Should we be using virtualenv (or pipenv) instead of conda?

[seawolf42]

Setting up the local environment for development includes installing conda as the virtual environment manager instead of lightweight tools like virtualenv. This has the following implications:

• Advantages
  • very powerful environment management, including handling of non-Python projects
  • extensive online documentation
• Disadvantages
  • extremely large download (can be an issue in bandwidth-limited teaching environments)
  • installation woes between multiple operating systems/versions
  • mentors may not be familiar with this tooling
  • many (most?) projects in the real world assume the user is using pip

Probably the best tooling to use is Python's built-in virtualenv command, as this exists for everyone with Python installed. A possible alternative would be to use pipenv, though I think this is not ideal as it entails installing additional tools that are directly relevant to this course.

[seawolf42]

On further thought, it probably makes a lot of sense to move virtual environments to the last step in the basic course, so that the focus is more on using git and GitHub in a project-agnostic way.

[seawolf42]

Creating a separate issue for changing the order, issue #157

[jlgimeno]

Pipenv with git 18.1 seems to be broken. See here: pypa/pipenv#2924. Workaround seems to be pipenv install git==18.0 before installing anything else in the virtual enviroment.