remove redundant attrs leaked from raw data

[rhaegar325]

Fix

Branch fix_outname_issue.

src/access_moppy/base.py

An allow-list of the table fields that describe the variable, and a shared helper applying it:

_CMOR_VARIABLE_ATTRIBUTES = frozenset({
    "standard_name", "long_name", "units", "cell_methods", "cell_measures",
    "comment", "positive", "valid_min", "valid_max",
    "flag_values", "flag_meanings", "_FillValue", "missing_value",
})

def _apply_cmor_variable_attributes(self, cmor_attrs):
    self.ds[self.cmor_name].attrs.update({
        k: v for k, v in cmor_attrs.items()
        if k in self._CMOR_VARIABLE_ATTRIBUTES and v not in (None, "")
    })

Two decisions carry the safety of this change.

The filter reads the table entry, not the variable's final attributes. Its reach is therefore one assignment: coordinates, units_metadata and anything inherited from the source file are set elsewhere and cannot be touched, so a name missing from the allow-list cannot silently delete them. A filter over the final attributes would have to enumerate them all, and a gap there deletes real metadata.

_FillValue and missing_value must be in the list. They are not table fields at all — _get_variable_entry injects them (defaulting to 1e20, since no CMIP6 or CMIP7 entry carries either), and base.py:2388 reads _FillValue back from the variable attributes to create the netCDF variable. wcrp_cmip7 requires both at severity HIGH with a constant of 1e20. Dropping them would write files with no fill value and two weight-3 compliance failures — worse than the leak.

The allow-list is a complete partition, not a guess: the CMIP6 and CMIP7 tables carry 16 distinct fields between them, plus the 2 injected ones; 13 are written and 5 are directives.

src/access_moppy/{atmosphere,ocean,sea_ice}.py

Each of the three attrs.update(...) blocks becomes self._apply_cmor_variable_attributes(cmor_attrs). _drop_stale_range_attributes still runs immediately afterwards and is unchanged.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.6%. Comparing base (43f4d94) to head (5f72cf3).

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #688   +/-   ##
=====================================
  Coverage   78.6%   78.6%           
=====================================
  Files         41      41           
  Lines       9018    9021    +3     
  Branches    1687    1687           
=====================================
+ Hits        7091    7094    +3     
  Misses      1592    1592           
  Partials     335     335           

Flag	Coverage Δ
unit	78.6% <100.0%> (+<0.1%)	⬆️

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.