includes/SearchUpdate.php

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