Summary

There are at least two places where matplotlib provides a message to the user about usage patterns that are potentially mistakes using logging functionality, rather than as a warning.

Two examples I've encountered recently:

• Passing string-formatted numbers as data
• [Done] Passing an RGB tuple to the c parameter of scatter

I don't think logging is the right mechanism here:

• The feedback provides very little information about the source of the problem, you just get some text in stderr.
• Related, you can convert a warning to an error and then drop into a debugger to figure out what's happening; I don't think that's the case with logging?
• Warnings can be intercepted by third party libraries that are doing something on purpose (i.e., passing numbers-as-strings to use a categorical axis), whereas logs are harder to muffle without intervening.

Proposed fix

It's possible I am wrong here, but warnings.warn feels like the right mechanism for both of these pieces of feedback.