includes/search/SearchUpdate.php

   1 <?php
   2 /**
   3  * Search index updater
   4  *
   5  * See deferred.txt
   6  *
   7  * @file
   8  * @ingroup Search
   9  */
  10
  11 /**
  12  * Database independant search index updater
  13  *
  14  * @ingroup Search
  15  */
  16 class SearchUpdate implements DeferrableUpdate {
  17
  18         private $mId = 0, $mNamespace, $mTitle, $mText;
  19         private $mTitleWords;
  20
  21         function __construct( $id, $title, $text = false ) {
  22                 $nt = Title::newFromText( $title );
  23                 if( $nt ) {
  24                         $this->mId = $id;
  25                         $this->mText = $text;
  26
  27                         $this->mNamespace = $nt->getNamespace();
  28                         $this->mTitle = $nt->getText(); # Discard namespace
  29
  30                         $this->mTitleWords = $this->mTextWords = array();
  31                 } else {
  32                         wfDebug( "SearchUpdate object created with invalid title '$title'\n" );
  33                 }
  34         }
  35
  36         function doUpdate() {
  37                 global $wgContLang, $wgDisableSearchUpdate;
  38
  39                 if( $wgDisableSearchUpdate || !$this->mId ) {
  40                         return;
  41                 }
  42
  43                 wfProfileIn( __METHOD__ );
  44
  45                 $search = SearchEngine::create();
  46                 $lc = SearchEngine::legalSearchChars() . '&#;';
  47
  48                 if( $this->mText === false ) {
  49                         $search->updateTitle($this->mId,
  50                                 $search->normalizeText( Title::indexTitle( $this->mNamespace, $this->mTitle ) ) );
  51                         wfProfileOut( __METHOD__ );
  52                         return;
  53                 }
  54
  55                 # Language-specific strip/conversion
  56                 $text = $wgContLang->normalizeForSearch( $this->mText );
  57
  58                 wfProfileIn( __METHOD__ . '-regexps' );
  59                 $text = preg_replace( "/<\\/?\\s*[A-Za-z][^>]*?>/",
  60                         ' ', $wgContLang->lc( " " . $text . " " ) ); # Strip HTML markup
  61                 $text = preg_replace( "/(^|\\n)==\\s*([^\\n]+)\\s*==(\\s)/sD",
  62                   "\\1\\2 \\2 \\2\\3", $text ); # Emphasize headings
  63
  64                 # Strip external URLs
  65                 $uc = "A-Za-z0-9_\\/:.,~%\\-+&;#?!=()@\\x80-\\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( __METHOD__ . '-regexps' );
 106
 107                 wfRunHooks( 'SearchUpdate', array( $this->mId, $this->mNamespace, $this->mTitle, &$text ) );
 108
 109                 # Perform the actual update
 110                 $search->update($this->mId, $search->normalizeText( Title::indexTitle( $this->mNamespace, $this->mTitle ) ),
 111                                 $search->normalizeText( $text ) );
 112
 113                 wfProfileOut( __METHOD__ );
 114         }
 115 }
 116
 117 /**
 118  * Placeholder class
 119  *
 120  * @ingroup Search
 121  */
 122 class SearchUpdateMyISAM extends SearchUpdate {
 123         # Inherits everything
 124 }