closes #12967 fix [en|de]coding of HashMap<K,V> where K is a numeric type

[olsonjeffery]

Details are in #12967.

This is just w/i the serialize::json module. If you have feedback for how to improve this impl, I'm all ears. This is to work around a blocking issue I've encountered with an application I'm working on.

[huonw]

This will fail if out has a multibyte trailing character. It should be one of

out.char_at_reverse(out.len());
// or
out.as_bytes()[out.len() - 1] != '"' as u8;

(The last is valid because all 7-bit ASCII bytes in a UTF-8 string represent that ASCII value.)

[emberian]

This is going to be correct because if there are multibyte characters, it must start with a " and end with a ". That is, it's going to fail anyway because if you have a multibyte character there, it's invalid JSON.

[huonw]

Oh, hm; I guess so. (It's worth having a comment about why we can just wilfully fly in the face of unicode.)

[emberian]

This raises the larger point that our JSON support is really fragile,
though.

On Mon, Mar 17, 2014 at 9:24 AM, Huon Wilson notifications@github.comwrote:

In src/libserialize/json.rs:

     if idx != 0 { try!(write!(self.wr, ",")) }

•    f(self)
  

•    // ref #12967, make sure to wrap a key in double quotes,
  

•    // in the event that its of a type that omits them (eg numbers)
  

•    let mut buf = MemWriter::new();
  

•    let mut check_encoder = Encoder::new(&mut buf);
  

•    f(&mut check_encoder);
  

•    let buf = buf.unwrap();
  

•    let out = from_utf8(buf).unwrap();
  

•    let needs_wrapping = out.char_at(0) != '"' &&
  

•        out.char_at(out.len() - 1) != '"';
  

Oh, hm; I guess so. (It's worth having a comment about why we can just
wilfully fly in the face of unicode.)

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/12973/files#r10655371
.

http://octayn.net/

[olsonjeffery]

@cmr you know I agree but, as I said in reply to someone making the same point in IRC last night: let not the perfect by the enemy of the good.

This is a specific fix to an issue I encountered in the wild and counts as progress. By all means let us open up a follow-on issue to address this, but let's have that discussion somewhere else and not in this PR, whose specific issue is blocking me :P

[huonw]

r=me with a squash (and working code ;P ).