Fix ray test

[TomTranter]

No description provided.

[codecov]

Codecov Report

Merging #87 (9251a1e) into main (6db7aa0) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #87      +/-   ##
==========================================
+ Coverage   97.38%   97.39%   +0.01%     
==========================================
  Files          21       21              
  Lines        1299     1307       +8     
==========================================
+ Hits         1265     1273       +8     
  Misses         34       34              

[Saransh-cpp]

Debugging GitHub Actions is weird 😅

Summary -

1. For some reason, ray requires liionpack to be installed in the given environment as a library (I have no idea why).
2. Now the workflows install liionpack using python setup.py install before running the tests.
3. Same for the conda environment.
4. Conda doesn't care about resolving conflicts between dependency versions while installing a library from source (setup.py).
5. distributed requires dask==2021.10.0 and because of the above point, dask had to be pinned everywhere.

Edit: 4th and 5th point -> Use "dask[complete]"

[TomTranter]

Thanks for your work on this. I understand that when installing dask in conda the distributed package is included. I don't particularly like the idea of pinning dask so would we fix all our problems if we just didn't test the pip install?

[wigging]

Do you need to explicitly provide distributed for the pip install? Running python -m pip install "dask[complete]" should install everything needed for Dask. More info at https://docs.dask.org/en/latest/install.html

[Saransh-cpp]

Thanks, @wigging @TomTranter! Using "dask[complete]" worked!

[wigging]

Why is dask[complete] needed for the conda environment? Just using dask in the environment file works fine for me. I think dask[complete] is just needed for the pip install.