MAINT: Trying uv for faster CI

[MartinThoma]

I want to try to use uv for pypdf in CI. The main reason is that I'm curious and I want to experiment. It might be that it's just not the right fit for pypdf :-)

Potential benefit

Speed: -17s (?)

• https://docs.astral.sh/uv/guides/integration/github/#installation
• https://javaguirre.me/blogs/2024-08-15-speeding-up-python-deployments-with-uv/

Looking at https://github.com/py-pdf/pypdf/actions/runs/10873983099/job/30170754308 that could mean:

• Additional time for installing UV, but hopefully equivalent to the 16s of updating pip
• 17s for installing requirements ... ok, not too much 😅 But maybe we can reduce that

Modernized Tooling

• uv is by the same people who created ruff. They are innovating a lot of the ecosystem.
• pip-tools is super robust and does what it should. I'm happy with it. The tool feels mature.

However:

• ci.in / dev.in / docs.in are non-standard
• We use pyproject.toml with optional dependencies (including docs, but not including ci) already. So it might be a chance to simplify how we store dependencies.

[stefan6419846]

Just my 2 cents: As long as we keep compatibility with pip, I am fine with it. ruff and uv (as most Rust applications) are a license compliance nightmare due to tons of dependencies to look into, thus I would like to avoid solely relying on them.

[MartinThoma]

ruff and uv (as most Rust applications) are a license compliance nightmare due to tons of dependencies

Interesting! I wasn't aware of that