UV as an alternative to Poetry

There are somehow too many package managers for Python. This, of course, has an obvious reason - the paucity of functionality of pip, the built-in package manager. But still, the inner perfectionist wants a simple solution out of the box. Install Python - and you immediately get a fast and convenient package manager, and preferably a Python version manager. But instead you get pip. Some people, of course, actually use it, but still poetry, pdm, conda, pipenv or at least pip-tools are much more convenient.

Well, okay, it would seem that poetry is good for everyone. I use it myself in most projects. But installing dependencies starts to seem slow - especially when rebuilding the docker container. Plus there is a hassle with installing poetry itself or changing versions of python - you change the version through, for example, pyenv, and poetry gives an error when you try to recreate the environment. Although it is assumed that it can work with different versions of Python. This can be solved, of course, easily - by specifying the full path to the interpreter, but it’s still a crutch. As well as installing the same pyenv, and generally working with it. And there seems to be nothing more to control python versions.

And so, in February, an interesting solution appeared from the creators of Ruff. UV package manager written in Rast. Purely console-based, of course, and the syntax is very reminiscent of poetry. In terms of functionality, these are almost the same thing, but with a bunch of goodies and several times faster. The doc shows this diagram of the installation time for the same set of dependencies:

The syntax is really similar to poetry. This is how, for example, a project is created (go straight to the directory):

uv init project
cd project

There we have this structure:

project
├── .python-version
├── hello.py
├── pyproject.toml
└── README.md

Dependencies, like poetry, are saved in the pyproject.toml config, the python version is saved in .python-version.

Let's create a virtual environment:

uv venv

Add SQLAlchemy depending on:

uv add sqlalchemy

Or we can add a specific version:

uv add sqlalchemy@2.0.32

Now let's delete:

uv remove sqlalchemy

Now sugar - python version management. Let's install 3.11.9 and create an environment with such an interpreter.

uv python install 3.11.9
uv venv --python 3.11.9

The Python version will be saved in .python-version, and you don’t need to point this out to the package manager every time you change the interpreter, because UV is the package manager. Moreover, you don’t even have to install the version manually, but immediately create an environment with the desired version: if it is not installed, then the UV itself will pull it, that is, you don’t have to think about it at all - well, it’s a thrill.

Like poetry, there is functionality for building and publishing packages on PyPI. You can build a container and publish a package using two commands:

uv build
uv publish

You can also use UV as a supervisor, and run scripts and applications using uv run

And the cherry on the cake is the docker image.

FROM ghcr.io/astral-sh/uv:python3.12

WORKDIR /app

RUN uv venv

CMD ["run", "app"]

You don’t need to pull the python image and install UVs there via pip, you can immediately pull the UV image and have fun. Plus there are a lot of other tricks there, but about this in the dock (which, by the way, is very clear) - those who need these chips in the dock will not be afraid to go into the dock.

In general, a really convenient tool and a good alternative to poetry. There is no UV support in any IDE yet, but it's a matter of time. It’s cool that there are so many features, I want the project to develop. Share this post and write what you use.

P.S. And you need to indulge in TGC: https://t.me/dmkjfss

The above is the detailed content of UV as an alternative to Poetry. For more information, please follow other related articles on the PHP Chinese website!