How does uxarray compares with other ugrid libraries?

[ocefpaf]

This is not really a feature request, maybe it is better as a topic discussion? Anyway here is the gist of it:

I'm taking a tour at the many ugrid libraries out there

• ugridpy: https://github.com/Deltares/UGridPy
• xugrid: https://github.com/Deltares/xugrid
• gridded: https://github.com/NOAA-ORR-ERD/gridded
• pyugrid: https://github.com/pyugrid/pyugrid

Ugridpy depends on two c++ libraries that are quite hard to compile and upstream only provides Windows binaries.

Xugrid seems to be the most "feature complete" but it failed to load a compliant dataset (I'm still investigating that).

Gridded has almost all the functionalities of xugrid but operates on netcdf4-python objects instead of xarray.

Pyugrid is the oldest and also only operates on netcdf4-python.

Uxarray can read from multiple grid formats, only from files for now (PR #68 fixes that), but lacks some of the xugrid functionalities like plotting and high level slices.

I'd love to hear thoughts from the devs and/or comments/corrections on my assent above.

[paullric]

I'd say that UXARRAY is most similar to xugrid. The expectation is that all code will be in Python and built upon the xarray library. A key focus of this effort is on high performance algorithms for very large unstructured grids (i.e., more than 1M faces). Ultimately we want users to be able to interact with unstructured data without necessarily being aware that it is unstructured -- that is, providing all of the functionality one would expect from structured data operators. As such we're trying to be fairly ambitious with our suite of operators (pointwise/areal sampling, zonal/meridional means, cross-sections, regridding, visualization, shapefile support, etc.).

[dcherian]

Iris has experimental ugrid support too now: https://scitools-iris.readthedocs.io/en/stable/generated/api/iris/experimental/ugrid.html#module-iris.experimental.ugrid