audit datahub `environment.yml` files for correctness

[shaneknapp]

Bug description

this will pertain only to the conda envs, not pip.

looking at datahub/deployments/datahub/images/default/envionment.yml, i discovered a wide range of how we define package versions. all of this variation can potentially create confusion and behave unexpectedly. package=version is different than package==version and in fact = puts a wildcard after package.

https://conda.io/projects/conda/en/latest/user-guide/concepts/pkg-specs.html#package-match-specifications

for all package requests that have a specific version suppled, i would strongly recommend we move to == to pin exactly to that version. for those packages with wildcards in their versioning, leave the = operator as is.

we should check these out, and clean up/fix if necessary:

./deployments/stat20/image/environment.yml
./deployments/datahub/images/default/environment.yml
./deployments/stat159/image/environment.yml
./deployments/biology/image/environment.yml
./deployments/data8/image/environment.yml
./deployments/cee/image/environment.yml
./deployments/a11y/image/environment.yml
./deployments/data100/image/environment.yml
./deployments/eecs/image/environment.yml
./deployments/data8xv2/image/environment.yml
./deployments/data8x/image/environment.yml
./deployments/astro/image/environment.yml
./deployments/julia/image/environment.yml

Environment & setup

all hubs

How to reproduce

here be dragons

[ryanlovett]

I totally agree. The versions fuzzified by = will have to be reconciled with what actually got installed in those images and then converted to ==.

The downside may be that fuzziness is actually necessary to resolve some dependencies, and we have to see how it all turns out in the built image.