I'm using Mocha and SuperTest to test my Express API. However my first test always seems to pass when inside the .then() of my request().
I'm passing in a String to a test that is expecting an Array. So should definitely fail the test.
It fails outside of the then() as expected, but I won't have access to the res.body there to perform my tests.
Here is my code:
const expect = require('chai').expect;
const request = require('supertest');
const router = require('../../routes/api/playlist.route');
const app = require('../../app');
describe('Playlist Route', function() {
// before((done) => {
// }
describe('Get all playlists by user', function() {
it('Should error out with "No playlists found" if there are no Playlists', function() {
request(app).get('/api/playlists/all')
.then(res => {
const { body } = res;
// Test passes if expect here
expect('sdfb').to.be.an('array');
})
.catch(err => {
console.log('err: ', err);
});
// Test fails if expect here
expect('sdfb').to.be.an('array');
})
})
});
I found this article but I'm not using a try catch block, but I thought maybe it could have something to do with the promise.
Quick reponse
it('decription', function(done) {
asyncFunc()
.then(() => {
expect(something).to.be(somethingElse)
done()
})
})
Detailed response in the comment of #jonrsharpe
Rather than using done, simply return request(app).get('/api/playlists/all') since request() returns a promise. Since you have expect('sdfb').to.be.an('array'); twice, remove the one that's not in the .then callback. When using asynchronous code, remember that synchronous code that appears to come after the async chain will execute before the promise .then handlers. This is counterintuitive.
Here's the .then approach:
it('should ...', () => {
return request(app)
.get('/api/playlists/all')
.then(res => {
const {body} = res;
// assert here
});
});
The other approach is to await the promise yourself in the test case function, then make assertions on the resolved response object. In this case, drop the then chain. This approach is generally preferred as it reduces nesting.
it('should ...', async () => {
const res = await request(app).get('/api/playlists/all');
const {body} = res;
// assert here
});
If you don't let Mocha know you're working with asynchronous code by returning a promise, awaiting the promises, or adding and calling the done parameter, the assertions occur asynchronously after the test is over and disappear into the void, creating a false positive.
Skip .catch either way. Since you've informed Mocha of the promise, if it rejects, it'll let you know.
Related
I have a set of integration tests for message queueing code (over a real RabbitMQ). The tests are verifying that the correct use case is triggered when a specific message arrives.
it("should trigger the use case", () => {
const stub = sinon.stub(app, 'useCaseToTrigger').resolves();
await publishMessage({ id: "SOME_ID" })
// FAIL - Message isn't processed yet.
stub.should.have.been.calledWith(match({id: "SOME_ID" }))
})
This doesn't work. The await statement waits for the message to be published, but says nothing of when the message is processed.
I can add a delay, but that quickly becomes a huge waste of time (if there are multiple tests).
it("should trigger the use case", () => {
const stub = sinon.stub(app, 'useCaseToTrigger').resolves();
await publishMessage({ id: "SOME_ID" })
await new Promise(r => setTimeout(r, 100));
// PASS - the message has been processed before the verification
stub.should.have.been.calledWith(match({id: "SOME_ID" }))
})
But if I make the stubbed function resolve a promise when called, I can get my test to work without any undue delay.
it("should trigger the use case", () => {
const promise = new Promise(r => {
sinon.stub(app, "useCaseToTrigger").callsFake(() => {
setImmediate(() => { r(); }) // Push to the end of the event loop
return Promise.resolve();
})
})
await publishMessage({ id: "SOME_ID" })
await promise;
// PASS - Message processed before continuing test
app.useCaseToTrigger.should.have.been.calledWith(match({id: "SOME_ID" }))
})
This does become a bit cumbersome to deal with (and probably not the easiest to read either). But it also gives eslint floating promises warnings. So to avoid those warnings, I need to rewrite it to:
it("should trigger the use case", () => {
await Promise.all([
new Promise(r => {
sinon.stub(app, "useCaseToTrigger").callsFake(() => {
setImmediate(() => { r(); }) // Push to the end of the event loop
return Promise.resolve();
})
}),
publishMessage({ id: "SOME_ID" })
]);
// PASS - Message processed before continuing test
app.useCaseToTrigger.should.have.been.calledWith(match({id: "SOME_ID" }))
})
Is there a way to make this cleaner.
I think you can use an empty promise to await, so the app logic eventually resolves that promise when it is called. Sinon provides a promise fake that makes this a bit easier than creating a new Promise with an external reference to the resolve function.
const p = sinon.promise()
const stub = sinon.stub(app, 'useCaseToTrigger').callsFake(p.resolve)
await publishMessage({ id: "SOME_ID" })
await p
sinon.assert.calledWith(stub, { id: "SOME_ID" })
It looks like sinon .resolves() uses Promise.resolve at the time of the stub call, so it doesn't have a reference to the eventual promise at stub creation, otherwise a stub.promise type of reference might have been useful here to avoid the additional promise setup.
