includes/SearchUpdate.php

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