Get hooked on Git hooks

Last updated: 16.12.2022

If you're like me, you're crazy about automating boring stuff. One of the things I got hooked on (pun intended) during the last year, and which helps in that automation process, is Git Hooks. If you haven't heard of Git Hooks and want to see some cool ways of improving your daily git workflow stay tuned!

What are Git Hooks? 🎣

This page from Git documentation sums it up pretty well but in general Git Hooks are Git's answer on firing custom events when some Git related action occurs. We will focus on client-side pre-commit and commit-msg hooks today but the following options are available:

• Client-Side Hooks

  • pre-commit - runs before we even type the commit message.
  • prepare-commit-msg - runs before the commit message editor is opened up but after the default message is created.
  • commit-msg - a good place to validate the project state or the commit message before allowing the commit to proceed further.
  • post-commit - runs after the entire commit process is completed, used mostly for notifications.
  • pre-rebase - runs before the rebase.
  • post-merge - runs after the successful merge.
  • pre-push - runs during the Git push.
  • pre-auto-gc - runs before Git triggers a garbage collector.

• Server-Side Hooks

  • pre-receive- the first script that is run on the client-side push, if it exits non-zero, the push is not accepted.
  • update - pretty similar to the pre-receive except it runs once for every branch that the client-side wants to update. For example, if we're pushing to five branches at the same time, pre-receive will run once, and update will run five times.
  • post-receive - similar to the client-side post-commit just on the server side.

Talk is cheap, show me the code

Since Git hooks don't have the best out-of-the-box experience, we'll use the Husky library to make stuff easier:

Alternatively, you can just run npx husky-init && npm install which will do the same.

pre-commit

In most of cases we want to run the pre-commit hook only on the staged files, lint-staged library helps us with that:

After we've added the lint-staged we need to extend package.json file like this:

To run scripts defined inside of the lint-staged object we need to modify our pre-commit file that was generated by Husky earlier:

Now that we know the basics, it's time to start adding scripts that will help our repository become a better place ✨.

First, let's add prettier - hope you've heard of it since it's the best thing that happened to code formatting in a while.

We can pass arguments to the prettier script directly but I'm a fan of config files, so we'll create a .prettierrc file in the project root directory:

Prettier will format all staged files on the commit so they follow a code convention defined inside the .prettierrc.

Time to lint our .js files, we can easily do that with eslint.

We will define a config file again, this time the eslintrc.json:

We need to define a special rule that will be triggered for .js files only. eslint will prevent committing if error is thrown.

As the final step I'll show you how to run relevant unit tests (relevant to committed files) and prevent commit if some of them are failing.

We should add the previously installed jest plugin to our eslint config file so we eliminate eslint errors on .spec.js files.

Now extend lint-staged script:

--bail will skip execution of other tests when the first test fails and --findRelatedTests is pretty self-explanatory 😁.

To demonstrate how this works we can create two files test-file.js and test-file.spec.js

We're intentionally making the unit test fail so we can see the commit failing:

commit-msg

There are only two hard things in Computer Science: cache invalidation and naming things

This rule applies to commit messages also, we've all seen or written commits like this in past:

This is an extreme example but it perfectly shows how we can't make a clear conclusion about what is going on in a particular commit.

If we check the history of commit messages created during previous examples:

Much cleaner right? These commits follow Conventional Commit convention created by Angular team.

In general, the pattern that a commit message should follow mostly look like this:

Some of the common types are:

• feat - commit adds a new feature.
• fix - commit fixes a bug.
• docs - commit introduces documentation changes.
• style - commit introduces code style change (indentation, format, etc.).
• refactor - commit introduces code refactoring.
• perf - commit introduces code performances.
• test - commit adds test to an existing feature.
• chore - commit updates something without impacting the user (ex: bump a dependency in package.json)

So, now that we know this, it's the perfect time to introduce commit-msg hook where we'll check if the commit message respect these rules before we commit.

First, we want to install commitlint, something like eslint just for commit messages.

Of course, we need to create another config file, .commitlintrc.json, the last one I promise! 🤞

Last but not least, we need to add commit-msg hook using husky add command:

Quick recap of what we learned today:

lint-staged inside the pre-commit hook will take care of:

• formatting all staged files via Prettier.
• check all staged .js files for syntax errors via Eslint
• check if relevant .spec.js unit test files are failing before we commit via Jest

commitlint inside the commit-msg hook will take care of:

• enforce commit message to follow Conventional Commit rules via Commitlint.

See also

• cz-cli - The commitizen command line utility.
• husky-sandbox - Code samples from this post.