How to stub a service with Sinon in an integration test? - node.js

Question

I'm trying to do some integration tests for my api in express.
My API's structure is something like:
app -> routes -> controllers -> services
Because I already have unit tests, my idea is only test that all that components are connected in the correct way.
So my idea was created an stub with Sinon for the service, and only check the responses of the controller with supertest.
When I run a single test everything is ok. The problem is when I run more than one unit test for different controllers, the stub doesn't work in the second run.
I think it's because the app is already saved in cache as a module, so sinon can't stub the service.
Some examples of my code:
controller.js
const httpStatus = require('http-status');
const { service } = require('../services/croupier');
/**
* Execute lambda tasks for candidates
* #public
*/
exports.task = async (req, res, next) => {
try {
const result = await service({
body: req.body,
authorizer: req.authorizer
});
console.log('res', result);
res.status(httpStatus.OK).json(result);
} catch (error) {
next(error);
}
};
foo.integration.test.js
const request = require('supertest');
const httpStatus = require('http-status');
const sinon = require('sinon');
const mongoose = require('../../../database');
const deleteModule = module => delete require.cache[require.resolve(module)];
const requireUncached = module => {
deleteModule(module);
return require(module);
};
describe('Foo - Integration Test', async () => {
describe('POST /v1/foo', () => {
const fooService = require('../../services/foo');
const stub = sinon.stub(fooService, 'service');
let db;
before(async () => {
db = await mongoose.connect();
});
afterEach(async () => {
sinon.restore();
});
after(async () => {
await db.close();
});
it('the api should response successfully', async () => {
stub.returns({});
const payload = { task: 'task', payload: [{ pathParameters: {}, body: {} }] };
const app = requireUncached('../../../app');
await request(app)
.post('/api/foo')
.send(payload)
.expect(httpStatus.OK);
});
it('the api should response with an error', async () => {
stub.throwsException();
const payload = { task: 'task', payload: [{ pathParameters: {}, body: {} }] };
const app = requireUncached('../../../app');
await request(app)
.post('/api/foo')
.send(payload)
.expect(httpStatus.INTERNAL_SERVER_ERROR);
});
});
});
The other integration tests have the same structure. I've also tried using proxyquire but didn't work.
Also I tried deleting cache of de app.js with any success.
Any ideas?

Answer 1

Context: integration test.
I agree with your idea: "test that all that components are connected in the correct way". Then what you need is spy, not stub. When there is a case / condition, you need to setup preconfigured/dummy data (up mongodb with specific data), turn on HTTP server, call HTTP request with specific data (post / get with specific query), and check the HTTP response for correct status, etc. The spy needed to check/validate/verify whether your service get called with correct parameter and response with correct result. This test validate you have correctly configured route - controller to a service for specific HTTP request.
You must have question: How to test negative scenario? For example: 404, 500. Then you need to know which specific scenario do what, which result negative condition. For example: if request come with unknown ID query, then response will be 404. Or if express not connected to database, then response will be 500. You need to know the real scenario, and again provide the require setup to produce the negative response.
For problem: "When I run a single test everything is ok. The problem is when I run more than one unit test for different controllers, the stub doesn't work in the second run.". There are several possible solutions, the main point is: you must make sure that the conditions for specific scenario/case are correctly prepared.
You can do:
create sandbox, to make sure no other stub service run between test cases.
start up fresh http (and or db) server before and shut down the server after the test run for each services, (for example: start the app and use real http client - as alternative to supertest)
run on debug mode to find out why the second stub not run or not get called or not work,
change implementation from stub to spy, you have already had a unit test, you just need to check whether the service get called or not, and then check the overall response.
Hope this helps.