Will Benfold

I don't know enough to tell everything that's going on in #3171 but the stated intent seems like a good resolution

I'm currently intending that it can handle being between two thresholds (or any other criterion you can write as a lambda) but only one condition at a time, yes

Can you confirm the versions Iris and other libraries that you're using @valeriupredoi ? I can't reproduce this with Iris as it stands on main (`3.3.dev0`), indeed I get a...

I hadn't tried deriving a `ccrs.Globe` like that, but I don't think it works? ``` >>> import cartopy.crs as ccrs >>> import pyproj >>> proj_crs = pyproj.CRS.from_user_input(26915) >>> proj_crs Name:...

I think I'd then struggle to plot data using the `cp_crs` as it wouldn't have any of the attributes that Cartopy uses to distinguish plotting a `PlateCarree` and a `Mercator`?...

Ah, sorry, I think my example was misleading because it included a CRS with an intrinsic projection. Iris has some CRSes without an intrinsic projection which this doesn't work for,...

This happens because doing `cf_var.dtype` gives us `str` rather than a numpy datatype. This causes issues in the save code, when it's determining the default fill value and when it's...

Hi @valeriupredoi, sorry for the delay getting back to you on this. In terms of what Iris should do here, what's the ideal behaviour for you? My go-to list of...

> How easy that would be to implement I've no idea! My immediate thought is that we'd at least have to realise the data to do that check, which might...

What do you reckon would be more helpful @rcomer? Given that a failure of the cubes to all look the same will, from my understanding, fail inside concatenate because of...