Managing Virtual Environments on Windows

Posted on Thu 30 April 2020 in Python • 3 min read

Managing virtual environments for Python on Windows was never straightforward in my experience. There is so many different tools available (which is fantastic!), it's difficult to find the right combination for Python projects. This post is for the combination of tools that work for my application, if there are any recommendations or improvements on this post, please let me know!

The Tools

Prior to using these tools, Anaconda was the go-to. Admittedly, the only reason to stop using was disk space. After using Anaconda on every project for a few months, around 30GB of space was being taken up for conda environments. As a lot of my projects involve using Pandas, Gooey & PyInstaller, when packaging these executables up they would come out bigger than expected (250MB vs 25MB). This is a well documented issue across the internet.

The Workflow

Pyenv-win

Pyenv (specifically pyenv-win), is used in this workflow to manage multiple versions of Python. To make interfacing with pyenv-win & subsequent tools, let's include system environment variables. This let's us call for the enabled Python version executable & installed packages (virtualenv, poetry, etc) from anywhere on the PC.

Ensure to setup pyenv-win as per the instructions at: https://github.com/pyenv-win/pyenv-win#installation. I personally installed via the zip file method. Thank you to read Julian for raising this.

Firstly, let's set up a variable of where the pyenv-win installation lives as PYENV, which I placed in C:\Users\%USER%\.pyenv\pyenv-win. Followed by a variable which will hold the current version of Python that pyenv will use as PYENV_VERSION, the value of this variable will be dependant on the installed version folder name in User/.pyenv/pyenv-win/versions/, which is currently 3.8.1-amd64. These are set up to use with the PATH variable, which is integrate with Windows and makes all files included in the paths listed in PATH available to anywhere on the PC.

The result is:

Pyenv Variables

Now we need to add in new variables to the PATH variable. The two variables we want to add are %PYENV%\versions\%PYENV_VERSION% and %PYENV%\versions\%PYENV_VERSION%\Scripts. This enables us to access both the python.exe & pip.exe of the selected version from anywhere, and then when we install packages with pip, we can access packages that have been installed. This ends up looking like:

Path Variables

Following this, open command prompt, check that python --version matches the version variable selected, and then install packages required. The packages that I install are virtualenv & poetry. All project specific packages are installed in their own virtual environment (more on this below).

Virtualenv

Once virtualenv is installed and accessible from anywhere, whenever a new folder is created for a project, you can run virtualenv --prompt folder_name .env. This command will create a new virtual environment in a folder named .env, when using with VS Code, you can select this as the current environment with the Python extension. This setting lives in it's own folder .vscode/settings.json.

To install new packages inside the virtual environment, either run .env/Scripts/activate.bat or open a new terminal in VS Code once selected. That's it!

Poetry

The use-case for Poetry is solely for packaging & distributing packages, something personally I don't do very much. To read further around packaging projects with Poetry, I've written a post at Packaging Python Packages with Poetry