Float32 arrays don't plot reliably on 0.5.0

[ssfrr]

It looks like Julia 0.5 changes the way that single-precision floating-point numbers are displayed in scientific notation. Instead of 3.05e-5 now it's 3.05f-5. Looks like if any of these show up in the data it hoses the display:

arr = Float32[-0.00735474, 3.05176f-5, 0.00769043]
plot(scatter(y=arr))

causes jupyter to say:

Javascript error adding output!
SyntaxError: Invalid or unexpected token
See your browser Javascript console for more details.

Related:
JuliaIO/JSON.jl#163
JuliaLang/julia#17720
JuliaLang/julia#18053
JuliaLang/julia#18682

Seems that it should be fixed in master, but hopefully it'll be backported to the next 0.5.x release. This probably shouldn't be worked-around in PlotlyJS, but I wanted to file the issue here in case somebody else came looking for an explanation of the error message.

[sglyon]

Thanks for letting is know here -- this is good to know.

Do you think it is an issue that can/should be resolved in JSON.jl?

[ssfrr]

OK. the fix PR just got merged into JSON.jl master. Hopefully it will be tagged pretty soon.

[ssfrr]

And now it's tagged and merged into METADATA. JSON version 0.7.1 should be good.

[sglyon]

Thanks for following up. Next time I tag plotlyjs I'll bump JSON in REQUIRE