I'm not sure how to add the check for whetherc is indicating a color in a color cycle. I think it requires checking whether the first element of the string is'C' and if the next elements represent a digit.

Also: I'm not sure why those tests are failing, but it seems like the test server isn't connected to the internet..?

Looks like the CI failures were largely due to network problems on GHA/Azure so cycling to rerun

Background

I opened this PR because of my work on PR#27304, which I though would benefit from a check on whether the user provided a color with an alpha value to determine how the violinplot function handles alpha values. However, after@story645 explained, the point of allowing alpha to be included in a color was to remove unnecessary arguments, which I agree with, so that sort of makes my work here less relevant (except for the fact that the previous_has_alpha_channel() was out of date).

Current use of_has_alpha_channel()

The_has_alpha_channel() method is only used in one place throughout the entire library:lib/matplotlib/axis.Tick. In that location, it just checks whether the provided grid color has an alpha channel specified, and if not, usesmpl.rcParams["grid.alpha"] instead.

Tick() is the parent ofXTick() andYTick(), which are the parents ofSkewXTick(). However, nowhere in the entire library do those methods set the kwargsgrid_color orgrid_alpha.

Proposal

It seems like_has_alpha_channel() is obsolete, and should be removed from the library. I propose:

1. Removing it (and it's tests intest_colors.py).
2. InTick(), removinggrid_alpha as an argument and lettinggrid_color specify the alpha setting.

I'm +0.5 on removinggrid_alpha. There's value in reducing the number of parameters an dependency between them. While we should not remove primaryalpha parameters from functions because they are used too widely, removing*_alpha parameters for secondary color attributes (typically pass through) seems viable.

Before we can removegrid_alpha, it has to go through our deprecation cycle.

It should also be ensured, thatgrid_color=(None, 0.5) works and only modifies the alpha. Otherwise, only modifying the alpha becomes cumbersome.

@landoskape are you still interested in working on this?

Ah, yes! Lost track of it but I can continue in early January. Thanks for the reminder.

See my responses to@timhoffm. I believe this is all that is needed now since you suggested deferring gridalpha discussion to another time?

Ah -@timhoffm good point! The function was introduced because it would have helped the violinplot update that is in PR#27304. It's now removed entirely.

I suggest starting by checking whether we have a valid color withis_color_like and if not returnFalse. Then the remaining logic can be confident about howc is structured.

Smart. I incorporated this.

Of course, no worries. In agreement with all and suggestions all accepted.

I do not know what the doc failure is about, but it cannot possibly be caused by this change since we are only modifying a private function.

Thanks@landoskape and congratulations on your first PR merged into Matplotlib! We hope to hear from you again.

Woohoo! Thanks :)