We should closely monitor the state of tests in our NestJS application. In a few previous parts of this series, we’ve used Docker and GitHub actions to design a CI/CD pipeline that automates our deployments. In this article, we avoid deploying faulty code by running unit tests before each pull request automatically.

Preparing the Docker image

The first step is ensuring that our tests can run in our Docker image. To verify that, let’s look at the Dockerfile we’ve built so far.

Dockerfile

If you want to see how we’ve built the above configuration, check out the following articles:

We would fall into a significant pitfall if we used the above Docker image to run tests.

To install the dependencies, we run npm ci --omit=dev. Adding the --omit=dev flag tells NPM to skip installing all of the libraries listed in devDependencies. However, this is typically where we list libraries such as Jest that we need for testing.

package.json

Let’s modify our configuration to accommodate that.

Dockerfile

Since we need the packages listed in devDependencies to run our unit tests, we install all of them during the install-dependencies stage by skipping the --omit=dev flag.

This would make the resulting Docker image heavier if not for a modification to the run stage. By adding the npm prune --production command at the bottom, we eliminate the libraries from devDependencies before making the final Docker image.

Thanks to doing all of the above, the result of the create-build stage in the Dockerfile has all of the dependencies, but they are stripped when no longer necessary.

We also add RUN npm install -g [email protected] to ensure the latest version of NPM and avoid warnings in the terminal.

Running tests locally

We must run our tests with the Docker image targeting the create-build stage before removing the devDependevies. Therefore, let’s use the --target flag to build the image without running the last stage from our Dockerfile.

When we look at the create-build stage in our Dockerfile, we can see that it doesn’t have the CMD keyword. When running the image, we must provide Docker with the desired command. Let’s run npm run test inside the shell of a Docker container running with our image.

PASS src/authentication/authentication.service.test.ts
The AuthenticationService
when creating a cookie
✓ should return a string (17 ms)

Test Suites: 1 passed, 1 total
Tests: 1 passed, 1 total
Snapshots: 0 total
Time: 2.662 s
Ran all test suites.

Running Prettier

When checking the formatting, we usually set up Prettier using the --write flag to overwrite our files.

Instead, we want to be notified when our code does not meet the formatting standards. To do that, we need the --list-different flag.

If a particular file is not formatted correctly when running the above command, Prettier lists it. It also uses the 1 exit code to signify an error.

Let’s rebuild our Docker image to include the above command.

Now, we can run Prettier in our Docker container to verify our code.

Running ESLint

The official NestJS starter repository comes with ESLint configured and ready to go. We can easily use it with our existing configuration.

Running the tests with GitHub Actions

In the previous part of this series, we created a GitHub Actions workflow that runs every time we push our change to the master branch. Before that happens, we usually make a pull request to merge our changes. Let’s create a new GitHub Actions workflow that runs every time a PR is created.

.github/workflows/test.yml

Thanks to adding on: [pull_request] above, there is a set of steps happening above every time we create a pull request:

1. For GitHub Actions to access our code, we need to check out our repository. To do that, we use the actions/checkout@v3 action provided by GitHub.
2. We build the Docker image targeting the create-build stage.
3. Since our environment is ready, we run Prettier, ESLint, and unit tests.

Let’s create a Pull Request and verify that everything works as expected.

We can see the information saying that “all checks have passed“. When we click “Details”, we can see more job information.

Protecting our master branch

Even though we have the tests in place, nothing stops us from disregarding them and merging the changes even if they fail. Let’s change that.

To do that, we need to open our GitHub repository and go to Settings -> Branches. Here, we can add a branch protection rule prohibiting merging changes if our status checks haven’t succeeded.

Summary

In this article, we established a process of testing our application whenever we want to merge new changes to the master branch. To do that, we had to modify our Docker image slightly to install all necessary dependencies and remove them when they were no longer needed. We’ve also created a new GitHub Actions workflow to run tests on every pull request.

Thanks to the above, we couldn’t merge our changes if any of our tests failed. This gives us an additional layer of security and prevents us from deploying a faulty version of our application.