Sections using Base.Grisu need to be rewritten

[dmbates]

As of JuliaLang/julia#32859 Base.Grisu has been removed from Julia and the sections of Showoff that depend on Grisu will need to be rewritten, I believe.

[dmbates]

Actually, I see from the News file changes that Grisu was replaced by Ryu in version 1.4

The `Base.Grisu` code has been officially removed (float printing was switched to the ryu algorithm code in 1.4)

[quinnj]

Luckily, it looks like a lot of the code can be simplified with the new Printf.format(::Printf.format, args...) functionality. It allows avoiding the @sprintf macro usage and dynamically formatting things much easier.

[jw3126]

@quinnj can Printf also help with the following:

Showoff.jl/test/runtests.jl

Line 47 in f5e5329

@test showoff([1, 1e39]) == ["1×10⁰", "1×10³⁹"]

[quinnj]

What are you hoping Printf can do for you there? (sorry, I'm not familiar w/ showoff at all)

[jw3126]

Something like this:

function showoff(x)
    # trivial code
    intermediate = Printf.nontrivial(...)
    # trivial code
end
@test showoff(1e-39) == "1×10³⁹"

For instance if nontrivial could decompose 1.234e-39::Float64 into e.g. (mantissa=[1,2,3,4], exponent=-39, sign=:+) that would be great.

[quinnj]

Not that I can tell; it's really just for printing things to a buffer and it's pretty highly optimized for that, so it doesn't really do any decomposing.

[jw3126]

So one way or another we need to pass from 1.234e-39 to the decimal representation of the mantissa and the exponent. AFAICT Grisu.grisu was the non trivial piece in doing this and now I am sure, what it should be replaced by?

[KristofferC]

Grisu.grisu was the non trivial piece in doing this and now I am sure, what it should be replaced by?

Copy paste whatever you need from Grisu into here?

[KristofferC]

See https://github.com/JuliaAttic/Grisu.jl and #25.

[jw3126]

Thanks @KristofferC !