Question

I’m quite fond of Python’s virtualenv, which facilitates maintenance of separate Python configurations. I’m considering embedding Python into a C++ application and was wondering how the embedded Python would behave with respect to virtual environments.

In particular, I’m intersted in knowing if it’s possible to “select” a virtual environment based on some user-defined setting (e.g. by naming the virtual environment of interest in a configuration file).

Answer 1

The virtualenv documentation includes a Using virtualenv without bin/python section that hints at how to configure a virtual environment once the interpreter is already running.

To avoid hardcoding the path to the activate_this.py script, I use the following snippet:

def resolve_virtual_environment(override=None):
    """Fetch the virtual environment path in the
       process' environment or use an override."""
    path = os.getenv('VIRTUAL_ENV')
    if override:
        path = os.path.join(os.getcwd(), override)
    return path

def activate_virtual_environment(environment_root):
    """Configures the virtual environment starting at ``environment_root``."""
    activate_script = os.path.join(
        environment_root, 'Scripts', 'activate_this.py')
    execfile(activate_script, {'__file__': activate_script})

And you can use it like so:

if __name__ == '__main__':
    # use first argument is provided.
    override = None
    if len(sys.argv) > 1:
        override = sys.argv[1]
    environment_root = resolve_virtual_environment(override)

You can fetch the override value from a configuration file or something instead of from the command-line argument.

Note that you can only still use a single virtual environment pre-process.

Note: in contrast with using the interpreter bundled in the virtual environment, you have access to the packages installed for the interpreter you started. For example, when using a globally-installed Python, you will have access to the globally-installed packages.

Also make sure that you use a Python interpreter with a version that matches whatever version you used to create the virtual environment to make sure that the standard library (copied into the virtual environment) version matches the Python interpreter version.