Testing Google Cloud Functions with Vitest

Recently I was tasked with building a system for logging client side errors from the suite of JavaScript widgets that my team maintains. To keep things simple I thought a serverless architecture would fit this use-case perfectly. In essence, when an error occurs in a widget, the widget catches the error and sends a POST request to the serverless function (in this case, a Google Cloud Function). From there the error log can be stored for analysis later. One goal of this project was to have decent test coverage over this relatively small service. Since all of our other production code is tested with Vitest, I wanted to setup the same testing tools for the serverless function. In this article I’ll show a minimal example of how I tested the Google Cloud function using Vitest.

Setup

For the sake of example, below is a “Hello World” example of a Google Cloud Function, implemented in TypeScript. To follow along in your own project, you should have TypeScript, Vitest, and the functions-framework package (@google-cloud/functions-framework) installed as dependencies.

// helloWorld.ts
import { type HttpFunction } from '@google-cloud/functions-framework';

export const helloWorld: HttpFunction = async (req, res): Promise<void> => {
	const { body, method } = req;

	if (method === 'GET') res.send('Hello, World!');
	if (method === 'POST' && body.name) res.send(`Hello, ${body.name}!`);
};

This function can accept GET & POST requests, and will respond with either “Hello, World!” or “Hello, name!” where name is a property on the requests body.

Writing the test

Now that we have our serverless function, lets setup our test file and some utilities we’ll need for testing. First we need to import a few things from Vitest and the Google functions-framework package.

// helloWorld.test.ts
import { describe, expect, vi } from 'vitest';
import { Request, Response } from '@google-cloud/functions-framework';
import { helloWorld } from '.'; // Also import the serverless function itself

The classes we’re importing from the functions-framework package are very loosely typed versions of the Request and Response objects from the very popular Express framework. To test our function, we need to create mocked versions of these objects, so let’s write a utility function to generate those mocks in our tests.

// helloWorld.test.ts
import { describe, expect, vi } from 'vitest';
import { Request, Response } from '@google-cloud/functions-framework';

const getMocks = (method: string) => {
	return {
		req: { body: { name: 'Jacob' }, method, query: {} },
		res: {
			send: vi.fn() // make send a spy
		}
	} as { req: Request; res: Response };
};

Notice that we are type casting the returned object; this will make our tests cleaner and a bit more terse. Now when we call getMocks, the returned object contains the two objects we need to call our serverless function. At this point, we’re ready to start writing the test.

describe('helloWorld', () => {
	// First, we test the GET method code path in our function
	it("should call the response's send method", async () => {
		const mocks = getMocks('GET');
		const spy = vi.spyOn(mocks.res, 'send');
		await helloWorld(mocks.req, mocks.res);
		expect(spy).toHaveBeenCalledWith('Hello, World!');
	});

	// Second, we test the POST method code path in our function
	it("should call the response's send method and return a message containing the value of body.name", async () => {
		const mocks = getMocks('POST');
		const spy = vi.spyOn(mocks.res, 'send');
		await helloWorld(mocks.req, mocks.res);
		expect(spy).toHaveBeenCalledWith('Hello, Jacob!');
	});
});

In these tests we’re testing both code paths (GET & POST requests) by calling getMocks, spying on the send() method of the mocked response object, and calling our serverless function while passing the mocks as arguments. We then expect the spy to have been called with the correct value based on the request method used in the test.

That’s it! We now have full test coverage of this basic Google Cloud function using Vitest. Obviously in a production example you would have much more complicated business logic, but this pattern should help you get started. Happy testing!