for Python 2

Python 2 has no built-in mechanism for virtual environments, we use mkvirtualenv.

To install it, in the default python environment (which means grab a shell, and it's done) run pip install virtualenv and also pip install virtualenvwrapper.

Virtualenv short howto

First, have a $WORKON_HOME env var. Better put this in your .bashrc

export WORKON_HOME=~/yourpath-to-python-envs

Useful basic commands

First the super-basics:

• Create an environment: mkvirtualenv data-analysis
• List environments: lsvirtualenv
• Switch to an environment: workon data-analysis
• Search and install packages: pip search ... and pip install ...
• When in an environment, list packages: lssitepackages

Less basic stuff

It's more about automating:

Automatically install package

Add the command to be executed after each mkvirtualenv to the appropriate postmkvirtualenv:

echo 'pip install sphinx' >> $WORKON_HOME/postmkvirtualenv

Prompt configuration

function _prompt_pyvenv() {
  if [ -n "${VIRTUAL_ENV}" ]; then
    echo "(`basename ${VIRTUAL_ENV}`)"
  else
    echo '' 
  fi
}

and set your prompt to be

PROMPT='%{$fg[green]%}$(_prompt_pyvenv)%{$reset_color%}'

I personally like it as right prompt (thank you zsh)

for Python 3.3 and above

Python 3.3 has a built-in venv (virtualenvs). Let's use it.

Create a venv

I've create a small custom zsh function called mkvenv which does the following:

1. create (using pyvenv) and activate the environment
2. install distribute and pip
3. deactivate and cleanup the setuptools-?.?.?.tar.gz source

Usage:

mkvenv my-nice-environment

activate the venv

A custom shell function acton (reminiscent of workon) activates an environment and sets up the right prompt.

Usage:

acton my-nice-environment

provision the env

You can use the ordinary

pip install ipython

or use the package list I prepared:

venv-provision data-science

Both of these commands autocomplete, so check for your options