I’ve been told that
python package management is bad. I have seen some really bad practices online, asking you to run commands here and there without an understanding of the bigger picture, what they do and sometimes with escalated privileges.
Along the years, I have compiled a list of practices I follow, and a list of tools I use. I hope to be able to share some of the knowledge I’ve acquired and show you a different way of doing things. You might learn about a new tool, or a new use for a tool. Come along for the ride !
As most know, Python is an interpreted programming language. I am not going to go into the details of the language in this post, I will only talk about management.
If you want to develop in Python, you need to install libraries. You can find some in your package manager but let’s face it
pip is your way.
The majority of Linux distributions will have Python installed as a lot of system packages now rely on it, even some package managers.
Okay, this is the last time I actually use the system’s Python. What ? Why ? You ask !
I introduce you to pyenv. Pyenv is a Python version management tool, it allows you to install and manage different versions of Python as a user.
Beautiful, music to my ears.
Let’s get it from the package manager, this is a great use of the package manager if it offers an up to date version of the package.
sudo pacman -S pyenv
If you’re not using an Archlinux based distribution follow the instructions on their webpage.
Alright ! Now that we’ve got ourselves pyenv, let’s configure it real quickly.
Following the docs, I created
~/.config/fish/config.d/pyenv.fish and in it I put the following.
# Add pyenv executable to PATH by running # the following interactively: set -Ux PYENV_ROOT $HOME/.pyenv set -U fish_user_paths $PYENV_ROOT/bin $fish_user_paths # Load pyenv automatically by appending # the following to ~/.config/fish/config.fish: status is-login; and pyenv init --path | source
Open a new shell and you’re all ready to continue along, you’re all locked, loaded and ready to go!
Setup the environment
This is the first building block of my environment. We first start by querying for Python versions available for us.
pyenv install --list
Then, we install the latest Python version. Yes, even if it’s an upgrade, I’ll handle the upgrade, as well, as we go along.
Set everything up to use the new installed version.
First, we set the global Python version for our user.
pyenv global 3.9.5
Then, we switch our current shell’s Python version, instead of opening a new shell.
pyenv shell 3.9.5
That was easy. We test that everything works as expected by checking the version.
Now, if you do a
which on the
python executable, you will find that it is in the
pyenv shims' directory.
In the future, the upgrade path is exactly the same as the setup path shown above. You query for the list of Python versions available, choose the latest and move on from there. Very easy, very simple.
pip is the package installer for Python.
At this stage, you have to understand that you are using a Python version installed by pyenv as your user. The pip provided, if you do a
which, is also in the same shims directory.
pip at this stage as a user is better than running it as root but it is also not touching your system; just your user. We can do one better. I’m going to use
pip as a user once !
I know, you will have a lot of questions at this point as to why. You will see, patience is a virtue.
Meet pipx, this tool is the amazing companion for a DevOps, and developer alike. Why ? You would ask.
It, basically, creates Python virtual environments for packages you want to have access to globally. For example, I’d like to have access to a Python LSP server on the go. This way my text editor has access to it too and, of course, can make use of it freely. Anyway, let’s cut this short and show you. You will understand better.
Let’s use the only
pip command as a user to install
pip install --user pipx
You are setting yourself up for a world of hurt if you use
pip or run it as
root. ONLY run commands as
root or with escalated privileges when you know what you’re doing.
As I gave the LSP server as an example, let’s go ahead and install it with some other Python packages needed for global things like emacs.
pipx install black pipx install ipython pipx install isort pipx install nose pipx install pytest pipx install python-lsp-server
Now each one is in it’s own happy little virtual environment separated from any other dependency but its own. Isn’t that lovely ?
If you try to run
ipython, you will see that it will actually work. If you look deeper at it, you will see that it is pointing to
~/.local/bin/ipython which is a symlink to the actual package in a pipx virtual environment.
After you set a new Python version with pyenv, you simply reinstall everything.
And like magic, everything get recreated using the new version of Python newly set.
Now that pipx is installed, let’s go head and install something to manage our Python virtual environments on-demand, for use whenever we need to, for targeted projects.
Some popular choices people use are Pipenv, Poetry, virtualenv and plain and simple python with the
You’re welcome to play with all of them. Considering I use fish as my default shell, I like to use virtualfish.
Let’s install it.
pipx install virtualfish
This offers me a new command;
vf, I can create Python virtual environments and they will all be saved in a directory of my choosing.
Let’s create one for Ansible.
vf new ansible
This should activate it. Then, we install Ansible.
pip install ansible molecule docker
At this stage, you will notice that you have
ansible installed. You will also notice that all the pipx packages are also still available.
If you want to tie virtualfish to a specific directory, use
To upgrade the Python version of all of our virtual environments, virtualfish makes it as easy as
And we’re done !
At this stage, you have an idea about the tools I use and where their scope falls. I like them because they are limited to their own scope, each has its own little domain where it reigns.
- I use pyenv to install and manage different versions of Python for testing purposes while I stay on the latest.
- I use pipx for the commands that I need access to globally as a user.
- I use virtualfish to create one or more virtual environment per project I work on.
With this setup, I can test with different versions of Python by creating different virtual environments with different version each, or two versions of the tool you’re testing as you keep the Python version static. It could also be different versions of a library, testing forward compatibility for example.
At each step, I have an upgrade path to keep all my environments running the latest versions. I also have a lot of flexibility by using
requirements.txt files and others for development sometimes or even testing.
As you can see, with a little bit of knowledge and by standing on the shoulders of giants, you can easily manage a Python environment entirely as a user. You have full access to a wide array of Python distributions to play with. Endless different versions of packages, globally and locally installed. If you create virtual environments for each of your projects, you won’t fall in the common pitfalls of versioning hell. Keep your virtual environments numerous and dedicated to projects, small sets, and you won’t face any major problems with keeping your system clean yet up to date.