An Introduction to Python Virtualenvs

When you are developing python you’ll end up using third party libraries. The preferred way to acquire these is by using pip.

pip out of the box will install things as system packages. While this is helpful there are two problems with this:

  • Not every library you’ll download should be installed as a system package
    • You might need specific versions of specific libraries
    • You might have bad packages that don’t uninstall cleanly
  • You will need to install every package using sudo which can be annoying and cumbersome.

Luckily there’s an easy way around this. virtualenv creates virtual python environments with their own isolated set of packages.

If you are on a Mac, I suggest the following for your environment:

Note: If you are not on a Mac you can omit the Homebrew step. Unfortunately if you are on Windows, I’m not sure what good options are other than to use something like VirtualBox.

Installing Homebrew

Homebrew is a commandline package manager for OS X. It’s more often than not what you want to use to install any system level packages and tools like, python, ruby or wget.

The installations is fairly simple:

ruby -e "$(curl -fsSL"

More details are available on the Homebrew homepage.

Once this is installed you’ll want to brew a python installation (in favor over the default python that ships with OS X):

brew install python

Installing virtualenv_wrapper

virtualenv has a nice companion package called virtualenv_wrapper. Installing it will give you a rich set of tools to switch between virtual python environments.

To install this tool:

sudo pip install virtualenvwrapper

Note: This is the one time where you want to install something with pip while using sudo. The reason being is that virtualenvwrapper is something that you do want as part of your system packages.

You’ll want to add the following to your .profile (or .bashrc or .zshrc):

export WORKON_HOME=$HOME/.virtualenvs
source /usr/local/bin/

More details are found here.

You’ll want to log out and log back in, or run a command like: source ~/.bashrc. This will give you all the special features of virtualenv_wrapper into your shell.

Working on projects

Most of this is covered in the introduction to virtualenv. Let’s say you have a project called ponies and that project lives in ~/Projects/ponies you can now create a virtualenv just for ponies:

mkvirtualenv ponies -a ~/Projects/ponies

Now, whenever you need to work on ponies you can do the following:

workon ponies

This will automatically use a special ponies-only version of python, any packages that you pip install will only be available to ponies and you won’t need to use sudo.

Furthermore if things go south you can always do the following:

deactivate  # This takes you out of the ponies virtualenv
rmvirtualenv ponies
mkvirtualenv ponies -a ~/Projects/ponies

Good as new.

Final Thoughts.

This may seem like a lot of work at first. After all you just want to write code. That’s fine, you can probably skip this. As you work with more python code, and with multiple projects, you may then see the need for something like virtualenv. You can more than likely retroactively virtualenv-ize your development environment.

This tutorial is something I recommend for all new python engineers as it generally will cause hardships at inconvenient times. May as well take care of this stuff now.

The world is your oyster.