Restructure plotting

[ahartikainen]

Restructure plotting.

backend related defaults to backend code.

[canyon289]

@ahartikainen Do you need help with this? Seems like theres a lot going on

[ahartikainen]

Sure if you have time.

I will continue when I have some time.

Basically there is couple of lines in each main plot function that is moved to backend functions, and then some of the inner kw are changed.

Then there is still a lot to do. I will add discussion here when the time is right.

[codecov]

Codecov Report

Merging #1312 into master will decrease coverage by 0.05%.
The diff coverage is 96.47%.

@@            Coverage Diff             @@
##           master    #1312      +/-   ##
==========================================
- Coverage   93.00%   92.94%   -0.06%     
==========================================
  Files         101      101              
  Lines       10115    10184      +69     
==========================================
+ Hits         9407     9466      +59     
- Misses        708      718      +10     

Impacted Files	Coverage Δ
arviz/plots/autocorrplot.py	100.00% <ø> (ø)
arviz/plots/backends/bokeh/distcomparisonplot.py	100.00% <ø> (ø)
arviz/plots/backends/bokeh/kdeplot.py	93.33% <ø> (ø)
arviz/plots/backends/matplotlib/forestplot.py	97.93% <ø> (ø)
arviz/plots/backends/matplotlib/kdeplot.py	97.46% <ø> (ø)
arviz/plots/bpvplot.py	79.24% <ø> (+2.10%)	⬆️
arviz/plots/densityplot.py	90.00% <ø> (-1.94%)	⬇️
arviz/plots/distcomparisonplot.py	92.30% <ø> (-2.94%)	⬇️
arviz/plots/elpdplot.py	100.00% <ø> (ø)
arviz/plots/essplot.py	100.00% <ø> (ø)
... and 58 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ae0138c...8edd4e7. Read the comment docs.

[canyon289]

Wow @ahartikainen you did it. Is this still in draft stage? I'm dying to hit that merge button

[ahartikainen]

It still in draft. There are some normalization to be done

[ahartikainen]

Ready for review.

This PR enables mpl=>3.3 and doesn't change the user-interface.

[OriolAbril]

Thanks for all the work!

[OriolAbril]

not sure if we should use alpha or line_alpha, in fact, not even sure if we should create a bokeh dealiaser too

[ahartikainen]

bokeh kwargs are currently a mess, so I didn't touch them for now.

Also most bokeh kwargs are currently called explicitly

[OriolAbril]

Let's leave it like this then

[ahartikainen]

@OriolAbril @canyon289 I'm starting to think that maybe we should move closer to "native" bokeh settings and colors.

This would mean that we don't change "red" to mpl red but use bokeh red. We could have handling for special colors

C0-C9 --> let's fix this by our selves
`red_matplotlib` or `red_mpl` --> force mpl colors? 

[OriolAbril]

For base colors, both libraries use CSS colors so red is exactly the same color for mpl and for bokeh. https://docs.bokeh.org/en/latest/docs/reference/colors.html#bokeh-colors-named and https://matplotlib.org/3.1.0/gallery/color/named_colors.html

The conversion using mpl helpers has the advantage that then, in addition to CSS colors, all the following are also supported:

• C0... so that defaults easily mute when changing the stylesheet
• single letter colors
• tableau colors
• xkcd colors

Note: xkcd:red can be different than CSS red

[ahartikainen]

Ok, if the colors are the same, I'm fine with the helper function.

[canyon289]

I'll take a look at this in next ~24 hours but dont let me block a merge either

[canyon289]

I dont want there to be a merge conflict so I'm going to merge this and keep reviewing. If stuff comes up I'll make an issue ticket