Why would json_encode return an empty string

I have a simple php structure with 3 nested arrays.

I do not use particular objects and I build myself the arrays with 2 nested loops.

Here is a sample of the var_dump of the array I want to convert to Json.

array (size=2)
  'tram B' => 
    array (size=2)
      0 => 
        array (size=3)
          'name' => string 'Ile Verte' (length=9)
          'distance' => int 298
          'stationID' => int 762
      1 => 
        array (size=3)
          'name' => string 'La Tronche Hôpital' (length=18)
          'distance' => int 425
          'stationID' => int 771
  16 => 
    array (size=4)
      0 => 
        array (size=3)
          'name' => string 'Bastille' (length=8)
          'distance' => int 531
          'stationID' => int 397
      1 => 
        array (size=3)
          'name' => string 'Xavier Jouvin' (length=13)
          'distance' => int 589
          'stationID' => int 438

In another script I have a similar structure and json_encode works fine. So I don't understand why json_encode won't work here.

Edit : there seems to be a problem with the encoding. When mb_detect_encoding returns ASCII, the json_encode works but when it returns UTF8, it doesn't work anymore.

Edit2 : json_last_error() returns JSON_ERROR_UTF8 which means : Malformed UTF-8 characters, possibly incorrectly encoded.

Well after 2 hours of digging (cf Edits)

I found out following :

• In my case it's a encoding problem
• mb_detect_encoding returns probably a faulty response, some strings were probably not UTF-8
• using utf8_encode() on those string solved my problem, but see note below

Here is a recursive function that can force convert to UTF-8 all the strings contained in an array:

function utf8ize($d) {
    if (is_array($d)) {
        foreach ($d as $k => $v) {
            $d[$k] = utf8ize($v);
        }
    } else if (is_string ($d)) {
        return utf8_encode($d);
    }
    return $d;
}

Use it simply like this:

echo json_encode(utf8ize($data));

Note: utf8_encode() encodes ISO-8859-1 string to UTF-8 as per the docs so if you are unsure of the input encoding iconv() or mb_convert_encoding() may be better options as noted in comments and other solutions.

Matthieu Riegler presented really good solution however I had to slightly modify it to handle objects too:

function utf8ize($d) {
    if (is_array($d)) 
        foreach ($d as $k => $v) 
            $d[$k] = utf8ize($v);
        
     else if(is_object($d))
        foreach ($d as $k => $v) 
            $d->$k = utf8ize($v);
        
     else 
        return utf8_encode($d);
   
    return $d;
}

One more note: [json_last_error()][1] may be helpful in debugging json_encode()/json_encode() functions.

[1]: http://php.net/manual/en/function.json-last-error.php "json_last_error()"

For me, the answer to this problem was setting charset=utf8 in my PDO connection.

$dbo = new PDO('mysql:host=localhost;dbname=yourdb;charset=utf8', $username, $password);

Adam Bubela also presented really good solution who helped me solved my problem, and here is the simplified function :

function utf8ize($d)
{ 
    if (is_array($d) || is_object($d))
		foreach ($d as &$v) $v = utf8ize($v);
	else
		return utf8_encode($d);

	return $d;
}

I have exactly the same problem on PHP 5.6. I use Open Server + Nginx on Windows 7. All charsets are set to UTF-8. In theory, according the official documentation, flag

> JSON_UNESCAPED_UNICODE

should solve this. Unfortunately this is not my case. I do not know, why. All snippets above do not solve my problem, thus I have found my own implementation. I believe it could help someone. At least, Russian letters pass the test.

function utf8ize($d) {
	if (is_array($d) || is_object($d)) {
		foreach ($d as &$v) $v = utf8ize($v);
	} else {
		$enc   = mb_detect_encoding($d);

		$value = iconv($enc, 'UTF-8', $d);
		return $value;
	}

	return $d;
}

This accepted answer works. But in case you are getting your data from MySQL (as I was) there is an easier way.

Once you open you database, before you query you can set the character set using mysqli as follows:

/* change character set to utf8 | Procedural*/
if (!mysqli_set_charset($link, "utf8")) {
    printf("Error loading character set utf8: %s\n", mysqli_error($link));
    exit();
}

OR

/* change character set to utf8 | Object Oriented*/
if (!$mysqli->set_charset("utf8")) {
        printf("Error loading character set utf8: %s\n", $mysqli->error);
        exit();
 }

LINK: http://php.net/manual/en/mysqli.set-charset.php

I ran into this issue on a server running an older version of PHP (5.2). I was using the JSON_FORCE_OBJECT flag, and apparently that isn't supported until 5.3

So if you're using that flag be sure to check your version!

A workaround appears to be just casting to an object before encoding, like:

json_encode((object)$myvar);

I was getting data from ob_get_clean() and had the same problem, but above solutions don't work for me. In my case solution was this, maybe it will help somebody.

$var = mb_convert_encoding($var, 'UTF-8');

The return of mb_detect_encoding may not be correct:

$data = iconv('UTF-8', 'ISO-8859-1', 'La Tronche Hôpital');
var_dump(
    mb_detect_encoding($data),
    mb_detect_encoding($data, array('ISO-8859-1', 'UTF-8'))
);

Depending on the default detection order, the above may return different results, so the encoding is falsely being reported as UTF-8. (Here's a larger example.)

It is likely that your data is not encoded as UTF-8, so json_encode is returning false. You should look at converting your strings to UTF-8 before JSON-encoding:

$fromEncoding = 'ISO-8859-1'; // This depends on the data

array_walk_recursive($array, function (&$value, $key, $fromEncoding) {
    if (is_string($value)) {
        $value = iconv($fromEncoding, 'UTF-8', $value);
    }
}, $fromEncoding);

I have improved Adam Bubela's answer. I just hate it when blocks are not closed by { and }. It's cleaner and you don't introduce bugs or maybe it's that I did use Perl in the past :)

<?php

class App_Updater_String_Util {    
    /**
     * Usage: App_Updater_String_Util::utf8_encode( $data );
	 *
     * @param mixed $d
     * @return mixed
     * @see http://stackoverflow.com/questions/19361282/why-would-json-encode-returns-an-empty-string
     */
    public static function utf8_encode($d) {
        if (is_array($d)) {
            foreach ($d as $k => $v) {
                $d[$k] = self::utf8_encode($v);
            }
        } elseif (is_object($d)) {
            foreach ($d as $k => $v) {
                $d->$k = self::utf8_encode($v);
            }
        } elseif (is_scalar($d)) {
            $d = utf8_encode($d);
        }

        return $d;
    }
}

?>

using utf8_encode() on those string solved my problem.

If you get this data from a database, use mysqli_set_charset($connection, "utf8"); in connection when get the params from database

Since the output is going to end up as JSON (string) anyway, here is a function that can handle all kind of variable type:

    function utf8ize($arg)
    {
        if (is_array($arg))
            foreach ($arg as $k => $v)
                $arg[$k] = utf8ize($v);
         else if(is_object($arg))
            return utf8ize((array) $arg);
         else
            return utf8_encode(strval($arg));
        return $arg;
    }

This answer is improvement from Matthieu Riegler and Adam Bubela's answer, but this one can handle all type of variable / data type (e.g.: resource like file handle/stream/phpgd/sql connection)

I'm using cast as array when the argument is object in order to prevent error if the object property is read only or private/protected.

Also, on else case, I strval it to handle other data type like resource

I'm using this function to 'force' a variable into string (e.g.: when var_dump-ing or logging to text file)