I think it's a depends what you're trying to do situation?

Like I really don't like splitting the grays and I'm partial to sorting on hues cause I'm usually looking at that table to find/build a gradient?

Agree with@story645 - by-hue makes more sense to me than by-saturation. Usually you are looking for a shade of green, not any color with a certain saturation.

As an aside, it would nice if this gallery were linked athttps://matplotlib.org/stable/users/explain/colors/colors.html, and/or merged. I also think it would be nice if we added the full xkcd colortable - sure 954 colors is a lot, but its not that many

.. should add, I don't see why we couldn't include both sortings, if both seem useful.

If we are doing multiple sortings we should also include the alphabetical one as well.

If we are doing multiple sortings we should also include the alphabetical one as well.

We don't really use sphinx design in the gallery but maybe makes sense to put in a tab grouping - hue, saturation, alphabetical-which I wonder if it'd be good to move this table into the color reference kinda like@jklymak suggested

TL;DR: Let's table this for later.

Still, I want to comment on your remarks.

by-hue makes more sense to me than by-saturation. Usually you are looking for a shade of green, not any color with a certain saturation.

I believe a fundamental partition by-saturation makes sense. Low saturation is more suited for large areas (facecolor), high saturation is more suited for small objects (text, marker, ...). Note that for examplehttp://davidbau.com/archives/2013/09/28/css_color_names.html also aggregates low-saturation colors together.

Like I really don't like splitting the grays

Thathas always been there, but I agree it should be improved.

I'm partial to sorting on hues cause I'm usually looking at that table to find/build a gradient?

I would say, that the old one is not at all better for this:

old: new:

The strong variation on saturation makes the gradient uneven.

The strong variation on saturation makes the gradient uneven

This is making me think that maybe the sort should be (hue, saturation) table, sorta like Google docs:

You could try, but google has the advantage that they can choose the colors, and thus choose a grid in (hue, saturation) space.

The CSS4 colors are unevenly spaced, so that it may be quite hard pressing them in a (hue, saturation) grid:

import matplotlib.pyplot as pltimport matplotlib.colors as mcolorshsv = np.array([mcolors.rgb_to_hsv(mcolors.to_rgb(c)) for c in mcolors.CSS4_COLORS])plt.scatter(hsv[:,0], hsv[:,1], c=mcolors.CSS4_COLORS)plt.xlabel("hue")plt.ylabel("sat")

Note: Prior to this PR all colors are sorted by hue only, i.e. in the above picture, strictly go from left to right. This results in quite stong saturation variations. This PR separtes the 2D hue/sat space in two groups, splitting at sat=0.2, and for each group draw by hue, i.e. go from left to right.