Goal: Suggest Bringing together concepts from the workshop, list suggested best practices and summarize workshop.

...

• If not using Miniconda, use a module version of Python. Do not rely on a the general System/compute node version.

• Install packages within the Conda environment. Do not rely or try and manage packages installed under ~/./local/lib/pythonX.Y.

• Be aware if relying on what is under ~/./local/lib/pythonX.Y then updates to packages could effect other Conda environments that use them.

• After creating/activating an environment to setup/update, set the PYTHONUSERBASE environment variable to force to only using what is installed directly within the Conda environment.

• For individual research tasks, create self-contained Conda environments to enable reproducibility and sharing.

Summary

 Looked at:

• How using different versions of Python pip installs into different locations.

• How Conda environments and pip installs interplay together and using the PYTHONUSERBASE environment variable to force installs into the Conda environment.

• A general process for creating a shared Conda environment with pip installs within a project.

• The basics on creating a module file to expose a Conda environment for general use across a project.  

...

...