In this post I’ll go over my attempt to setup virtual environments for Python development. Most Python users probably don’t want to use virtual environments and should just set up a single user environment that works for their needs. However, if you are working on writing one or more Python modules or packages, these tools are geared towards creating isolated Python environments for each project. This is very useful for keeping track of such things as the minimal Python requirements for each project.
Assuming you want to proceed, my goal is to setup and usevirtualenvwrapper (see virtualenvwrapper docs for more info), a set of shell tools wrapped around the virtualenv package. Both the Doug Hellman post and the simonsoftware post provide some motivation for the development and use of the wrapper formulation– in simple termsvirtualenvwrapper provides some shortcuts for common actions. So, I’ve decided to use the wrapped version.
Basic install
Following the virtualenvwrapper basic install, the installation process is to install virtualenv and virtualenvwrapper using pip. I have already installed virtualenv, as can be seen by using pip show:
$ pip show virtualenv
---
Name: virtualenv
Version: 1.11.6
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires:
If you need to install, the command is – I install as a user:
$ pip install --user virtualenv
Next, install the virtualenvwrapper:
$ pip install --user virtualenvwrapper
Now a pip show should show something like:
$ pip show virtualenvwrapper
---
Name: virtualenvwrapper
Version: 4.3.1
Location: /home/cstrelioff/.local/lib/python2.7/site-packages
Requires: virtualenv, virtualenv-clone, stevedore
Finally, we add some information to our ~/.bashrc file – I added to the end of the file, as usual for such things. The actual contents for you will be different.
- I wanted my virtual environments in ~/virtenvs/ and I had already made that directory.
- I want to keep my active projects in ~/Projects-Active/, an existing directory.
- Finally, because I installed as a user, the path to myvirtualenvwrapper.sh is in ~/.local/bin/, as indicated by usingwhich:
$ which virtualenvwrapper.sh
/home/cstrelioff/.local/bin/virtualenvwrapper.sh
Putting all of that specific information together, I added the following to ~/.bashrc:
# where to store our virtual envs
export WORKON_HOME=$HOME/virtenvs
# where projects will reside
export PROJECT_HOME=$HOME/Projects-Active
# where is the virtualenvwrapper.sh
source $HOME/.local/bin/virtualenvwrapper.sh
After saving the changes, I sourced the file to make the changes active:
$ source ~/.bashrc
Working with virtualenvs
Next up, let’s figure out how to use all this– you should also look at the simonsoftware post for another take on this. The main command to remember is workon, as in I’m going to work on this project. However, if we try it now we get nothing:
$ workon
$
We need to make a virtual environment. So, let’s make one:
$ mkvirtualenv test_env01
New python executable in test_env01/bin/python
Installing setuptools, pip...done.
We can use pip list to see the packages available:
(test_env01)$ pip list
argparse (1.2.1)
pip (1.5.6) setuptools (3.6) wsgiref (0.1.2)
Notice that the command prompt has changed to include the environment name. If we want to install a package in this environment we use pip:
(test_env01)$ pip install pyaml
Now, a pip list gives:
(test_env01)$ pip list
argparse (1.2.1)
pip (1.5.6) pyaml (14.05.7) PyYAML (3.11) setuptools (3.6) wsgiref (0.1.2)
To deactivate the virtual environment, we type exactly what you’d expect:
(test_env01)$ deactivate
$
and we get back to the normal command prompt. However, now theworkon command will show the virtual environment that we created:
$ workon
test_env01
$
To start working on it again, simply try out the following to see everything is there:
$ workon test_env01
(test_env01)$ pip list
argparse (1.2.1) pip (1.5.6) pyaml (14.05.7) PyYAML (3.11) setuptools (3.6) wsgiref (0.1.2) (test_env01)$ deactivate
Projects in virtualenvwrapper
Finally, let’s talk about projects in virtualenvwrapper. This creates both (i) a virtual environment and (ii) a project directory in the location specified by PROJECT_HOME variable in the additions to the~/.bashrc file. Let’s try it out:
$ mkproject test_project02
New python executable in test_project02/bin/python
Installing setuptools, pip...done.
Creating /home/cstrelioff/Projects-Active/test_project02
Setting project for test_project02 to
/home/cstrelioff/Projects-Active/test_project02
(test_project02)~/Projects-Active/test_project02$
Notice that this also creates a directory and cd’s to directory– very nice! Now, if we try pip list we’ll see only the packages for a new environmnet:
(test_project02)~/Projects-Active/test_project02$ pip list
argparse (1.2.1)
pip (1.5.6) setuptools (3.6) wsgiref (0.1.2)
Switching between environments
Now we have two virtual environments setup, but only one is setup as a project. We can see both with workon:
$ workon
test_env01
test_project02
To get a sense of how this all works, let startup test_env01 and usepip list to see that PyYAML is installed:
$ workon test_env01
(test_env01)$ pip list
argparse (1.2.1) pip (1.5.6) pyaml (14.05.7) PyYAML (3.11) setuptools (3.6) wsgiref (0.1.2)
Next, while in test_env01, let’s switch to test_project02 usingworkon and look at the installed packages (no PyYAML):
(test_env01)$ workon test_project02
(test_project02)~/Projects-Active/test_project02$ pip list argparse (1.2.1) pip (1.5.6) setuptools (3.6) wsgiref (0.1.2)
Notice that the workon cd’s to the project directory. This happens because we setup test_project02 as a project and not just avirtualenv. If you use workon to switch back to test_env01 there will be no cd because there is no project file associated with that virtual environment. In practice I imagine I will always use mkproject to set things up.
Cleaning up
Finally, to clean up the example above we can use rmvirtualenv:
$ workon
test_env01
test_project02
$ rmvirtualenv test_env01
Removing test_env01...
$ rmvirtualenv test_project02
Removing test_project02...
$ workon
$
With the final workon we can see that all of our environments are gone. However, note that the directory created in PROJECT_HOMEwill not be deleted by the above– this is probably a good default behaviour. You’ll have to go delete the directory (if you want).
That’s it, hopefully some will find this useful post useful. If you have cool/better ways to use these tools leave a comment below.