How do I detect if have to apply UTF-8 decode or encode on a string?

I can't say I can rely on mb_detect_encoding(). I had some freaky false positives a while back.

The most universal way I found to work well in every case was:

if (preg_match('!!u', $string))
{
   // This is UTF-8
}
else
{
   // Definitely not UTF-8
}

function str_to_utf8 ($str) {
    $decoded = utf8_decode($str);
    if (mb_detect_encoding($decoded , 'UTF-8', true) === false)
        return $str;
    return $decoded;
}

var_dump(str_to_utf8("« Chrétiens d'Orient » : la RATP fait marche arrière"));
//string '« Chrétiens d'Orient » : la RATP fait marche arrière' (length=56)
var_dump(str_to_utf8("« Chrétiens d'Orient » : la RATP fait marche arrière"));
//string '« Chrétiens d'Orient » : la RATP fait marche arrière' (length=56)

You can use

• mb_detect_encoding — Detect character encoding

The character set might also be available in the HTTP response headers or in the response data itself.

Example:

var_dump(
    mb_detect_encoding(
        file_get_contents('http://stackoverflow.com/questions/4407854')
    ),
    $http_response_header
);

Output (codepad):

string(5) "UTF-8"
array(9) {
  [0]=>
  string(15) "HTTP/1.1 200 OK"
  [1]=>
  string(33) "Cache-Control: public, max-age=11"
  [2]=>
  string(38) "Content-Type: text/html; charset=utf-8"
  [3]=>
  string(38) "Expires: Fri, 10 Dec 2010 10:40:07 GMT"
  [4]=>
  string(44) "Last-Modified: Fri, 10 Dec 2010 10:39:07 GMT"
  [5]=>
  string(7) "Vary: *"
  [6]=>
  string(35) "Date: Fri, 10 Dec 2010 10:39:55 GMT"
  [7]=>
  string(17) "Connection: close"
  [8]=>
  string(21) "Content-Length: 34119"
}