Remove "edition" from GLOBAL_ATTRIBUTES_KEYS

[StephanSiemen]

As identified in issue #232, the GRIB edition should not be an global attribute key.

This is a duplicate of #233, because more changes are necessary to not have tests fail.

[codecov-commenter]

Codecov Report

Merging #234 (aa4a7bf) into master (843bf91) will decrease coverage by 37.64%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           master     #234       +/-   ##
===========================================
- Coverage   92.24%   54.59%   -37.65%     
===========================================
  Files          25       25               
  Lines        1754     1740       -14     
  Branches      206      206               
===========================================
- Hits         1618      950      -668     
- Misses        112      774      +662     
+ Partials       24       16        -8     

Impacted Files	Coverage Δ
tests/test_20_messages.py	99.25% <ø> (-0.02%)	⬇️
tests/test_30_dataset.py	100.00% <ø> (ø)
tests/test_40_xarray_store.py	4.91% <ø> (-95.09%)	⬇️
cfgrib/dataset.py	90.81% <100.00%> (-8.13%)	⬇️
cf2cdm/cfunits.py	0.00% <0.00%> (-100.00%)	⬇️
cf2cdm/__init__.py	0.00% <0.00%> (-100.00%)	⬇️
cf2cdm/cfcoords.py	0.00% <0.00%> (-100.00%)	⬇️
cf2cdm/datamodels.py	0.00% <0.00%> (-100.00%)	⬇️
tests/test_20_cfcoords.py	5.95% <0.00%> (-94.05%)	⬇️
tests/test_50_xarray_getitem.py	7.31% <0.00%> (-92.69%)	⬇️
... and 12 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 843bf91...aa4a7bf. Read the comment docs.

[iainrussell]

@StephanSiemen , do you get any test failures when you run pytest? I get the same errors as are shown un the Ubuntu py3.8 tests on github.

[FussyDuck]

All committers have signed the CLA.

[iainrussell]

I don't really like just removing tests, and some of those were testing something that was not being tested elsewhere, so I put some substitute tests in place. These are not related to the failing tests, they are only in test_40_xarray_store.py

[iainrussell]

My fear is that cfgrib is storing and using the edition so that it can be used when writing an xarray to GRIB. And it could be important.... that said, the functionality to write GRIB is broken anyway with modern xarray/pandas versions.

[iainrussell]

@alexamici , what do you think about this PR, and is the GRIB edition important for writing xarrays back as GRIB (it seems to be)?

[alexamici]

@iainrussell if you want to save the same GRIB file as the input it is needed to keep the GRIB_edition attribute. Otherwise it may be added back inside to_grib I think.

It is not strictly needed by xarray, but cfgrib probably assumes that it is present at save time.