fix: handle incompatible coords in merge_cubes

[clausmichele]

@LukeWeidenwalker, in load_stac, stackstac loads many additional metadata as data variables/coordinates. It is an advantage when we want to have a complete overview on the dataset but in this case creates an issue when trying to merge different datasets, An example is in this notebook I recently created:
https://github.com/clausmichele/openeo-python-client/blob/add_stac_notebook/examples/notebooks/Client_Side_Processing/client_side_processing_STAC_data_fusion.ipynb

This fix removes the unnecessary coordinates if the merge fails and tries to merge the datacubes again. I spent a lot of time trying to find a more elegant solution, but I couldn't.

[codecov]

Codecov Report

Merging #148 (a3d3559) into main (dfee6f6) will not change coverage.
Report is 1 commits behind head on main.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #148   +/-   ##
=======================================
  Coverage   78.80%   78.80%           
=======================================
  Files          27       27           
  Lines        1175     1175           
=======================================
  Hits          926      926           
  Misses        249      249           

Files Changed	Coverage Δ
...cesses_dask/process_implementations/cubes/merge.py	96.10% <ø> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[LukeWeidenwalker]

Ah, thanks for reporting this - I had a similar problem in #147, but there I could just use xr.merge, which doesn't seem to have this problem. I'm still a bit confused on why this wouldn't work though, I would have imagined combine_by_coords to just create the missing coord - do you think this is something worth raising with xarray?

Regardless, happy to go for a workaround like you propose - if you could please also add a test case to cover this failure mode, that would be much appreciated - thanks!

[clausmichele]

I tried also to write the two datasets to netCDFs, read them again with xArray and then it worked fine, which is so strange.
I agree that it would be good to create a minimal working example and open an issue for xarray.

[clausmichele]

@LukeWeidenwalker the special case causing issues is when there's an additional coordinate which has different values in the two cubes we want to merge. I found out that adding compat="override" solves the issue without the need to remove all the coordinates that are not in common. I also added a test for this!

[LukeWeidenwalker]

Ah, good find - thanks for the fix!