Brett Zamir
28 Oct '09

  @Joris: Thanks for your code (it's helpful to have here for reference and makes a good complement to the previous code for urlencode and reposted just now in the comments at http://phpjs.org/functions/urlencode ), but I went ahead with the faster use of decodeURIComponent().

Brett Zamir
26 Oct '09

  @Joris: Sorry I haven't gotten to your post yet; that looks great! I'd like to test it out a little first and then commit, if you can bear with me a little...

Joris van der Wel
29 Sep '09

  Incase anyone is interested, here is a version with full UTF-8 support written without decodeURIComponent or any maps.

function rawurldecode(url)
{
	// This function mimmicks PHP's rawurldecode under UTF-8
	// Any percentage notation is converted to its UTF-16 character.
	// Only tested on Mozilla browsers (Firefox 3.5)
	// Does NOT use any of decodeURIComponent, decodeURI, unescape, etc
	// Supports 4 byte characters (so unicode characters 0x0000 through 0x10FFFF)
	//
	// Original by Joris van der Wel
	var chr, a, len, ret, c, c2, c3, c4, hi, low;
	
	ret = '';
	for (a = 0, len = url.length; a < len; a++)
	{
		chr = url.charAt(a);
		if (chr != '%')
		{
			ret += chr;
			continue;
		}
		
		c = parseInt(url.charAt(a+1) + url.charAt(a+2), 16);
		if (isNaN(c))
		{
			ret += '%'; // If php comes across something invalid, it just shows it without parsing 
			continue;
		}
		
		a += 2; // skip 2
		
		ret += String.fromCharCode(c);
	}
	
	// second pass, convert UTF-8 to UTF-16 (Strings in javascript (ECMA-262 to be exact) are UTF-16)
	url = ret;
	ret = '';
	for (a = 0, len = url.length; a < len; a++)
	{
		c = url.charCodeAt(a);
		
		//        c & 1000 0000  === 0000 0000
		if(      (c &      0x80) === 0        ) // 0xxxxxxx
		{
			ret += url.charAt(a);
		}
		//        c & 1110 0000  === 1100 0000
		else if ((c &      0xE0) ===      0xC0) // 110y yyxx	10xx xxxx
		{
			a++;
			c2 = url.charCodeAt(a);
			ret += String.fromCharCode(
					((c  & 0x1F) << 6) | 
					((c2 & 0x3F) << 0)
				);
		}
		//        c & 1111 0000  === 1110 0000
		else if ((c &      0xF0) ===      0xE0) // 1110 yyyy	10yy yyxx	10xx xxxx
		{
			a++;
			c2 = url.charCodeAt(a);
			a++;
			c3 = url.charCodeAt(a);
			ret += String.fromCharCode(
				       ((c  & 0x0F) << 12) |
				       ((c2 & 0x3F) << 6 ) |
				       ((c3 & 0x3F) << 0 )
			       );
		}
		//        c & 1111 1000  === 1111 0000
		else if ((c &      0xF8) ===      0xF0) // 1111 0zzz	10zz yyyy	10yy yyxx	10xx xxxx 
		{
			a++;
			c2 = url.charCodeAt(a);
			a++;
			c3 = url.charCodeAt(a);
			a++;
			c4 = url.charCodeAt(a);
			
			c =	((c  & 0x07) << 18) |
				((c2 & 0x3F) << 12) |
				((c3 & 0x3F) << 6 ) |
				((c4 & 0x3F) << 0 ) ;
			
			if (c >= 0x10000) // split it up using surrogates
			{
				c -= 0x10000;
				
				hi  = (c & 0xFFC00) >> 10; // first 10 bits
				low = c & 0x003FF; // last  10 bits
				
				hi  += 0xD800; // high surrogate range
				low += 0xDC00; // low surrogate range
				ret += String.fromCharCode(hi, low);
			} 
			else
			{
				ret += String.fromCharCode(c);
			}
		}
	}
	
	return ret;
}

You could probably rewrite it to use only one loop, but that would turn into spaghetti code very fast

Gr.

Kevin van Zonneveld
3 Apr '09

  @ travc & Brett Zamir: Fixed in SVN: Thanks!

Brett Zamir
30 Mar '09

  Sorry, it looks like to fully reflect PHP's behavior, you have to add this to the histogram (all of our other related functions should be converted accordingly as well):

histogram['\u20AC'] = '%80';
histogram['\u0081'] = '%81';
histogram['\u201A'] = '%82';
histogram['\u0192'] = '%83';
histogram['\u201E'] = '%84';
histogram['\u2026'] = '%85';
histogram['\u2020'] = '%86';
histogram['\u2021'] = '%87';
histogram['\u02C6'] = '%88';
histogram['\u2030'] = '%89';
histogram['\u0160'] = '%8A';
histogram['\u2039'] = '%8B';
histogram['\u0152'] = '%8C';
histogram['\u008D'] = '%8D';
histogram['\u017D'] = '%8E';
histogram['\u008F'] = '%8F';
histogram['\u0090'] = '%90';
histogram['\u2018'] = '%91';
histogram['\u2019'] = '%92';
histogram['\u201C'] = '%93';
histogram['\u201D'] = '%94';
histogram['\u2022'] = '%95';
histogram['\u2013'] = '%96';
histogram['\u2014'] = '%97';
histogram['\u02DC'] = '%98';
histogram['\u2122'] = '%99';
histogram['\u0161'] = '%9A';
histogram['\u203A'] = '%9B';
histogram['\u0153'] = '%9C';
histogram['\u009D'] = '%9D';
histogram['\u017E'] = '%9E';
histogram['\u0178'] = '%9F';

and then add this line right before the call to decodeURIComponent():

ret = ret.replace(/%([a-fA-F][0-9a-fA-F])/g, function (all, hex) {return String.fromCharCode('0x'+hex);}); // These Latin-B have the same values in Unicode, so we can convert them like this

travc
30 Mar '09

  Found an apparent bug... I'll try to track it down, but I'm a javascript noob.

<?  print rawurldecode('-22%97bc%2Fbc'); ?>

<script type="text/javascript">
  var foo = rawurldecode('-22%97bc%2Fbc');
  alert(foo);
</script>

php part works fine, js breaks.

firebug reports:
malformed URI sequence
rawurldecode("-22%97bc%2Fbc")

And, yes, this string comes from encoding in php with rawurlencode (from a big nasty db response).