Add pre-commit hook to lint?

[shawnbot]

I just stumbled across lint-staged, which allows you to run (if I'm reading it correctly) any linter on staged files in a pre-commit hook. In other words, if you did this:

perl -pi -e 's/font-size: 12px;/font-size: \$font-size-small;/' **/*.scss
git commit -am 'use $font-size-small instead of 12px'

Then we'd run stylelint on only the staged files, and if any errors were found the commit would be aborted. ✨

We'd either need to offer instructions for adding the relevant commit hook or add husky to our dev dependencies.

[shawnbot]

And then, once we were sure that was working, we could run stylelint --fix (introduced in 7.11.0) to automatically fix our errors. 💥

Also, TIL that prettier can format SCSS! 😮

[broccolini]

I want to make local development easier and for people not to have to hit the CI before seeing fails, but I am a bit wary of pre-commit hooks, and this is why:

I think it can lead to large commits and big pr's, and be a pretty frustrating dev experience. I think this would be particuarly frustrating for new contributors and potentially them. In cases when we're exploring new styles, it's not always possible to easily stick with the linters. I use the atom plugin that @jonrohan made, which gives me stylelint warnings in the editor, and often disable it while I build out new styles because it's too disruptive, and then re-enable and fix things up. I've usually made commits in that time to track the parts of the component I'm building out and I value that history. The atom plugin works a bit differently to pre-commit hooks but my gut says it will still be frustrating.

This is my opinion without having tried it of course and I'm down to test it out if folks are keen, but I'd be more excited if we could find a method of warning and correcting folks that didn't abort commits or prevent iterative development... let me know if you have any ideas! 🙃

[shawnbot]

Yeah, I think I'm with you @broccolini. Maybe instead we could consider adding Sentinel or @houndci to this repo to "enforce" the same rules that we impose over in github/github?

[shawnbot]

Hey, we're not doing this for now! ✌️