Bivariate mapping not working as a recipe

[tpoisot]

This needs add SimpleSDMLayers#interfaces-iteration to work

@mkborregaard, following your advice in #120, I tried to put the bivariate plot to a recipe. When I do this, it "works" (I get no error message), but the plot is what you get when only calling plot().

Any idea? Can it be because there is a plot(x, y) method for two layers already defined?

using Plots
using RecipesBase
using SimpleSDMLayers
using Colors, ColorBlendModes

layer1, layer2 = SimpleSDMPredictor(WorldClim, BioClim, [1,12]; left=-20.0, right=55.0, top=42.0, bottom=-36.0)

q1 = rescale(layer1, collect(LinRange(0.0, 1.0, 25)))
q2 = rescale(layer2, collect(LinRange(0.0, 1.0, 25)))

@shorthands bivariate

@recipe function f(::Type{Val{:bivariate}}, plt::AbstractPlot; classes=3, p0=colorant"#e8e8e8", p1=colorant"#64acbe", p2=colorant"#c85a5a")

    # Get the palettes
    c1 = palette([p0, p1], classes)
    c2 = palette([p0, p2], classes)
    breakpoints = LinRange(0.0, 1.0, classes+1)

    xlims --> extrema(longitudes(x))
    ylims--> extrema(latitudes(x))
    legend --> false
    aspectratio --> 1
    frame --> :box
    subplot --> 1
    seriestype --> :heatmap

    for i in 2:length(breakpoints)
        m1 = broadcast(v -> breakpoints[i - 1] <= v <= breakpoints[i], x)
        for j in 2:length(breakpoints)
            m2 = broadcast(v -> breakpoints[j - 1] <= v <= breakpoints[j], y)
            m = reduce(*, [m1, m2])
            replace!(m, false => nothing)
            replace!(m, nothing => NaN)
            @series begin
                color --> BlendMultiply(c1[i - 1], c2[j - 1])
                longitudes(m), latitudes(m), convert(Matrix{Float64}, m.grid)
            end
        end
    end
end

bivariate(q1, q2)

[tpoisot]

Turns out that yes, defining the bivariate map in the plot recipe solved everything

[mkborregaard]

Cool - is that the ifnal recipe?