Is there some actual use-case for these? I don't see anything using them, and I'm not sure we want to be a general-purpose colour-processing library.

This is based on my initiative and has two main use-cases:

1. Creating matching shades of color for fill color and edge color, e.g. to easily create https://matplotlib.org/devdocs/gallery/text_labels_and_annotations/fancytextbox_demo.html#sphx-glr-gallery-text-labels-and-annotations-fancytextbox-demo-py based on any color

2. Creating similar colors for related data, e.g. if I have datasets a_before, a_after, b_before, b_after I could

plot(a_after)
plot(a_before, color=lighter('C0', factor=1.5))
plot(b_after)
plot(b_before, color=lighter('C1', factor=1.5))

It seems there is no standard package for colors in python; is that correct? Because ideally, such color arithmetics would be provided by a standard library and in turn matplotlib would provide an interface in the sense of

import python_colors as colors
plt.plot(..., color=colors.Color("red").darker())

This has been discussed for while (dating back to 2013):

#2159, #2745, #8895, #9985, #9995

and a pretty well up-voted SO question https://stackoverflow.com/questions/37765197/darken-or-lighten-a-color-in-matplotlib . So there is clearly some desire for this among our users (and I suspect that a non-zero number of our users abuse alpha for this).

There exists several packages specific to color changes (https://colorspacious.readthedocs.io/en/latest/, https://docs.python.org/3.7/library/colorsys.html, https://github.com/vaab/colour) and we are in the process of standing up a new plotting-specific color library (https://github.com/pyviz/contrib_colormaps) . I think that this functionality would be a much better fit for one of those libraries.

I think a PR adding a (dependencyless) way of consuming the colour objects in our internal color normalization code would be a good thing.

However, I still think that the consensus from #9985 still holds that we don't want do this piecemeal and it looks like since that discussion at least one color specific package has arisen.

Thank you for your work @chaoyi1 , but I am not in favor of merging this as-is. I would rather see the colour integration above and a good example showing how to use it for the use-cases @timhoffm mentioned above.

colorspacious and pyviz/contrib_colormaps have different scopes. Only colour seems to be a possible color representation package. But I have two concerns with it:

• I would expect such a package to accept the same colors as we do. Otherwise, there'll be a lot of confusion that I can do Color('blue').lighter() but not Color('mediumslateblue').lighter(). This is something that could be worked upon, but needs coordination with an external library (and their willingness to accept these). In particular alpha and CSS3 colors are currently not supported (RGBA support vaab/colour#42).
• The colour package hasn't seen any activity for two years. I would only be willing to introduce support for such a library if long-term support is guaranteed. That would seem safer under a matplotlib or pyviz umbrella organization compared to a single-developer project.

This was discussed on this weeks call: https://paper.dropbox.com/doc/In-Matplotlib-2019-meeting-agenda--AoeL8YHb2wWKxPGdvfQ0hkolAg-aAmENlkgepgsMeDZtlsYu#:h2=color-manipulation-tools

The consensus we came to is that we do want this functionality, but want to think a bit more about the algorithm and parameters. We should do some research to see if there is a "standard" way to do this across CSS / Qt / ... and go with that rather than inventing our own thing.

I'll close as not being actively worked on