Minimize Build Time: How to run Jest Tests on Changed Code Only
Efficiency is a cornerstone of any CI/CD pipeline. Running the full suite of Jest tests for every small modification can be a resource-intensive task that slows down the development process. Normally, your package.json might contain a script to run tests like this:
{
"scripts": {
"test:ci": "jest --watchAll=false --silent --reporters=default --reporters=jest-junit"
}
}
To optimize this, we can create a test running script called run-tests.js. This script will allow us to run only those Jest tests that are impacted by the changes in a GitHub Pull Request. This approach, while demonstrated on CircleCI, is designed to be agnostic of the build pipeline infrastructure.
The first step in our script is to import the modules needed for HTTP requests, process spawning, and path manipulation:
const fetch = require('node-fetch');
const { spawn } = require('child_process');
const path = require('path');
The Jest command and its arguments are set up in the following block. This includes base Jest command that needs to run:
const codeRoot = '.'; // in case your package.json is not in the repo root put the path here
const sourceCodeBase = 'src';
const jestCmd = 'jest';
const jestArgs = [
'--watchAll=false',
'--silent',
'--reporters=default',
'--reporters=jest-junit',
];
We then define a function called runJest that will execute the Jest tests with these arguments:
const runJest = (args) => {
return new Promise((resolve, reject) => {
const jestProcess = spawn(jestCmd, args, { stdio: 'inherit', shell: true });
jestProcess.on('close', (code) => {
if (code === 0) {
resolve();
} else {
reject(new Error(`Jest exited with code ${code}`));
}
});
});
};
The next function, fetchChangedFiles, communicates with GitHub’s API to determine which files have been altered in the pull request:
const fetchChangedFiles = async (repo, prNumber, token) => {
const url = `https://api.github.com/repos/${repo}/pulls/${prNumber}/files`;
const response = await fetch(url, {
headers: {
Accept: 'application/vnd.github+json',
Authorization: `Bearer ${token}`,
},
});
if (!response.ok) {
throw new Error(`Failed to fetch changed files: ${response.statusText}`);
}
const files = await response.json();
return files.map((file) =>
path.relative(codeRoot, file.filename)
);
};
Depending on the files that have changed, we decide to either run all tests or only the impacted ones:
const runImpactedTests = async (repo, prNumber, token) => {
let changedFiles;
try {
changedFiles = await fetchChangedFiles(repo, prNumber, token);
} catch (error) {
console.log(
'Error while fetching PR diff, running all tests:',
error.message
);
return runAllTests();
}
console.log('Changed files:', changedFiles);
const otherChanges = changedFiles.some((file) => !file.startsWith(sourceCodeBase));
if (otherChanges) {
console.log(`Changes detected outside of ${sourceCodeBase}, running all tests`);
return runAllTests();
} else {
console.log('Running only impacted tests');
return runJest([...jestArgs, '--findRelatedTests', ...changedFiles]);
}
};
const runAllTests = () => {
console.log('Running all tests in main-client');
return runJest([...jestArgs]);
};
The main function combines all the above functionalities:
const main = async () => {
const prUrl = process.env.CIRCLE_PULL_REQUEST;
const prNumber = prUrl ? prUrl.split('/').pop() : null;
const repo =
process.env.CIRCLE_PROJECT_USERNAME && process.env.CIRCLE_PROJECT_REPONAME
? `${process.env.CIRCLE_PROJECT_USERNAME}/${process.env.CIRCLE_PROJECT_REPONAME}`
: null;
const token = process.env.GITHUB_TOKEN;
if (!prNumber || !repo || !token) {
console.log(
'PR number, repository information, or GitHub token is not available.'
);
await runAllTests();
return;
}
const isPullRequest =
process.env.CI_PULL_REQUEST && process.env.CI_PULL_REQUEST !== 'false';
if (isPullRequest) {
await runImpactedTests(repo, prNumber, token);
} else {
await runAllTests();
}
};
Finally, the script calls the main function and handles any potential errors:
(async () => {
try {
await main();
} catch (error) {
console.error('Error while running tests:', error.message);
process.exit(1);
}
})();
After dissecting each part, here’s the complete run-tests.js script:
const fetch = require('node-fetch');
const { spawn } = require('child_process');
const path = require('path');
const codeRoot = '.'; // in case your package.json is not in the repo root put the path here
const sourceCodeBase = 'src';
// Jest configuration
const jestCmd = 'jest';
const jestArgs = [
'--watchAll=false',
'--silent',
'--reporters=default',
'--reporters=jest-junit',
];
// Run Jest with the provided arguments
const runJest = (args) => {
return new Promise((resolve, reject) => {
const jestProcess = spawn(jestCmd, args, { stdio: 'inherit', shell: true });
jestProcess.on('close', (code) => {
if (code === 0) {
resolve();
} else {
reject(new Error(`Jest exited with code ${code}`));
}
});
});
};
// Fetch the list of changed files in a pull request
const fetchChangedFiles = async (repo, prNumber, token) => {
const url = `https://api.github.com/repos/${repo}/pulls/${prNumber}/files`;
const response = await fetch(url, {
headers: {
Accept: 'application/vnd.github+json',
Authorization: `Bearer ${token}`,
},
});
if (!response.ok) {
throw new Error(`Failed to fetch changed files: ${response.statusText}`);
}
const files = await response.json();
return files.map((file) =>
path.relative(codeRoot, file.filename)
);
};
// Run all tests in main-client
const runAllTests = () => {
console.log('Running all tests');
return runJest([...jestArgs]);
};
// Run only the tests impacted by the changes in the pull request
const runImpactedTests = async (repo, prNumber, token) => {
let changedFiles;
try {
changedFiles = await fetchChangedFiles(repo, prNumber, token);
} catch (error) {
console.log(
'Error while fetching PR diff, running all tests:',
error.message
);
return runAllTests();
}
console.log('Changed files:', changedFiles);
const otherChanges = changedFiles.some((file) => !file.startsWith(sourceCodeBase));
if (otherChanges) {
console.log(`Changes detected outside of ${sourceCodeBase}, running all tests`);
return runAllTests();
} else {
console.log('Running only impacted tests');
return runJest([...jestArgs, '--findRelatedTests', ...changedFiles]);
}
};
// Main function to determine whether to run all tests or only impacted tests
const main = async () => {
const prUrl = process.env.CIRCLE_PULL_REQUEST;
const prNumber = prUrl ? prUrl.split('/').pop() : null;
const repo =
process.env.CIRCLE_PROJECT_USERNAME && process.env.CIRCLE_PROJECT_REPONAME
? `${process.env.CIRCLE_PROJECT_USERNAME}/${process.env.CIRCLE_PROJECT_REPONAME}`
: null;
const token = process.env.GITHUB_TOKEN;
if (!prNumber || !repo || !token) {
console.log(
'PR number, repository information, or GitHub token is not available.'
);
await runAllTests();
return;
}
const isPullRequest =
process.env.CI_PULL_REQUEST && process.env.CI_PULL_REQUEST !== 'false';
if (isPullRequest) {
await runImpactedTests(repo, prNumber, token);
} else {
await runAllTests();
}
};
// Execute the main function and handle errors
(async () => {
try {
await main();
} catch (error) {
console.error('Error while running tests:', error.message);
process.exit(1);
}
})();
We’ve created a script that selectively runs Jest tests based on the code changes in a GitHub Pull Request. This not only makes your CI/CD pipeline more efficient but also speeds up your development cycle.
Do you have thoughts or questions? Maybe a different approach to optimizing test runs?
Feel free to leave a comment below.