) {
  20                         $this->mId = $id;
  21                         $this->mText = $text;
  22
  23                         $this->mNamespace = $nt->getNamespace();
  24                         $this->mTitle = $nt->getText(); # Discard namespace
  25
  26                         $this->mTitleWords = $this->mTextWords = array();
  27                 } else {
  28                         wfDebug( "SearchUpdate object created with invalid title '$title'\n" );
  29                 }
  30         }
  31
  32         function doUpdate() {
  33                 global $wgDBminWordLen, $wgLang, $wgDisableSearchUpdate;
  34
  35                 if( $wgDisableSearchUpdate || !$this->mId ) {
  36                         return false;
  37                 }
  38                 $fname = 'SearchUpdate::doUpdate';
  39                 wfProfileIn( $fname );
  40
  41                 require_once( 'SearchEngine.php' );
  42                 $lc = SearchEngine::legalSearchChars() . '&#;';
  43                 $db =& wfGetDB( DB_MASTER );
  44                 $searchindex = $db->tableName( 'searchindex' );
  45
  46                 if( $this->mText == false ) {
  47                         # Just update the title
  48                         $lowpri = $db->lowPriorityOption();
  49                         $sql = "UPDATE $lowpri $searchindex SET si_title='" .
  50                           $db->strencode( Title::indexTitle( $this->mNamespace, $this->mTitle ) ) .
  51                           "' WHERE si_page={$this->mId}";
  52                         $db->query( $sql, "SearchUpdate::doUpdate" );
  53                         wfProfileOut( $fname );
  54                         return;
  55                 }
  56
  57                 # Language-specific strip/conversion
  58                 $text = $wgLang->stripForSearch( $this->mText );
  59
  60                 wfProfileIn( $fname.'-regexps' );
  61                 $text = preg_replace( "/<\\/?\\s*[A-Za-z][A-Za-z0-9]*\\s*([^>]*?)>/",
  62                   ' ', strtolower( " " . $text /*$this->mText*/ . " " ) ); # Strip HTML markup
  63                 $text = preg_replace( "/(^|\\n)\\s*==\\s+([^\\n]+)\\s+==\\s/sD",
  64                   "\\2 \\2 \\2 ", $text ); # Emphasize headings
  65
  66                 # Strip external URLs
  67                 $uc = "A-Za-z0-9_\\/:.,~%\\-+&;#?!=()@\\xA0-\\xFF";
  68                 $protos = "http|https|ftp|mailto|news|gopher";
  69                 $pat = "/(^|[^\\[])({$protos}):[{$uc}]+([^{$uc}]|$)/";
  70                 $text = preg_replace( $pat, "\\1 \\3", $text );
  71
  72                 $p1 = "/([^\\[])\\[({$protos}):[{$uc}]+]/";
  73                 $p2 = "/([^\\[])\\[({$protos}):[{$uc}]+\\s+([^\\]]+)]/";
  74                 $text = preg_replace( $p1, "\\1 ", $text );
  75                 $text = preg_replace( $p2, "\\1 \\3 ", $text );
  76
  77                 # Internal image links
  78                 $pat2 = "/\\[\\[image:([{$uc}]+)\\.(gif|png|jpg|jpeg)([^{$uc}])/i";
  79                 $text = preg_replace( $pat2, " \\1 \\3", $text );
  80
  81                 $text = preg_replace( "/([^{$lc}])([{$lc}]+)]]([a-z]+)/",
  82                   "\\1\\2 \\2\\3", $text ); # Handle [[game]]s
  83
  84                 # Strip all remaining non-search characters
  85                 $text = preg_replace( "/[^{$lc}]+/", " ", $text );
  86
  87                 # Handle 's, s'
  88                 #
  89                 #   $text = preg_replace( "/([{$lc}]+)'s /", "\\1 \\1's ", $text );
  90                 #   $text = preg_replace( "/([{$lc}]+)s' /", "\\1s ", $text );
  91                 #
  92                 # These tail-anchored regexps are insanely slow. The worst case comes
  93                 # when Japanese or Chinese text (ie, no word spacing) is written on
  94                 # a wiki configured for Western UTF-8 mode. The Unicode characters are
  95                 # expanded to hex codes and the "words" are very long paragraph-length
  96                 # monstrosities. On a large page the above regexps may take over 20
  97                 # seconds *each* on a 1GHz-level processor.
  98                 #
  99                 # Following are reversed versions which are consistently fast
 100                 # (about 3 milliseconds on 1GHz-level processor).
 101                 #
 102                 $text = strrev( preg_replace( "/ s'([{$lc}]+)/", " s'\\1 \\1", strrev( $text ) ) );
 103                 $text = strrev( preg_replace( "/ 's([{$lc}]+)/", " s\\1", strrev( $text ) ) );
 104
 105                 # Strip wiki '' and '''
 106                 $text = preg_replace( "/''[']*/", " ", $text );
 107                 wfProfileOut( "$fname-regexps" );
 108                 $db->replace( $searchindex, array(array('si_page')),
 109                   array(
 110                     'si_page' => $this->mId,
 111                     'si_title' => $db->strencode( Title::indexTitle( $this->mNamespace, $this->mTitle ) ),
 112                     'si_text' => $db->strencode( $text )
 113                   ), 'SearchUpdate::doUpdate' );
 114                 wfProfileOut( $fname );
 115         }
 116 }
 117
 118 ?>