Python virtual environments

Python comes with a lot, but for many things you rely on packages that others have written: requests for requests to the web, pandas for tables, matplotlib for charts. You install them with pip, Python's package manager. Without further precautions they end up in the one Python installation on your computer, which all your projects share.

The problem: all projects share everything

This works well until two projects need different things. An old project only runs with an older version of a package, while a new one needs the current version. If you install the current version for the new project, the old one stops working. On top of that, packages pile up over time, and nobody knows any more which project needs them.

A virtual environment (venv for short) solves both problems. It is a folder in the project that contains its own Python and its own place for packages, and whatever you install while it is active ends up only there.

Creating an environment

Change into the project folder and let Python create the environment. The folder is usually called .venv:

python -m venv .venv

python3 -m venv .venv

-m venv means: run the module venv, which comes with Python. Afterwards the project contains a folder .venv with a few thousand files. You do not need to look inside, and you should not change anything in it by hand.

Activating and leaving

So that python and pip use the environment, you activate it, and you do so again in every new command line window:

.venv\Scripts\Activate.ps1

source .venv/bin/activate

The prompt now starts with (.venv). As long as that is there, the environment is active, and inside it the command is simply python, on the Mac too. With deactivate you leave it again.

Installing packages

In the active environment you install packages with pip. pip list shows what is installed:

pip install requests
pip list

You do not pass the environment itself on, and you do not put it in the cloud either, because it is large and only fits your computer. What you pass on is the list of packages. pip freeze writes it to a file, and whoever gets the project creates an environment of their own and installs exactly the same versions from it:

pip freeze > requirements.txt
pip install -r requirements.txt

If something goes wrong, you simply delete .venv and create the environment again; with requirements.txt it is restored in a minute.