MeshCoord class [AVD-1591]

[pp-mo]

🚀 Pull Request

Description

Adding MeshCoord.

Notes:

1. the characteristic points and bounds indexing calculations are not yet implemented here.
   I have other code prototyped for that, but this PR is simpler if I leave it out for now.
2. the implementation isn't quite minimal, but I have chosen to control copying and slicing to make it a 'good citizen' within a cube (i.e. not blocking cube copy or slicing).
3. there is as yet no sensible __str__ or __repr__. We have another task pending to address that.

---

Consult Iris pull request check list

[pp-mo]

I think this is all that is needed, but you may well decide there is something missing @trexfeathers.
The testing is using specific real things rather than mock-ist, so you might well uncover some possibility I have missed due to that.

I am a little surprised there is not more to this, but that is probably because we have decided now to defer implementing the "bounds but no points" ability. Which I thought might be tricky, and now think even more so!

@stephenworsley it will be good if you can cast an eye over this, but I think @trexfeathers is steering at this point.

[pp-mo]

Update: I'm now working to reduce the __init__ args to just mesh/location/axis, and thus enforce the intended content wrt the provided args.

I'll "WIP" / draft this : please don't merge till those changes are up

[pp-mo]

Failing docs tests -- probably needs another rebase ?

Meantimes, since discussing with @trexfeathers today, I am now intending to fix the copy/init to support the cube slicing needs in a nicer way (i.e. without bodging the allowed copy args as at present).
Hence this is still WIP

[trexfeathers]

I've been asked to stop reviewing this and work on something else. So here's what I've got so far. Mostly optional minor improvements, but I definitely think you should remove those setter methods (see individual comments). Thanks 🙂

[pp-mo]

Latest:

• rebased to fix docs test failures (I hope)
• addressed all outstanding comments (I think)
• revised the copy and __getitem__ overrides to support minimum needs, as promised above
• added testing for the result points+bounds arrays (preliminary, pending "dynamic" versions of the calcs)

Please review @bjlittle !

[bjlittle]

@pp-mo Lovely stuff 👍

This is the super interesting part (for me) now... the interaction of MeshCoord, Mesh and Cube

Okay, so I've raised a few review questions and issues, but nothing controversial . I'm sure you can easily address them.

Otherwise, this is looking very promising 👍

[pp-mo]

Thanks @bjlittle
Rebased to use your recent addition of Mesh.LOCATIONS.
Hope I've addressed everything raised so far.
See if you still don't like the setters for coord_system + climatological . If you really dislike this, I guess we can do something different with these somehow, but it's a bit "tail wagging dog" to modify Coord for the benefit of the subclass !

[pp-mo]

@bjlittle apologies, forgot to push.
Please re-review 😁

[pp-mo]

Further changes as requested (?mostly).
Please re-view @bjlittle

I'm still slightly uncomfortable about the shortcut on __eq__ comparison, but we clearly don't want it comparing points+bounds also.

[pp-mo]

Fixed the equality thing.
As discussed offline.
Ok @bjlittle !