`numba` isn't included in `v0.6.2` conda-forge feedstock recipe

[tomvothecoder]

numba is introduced as a new dependency in v0.6.2.

The setup.py file includes numba, so the PyPI release works fine: https://github.com/pangeo-data/xESMF/blob/860311345212c3e5f52d08335eb758cf903d6787/setup.py#L23-L31

However, the conda-forge feedstock recipe does not include numba: https://github.com/conda-forge/xesmf-feedstock/blob/fe72c86a62ed48b698eeb8e58acc50817bd954fc/recipe/meta.yaml#L16-L28

This results in environments and other packages that use xesmf to break with the following error:

File ~/miniconda3/envs/xcdat_test_dev/lib/python3.10/site-packages/xesmf/smm.py:7, in <module>
      4 import warnings
      5 from pathlib import Path
----> 7 import numba as nb
      8 import numpy as np
      9 import sparse as sps

ModuleNotFoundError: No module named 'numba'

This user has to manually install numba into their environment and packages that include xesmf must also include numba as a dependency (it should be installed implicitly via xesmf).

[aulemahal]

Sorry about that. Numba is direct dependency of recent versions of sparse, which is an explicit dependency of xESMF 0.6.2. However, I didn't know that older versions of sparse did not use numba. This issue was solved with #168, but we have not released a new xESMF since.

[tomvothecoder]

No problem @aulemahal, I just wanted to make sure this was noted! I appreciate the quick update.

[jbusecke]

I am running into the same issue over at https://github.com/jbusecke/cmip6_preprocessing/runs/7102175665?check_suite_focus=true. Any chance for a bugfix release + conda-forge bump? Or is anything blocking that. Happy to help.

[aulemahal]

@raphaeldussin I would suggest to publish a new release of xESMF (0.6.3), now that #175 is merged? This issue was fixed in #168, but I wanted to wait a successful conda release to close it.

[jbusecke]

Much appreciated.

[raphaeldussin]

@aulemahal go for it!

[huard]

0.6.3 now available on conda.