includes/SearchUpdate.php

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