@l-johnston Thanks again for your work on this. On the dev call @anntzer and @efiring pointed out that a) this is the correct functionality, so please forgive my concern, b) this PR could likely be made a lot simpler by just calling self.set_xlim or self.set_ylim, which have all the axis view logic in them already. ie. no need to do the loops in this PR.

I'm not sure it'd be so simple, as this is in the Axis class, not XAxis or YAxis, so it doesn't know which one to call. Even the implementation here is not quite correct, as it doesn't handle z axis sharing on 3D plots.

One possible way would be to update _axis_method_wrapper so that Axes.set_xticks used sharing information to properly update everything. But this means that Axes.xaxis.set_ticks would still be out of sync.

so it doesn't know which one to call.

You can do something along the lines of

set_xlim(...) if self is self.axes.xaxis else
set_ylim(...) if self is self.axes.yaxis else
set_zlim(...) if self is getattr(self.axes, "zaxis") else
(fall back on the old way of using set_view_interval)

or

try:
    axis_name, = [k for k, v in self.axes._get_axis_map() if v is self]
except ValueError:
    (fall back)
getattr(self, f"set_{axis_name}lim")(...)

btw instead of set_x/y/zlim you can also use set_x/y/zbound which will auto-handle inverted axises for you.

Sorry @l-johnston I think your rebase got messed up. Sorry, for the noise - it seems fine now...

@jklymak I implemented @anntzer 's approach of using set_x/y/zbound, but this breaks 20 tests such as:
_______________________________________________ test_set_ticks_inverted ________________________________________________

    def test_set_ticks_inverted():
        fig, ax = plt.subplots()
        ax.invert_xaxis()
        ax.set_xticks([.3, .7])
>       assert ax.get_xlim() == (1, 0)
E       assert (0.7, 0.3) == (1, 0)
E         At index 0 diff: 0.7 != 1
E         Use -v to get the full diff

lib/matplotlib/tests/test_axes.py:6520: AssertionError

Well, perhaps my suggestion was wrong, I didn't actually try it :)

Well, I'm confused - does setting the xticks set the limits to the first and last tick or does it insert a margin?

@jklymak setting ticks inserts some margin since the call to set_view_interval has default ignore=False and this behavior is described in the set_view_interval docstring. I can cause the exact same failures by setting ignore=True. So, to prevent breaking existing tests, I think we have to stay with the set_view_interval approach. I have updated it to accommodate @QuLogic 's 3D case and added a test.