Uns categorical colour maping as a dictionary

• [ x] Additional function parameters / changed functionality / changed defaults?
• New analysis tool: A simple analysis tool you have been using and are missing in sc.tools?
• New plotting function: A kind of plot you would like to seein sc.pl?
• External tools: Do you know an existing package that should go into sc.external.*?
• Other?

When determining colours of categorical variables in uns they are based on alphabetical order (if I am not mistaken) - being represented just as an ordered list. Thus it is a bit inconvenient to change colours, especially if categories change during the analysis - the whole order changes and the mapping breaks. Would it be possible to use a dictionary of colours as values and categories as keys (with a default for any categories missing colours)?

[ivirshup]

I think we should allow for categorical colors along either axis, and right now it’s becomes ambiguous. A good example of an annotation that can apply to both observations and variables is species.

I’d like to shift to a nested model to limit the amount of reserved keys in .uns. It reduces that chance of unintentional naming collisions.

As for the amount of things that would need to change, a lot has to change anyways. Hardly any code that works with the current setup will work with mappings (len is all I can think of). If we’re already making a breaking change, might as well take advantage and future proof it a bit.

[fidelram]

I agree with @ivirshup that the best would be to have a single function that returns the colors. This also facilitates documentation. The actual implementation afterwards could be modified without much effort.