ENH: make ax.get_position apply aspect

[jklymak]

PR Summary

Currently, ax.get_position(original=False) doesn't tell us the position of an Axes if its aspect ratio has been set, unless we call ax.apply_aspect() or fig.draw() beforehand. This is a bit annoying, because its hard to imagine why we would want to ask for the axes position without the proper aspect ratio already having being applied, particularly if we can get the original position by calling ax.get_position(original=True).

This PR, simply calls self.apply_aspect() if original=False.

The doc string was also updated to what I hope is numpydoc compliance.

Closes #9207

PR Checklist

• Has Pytest style unit tests
• Code is PEP 8 compliant
• New features are documented, with examples if plot related
• Documentation is sphinx and numpydoc compliant
• Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

[tacaswell]

Nominally 👍.

Share @efiring 's concern about performance, but this seems a good a place as any in the code to force application of the aspect. There is still a path confusion (create axes, set aspect, ask position, change limits) but is a more limited path that the current one (create axes, set aspect, set limits, ask position, draw).

[jklymak]

I'm less concerned about performance. I think if the user asks for the position, and they specify "original=False" they should get what they ask for. If its a performance issue for them, then they can figure things out, and then hardcode.

A thornier question, however, is if the default should be "original=True", since thats what most users were getting when calling get_position(). That is if they didn't know the trick of calling ax.apply_aspect(). But for all I can tell that trick would only be discernible from reading the source code. I would never have thought about calling that, and I got pretty heavily into the source code in #9082.

[efiring]

From the standpoint of internals, the api_changes is a little bit misleading; with original=False it was returning the active position, which at that point might or might not be identical to the original, and which might or might not be the position after the draw. The latter is still true, but this PR makes it more likely that get_position(False) returns what the user might reasonably expect. So I think it is a net gain, and worth trying.

[jklymak]

I'm sticking this in 2.2. Feel free to re-milestone, but its a pretty small change.

[dstansby]

I'd be in favour of putting this in 3.0 since it's a change that should probably go out in an rc before the final release.

[jklymak]

Now that we are on 3 can this go in?

[phobson]

Looks good