Allow masked data in coords

[ukmo-ccbunney]

🚀 Pull Request

Closes: #3584

Description

Fixes two related issues:

• Allows masked data in a Cell to be hashed
• Correctly handles masking and dtype when merging scalar AuxCoords with masked data

Whilst adding test, the following files have been converted from unittest to pytest:

• unit/coords/test_Cell.py
• test_merge.py [moved to unit/merge/test_merge.py]

---

Consult Iris pull request check list

Add any of the below labels to trigger actions on this PR:

• benchmark_this Request that this pull request be benchmarked to check if it introduces performance shifts

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.82%. Comparing base (9017fd2) to head (802eeca).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6468   +/-   ##
=======================================
  Coverage   89.81%   89.82%           
=======================================
  Files          90       90           
  Lines       23761    23782   +21     
  Branches     4422     4427    +5     
=======================================
+ Hits        21340    21361   +21     
  Misses       1672     1672           
  Partials      749      749           

☔ View full report in Codecov by Sentry.
📢 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.

[ukmo-ccbunney]

Mostly working, but outstanding problem with loading a scalar coord with a single masked value.
Whilst this does load ok and get's added to the cube, the dtype of the AuxCoord is always np.float64, regardless of the original dtype of the scalar coord in the netCDF file.

I am not sure yet why this works for an array of all masked values but not a scalar masked value...

[ukmo-ccbunney]

Commit dff7cb8 addresses issue where the original dtype of a scalar masked variable was lost.

The dtype of the AuxCoord was being set to np.float64 in all cases as this is the dtype of the np.ma.masked singleton value. Added special handling for this case to return a new 1-element masked array of the correct dtype, rather than the np.ma.masked instance.

[ukmo-ccbunney]

There is only one small snag now w.r.t. the dtype returned when working with Cells:

# Scalar coord points now have correct dtype
coord = cube.coord('realization')
coord.points
>> masked_array(data=[--],
>>           mask=[ True],
>>      fill_value=np.int64(999999),
>>            dtype=int16)

# ... but accessing them via the `cell` method still looses it:
coord.cell(0).point
>> masked

coord.cell(0).point.dtype
>> dtype('float64')   # <-- This is the dtype of np.ma.masked, not the coord

[trexfeathers]

Thanks for looking into this @ukmo-ccbunney!

[trexfeathers]

Awesome work @ukmo-ccbunney. Just one change left I think

[trexfeathers]

Superb! Thanks again 🚀