Use fixed clang-format version

[faaxm]

Different clang-format versions format code differently. This PR makes sure that we default to version 11.

[s-weigand]

You could also just use pre-commit together with the python distribution of clang-format. This is the easiest way to have a specific version of clang-format run on only matching files cross platform. Plus you can install the pre-commit hook so it runs before each commit and intercepts it if needed. ^^

[faaxm]

@s-weigand Oh, I didn't expect anyone to actually look at my pull request! Thanks for the info! 🙂

Using pre-commit would mean that anyone contributing to spix would have to have clang-format setup in that way, right? I think for an opensource project, I would like to keep the process as open and flexible as possible, to not add any entry barriers for new contributors. But maybe would be worth having a note on this somewhere in the readme or additional docs? Will think about it some more...

[s-weigand]

pre-commit is a python package and a framework to run QA tools with a specific version based on the git tag of a repository containing the hook configuration. Each tool gets its own environment so it isn't polluting you system path with those tools (e.g. you could use clang-format 11 on one project and clang-format 16 on another project without them interfering or even knowing about each other). It supports multiple languages natively (e.g. python, nodejs, go, rust) and can bootstrap isolated environments for them without the need for a system-wide installation. The configuration (including which version of a tool to use) is done .pre-commit-config.yaml which defines the version of a hook, additional dependencies, CLI flags, and which files it should be applied to. By default pre-commit only runs the hooks on tracked files with changes, but you can also run it on all tracked files (pre-commit run --all), so there is no need to manually filter gitignored files (e.g. in the build or dep-libs folders).

Since most of the good tooling for c++ is at least in part written in python (e.g. gcovr or clang-tidy and clang-format) contributor most likely already have python installed so that shouldn't be a barrier.

The goal of clang-format-wheel is to have an easy way to get clang-format installed and running cross platform without the need to compile half of the llvm-project.

Installing pre-commit as a git hook (pre-commit install) allows you to automatically run all your tools on the staged files of each commit, without the need to think about it ("Did I already run clang-format?") or polluting the commit history with "run clang-format" commits.

It also has a github-action that uses caching out of the box.

I have been using it for a few years now and can't imagine working without it 😅