Heatmap colorbar with non-finite values about unicodeplots.jl HOT 3 CLOSED

Yes, but it doesn't help if the entire field is NaN/Infs

julia> A = rand(2,2)
julia> A[1] = NaN
julia> NaNMath.extrema(A)
(0.18783010303710246, 0.9637507750862072)

julia> NaNMath.extrema(A*NaN)
(NaN, NaN)

I'm coming from a perspective where I run a simulation, hopefully everthing works fine and heatmap gives me a quick look that things look reasonable. However, when the simulation blows up I just see white stripes on the heatmap (which is good because then I know I did something wrong). But in that case I can't use the colorbar at the moment.

from unicodeplots.jl.

I suggest to change that function to

function ceil_neg_log10(x)
    val = -log10(x)
    isfinite(val) || return typemin(Int)
    roundable(val) ? ceil(Int, val) : floor(Int, val)
end

such that

julia> ceil_neg_log10(Inf)
-9223372036854775808

julia> ceil_neg_log10(1e300)
-300

the output is of the same type (Int) and the smallest representable to match the -log10(x) idea as best as possible. In the end it seems to be used here

where this works fine now

julia> round(Inf,digits=typemin(Int)+1)
Inf

julia> round(-Inf,digits=typemin(Int)+1)
-Inf

julia> round(NaN,digits=typemin(Int)+1)
NaN

If there are no objections, I'll create a pull request replacing the ceil_neg_log10 function

from unicodeplots.jl.

Thanks for reporting.

First, I think the colorbar limits should not take NaNs into account. That is an easy fix replacing

Then, I think we should ignore those NaN values when printing Crayons.

from unicodeplots.jl.