This weekend I upgraded one of my machines to Mountain Lion. My system-level Python is pretty bare: I typically just keep pip, virtualenv, virtualenvwrapper, and a few associated libraries installed there. So, as I worked through a few errors I decided to follow this advice from the Hitchhiker’s Guide to Python and begin using Homebrew to manage my Python installation. What follows is a brief list of steps I used to troubleshoot the issues I ran into during the process:
First, we’ll want to install Python:
$ brew install python
Homebrew installs python, easy_install, and pip, so you’ll bootstrap quickly.
Next, add “
export PATH=/usr/local/bin:/usr/local/share/python:$PATH” to your
~/.bash_profile. Homebrew describes
/usr/local/share/python as an optional addition, but things like virtualenv, virtualenvwrapper, etc., get added there, so you’ll definitely want to include it.
Now, if you previously installed virtualenv, it’s probably pointing to
/usr/bin/python. You can verify this by just taking a look at the hashbang in, for example,
/usr/local/bin/virtualenv. Let’s get rid of those so they don’t shadow the versions we’re about to install via our brewed pip:
$ rm /usr/local/bin/virtualenv*
Now we can install the virtualenv ecosystem. You can do this pretty quickly with just:
$ pip install virtualenvwrapper
This will install virtualenv, virtualenvwrapper, and virtualenv-clone.
Finally, add “
source /usr/local/share/python/virtualenvwrapper_lazy.sh” to your
~/.bash_profile. You can use the non-lazy version if you want, but on my machine new shells spin up a lot faster with the lazy version.
This process solved an issue I was experiencing where virtualenvwrapper seemed to have run correctly on login, but I was unable to use
virtualenv itself. That problem presented itself as a
pkg_resources.DistributionNotFound exception (traceback). It was apparent that virtualenv was relying on the system Python, but I couldn’t figure out why. The legacy
/usr/local/bin/virtualenv was the culprit.
Thanks to Thomas Wouters (Yhg1s on #python) for his help in figuring this out.