unexpected and undocumented behaviour in dim coord assoc.

[cpelley]

        if not matches:
            matches = [(dim,) for coord_, dim in self._dim_coords_and_dims if
                       coord_.name() == coord.name()]

        # Search aux coords
        if not matches:
            matches = [dims for coord_, dims in self._aux_coords_and_dims if
                       coord_.name() == coord.name()]

https://github.com/SciTools/iris/blob/master/lib/iris/cube.py#L791

Currently we are looking at whether the given coord shares its name with a coordinate on the cube. I really dont think this belongs in coord_dims at all and may result in unexpected results to users.

Effected tests:
test_contrived_differential1 (iris.tests.test_analysis_calculus.TestCalculusWKnownSolutions)
test_contrived_spherical_curl1 (iris.tests.test_analysis_calculus.TestCalculusWKnownSolutions)
test_contrived_sphrical_curl2 (iris.tests.test_analysis_calculus.TestCalculusWKnownSolutions)

Since this is not documented, I would at the very least label it as a bug, further more, the test failures based on its removal require investigation, determining why its necessary.

Happy to look into this myself when I have time.

[ajdawson]

Was this issue fixed by #587? If so can we close this?

[rhattersley]

Was this issue fixed by #587? If so can we close this?

Unfortunately not. The lines in question are still present:

iris/lib/iris/cube.py

Line 924 in 6c8144a

### Search by coord name, if have no match

[pelson]

I've put this in the v1.7 milestone as it sounds like there is a bug here, I just don't know what it is - can anyone shed any light?

[pelson]

Ah, I see (it has taken a long time to understand the problem, but I has finally clicked - perhaps the issue description could have been a little cleared in describing the problem).

Essentially I believe the complaint is that cube.coord_dims() takes a coord instance, first checking whether the given coord actually exists on the cube. If not, the coord definition will be used to find the appropriate coordinate and its dimensions. If there is still no luck, then the coordinate name is used instead.

It is the ambiguity of using the coordinate name that is the complaint/bug here.

I'm going to push this out of v1.7 as I don't think it is a show stopped.

[cpelley]

It is the ambiguity of using the coordinate name that is the complaint/bug here.

Thank you for your interest, yes this is what was intended as the 'issue' raised.

Cheers @pelson, still on my radar! will look into it!

[rcomer]

No idea when they were removed, but those name matching lines are no longer present.