support of heatmats

[bgctw]

Are heatmaps still possible with the current version? Where can I find an example?

[jacobusmmsmit]

I'm afraid I removed it from the current version because it was so janky. I'm not sure what the best way to go about doing ternary heatmaps would be.

[bgctw]

ok. I found some examples with the GMT.jl.

There, however, I cannot specify my own function (e.g. a Distribution) but need to generate a grid which will then be interpolated. I share my code for the grid construction. Maybe it can be useful, as well as the code in GMT.jl.

"""
    simplex_grid(m,n=3)

Create a grid across the n-dimesional unit simplex with m points on each dimension.
If `unit=false` then return an Integer grid ranging from 0 to (m-1).

"""
function simplex_grid(m,n=3; unit=true)
    n < 2 && error("n needs to be at least 2 but got ", n)
    m < 2 && error("m needs to be at least 2 but got ", m)
    L = num_compositions(m-1,n)
    A = Array{Int}(undef, L, n)
    simplex_grid!(A,m-1,n; L)
    unit ? A ./(m-1) : A
end

# number of cases for simples_grid
num_compositions(mstep,n) = binomial(n+mstep-1, n-1)

"""
Update matrix A with positions of simplex grid with msteps, i.e. mstep+1 records
for all other dimesnions being 0
"""
function simplex_grid!(A, mstep, n; L=num_compositions(mstep,n))
    if n==2
        A .= cat(0:mstep,  mstep:-1:0; dims=2)
    else        
        istart0 = 0
        for im = 0:mstep
            Lsub = num_compositions(mstep-im, n-1)
            i = istart0.+(1:Lsub)
            A[i,1] .= im
            # update the remaining columns recursively with decreased mstep
            B = @view A[i, 2:n]
            simplex_grid!(B, mstep-im, n-1; L=Lsub)
            # keep track where to update the next rows
            istart0 = istart0 + Lsub
        end
    end
    nothing
end