The End of Python's Package Management Headaches: uv

Having worked with Python for years as a data engineer, I've had the pleasure of helping many developers get started with Python. This always starts with a too long and confusing talk by me on how to manage Python project dependencies.

But now there is a better way: uv

This makes the introduction talk a bit shorter and maybe less confusing 🎉

The old maze of tools and terms

If you've ever tried to help someone get started with Python, this scenario might sound familiar:

"Okay, first install Python. No, not from the website - use pyenv. Oh, you're on Windows? Maybe try the Windows Store version. Now we need pip. Wait, why does it say 'pip not found'? Right, we need to add it to PATH. Now create a virtual environment... No, don't use venv, use virtualenv instead. Or maybe we should use conda? Actually, let's try poetry... it has lockfiles"

This maze of tools and commands has been causing problems like:

• Different ways to install Python itself (pyenv, system Python, downloaded installers)

• Confusing package management commands and tools (pip vs pipenv, poetry)

• Multiple competing tools for virtual environments (venv, virtualenv, conda)

• Dependency resolution nightmares ("It works on my machine!")

• Complex project setup with requirements.txt, setup.py, pyproject.toml

For reference see this relevant XKCD:

Source: https://www.xkcd.com/1987/

This is the way

Here's what the same process looks like with uv:

• install uv

• Run uv venv or uv init

• Start coding

That's it. No PATH configurations, no pip vs poetry debates, no virtualenv vs venv discussions. Just a single tool that handles everything from python version management to virtual environment creation to dependency management.

By simplifying the setup process, uv allows you to focus on what really matters: writing code for cool stuff.

Need for speed

While simplicity is the main appeal, uv also brings impressive performance improvements:

• Creates virtual environments 80x faster than traditional tools

• Installs packages up to 10x faster than pip

• With caching, operations can be up to 115x faster

This is a huge improvement that will be especially felt by engineers working on projects with many dependencies.

Nice features

Beyond the basics, uv has some other good features:

Quickly running python tools: uvx is a command that allows you to run a package in a ephemeral virtual environment.

Global Module Cache: Install a package once, use it everywhere

Inline script dependencies: uv run --with <package> --python <python_version> <script.py> to run a script with dependencies. Also supports dependencies defined in the script itself.

Let’s go!

How to get started? Installation is straightforward:

# On macOS and Linux.
curl -LsSf <https://astral.sh/uv/install.sh> | sh

For full documentation, check out the docs or the GitHub repository.

After installation, be sure to add ruff as dependency to your project. This is the super fast Python linter created by the same team, and is really nice to work with.

Should You Switch?

Short answer: yes.

Long answer: If you're tired of explaining Python setup to new team members, dealing with "works on my machine" issues, or just want a more streamlined development experience, uv is worth your attention. Its focus on simplicity and reliability makes it particularly valuable for teams that want to spend less time on configuration and more time on actual development. So still yes i guess.

For the confused people

Did you read this article and think: "Hey, I know some of these words!" or "What is this guy talking about?"

Good job on reaching the end of this very dry blog post.

Want to know more about what the other words mean? Send me an email and i'll be happy to help.