Normally virtualenv creates environments in the current directory. Unless you’re intending to create virtual environments in C:\Windows\system32 for some reason, I would use a different directory for environments.
You shouldn’t need to mess with paths: use the activate script (in <env>\Scripts) to ensure that the Python executable and path are environment-specific. Once you’ve done this, the command prompt changes to indicate the environment. You can then just invoke easy_install and whatever you install this way will be installed into this environment. Use deactivate to set everything back to how it was before activation.
Example:
c:\Temp>virtualenv myenv New python executable in myenv\Scripts\python.exe Installing setuptools..................done. c:\Temp>myenv\Scripts\activate (myenv) C:\Temp>deactivate C:\Temp>
Notice how I didn’t need to specify a path for deactivate – activate does that for you, so that when activated “Python” will run the Python in the virtualenv, not your system Python. (Try it – do an import sys; sys.prefix and it should print the root of your environment.)
You can just activate a new environment to switch between environments/projects, but you’ll need to specify the whole path for activate so it knows which environment to activate. You shouldn’t ever need to mess with PATH or PYTHONPATH explicitly.
If you use Windows Powershell then you can take advantage of a wrapper. On Linux, the virtualenvwrapper (the link points to a port of this to Powershell) makes life with virtualenv even easier.
Update: Not incorrect, exactly, but perhaps not quite in the spirit of virtualenv. You could take a different tack: for example, if you install Django and anything else you need for your site in your virtualenv, then you could work in your project directory (where you’re developing your site) with the virtualenv activated. Because it was activated, your Python would find Django and anything else you’d easy_installed into the virtual environment: and because you’re working in your project directory, your project files would be visible to Python, too.
Further update: You should be able to use pip, distribute instead of setuptools, and just plain python setup.py install with virtualenv. Just ensure you’ve activated an environment before installing something into it.