add whatsnew and test coverage

[bjlittle]

🚀 Pull Request

Description

Hey @krikru,

As promised, here's some updates that's targeting your patch-1 branch, which is the source of the pull-request SciTools#4894.

It basically adds a suitable entry to our documentation whatsnew, plus we blow the trumpet that it's your first contribution to iris 😄

I've also added a couple of tests to exercise the change in behaviour, which is based on a pull-request (now closed and unmerged) by @rcomer, see SciTools#4136.

To be transparent and fair, I've also acknowledged @rcomer's contribution in the whatsnew along side yourself.

Here is proof that the ci-tests are passing for this change from my fork of iris.

If you're happy with my suggestions, then simply merge this pull-request, and it will automatically become part of your associated pull-request on iris.

At that point, I think we're good to bank this goodness in time for the iris 3.3 release 👍

---

Consult Iris pull request check list

[krikru]

@bjlittle The tests look good to me. Maybe you could also do self.assertIs(gcs.colorbar.ax, ax1) in test__with_axes (I guess a similar line could be added in test__without_axes, but that requires _plot to also return the Axes object that is returned by fig2.add_subplot(1, 1, 1))? I'm just assuming that gcs.colorbar.ax should be the Axes object you specified when calling qplt.contourf (if you do specify one), but maybe it's not.

It looks like test__with_axes is the test that checks for the bug. Have you tested that it triggers the bug that I described when you don't have the fix, and that it doesn't trigger it when you do have the fix?

[bjlittle]

Hey @krikru,

I like your thinking... but just for clarity, the gcs.colorbar.ax isn't the axes that you think it might be, it's an entirely different axes, but I totally get your intent.

So I've pushed a change which does as you suggested, but the axes information that we need is squirreled away somewhere differently within matplotlib. However, with a little bit of detective work it was reasonably straight forward to understand the internal plumbing.

And yeah, test_quickplot.TestSubplotColorbar.test__with_axes is specifically targeting your fix. Whereas test_quickplot.TestSubplotColorbar.test__without_axes is testing the status-quo (aka default behaviour).

Consequently, when I remove your fix to iris.quickplot._label the test_quickplot.TestSubplotColorbar.test__with_axes fails, as we might expect 👍

[bjlittle]

Also, here is the latest ci-tests run on my repo for the latest change.

[krikru]

Looks good

[bjlittle]

Just rerunning the tests after the last refactor, see https://github.com/bjlittle/iris/actions/runs/2828606527

[bjlittle]

@krikru Once you merge this, then I'll bank SciTools#4894 👍