Two negating pins for dependencies - libnetcdf and numpy

[trexfeathers]

🚀 Pull Request

libnetcdf !=4.9.1

Closes #5245, closes #5187. Thanks very much @zklaus for sorting this out in conda-forge/libnetcdf-feedstock#175 🌟. I've tested locally and stdout is no longer polluted with numerous HDF5 errors.

numpy !=1.24.3

Updating the lock files introduced errors in iris.tests.test_analysis.TestLatitudeWeightGeneration:

• test_cosine_latitude_weights_2d
• test_cosine_latitude_weights_2d_latitude_first
• test_cosine_latitude_weights_2d_latitude_last

Breadcrumb trail starts with numpy/numpy#23651. This relates to ordering in masked arrays, and having looked at the way the tests are constructed, I don't believe they are asserting for the wrong answer.

---

Consult Iris pull request check list

[trexfeathers]

numpy !=1.24.3

I can't be 100% sure if the below was originally written accounting for broken NumPy behaviour that has now been fixed. Anyone care to weigh in?

iris/lib/iris/analysis/cartography.py

Lines 559 to 569 in 1afae0d

	# Create weights for each grid point. This operation handles adding extra
	# dimensions and also the order of the dimensions.
	broadcast_dims = [x for x in lat_dims if x is not None]
	wshape = []
	for idim, dim in enumerate(lat_dims):
	if dim is not None:
	wshape.append(l_weights.shape[idim])
	l_weights = l_weights.reshape(wshape)
	broad_weights = iris.util.broadcast_to_shape(
	l_weights, cube.shape, broadcast_dims
	)

[codecov]

Codecov Report

Patch and project coverage have no change.

Comparison is base (1afae0d) 89.32% compared to head (c03fde8) 89.32%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5274   +/-   ##
=======================================
  Coverage   89.32%   89.32%           
=======================================
  Files          89       89           
  Lines       22390    22390           
  Branches     5374     5374           
=======================================
  Hits        20000    20000           
  Misses       1640     1640           
  Partials      750      750           

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[trexfeathers]

I can't be 100% sure if the below was originally written accounting for broken NumPy behaviour that has now been fixed. Anyone care to weigh in?

Although this would be rather odd, since NumPy's change is only for masked arrays, and if we had to put some sort of masked-vs-not type check in here that would seem like evidence that the problem is in NumPy and not in Iris.

[bjlittle]

@trexfeathers Just to confirm, from the PR description, are you saying that the lockfiles are generating issues with the new dependencies?

[trexfeathers]

@trexfeathers Just to confirm, from the PR description, are you saying that the lockfiles are generating issues with the new dependencies?

No, sorry for lack of clarity.

1. Sorting out the libnetcdf pin required updating the lock files.
2. Updating the lock files newly introduced NumPy v1.24.3, since it was released very recently.
3. NumPy v1.24.3 caused the test failures described above.

[bjlittle]

@trexfeathers We can watch numpy and see whether the 1.24.3 avoidance resurfaces in their next minor/patch release 👍

[trexfeathers]

Thanks @bjlittle 🙂