I am -1 on this. I have a philosophy of all magic being shallow. What is the usecase of having this function perform the expansion as opposed to having the user do it when it calls plt.style.use()?

Yeah I could see it either way. I think it's convenience. In most minds of most users, ~/blah is a valid path and so the function should accept it. Making the user convert is a mild annoyance.

The magic in this case is also fairly benign. The standard case is still attempted first, and the fallback only works on fnames which are never valid literal paths on any OS I've used.

This seems to me like a good idea. I don't see any disadvantage. We are using our own _get_home() extensively in mpl's startup, and allowing '~/config' in plt.style.use() is in that spirit.

ok, perhaps I can be convinced, however the error handling will need to be fixed. If neither attempt works, it will only report the failure of the second attempt. We would need proper error messaging.

Aha, yes that needs fixing. style.use handles it correctly b/c it just catches and returns its own message, I think, but when called on its own, this needs adjusting.

Why not always call expand_user?

That should be safe too...I was being overly conservative. With that change made, I guess the error handling is fine as is?

👍 from me