Implemented constraint equality.

[pp-mo]

Aiming to address #3616

[lbdreyer]

Although this is technically in progress, it isn't necessarily going to go into Iris 3.2. So to avoid it cluttering the "in progress" column, I will move it to to do. We can then pick it up if we time.

[wjbenfold]

If we do get to it then we'll need a rebase onto main before it can be reviewed I think

[stephenworsley]

I'm happy with this approach. I think the important thing here is that no two constraints with different behaviour can show up as equal.

I think the only changes needed here are changing references from defn to metadata.

[trexfeathers]

@pp-mo are you able to finish this one off before getting back into #5016? Sounds like this is pretty close.

[pp-mo]

Rebased.
but N.B. put in draft temporarily

It is not good to go, since the auto-merged result is still using Coord._as_defn()
And we now want to use common metadata instead.

[pp-mo]

Code now simplified ...
On closer inspection, I conclude that the 'coord_name' of a _CoordConstraint is only ever intended to be a string.
The only way it can be otherwise is if you construct Constraint(coord_values={something:value}). But whereas my earlier code suggested that 'something' could be a CoordDefn (or, now, a metadata), it actually can't since that is not a hashable type, so cannot act as a dict key.
It is still just about "possible" that you could pass a Coord instance as the identifier in this case, but we have no tests for it.
I simply don't believe it is useful, so I'm no longer supporting it as a special case.

[pp-mo]

@trexfeathers I think this is good to go. Stephen's comment with change request is outstanding, but I think you can dismiss that if you think it is now OK.

[trexfeathers]

Thanks @pp-mo! I'll let @stephenworsley take it over the line.

[stephenworsley]

Looking good, just a few minor changes and a whatsnew entry and I think this should be good to go

[pp-mo]

Thanks @stephenworsley
I hope this latest addresses all the points raised.
Sorry for the delay !

[stephenworsley]

Looking good, just some very minor changes and I think we're good to go.

[pp-mo]

Thanks @stephenworsley
I think I've now fixed them all to be the same.

[stephenworsley]

Looks good!

Oh, I think this might want a whatsnew also.

[stephenworsley]

I forgot to mention before, but this still needs a whatsnew entry.

[pp-mo]

I forgot to mention before, but this still needs a whatsnew entry.

Whoops sorry forgot.
Doing it now ...

[pp-mo]

Hang on now I need to rebase/resolve...

[pp-mo]

Ok, rebased with added whatsnew

• please see if that suits, @stephenworsley

[stephenworsley]

Looks good!

[pp-mo]

Thanks @stephenworsley for going the distance !

[rcomer]

Thanks guys! 😀