ENH: Added coordinate inversion functionality

[cpelley]

Addition to iris.util for convenient inversion of a specified coordinate on a cube or list of cubes.

[cpelley]

Test failures not relating to the changes of this PR

[rhattersley]

Test failures not relating to the changes of this PR

See #1730.

[rhattersley]

There's already a lat_lon_cube function in this module which does a very similar (albeit more limited) job. It'd be nicer if we could just extend lat_lon_cube.

[ajdawson]

If I reverse a dimension that also has a 1D AuxCoord describing it, will that be reversed too? And what about if there is a 2D AuxCoord overlapping the dimension, will it be reversed along the required axis?

[rhattersley]

This doesn't really do enough to warrant existing as a separate function.

[rhattersley]

There an iris.util... function for this ... demote_dim_coord_to_aux_coord. More on this topic later...

Perhaps this is a good time to consider whether iris.util is really the best place for this function (and a bunch of other functions that are already in there!). The documentation shows the module is a mess already - a real mix of arbitrary helpers (timers, create_temp_filename, ...), numeric helpers (monotonic, reverse, ...), Cube operations (as_compatible_shape, describe_diff, ...). Surely we can do better...?

(NB. This is somewhat related to @bjlittle's upcoming "add-ons"/"extensions"/"incubator" repo.)

[cpelley]

Thank you for your interest in this @rhattersley

..the module is a mess already - a real mix of arbitrary helpers..

I very much agree.

This is somewhat related to @bjlittle's upcoming "add-ons"/"extensions"/"incubator" repo.

I'm happy to hold off on this PR until the new repo appears and service this for that?

[rhattersley]

This is somewhat related to @bjlittle's upcoming "add-ons"/"extensions"/"incubator" repo.

I'm happy to hold off on this PR until the new repo appears and service this for that?

At the moment, I still think the main Iris repo is the place for this routine. I just wanted @bjlittle to be aware of this discussion since it's a related topic.

[rhattersley]

Nudge... is there still life in this PR?

[cpelley]

Hi @rhattersley thanks for prodding. Based on the feedback above (the limitations as well as fragilities, I decided to re-implement from an alternative approach yesterday, see #1983). This new implementation is also driven by a new requirement for in-place inversion without touching the data.