search

a-r-j / graphein

Protein Graph Library
https://graphein.ai/
MIT License
1.03k stars 131 forks source link

Simplify environment.yml to minimal set of packages required #23

Closed ericmjl closed 2 years ago

ericmjl commented 4 years ago

Issue

Currently environment.yml and ubuntu_environment.yml are duplicates of one another, and could be simplified into a single file that uniquely specifies environment dependencies.

Proposal

I'd like to propose unifying the two environment spec files into an environment-dev.yml file, which only specifies the minimal set of packages necessary for development/hacking on graphein. Deployment onto PyPI and conda forge later can allow for installation with dependencies automatically pulled in. This would be in tandem with the removal of the current *environment.yml files.

A proposed starter environment spec could be:

name: graphein-dev
channels:
- conda-forge
dependencies:
- python=3.8
- biopandas
- pandas
- networkx
- numpy

Conversion to DGL/PyTorch/JAX can be enabled once the library of code has been rewritten.

a-r-j commented 4 years ago

The two environment.yml files are legacy at the moment. The reason for there being two of them was a result of getting RTD to build the docs properly - some of the packages install different builds depending on platform.

That being said, I do agree with the need for simplification & think it's timely given the difficulties people are experiencing. Happy to move forward on this.