Comment by ericyd

Comment by ericyd 6 months ago

There's always a spectrum of preference with these things. Personally I don't like git hooks for my workflows because I know CI will enforce the relevant checks. In addition, I am fine running an extra test or lint or build command before opening my PR to verify it works locally. Git hooks are a nice solution but for me they just slow down common operations so I'd rather not use them. This tool looks like a nice improvement over project-level configurations like Husky (in the npm space) because it's an individual configuration so I could just skip it, whereas husky gets configured for everyone.

SOLAR_FIELDS 6 months ago

There are dozens of your type out there that I don’t understand. The “I don’t want no commit hooks running on my machine. I and only I control what happens on my local machine” purists. There’s at least one in every job I work.

If the thing is going to block in ci anyway, why are you opting for a push and pray approach? Why arbitrarily increase your feedback loop time and add waiting time for each loop in ci? Chances are the time to get feedback is at least a couple orders of magnitude faster locally, you’re paying not only the startup time to register the runner and bring it online, get dependencies installed etc but also the manual time to context switch to the CI window etc. Just do all your linting and auto formatting and whatever on commit. It’s all work you’re going to have to do anyway, why introduce some extra less efficient step on yourself to slow yourself down?

Reply View 8 replies

ericyd 6 months ago

Valid questions. For me it's two things:
1) sometimes hooks are configured on commit, and I prefer to have a very quick and lightweight commit action. If I'm changing branches sometimes I'll commit WIP changes so I can easily come back to it later. I know git stash accomplishes a very similar functionality but it's just a preferred workflow.
2) I don't like the feedback I get from git hooks when committing in the VS Code interface. For example, we have a "lint" hook that runs on pre-push. When it fails, I have to run "lint" manually in my terminal anyway to actually see the errors, because VS Code doesn't show me the actual errors. I believe the hook results are available in some other tab or something but I haven't bothered to learn it.
Both of these are just personal preferences, and maybe a little bit of resistance to learning new workflows. I don't consider myself a purist about it and I've never argued about it at work but hopefully this sheds a little light on my perspective.
Edit: regarding the "push and pray" approach, I personally don't do that, I'll run it locally first, I just prefer to run it myself rather than via a hook.

Reply View | 7 replies
- Ezpie 6 months ago
  
  I can totally understand the point you're making there. I used to use VScode too few years ago, but now I use vim most of the time. So majority of the time I'm in my terminal so if the hook fails for me, it gives me the error, and I have zero problems with that. So I get what you mean there, but I mean, for those who live in their terminals git hooks kind of make sense, I mean after all, we see the error cause its the terminal after all.
  So yeah it just boils down to the fact how you would use git, in vscode, hooks aren't great, cause its not going to display the error, but if you use the terminal, even in vscode, you'll see the error getting displayed on the terminal.
  
  Reply View | 0 replies
- jmholla 6 months ago
  
  1) You can pass `--no-verify` to `git commit` and it will skip the hooks. An alias would make it even more ergonomic.
  2) You can write hooks that make the changes for you where possible. This is what I do for lints from auto-formatters. Don't know how to handle the rest of your VS Code issues though.
  
  Reply View | 4 replies
  
  SOLAR_FIELDS 6 months ago
  
  Unfortunately —-no-verify is all or nothing. If I have 10 hooks and only want to skip one of them it won’t do the job
  
  Reply View | 3 replies
- SOLAR_FIELDS 6 months ago
  
  Yeah it seems a lot of this boils down to the myriad of ways people use git. I can imagine some ways are easier than others when using hooks. And githooks are “all or nothing” without an easy way to enable/disable them per hook - which I believe is one of the problems that OP aims to solve.
  
  Reply View | 0 replies