mochiweb_html unicode breakage about mochiweb

[etrepum]

The intention is that the input is already UTF-8, whether it is binary or iolist. This is consistent with what you would get by doing something like reading the request body from a mochiweb request

1> io:fwrite([mochiweb_html:to_html(mochiweb_html:parse(<<"<done>Matt\xe2\x80\x99s Breaking Line</done>">>)), "\n"]).
<done>Matt’s Breaking Line</done>
ok

Note the usage of io:fwrite/1 to send the output "directly" to stdout instead of io:format/2 which does the wrong thing. io:format/2 does some really broken shit by default if you are using non-ascii characters.

2> io:format("~s~n", [[8217]]).                                              
** exception exit: {badarg,[{io,format,[<0.25.0>,"~s~n",[[8217]]]},
                            {erl_eval,do_apply,5},
                            {shell,exprs,7},
                            {shell,eval_exprs,7},
                            {shell,eval_loop,3}]}
     in function  io:o_request/3

[dreid]

@mattsta it's also worth noting here that you can use mochiutf8:codepoints_to_bytes/1 to convert the list you have into a utf8 binary.

1> S = "<done>Matt’s Breaking Line</done>".
[60,100,111,110,101,62,77,97,116,116,8217,115,32,66,114,101,
 97,107,105,110,103,32,76,105,110,101,60,47,100|...]
2> mochiutf8:codepoints_to_bytes(S).       
<<60,100,111,110,101,62,77,97,116,116,226,128,153,115,32,
  66,114,101,97,107,105,110,103,32,76,105,110,101,60,...>>
3> mochiweb_html:parse(mochiutf8:codepoints_to_bytes(S)).
{<<"done">>,[],
 [<<77,97,116,116,226,128,153,115,32,66,114,101,97,107,
    105,110,103,32,76,105,110,101>>]}

[etrepum]

I believe that unicode:characters_to_binary(S, utf8) is faster than mochiutf8:codepoints_to_bytes/1 (some functions in the unicode module are BIFs). Different error behavior though.

[mattsta]

Thanks for the clarification. I was passing in a fetched HTML body from httpc:request without first running through unicode/utf8 conversion.

Everything works fine as long as I pre-convert:

 27> io:fwrite(mochiweb_html:to_html(mochiweb_html:parse(unicode:characters_to_binary("<done>Matt’s Breaking Line</done>")))).
 <done>Matt’s Breaking Line</done>ok

 28> io:format(mochiweb_html:to_html(mochiweb_html:parse(unicode:characters_to_binary("<done>Matt’s Breaking Line</done>")))).
 <done>Matt’s Breaking Line</done>ok

What can we do to help people not bug you about this in the future? The type system isn't elaborate enough to reject non-utf8 outright.

There's always the cheap:

 parse(Input) ->
     try
        parse_tokens(tokens(Input))
     catch
       error:badarg -> not_utf8
     end.

Or maybe even try to fix it ourselves, though it could hide performance issues by double-evaluating on every run:

 parse(Input) ->
     try
        parse_tokens(tokens(Input))
     catch
       error:badarg -> parse_tokens(tokens(unicode:characters_to_binary(Input)))
     end.

resolved: dumbmatt

[etrepum]

I'd accept a documentation patch that says the input must be UTF-8. I guess string() is a bit misleading, but I wrote all this stuff before Erlang had any attempt to support unicode (except for some functions hiding in xmerl).