TabularMakie

I wrote this package because I couldn't figure out how to fix some things in AlgebraOfGraphics.jl (see #136).

NOTE: Most of these things are fixed in AlgebraOfGraphics.jl 0.4.

An example

Generate Data

Generate data

using DataFrames, CategoricalArrays
using DataAPI: refarray

cs_df = let
	N = 100
	dummy_df = DataFrame(
		xxx = rand(N),
		yyy = rand(N),
		s_m  = rand(5:13, N),
		g_c  = rand(["c 1", "c 2", "c 3"], N) |> categorical,
		g_lx = rand(["lx 1", "lx 2", "lx 3"], N) |> categorical,
		g_m  = rand(["m 1", "m 2", "m 3"], N) |> categorical
		)

	dummy_df[:,:s_c] = 2 .* rand(N) .+ refarray(dummy_df.g_lx)
	dummy_df
end

This is how you would create a quick plot where the axis are automatically labelled and legends are automatically created.

using TabularMakie, CairoMakie

fig = lplot(Scatter, cs_df,
	:xxx,
	:yyy;
	color = :s_c,
	marker = :g_m,
	markersize = :s_m,
	layout_wrap = :g_lx
  )

You can rename and transform variables on the fly using the source => transformation => newname syntax from DataFrames.jl.

using CategoricalArrays: recode!

rec_1(x) = recode!(x, "m 1" => "Nice name", "m 2" => "Other")
rec_2(x) = recode!(x, "lx 1" => "Panel 1")
minus1(x) = x .- 1

fig = lplot(Scatter, cs_df,
	:xxx => minus1,
	:yyy => ByRow(x -> x + 1) => "the y plus one";
	color = :s_c => "hey there",
	marker = :g_m => rec_1 => "bla",
	markersize = :s_m => :tada,
	layout_wrap = :g_lx => rec_2
  )

What this package can do but AlgebraOfGraphics can't

• rename or transform variables on the fly (e.g. :xxx => "name of x" or :yyy => ByRow(log))
• supports layout_wrap (in addition to layout_x and layout_y)
• generates legend for continuous aesthetics (e.g. markersize and linewidth)
• generates a colorbar if color is provided with a continuous variable
• adds a non-incremental mode for creating a plot, this allows creation of grouped bar plots
• allows access of the legend to change position and attributes

All these issues were fixed in AlgebraOfGraphics 0.4

What this package can't do but AlgebraOfGraphics can

• combine different plots (e.g. visual(Scatter) + linear)
• use other inputs than tables (the "slicing context")

Open an issue if you find more.