geom_line with colors connects different traces

[jfb-h]

If you plot different traces, the start of each trace gets connected to the next. Here's a reproducer:

df = mapreduce(vcat, 1:4) do i 
	DataFrame(
		x=collect(1:100),
		y=cumsum(randn(100)),
		g=fill(string(i), 100)
	)
end

ggplot(df, aes(x=:x, y=:y, color=:g)) + geom_line()

[rdboyes]

Thanks for reporting this - it's another problem (similar to #86) that is downstream from the "ad hoc" way I've handled the idea of "grouping" so far in the package. Hoping to have a fix for this on the way to 0.8.0

[adknudson]

I have a MWE of grouping that may be useful. I'm not sure how it's being handled in TidierPlots (somewhere in draw.jl?).

using DataFrames, ColorSchemes, CairoMakie

import Makie.SpecAPI as S

df = DataFrame(
    a = rand(20),
    b = rand(20),
    c = rand(["L", "H"], 20),
    d = rand(["A", "B"], 20),
)

group_df = groupby(df, [:c, :d])

palette = ColorSchemes.tab10

ax = S.Axis()

for (i, key) in enumerate(keys(group_df))
    s = S.Lines(group_df[key].a, group_df[key].b, color = palette[i])
    push!(ax, s)
end

layout_spec = S.GridLayout(ax)
plot(layout_spec)

It seems like the way to do grouping in Makie is to make multiple plot specs.

[rdboyes]

Thanks @adknudson - this was essentially how I was imagining this would be fixed, but helpful to have a MWE. I added support for geoms that return multiple PlotSpecs when I was fixing geom_smooth (just need to have them return a Vector{PlotSpec} and they should draw automatically). All we need to do is have geoms conditionally return a vector of plotspecs when appropriate and this should work

[rdboyes]

This is working for geom_path in a dev branch:

geom_line isn't currently cooperating - the sort is conflicting with the grouping. Should be fixed soon!

[rdboyes]

geom_line fixed - will merge soon