Makes sense to add the same to

This may just as well go in 3.7? (I guess it would if there was a 3.7-doc? Edit: no, not in 3.7-doc as it also changes the lib-files, but still...)

Do we want to get this in prior 3.7.1, which we are looking to tag today?

Just thought that I'd highlight#17286 as it is sort of related to this. At least the suggestion to rename this attribute and/or have setters/getters for it. Doesn't have to be a show stopper for this PR though.

If I’m not mistaken, this is the answer tohttps://discourse.matplotlib.org/t/toolbar-parameters-in-show/25908/2

This can't get pulled back to the doc branch (it touches library code).

Recently we have documented other attributes using class level type hints (and we did backport one of these):#29965#28825

Would it make sense to do that here too?

I'm a bit torn how to do this properly. The fundamental issue is that attributes are a dynamic concept and it's hard to attach meta information to them. Actually I don't like either approach: Class level type annotations are a bit awkward because we have otherwise pushed all typing to the stubs. OTOH I also don't like to create class attributes for documentation when we actually have instance attributes.

If I have to choose between those, I'd go for the type annotations, but YMMV.

If we have two other instances of using the typing style and@timhoffm prefers it, go with that method!

Oh it seems it doesn't work for 3D. Is that because we have not typed the module more generally?