jannik2099 / pms-utils Goto Github PK

a couple parsers use C locale parsers. This is both slow and likely inaccurate.

The upcoming Boost.Parse would also alleviate this.

Hello!

I found this project through @thesamesam's activity hehe. And it looks like we have some shared interests. I wanted to have a go at dependency resolution, and I had to write my own routines for PMS and this exactly what this project aims to offer.

The routines are part of my dep solver project https://github.com/adelKS/quantumresolver: it also uses C++20 and the meson build system. If you have a look at its readme, so far there's the quantum cli program which offers a mix between equery y [atom] and emerge -qpvO [atom]. That should give an idea on which routines were implemented for it to work.

So maybe we can collaborate so I can rework and transfer the other routines I wrote to this library and make my project depend on it. So my own project focuses only on the dependency resolution.

One example of discussion: I took a quick look at what you wrote for version comparison, I wanted to use std::strong_ordering but in the end preferred specific optimizations for each case, using lexicographical_compare_three_way with an internal representation that works with it, so it could leverage SIMD instructions (my intent is to write the fastest dependency resolver I could come up with). See ebuild_version.h and ebuild_version.cpp

The example from the README (when replacing __repr__ with __str__) currently prints

blocker: Blocker.weak
category: foo
name: bar
slotExpr: :0
usedeps: <pms_utils.atom.Usedeps object at 0x7feb1e209eb0>
version: 1
verspec: VersionSpecifier.ge

__str__ for enums and vectors is missing