VIS: plot method cannot handle new 'C0'-like colors (matplotlib 2.0)

[jorisvandenbossche]

matplotlib 2.0 introduced a few 2-character color abbreviations: C0, C1, C2, ... to refer to the colors of the current color cycle
(very handy! this way you can refer to a color of the color cycle without hardcoding it, and use it as a shortcut for color='tab:blue' if you want the new blue instead of the ugly color='b')

But, it seems pandas cannot handle this at the moment, as it tries to interpret it as multiple colors:

df = pd.DataFrame(np.random.randn(10,3), columns=['a', 'b', 'c'])
df.plot(color='C0')

gives ValueError: 'C0' can be parsed as both single color and color cycle. Specify each color using a list like ['C0'] or ['C', '0']

I know the error message already gives the solution (using color=['C0']), but I think it would be nice to special case those to have to working by default.

[GuessWhoSamFoo]

This one is pretty old, but I'd love to revive it. I'm not sure how tests will cover this change. Is it alright if I give it a try anyway?

[jreback]

sure!

try adding. test with the code as above (u will check this plot works)

[jbrockmendel]

@jorisvandenbossche we're now getting warnings about upper-case letters. Do we want to support upper-case specifically? If not, its an easy edit to get rid of the warnings in the CI logs.

[jorisvandenbossche]

Answered in the other thread, but to have it here as well: yes we want to support those upper-case ones specifically. They are a specific color naming scheme of matplotlib.
I don't think they are what cause the warnings (although maybe how it is implemented to parse them causes warnings)