Implement `Grid.build_node_face_connectivity`

[anissa111]

Implement Grid.build_node_face_connectivity

[anissa111]

@paullric I'm working on this implementation now, could you provide a little clarification about what this function is supposed to do?

Does the function build node_face_connectivity from scratch? Or is this function just building a reverse relationship from a pre-existing face_node_connectivity?

(cc @erogluorhan for updates)

[paullric]

It should have to build it from scratch, if face_node_connectivity is unavailable. If it is available then you can leverage it and transpose. ~ Paul

…

On Sep 13, 2022, at 3:31 PM, Anissa Zacharias ***@***.***> wrote: @paullric <https://github.com/paullric> I'm working on this implementation now, could you provide a little clarification about what this function is supposed to do? Does the function build node_face_connectivity from scratch? Or is this function just building a reverse relationship from a pre-existing face_node_connectivity? — Reply to this email directly, view it on GitHub <#130 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIVPZET272MKO6NI3JW5RTV6D6DFANCNFSM6AAAAAAQL3SVLY>. You are receiving this because you were mentioned.

[anissa111]

It should have to build it from scratch, if face_node_connectivity is unavailable. If it is available then you can leverage it and transpose.

Oh, okay, I see thank you! Is there a prefered method for doing this calculation or is something similar implemented in tempestremap I should look at?

[Huite]

Hi @anissa111,

I butt in here occasionally, I've implemented this in xugrid if you're looking for inspiration:

https://github.com/Deltares/xugrid/blob/main/xugrid/ugrid/ugrid2d.py#L406

It mostly uses some utils defined here:
https://github.com/Deltares/xugrid/blob/main/xugrid/connectivity.py

Which I find make it relatively easy to these transformations on connectivity arrays.

In general, I found it convenient to use scipy.sparse functionality to build/represent these arrays here and there. In particular, constructing a CSR (https://docs.scipy.org/doc/scipy/reference/generated/scipy.sparse.csr_matrix.html) matrix does a lot of sorting (which is C++ behind the scenes). Alternatively, you require a lot of numpy sorting calls, which aren't that easy to do efficiently if the arrays are really not that rectangular; or you'd have to write a lot of numba code, I suppose.

[anissa111]

@Huite, thank you! I will definitely have a look at the links you've provided.