feat: Multi-histogram plot

[lenianiva]

Description

This PR adds a multi-histogram plotting function multihist! like seaborn's stacked histogram. Below is an example of stacked vs dodged histograms.

Vide https://discourse.julialang.org/t/stacked-histogram/133248

using CairoMakie

data1 = rand(100) .* 2.0 .- 1.0
data2 = rand(100) .* 2.0

fig = Figure(size=(600,300))

multihist!(
    Axis(fig[1, 1]),
    [data1, data2];
    colormap = :Set3_10,
    label = ["red", "blue"],
    positioning = :stack,
)
multihist!(
    Axis(fig[1, 2]),
    [data1, data2];
    colormap = :Set3_10,
    label = ["red", "blue"],
    positioning = :dodge,
)
save("/tmp/stacked.svg", fig)

Type of change

Delete options that do not apply:

• New feature (non-breaking change which adds functionality)

Checklist

• Added an entry in CHANGELOG.md (for new features and breaking changes)
• Added or changed relevant sections in the documentation
• Added unit tests for new algorithms, conversion methods, etc.
• Added reference image tests for new plotting functions, recipes, visual options, etc.

[lenianiva]

I also want to add an optional dodge positioning for the histogram, but this would require conditionally setting either :stack or :dodge based on another option :positioning, and map! does not seem to be able to do this.

    map!(plot, :points, [:flatbins, :stack]) do points
        flatbins = collect(Iterators.flatten(points))
        stack = [i for (i, bin) in enumerate(points) for _ in 1:length(bin)]
        return (flatbins, stack)
    end

[lenianiva]

The weights argument should be a vector of vectors as well, but this doesn't work:

    map!(plot, [:values, :edges, :normalization, :scale_to, :weights], :points) do values, edges, normalization, scale_to, weights
        points = []
        for (i, dataset) in enumerate(values)
            if weights === automatic
                wgts = automatic
            else
                wgts = weights[i]
            end
            centers, weights = _hist_center_weights(dataset, edges, normalization, scale_to, wgts)
            push!(points, Point2.(centers, weights))
        end
        return points
    end

Giving the error

Triggered by update of:
  arg1, scale_to, weights, normalization or bins
Due to ERROR: MethodError: no method matching weights(::Float64)
The function `weights` exists, but no method is defined for this combination of argument types.

[ffreyer]

I'd say this should be a change to hist instead, and follow the syntax we already have for barplot, violin, boxplot and crossbar

[lenianiva]

I'd say this should be a change to hist instead, and follow the syntax we already have for barplot, violin, boxplot and crossbar

so with an extra categories variable?

[jkrumbiegel]

so with an extra categories variable?

No, barplot for example has a dodge and a stack attribute and these take Vector{Int}s that assign groups for each.

[ffreyer]

I can also refactor this if you're not already working on it

[ffreyer]

I think this should be done now.

You can now do either:

# established syntax:
hist(rand(200), stack = vcat(fill(1, 100), fill(2, 100))

# vector of vectors syntax:
hist([rand(100), rand(100)], stack = [1, 2])

where stack can be replaced with dodge. weights work if they match the syntax used for the argument. color can be :stack or :dodge to use those indices as color values, a per-group color vector (one color per stacked/dodged histogram), the existing :values (uses heights as color values) or the existing per-bin color vector. All of these work with both input styles.

[lenianiva]

I think this should be done now.

You can now do either:

# established syntax:
hist(rand(200), stack = vcat(fill(1, 100), fill(2, 100))

# vector of vectors syntax:
hist([rand(100), rand(100)], stack = [1, 2])

where stack can be replaced with dodge. weights work if they match the syntax used for the argument. color can be :stack or :dodge to use those indices as color values, a per-group color vector (one color per stacked/dodged histogram), the existing :values (uses heights as color values) or the existing per-bin color vector. All of these work with both input styles.

Thanks! Sorry I didn't have time to work on this.