more general ordering of input vertical grids in fast_interp3D about croco HOT 6 OPEN

The code that performs the interpolation is fairly simple:

and consistent with the notebook.
The question that follows is then: what do we do about it?

Allowing for grids that are reversed is certainly feasible.
But do we want to also allow unsorted vertical grid points?

from croco.

My preference would be for the former.

from croco.

Ah OK! I could have read more carefully fast_insterp3D.c. Maybe the least we could do, if we don't want to modify the routine, is to add a mention in the docstrings or even add a test in interp2z?

from croco.

Agreed, could you please push a commit updating the doc?

Could you also modify the title of this issue with something like: "more general orders of input vertical grids for fast_interp3D " ?

We'll list this issue as a new feature request to remember a PR should be proposed at some point in order to allow interpolation with vertical grids with opposite orders.

from croco.

done

from croco.

There's a new method in xgcm to interpolate (see xgcm tutorial in crocosi : xgcmgrid.transform). The target coordinate must be monotonically increasing or decreasing. Performance seems quite good. May be it could help...
@NoeLahaye , @apatlpo

from croco.