Categorical color palettes

[axsk]

While the current gradient palettes :inferno, :heat, :blues are suited to plot sequential data, reflecting their ordererd structure, I propose adding additional "categorical" (name suggestions are welcome) palettes with presets of distuingishable colors.

Here are some suggestions that I stumbled upon looking for such palettes two years ago.

• Few (from Stephen Few, Show Me the Numbers)

• Set3 (from http://colorbrewer2.org)

• Paired (from http://colorbrewer2.org)

• Munsell interpretation of the Twenty-Two Colors of Maximum Contrast by K. L. Kelly

The RGB values can be found in this gist.

Does anyone know if there are copyrights on colors?

Edit: Concerning the API I would suggest just adding new symbols for the palettes to the palette kwarg, i.e. plot(..., palette=:set3). Given n series, I think these should get mapped to the first n sequential colors of the palette, since I consider the correct/distuingishable ordering to be the task of the palette (and as far as I remember above palettes took this into account)

[tbreloff]

I certainly like the idea of having some pre-defined palettes available. We do have to be weary of copying without a proper license. (yes, palettes can be copy-protected)

There are already several sources of "distinguishable palettes". I think I'd like to make sure those can easily be used, and if we have pre-defined ones we could store in a Dict.

So the logic would be roughly:

if haskey(_categorical_palettes, v)
    # force color list
elseif # v is a color vector
    # force color list
elseif haskey(_gradients, v)
    # build palette from gradient
else
    # do the "distinguishable_colors method"
end

[mkborregaard]

I have opened an issue for enhanced color suppor in PlotUtils (where I think this stuff belongs) JuliaPlots/PlotUtils.jl#3 . That issue is for gradients, but I agree that other sequential palettes also belongs naturally here, as part of color libraries (PR here JuliaPlots/PlotUtils.jl#5) .
With regards to @axsk question of copyright, many gradients come with a reasonably permissive license, but most require attribution. Putting gradient libraries such as colorbrewer in PlotUtils allows the license to be displayed there.

[mkborregaard]

Closed in favour of JuliaPlots/PlotUtils.jl#3