Elastic Beanstalk: Triple Deployment required for Numpy and PyMatgen

[tskluzac]

Doozy of a bug. If I deploy the service to elastic beanstalk on a fresh environment, it fails with an 'unable to cythonize' error.

Then if I uncomment pymatgen and numpy from the requirements and push to github (the service is automatically redeployed via CodePipeline), then the service runs, albeit is missing 2 reqs.

Then I can add back numpy and push (and it's redeployed).

Then I can add back pymatgen and push (and it's redeployed WITH all necessary components).

I don't want to have to push code 3 times for this to correctly deploy. Let's find the solution.

[tskluzac]

Update: I guess you can install both Numpy and Pymatgen in a second pass.

I wonder if there's a way to do a two-stage pip install for requirements? But this still seems like an odd bug that only services when trying to install via EB...