Sync non-zero exit on fix

[enadeau]

Is your feature request related to a problem? Please describe.
There is no way to tell if the sync command made some changes or not. This makes it hard to run the command in CI and fail the job if the project is not in sync.

Describe the solution you'd like
Add a flag like --exit-non-zero-on-fix to the sync command to fail with non zero exit code if some changed occurred while running the command.

Describe alternatives you've considered
The check commands seems like the command that should be used for that but it does not complain if I remove one of the component from the main pyproject file

[DavidVujic]

I didn't know that the poly sync command was used in CI! My initial idea was to run it locally, and then commit the changes.

I'm curious about why the poly check doesn't seem to identify the changes, that sounds like a bug. Or is it the top-most pyproject.toml file you mean? If so, then I understand. The check command does the verifying on the projects only, to be able to switch between implementations/components in development.

[enadeau]

The idea to run it in CI was to make sure nobody in the team forgets about it and thing falling out of sync.

Yes I was actually testing it on the root pyproject so it is probably the reason why it didn't catched it. If there was a way for check to catch that to that would also be a good way to fix the issue

[DavidVujic]

That makes sense! I have to think about this and how to solve it, because I also believe some teams (during development/feature branches) exclude some components and I don't want that to break the builds. It's not strictly a requirement to include all code in development, even if it is highly recommended.

There is the --verbose option, but that is mainly for local development too. Unfortunately I have annotated it incorrectly for that specific command, it should be a boolean flag and not a string. I will make an update to make it behave the same as the same flag for other commands.

[DavidVujic]

Would a pre-commit hook be a solution?

[enadeau]

I think a pre-commit hook would be a great solution 😃

I was actually thinking of this flag as some per-requisite for building a hook

[DavidVujic]

Would this simple pre-commit do what's needed?

#!/usr/bin/env bash
set -eo pipefail

uv run poly sync   # or, if using poetry: poetry poly sync

(I created a file named pre-commit in the .git/hooks folder)

There's of course also the pre-commit tool/library with a lot more features 😄

[enadeau]

I was more thinking about the pre-commit library when I mentioned that as I'm looking for a reliable, consistent way to deploy it accross the team (or in CI)

[DavidVujic]

I think it is possible without a dedicated published hook, by using the already installed poly tool. But I am not that experienced in how the pre-commit tool is meant to be used.

I experimented with the pre-commit library:

uv add pre-commit

touch .pre-commit-config.yaml

and added this content to the yaml:

repos:
-   repo: local
    hooks:
    -   id: poly-sync
        name: run poly sync
        entry: bash -c "uv run poly sync"
        pass_filenames: false
        language: system

As an alternative, you could also have a hook.sh file with the contents I previously shared, and set the entry to the hook instead.

What do you think about that?

(EDIT: this will probably be a problem for Windows users, when relying on bash)

EDIT 2: here's a 🤪 inline Python equivalent
python -c "import subprocess; subprocess.run(['uv', 'run', 'poly', 'sync'])"

[DavidVujic]

(moving this to the ideas section in discussions)