[FR] indexed image with `:auto` color

[goretkin]

Something like

julia> p = plot(); scatter!(p, rand(10), rand(10), color=:auto); scatter!(p, rand(10), rand(10), color=:auto)

works to produce two scatter plots with automatically chosen colors:

I would like the same thing for binary images.

This is possible:

heatmap(rand(0:1, 5, 10), c=cgrad([:black, :red], categorical=true), colorbar=false)

but this is not:

heatmap(rand(0:1, 5, 10), c=cgrad([:black, :auto], categorical=true), colorbar=false)

Without something like this, I do not know how to make a recipe for e.g. a "grid" type, that correctly handles colors. See https://discourse.julialang.org/t/handling-color-in-plots-user-recipe-that-produces-an-image-or-indexed-color/59769

[goretkin]

Another way to state the issue is to give a binary image the same features as, say, plotting Shapes. From memory, some backends do not support a heatmap or an image, and do plot them as a grid of rectangle shapes.

[goretkin]

This is the best I know how to do given the current situation. Note "the workaround is partial because [...]".

using Plots
using Plots: RGBA
using RecipesBase
using OffsetArrays
import PlotUtils

struct Grid
    x::OffsetArray{Bool, 2, Array{Bool, 2}}
end

# partial workaround for https://github.com/JuliaPlots/Plots.jl/issues/3448
# the workaround is partial because it will not apply the color cycling of `Plots.get_series_color` https://github.com/JuliaPlots/Plots.jl/blob/8cf268a3903f847ef3071019b3e77b99240ba9c4/src/args.jl#L1607-L1616
function _get_color_hack(plotattributes)
    full_color_default = RGB(0.0, 0.0, 1.0)
    full_alpha_default = 1.0
    empty = RGBA(0.0, 0.0, 0.0, 0.0)
    full = PlotUtils.plot_color(
        get(plotattributes, :seriescolor, full_color_default),
        get(plotattributes, :seriesalpha, full_alpha_default)
    )
    (;empty, full)
end

# partial workaround for https://github.com/JuliaPlots/Plots.jl/issues/2410
# this will work for `OffsetArray` and `Array`
# but might break for other types that e.g. wrap OffsetArray
_unwrap_offsetarray(a) = parent(a)

@recipe function f(s::Grid)
    axs = axes(s.x)
    f = first.(axs) .- 0.5
    l = last.(axs) .+ 0.5
    _range(start, step, stop) = start:step:stop
    extents = map(_range, f, [1.0, 1.0], l)
    (d1, d2) = extents
    colors = _get_color_hack(plotattributes)
    d1, d2, ifelse.(transpose(_unwrap_offsetarray(s.x)), colors.full, colors.empty)
end

a = fill(false, (-5:5, 0:10))
a[:] .= rand(Bool, 11, 11)[:]
plot(Grid(a))