RFC: Merge back unblob-native into this repo

[vlaci]

The Unblob rust extension started its life as an opt-in piece in this repo. We extracted it into its own package due to (1) limitations in Poetry packaging native extensions, and (2) the higher barrier of entry people needed to build Unblob.

1. I think it is easy to overcome this by now, we just need to use the maturin builder in pyproject.toml to build wheels
2. More and more Python applications/libraries require Rust, so I wouldn't consider this a niche now. Also, direnv integration helps greatly provisioning dev environments.

+1: Refactors, like dropping support for Python 3.8, wouldn't be missing the unblob-native part :o)

It looks like, we'd need to switch Poetry for maturin or uv for installing the project environment as well, if we want to avoid having a separate build output.

[qkaiser]

I'm all for it. The most sensitive part will be porting unblob-native CI here and converting the current CI/CD pipeline definitions to use maturin without losing anything in the process.

Personally I'm all for limiting duplicate work. Right now the flake updates, the dependency upgrades from dependabot, the project settings, everything is done twice and it's annoying.

[qkaiser]

Not sure what's your take but I'd be happy to try uv. My trust in tooling coming out of @astral-sh is rather strong.

[kissgyorgy]

I'm a fan of monorepos, they are a productivity boost. I would have not split the repo first place, but I wasn't here to prevent it 😀

[vlaci]

If we are to do this, we need to get off of Poetry for sure. Tried various ways, but there is still no way to build abi3 wheels with it, and there is no way to use a different builder.
Swithing to uv would be the sensible thing to do, but then we also need to integrate renovate for package updates, as dependabot doesn't handle uv locks.

Renovate would be nice, as we could integrate flake lock updates into it as well.

[kissgyorgy]

We don't have to immediately integrate everything, the first step can be to just merge the two repos, later we can move to uv or whatever needed for a more seamless workflow.

[vlaci]

We don't have to immediately integrate everything, the first step can be to just merge the two repos, later we can move to uv or whatever needed for a more seamless workflow.

IMO, it would be easier the other way around: migrate to uv, then merge. That way, we can get rid of one round of CI test/publisher job merge hell.

[kissgyorgy]

https://github.com/stvnksslr/uv-migrator

[vlaci]

https://github.com/stvnksslr/uv-migrator

Will look into it. I've tried poetry-to-uv which worked reasonably well.

[kissgyorgy]

Ok this is crazy. uv is so fast I needed to try it manually because I thought nothing happened.