to_geodataframe() and integrate() check size==n_faces, not data_location

[Sevans711]

Version

2026.6.0

How did you install UXarray?

Source

What happened?

UxDataArray.to_geodataframe() and .integrate() check self.values.size == self.uxgrid.n_face instead of self._face_centered() (or self.data_location=='face_centered' or self.mapping=='faces', I'm not quite sure which of these is the preferred syntax). Similarly, they check self.values.size == self.uxgrid.n_node instead of checking if the data is node centered.

I'm not sure if it is possible to have n_faces==n_nodes or n_faces==n_edges, or n_nodes==n_edges, but if it is, this code would lead to very confusing bugs for such a case.

What did you expect to happen?

I expected the code to check self._face_centered() (or self.data_location=='face_centered' or self.mapping=='faces') instead of just comparing values size and n_faces.

Can you provide a MCVE to repoduce the bug?

[erogluorhan]

Great catch! Actually, AFAIK, #nodes cannot be equal to #edges in any grid model output in our case, but #nodes might be equal to #faces for some. I remember HEALPix might be so. I think the changes suggested here make sense and should be the way forward.

[Sevans711]

Sounds good! I will plan to fix using the self._face_centered() syntax since that seems to be appropriate for internal use.

I spent a little bit of time looking into examples where this might occur. This will be useful for adding regression tests after this fix.

n_nodes == n_faces

Healpix seems to always have 2 more nodes than faces. However, looking at the Healpix grid and thinking "how could I make this into a case where n_faces == n_nodes" helped me realize it is possible by adding some nodes to subdivide faces. Even though the result isn't a proper healpix grid anymore, I believe it would still be supported by uxarray.

As a small example, here n_nodes == n_faces == 6:

Without the blue node, there are instead 5 nodes and 4 faces. With that 1 node added, the total number of faces went up by 2, causing n_nodes == n_faces == 6. Even though this example is small, the principle should be possible to apply to larger grids as well: subdividing faces might lead to n_nodes==n_faces, if the grid originally starts with n_faces <= n_nodes.

n_nodes == n_edges

Meanwhile, it is hard to think of a realistic model where n_nodes == n_edges. Although, uxarray does technically support single-face grids (such as the single triangle grid example in the docs), which will always have n_nodes == n_edges. A single-triangle grid may work well for regression testing purposes here.

n_faces == n_edges

Finally, I currently can't think of any examples where n_faces == n_edges. I'm not sure if it is actually impossible, or if I am just not being clever enough. If there is an example like that, it would make sense to add it for testing purposes, but I don't think it's necessary to figure that out in order to solve this issue.

[erogluorhan]

Thanks for further looking into this! Adding some regression tests makes sense to me.