advanced code snippet search

We Recommend

Wicked Cool PHP: Real-World Scripts That Solve Difficult Problems Wicked Cool PHP: Real-World Scripts That Solve Difficult Problems
Wicked Cool PHP contains a wide variety of scripts to process credit cards, check the validity of email addresses, template HTML, and serve dynamic images and text.

Posted By

daisuke103 on 03/28/08


Tagged

php textmate str utf-8 encord shft jis


Versions (?)


PHP 文字コード判定

Published in: PHP 

Plain TextExpand 
  1. function detect_encoding_ja( $str )
  2. {
  3.     $enc = @mb_detect_encoding( $str, 'ASCII,JIS,eucJP-win,SJIS-win,UTF-8' );
  4.  
  5.     switch ( $enc ) {
  6.     case FALSE   :
  7.     case 'ASCII' : 
  8.     case 'JIS'   : 
  9.     case 'UTF-8' : break;
  10.     case 'eucJP-win' :
  11.         // ここで eucJP-win を検出した場合、eucJP-win として判定
  12.         if ( @mb_detect_encoding( $str, 'SJIS-win,UTF-8,eucJP-win' ) === 'eucJP-win' ) {
  13.             break;
  14.         }
  15.         $_hint = "¥xbf¥xfd" . $str; // "¥xbf¥xfd" : EUC-JP "雀"
  16.  
  17.         // EUC-JP -> UTF-8 変換時にマッピングが変更される文字を削除( ≒ ≡ ∫ など)
  18.         mb_regex_encoding( 'EUC-JP' );
  19.         $_hint = mb_ereg_replace( "¥xad(?:¥xe2|¥xf5|¥xf6|¥xf7|¥xfa|¥xfb|¥xfc|¥xf0|¥xf1|¥xf2)"
  20.  
  21.         $_tmp  = mb_convert_encoding( $_hint, 'UTF-8', 'eucJP-win' );
  22.         $_tmp2 = mb_convert_encoding( $_tmp,  'eucJP-win', 'UTF-8' );
  23.         if ( $_tmp2 === $_hint ) {
  24.  
  25.             // 例外処理( EUC-JP 以外と認識する範囲 )
  26.             if (
  27.                 // SJIS と重なる範囲(2バイト|3バイト|iモード絵文字|1バイト文字)
  28.                 ! preg_match( '/^(?:'
  29.                     . '[¥x8E¥xE0-¥xE9][¥x80-¥xFC]|¥xEA[¥x80-¥xA4]|'
  30.                     . '¥x8F[¥xB0-¥xEF][¥xE0-¥xEF][¥x40-¥x7F]|'
  31.                     . '¥xF8[¥x9F-¥xFC]|¥xF9[¥x40-¥x49¥x50-¥x52¥x55-¥x57¥x5B-¥x5E¥x72-¥x7E¥x80
  32.                     . '[¥x00-¥x7E]'
  33.                     . ')+$/', $str ) && 
  34.  
  35.                 // UTF-8 と重なる範囲(全角英数字|漢字|1バイト文字)
  36.                 ! preg_match( '/^(?:'
  37.                     . '¥xEF¥xBC[¥xA1-¥xBA]|[¥x00-¥x7E]|'
  38.                     . '[¥xE4-¥xE9][¥x8E-¥x8F¥xA1-¥xBF][¥x8F¥xA0-¥xEF]|'
  39.                     . '[¥x00-¥x7E]'
  40.                     . ')+$/', $str )
  41.             ) {
  42.                 // 条件式の範囲に入らなかった場合は、eucJP-win として検出
  43.                 break;
  44.             }
  45.             // 例外処理2(一部の頻度の多そうな熟語は eucJP-win として判定)
  46.             // (珈琲|琥珀|瑪瑙|癇癪|碼碯|耄碌|膀胱|蒟蒻|薔薇|蜻蛉)
  47.             if ( mb_ereg( '^(?:'
  48.                 . '¥xE0¥xDD¥xE0¥xEA|¥xE0¥xE8¥xE0¥xE1|¥xE0¥xF5¥xE0¥xEF|¥xE1¥xF2¥xE1¥xFB|'
  49.                 . '¥xE2¥xFB¥xE2¥xF5|¥xE6¥xCE¥xE2¥xF1|¥xE7¥xAF¥xE6¥xF9|¥xE8¥xE7¥xE8¥xEA|'
  50.                 . '¥xE9¥xAC¥xE9¥xAF|¥xE9¥xF1¥xE9¥xD9|[¥x00-¥x7E]'
  51.                 . ')+$', $str )
  52.             ) {
  53.                 break;
  54.             }
  55.         }
  56.  
  57.     default :
  58.         // ここで SJIS-win と判断された場合は、文字コードは SJIS-win として判定
  59.         $enc = @mb_detect_encoding( $str, 'UTF-8,SJIS-win' );
  60.         if ( $enc === 'SJIS-win' ) {
  61.             break;
  62.         }
  63.         // デフォルトとして SJIS-win を設定
  64.         $enc   = 'SJIS-win';
  65.  
  66.         $_hint = "¥xe9¥x9b¥x80" . $str; // "¥xe9¥x9b¥x80" : UTF-8 "雀"
  67.  
  68.         // 変換時にマッピングが変更される文字を調整
  69.         mb_regex_encoding( 'UTF-8' );
  70.         $_hint = mb_ereg_replace( "¥xe3¥x80¥x9c", "¥xef¥xbd¥x9e", $_hint );
  71.         $_hint = mb_ereg_replace( "¥xe2¥x88¥x92", "¥xe3¥x83¥xbc", $_hint );
  72.         $_hint = mb_ereg_replace( "¥xe2¥x80¥x96", "¥xe2¥x88¥xa5", $_hint );
  73.  
  74.         $_tmp  = mb_convert_encoding( $_hint, 'SJIS-win', 'UTF-8' );
  75.         $_tmp2 = mb_convert_encoding( $_tmp,  'UTF-8', 'SJIS-win' );
  76.  
  77.         if ( $_tmp2 === $_hint ) {
  78.             $enc = 'UTF-8';
  79.         }
  80.         // UTF-8 と SJIS 2文字が重なる範囲への対処(SJIS を優先)
  81.         if ( preg_match( '/^(?:[¥xE4-¥xE9][¥x80-¥xBF][¥x80-¥x9F][¥x00-¥x7F])+/', $str ) ) {
  82.             $enc = 'SJIS-win';
  83.         }
  84.     }
  85.     return $enc;
  86. }
  87.  

Report this snippet 

You need to login to post a comment.