Improving ggplot2 theme

[greimel]

I would like to improve the ggplot2 theme.

This is the first step.

Old :gglot2 (pyplot())

New :ggplot2 (pyplot())

Main differences

1. added 'outside ticks'
2. no lines around markers (EDIT: added this bullet)
3. lighter gray background. I tried to match the shades of gray according to

R> library("ggplot2")
R> theme_get()

Does someone know the Juliaequivalent of grey30, grey90, etc? I translated it to grey$x == plot_color(:black, 1 - x).

Actual ggplot2

Unfortunately, it is not equally nice on all backends.

• A problem is that often setting axes and ticks independently is not possible
  (axis color = white and tick color = gray)

Still missing:

1. adding minor gridlines but no minor ticks (cannot be controlled independently)
2. adding minor gridlines at all doesn't seem to be supported for some backends
3. removing the axis altogether to remove the gap between grid and outside tick

Note: the file in the test folder will be removed eventually. I will just keep it as long as I am working on this.

[mkborregaard]

Great effort 👍 It's definitely not easy to get this to match on all backends, but it's a clear improvement already

[greimel]

with JuliaPlots/Plots.jl/pull/1787 it will look like this.

GR:

Problems:

• doesn't support gridlinewidth <1
• it's currently not possible to control "minor" ticks separately from "major" ticks

(before the PR the axes looked like this:

)

PyPlot is unchanged.

Problem: currently doesn't handle minor grid and ticks

[mkborregaard]

Should we just merge now, or do you have other things planned here?

[greimel]

This depends on JuliaPlots/Plots.jl/pull/1788.

It looks good on GR and PyPlot.
GR:

PyPlot:

It doesn't look good on PGFPlots. It probably cannot handle plot_color(:black, 1-0.92) (to get the equivalent of R's gray92) etc. Can anyone point me where to look in order to circumvent that?

PGFPlots;

I should probably fix the problem with pgfplots before this gets merged. I haven't checked with the other backends.

[mkborregaard]

Could you just check the rgb values of grey92 in R and specify the color as an rgb instead?

[greimel]

I removed the dependency on the minor-tick-PR in Plots and followed @mkborregaard's advice. The theme now look semi-nice on gr(), pyplot() and pgfplots()

PGF

EDIT: with JuliaPlots/Plots.jl#1850 the axis will have the correct color

Improvements (minor ticks, axis color on pgfplot) are left for future work.

[mkborregaard]

Nice - what do you say we merge this now?

[mkborregaard]

The blue color is a little off - not sure if that could be quickly fixed.

[greimel]

Which blue color?

I am fine with merging otherwise.

[mkborregaard]

Ah lol I see the palette (the colors of the points) is just Plots' default. If you see the other themes they also add palettes (see e.g. https://github.com/JuliaPlots/PlotThemes.jl/blob/94e6427befef28433bdc0dbc23d4adaa543e47aa/src/wong.jl ) so that the dots get the right color.
Do you feel like adding the ggplot2 colors? I think it would improve the impression a lot. Otherwise I'm fine with merging.

[greimel]

I see, I found this: https://stackoverflow.com/a/8197706

Is the hcl color wheel implemented somewhere already?

I suggest merging this and leaving the rest for a separate PR.

[mkborregaard]

Not as such but I think it corresponds to the :isolum gradient in :colorcet, not sure that's any help though.

[greimel]

Ok, I'll probably tackle this at a later point. Feel free to merge this ;-)

[mkborregaard]

Great!

[daschw]

Awesome!

[pkofod]

Awesome! Statisticians are now less likely to be mocked for using Julia ;)

[isentropic]

@greimel hi, pyplot now somewhat supports the minorgrids JuliaPlots/Plots.jl#2621
Would you mind testing the ggplot2 and see if everything works like you expect it to?