Use Husky pre-push hooks to run linters

[anubh-v]

There are often cases when we:

• Push our work to our remote repo
• Realise the CI checks failed due to a code style issue
• Resolve the style issues locally and push to the remote repo again

This process can be avoided if we automatically run the linters before git push is executed.
We can use a git hook for this

[anubh-v]

@JYLee-GIT @fzdy1914 what do you think about this suggestion?

[dcshzj]

Let's abandon this. I think that it is better if such errors appear on GitHub itself instead of locally as a Git hook, so that it is easier for senior developers to guide newcomers if necessary. In other cases, code style issues are minor and does not affect the review workflow.

Having this hook requires writing a custom script which adds additional maintenance burden.

[gerhean]

Pretty sure that when code style issues are detected the CI tests would just fail, and senior developers usually do not look at PR with failed tests. I think a hook which checks for code style would be nice since if I forgot to run the linter before I push, I get reminded of my mistakes immediately instead of 5min later when the CI fails.

[damithc]

We don't use this technique in our other projects. Which means we can survive without it. But I don't mind giving it a try just to see how it goes. There might be some value if not a lot. Let's keep it, but with a low priority.

[yhtMinceraft1010X]

I think one possible reason why this issue is relevant is because of the different abstraction level for Run environmental checks in integration.yml compared to other linters:

RepoSense/.github/workflows/integration.yml

Lines 155 to 156 in 25f4ce2

	- name: Run environmental checks
	run: ./config/gh-actions/run-checks.sh

Most other linters are run in Gradle (lintFrontend and checkstyleAll), so developers are likely to forget about the existence of (and need to run) ./config/gh-actions/run-checks.sh. If we make it a Gradle task (and mention it in the Developer Guide), developers are more likely to know about this particular check.

[ckcherry23]

Considering the adoption of git hooks in other NUS-OSS projects like MarkBind and CATcher, it's worth exploring implementing pre-commit and pre-push hooks in RepoSense for linting. Notably, Markbind uses the pre-commit package and CATcher uses Husky for this. Considering that Husky (1) is actively maintained, (2) is the more popular tool, and (3) has easier configuration scripts, let us introduce Husky pre-push hooks to lint (and auto-fix) code in RepoSense.

Furthermore, if we only want to run the linters on staged files, we can make use of lint-staged.

Source: npm trends for git hook packages