Bad argument on splitted float

[kpy3]

Hi,

just found incorrect decoder behaviour on splitted float number:

(bocore@tommy.local)40> {incomplete, F} = jsx:decode(<<"{"library": "jsx", "awesome":0">>).
{incomplete,#Fun<jsx_decoder.37.8673461>}
(bocore@tommy.local)41> F(<<".21}">>).
** exception error: bad argument
in function jsx_decoder:maybe_done/4
called as jsx_decoder:maybe_done(<<".21}">>,
{jsx_to_term,{[[{<<"awesome">>,0},{<<"library">>,<<"jsx">>}],
[]],
{opts,binary,false}}},
[object],
{opts,false,false,false,false,false,false,false,false,false,
false})

[kpy3]

Hm, jsx works well with explicit_end option.

[talentdeficit]

I'm happy you found the solution, but this behaviour is inconsistent with what I want jsx to do. I'm going to make it more intuitive. Thanks for finding it

[talentdeficit]

the decoder is now much smarter about what's unambiguously incomplete and what is not. you should no longer need the explicit_end flag for cases that aren't ambiguous

Eshell V5.9.1 (abort with ^G)
1> {incomplete, F} = jsx:decode(<<"{"library": "jsx", "awesome":0">>).
{incomplete,#Fun}
2> F(<<".21}">>).
[{<<"library">>,<<"jsx">>},{<<"awesome">>,0.21}]

[kpy3]

It works for integer() too, thanks a lot.
Is develop brach is stable enough for production use or should I wait for merge this patch into master?

[talentdeficit]

this is the only significant difference between develop and master so it should be safe to use but there could be lurking edge cases i haven't considered. if you don't use explicit_end ever it's definitely safe

[kpy3]

Ok, I'll try, thanks.