FYI - Just now INFRA rejected the request on the basis of "code write" access permissions the app needs.
I'd still love to get feedback though on the concept - I am not giving up that easily. We might still get it approved easily. We likely have some ways we can get "auto-fixing" working for us. 1) I believe Github Applications now can use a bit different mechanism to ask for permissions and it can have "required" and "optional" permissions and I believe "Pull request write" should be enough (and I might attempt to convince the maintainers of it to adapt it to our needs). 2) Also, there is a "Pre-commit Lite Github Action" that we **might** be able to use to achieve a similar effect (with some added complexity to our `Pull Request Target` workflow. So I would still love to hear from others :) J. On Fri, Dec 29, 2023 at 11:52 PM Jarek Potiuk <ja...@potiuk.com> wrote: > Hello everyone, > > TL;DRl; I'd like to propose that we enable the pre-commit-ci GitHub > application for Airflow repo. According to how I understand it works, it > should greatly reduce friction (especially for new contributors) for > passing the quality gates for our pre-commits. That is - if we get the > approval for pre-commit-ci application from the ASF infra team. > > Some more context: > > We use and love (well some of us do, some of us likely hate) the > pre-commit as a quality gate for our checks. We have been using it for > years for local checks and CI integration and we have ~60 custom precommits > and in total we use about 100 pre-commit checks as our "quality" gates > > However, using `standard` pre-commit (that is a de-facto standard in > Python world) has a nice property of 'standing on the shoulders of giants'. > There is one thing that few of us are aware of, that there is a way to > reduce friction for pre-commits that are not only flagging errors but can > also fix them. If we get the `pre-commit-ci` application ( > https://pre-commit.ci/) approved for our repo from the ASF infra team it > - in theory - should be able to AUTO-FIX PRs that are not passing the > pre-commits (and can be automatically corrected). > > Yep. You read it right. No more asking a new contributor "please fix > static checks" - PRs that have auto-fixable pre-commit failures will be > fixed automatically. > > For example when you make a PR that does not pass "ruff" formatting, the > application should automatically amend the PR and FIX IT. We have quite a > number of such PRs from first-time contributors, but also a number of > seasoned contributors (including myself) occasionally send a PR that does > not pass an auto-fixable static check. This can happen with a few scenarios > (rebasing, or correcting a PR by applying a suggestion from review and a > few other scenarios). > > It can be a little strange to see your PR corrected by a bot though, so I > am reaching out here to see if you think it is a good idea. I also opened a > JIRA request to approve the application (but I made a comment that it > should be pending the discussion here): > https://issues.apache.org/jira/browse/INFRA-25322 - it will likely > require to slightly change our workflows to make it works as well. > > Do you think it's a good idea to have it enabled? Maybe it will be too > much for our contributors and they will be surprised to see it happening? > > WDYT? > > J. >
