fix(writers): accept fmt as an alias for to_netcdf format

[Manny7717]

What

Fixes #2685: save_datasets(writer='cf', fmt='nc', ...) raised TypeError: Dataset.to_netcdf() got an unexpected keyword argument 'fmt'. The fmt kwarg was forwarded verbatim into xarray's Dataset.to_netcdf(), which expects format, not fmt.

Change

CFWriter.save_datasets now pops fmt from the forwarded kwargs and passes it through as format, mapping the common 'nc' shorthand to 'NETCDF4'. Passing both fmt and format with different values raises a clear error.

Tests

• New test_save_array_fmt_nc_alias: saves with fmt='nc' and verifies the file round-trips. Fails on the previous implementation (TypeError; verified) and passes with the fix.
• Full test_cf.py: 32 passed. ruff clean.

[djhoese]

I'm confused by this pull request and even @mraspaud's original issue. Where is the format kwarg supposed to go? Why was it specified in @mraspaud's use case? Was he trying to pass the netcdf model to use like in xarray's to_netcdf?

https://docs.xarray.dev/en/latest/generated/xarray.Dataset.to_netcdf.html#xarray.Dataset.to_netcdf

Because in the issue Martin was passing format="nc" which is not an accepted value to that function in xarray anyway.

Regardless, why does Satpy need to use fmt instead of format? Also, maybe we shouldn't pass arbitrary kwargs through to xarray with **kwargs but instead have an netcdf_kwargs= or xarray_kwargs or to_netcdf_kwargs that gets passed along?

CC @pnuu @mraspaud @gerritholl and maybe @sfinkens (I think you did CF writer stuff?)

[codecov]

Codecov Report

❌ Patch coverage is 92.85714% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 96.33%. Comparing base (17bf08d) to head (116a299).
⚠️ Report is 39 commits behind head on main.

Files with missing lines	Patch %	Lines
satpy/writers/cf_writer.py	80.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3453      +/-   ##
==========================================
- Coverage   96.34%   96.33%   -0.02%     
==========================================
  Files         465      465              
  Lines       59151    59153       +2     
==========================================
- Hits        56990    56983       -7     
- Misses       2161     2170       +9     

Flag	Coverage Δ
behaviourtests	3.58% <0.00%> (-0.01%)	⬇️
unittests	96.42% <92.85%> (-0.02%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pnuu]

The original original issue was in Trollflow2, where we pass the format among **kwargs to scn.save_dataset() along with other kwargs that are not pop'd. I think we should be more explicit there what are passed to Satpy.

Some relevant points in Trollflow2:

• https://github.com/pytroll/trollflow2/blob/main/trollflow2/plugins/__init__.py#L367-L370
• https://github.com/pytroll/trollflow2/blob/main/trollflow2/plugins/__init__.py#L277

I actually don't follow anymore what's happening in the Trollflow2 code, and one of my goals with the Claude subscription would be to simplify it as much as possible. Remove all in-place product list operations and such.