In my view, neither performance nor trying to be Python-aligned is what sets uv apart. Don’t get me wrong — try switching from uv to Poetry, and you’ll quickly notice how sluggish it (poetry) feels. uv goes extra miles to adhere to PEPs, and IMHO it’s the go-to package manager for Python these days. But these aren’t the features that surprised me most.
There’s one small feature I initially overlooked that truly makes uv intriguing:
Imagine doing some ad-hoc scripting in Python 3.12. You run python, and you’re in the REPL, ready to go. But what if you need to pull in a dependency, like Pandas? Here’s where it gets interesting.
You would either:
- Run
pip install pandas, potentially modifying your global environment, or - Take the proper route:
- Create a virtual environment
- pip install pandas
- Activate the virtual environment
- Run python
This gets even more interesting if you need some other Python version than your global one. You would then use something like pyenv, install the version and set that version as local.
So, worst case:
pyenv install 3.12.pyenv local 3.12.python -m venv .venv.source .venv/bin/activate.pip install pandas.python.
With uv, it’s just 1 command:
uv run --python 3.12 --with pandas python
easy to remember, and no trace left behind. Happy scripting!