I'm running tests and I'm stubbing a function that calls the AWS sqs.deleteMessage function. .promise() is called on the call to this function. Every time I run my tests with the coverage I notice that it jumps to the catch block thus an error must be occurring on my .promise() call.
I've tried stubbing the function to resolve the promise but that doesn't seem to work. I've tried returning data as well and still have the same issue.
Below is an example of the code I'm trying to test. It never reaches the logger.info() line
fooObj.js
const foo = async (req) => {
try{
let res = await bar.deleteMessage(handle).promise();
logger.info("Sqs result message " + JSON.stringify(res));
} catch(error){
#catch block code
}
}
Below is the code for bar.deleteMessage()
bar.js
const aws = require('aws-sdk');
const sqs = new aws.SQS();
deleteMessage = function(handle){
return sqs.deleteMessage({
ReceiptHandle: handle
});
}
And finally here is the test code
const fooObj = require('foo')
const barObj = require('bar')
jest.mock('bar')
describe('foo test', ()=>{
test('a test' , ()=>{
barObj.deleteMessage.mockImplementation(()=>{
return Promise.resolve({status:200})
});
return fooObj.foo(req).then(data=>{
#Expect statements here
})
}
}
So I would like the logger.info line to be reached in coverage but I assume the issue has to do with how I'm stubbing the bar.deleteMessage function. I could use the aws-sdk-mock but I feel like I'm violating unit testing principles by mocking the sqs call that is in another file and the proper way to do it would simply be to properly stub the bar.deletemessage() function
You just need one change:
bar.deleteMessage needs to return an object with a promise property set to the function that returns the Promise:
barObj.deleteMessage.mockImplementation(() => ({
promise: () => Promise.resolve({ status: 200 })
}));
...or you can shorten it to this if you want:
barObj.deleteMessage.mockReturnValue({
promise: () => Promise.resolve({ status: 200 })
});
How would I test this code in jest? I'd like to make sure that the error and success of the passed promise is being called as needed. I'm sure it's something sorta simple, but it's driving me crazy. Thanks very much.
handleStatusChangeRequest (changeEntryStatus) {
return changeEntryStatus().then(() => {
this.handleStatusChangeSuccess()
}).catch(err => {
this.handleErrorDisplay(err)
})
}
If your code uses promises, there is a nice way to handle asynchronous tests. Just return a promise from your test, and Jest will wait for that promise to resolve.
If the promise is rejected, the test will automatically fail.
For example, let's say that changeData, instead of using a callback, returns a promise that is supposed to resolve to the string "status has been successfully modified".
Be sure to return the promise - if you omit this return statement, your test will complete before your changeData() -[async function] completes.
Here's a convenient and easy to follow pattern
test('if the data is changed', () => {
return changeData().then((data) => {
expect(data).toBe('status has been successfully modified');
});
})
Happy testing :)
This could be refactored, but for the sake of demonstration, I left the repeating bits in.
In example.spec.js, the callback, changeEntryStatus, is stubbed to return a promise. In order to check if other instance methods (this.method) were called, they are first mocked, then assertions are called on the mock after running the method being tested. Learn more in the Jest docs. (See my thoughts on mocking methods of the unit being tested at the bottom.)
Run the example on repl.it.
example.js:
class Example {
handleStatusChangeRequest(changeEntryStatus) {
return changeEntryStatus().then(() => {
this.handleStatusChangeSuccess()
}).catch(err => {
this.handleErrorDisplay(err)
})
}
handleStatusChangeSuccess() {
console.log('stubbed handleStatusChangeSuccess')
}
handleErrorDisplay(error) {
console.log('stubbed handleErrorDisplay:', error)
}
}
module.exports = Example;
example.spec.js:
const Example = require('./entryStatus')
describe('handleStatusChangeRequest', () => {
it('should run the changeEntryStatus callback', () => {
const {handleStatusChangeRequest} = new Example()
const stub = jest.fn().mockResolvedValue()
handleStatusChangeRequest(stub)
// must return because handleStatusChangeRequest is asynchronous
return expect(stub).toHaveBeenCalled()
});
it('should call example.handleStatusChangeSuccess', async () => {
const example = new Example()
const stub = jest.fn().mockResolvedValue()
example.handleStatusChangeSuccess = jest.fn()
await example.handleStatusChangeRequest(stub)
expect(example.handleStatusChangeSuccess).toHaveBeenCalled();
})
it('should call example.handleErrorDisplay', async () => {
const example = new Example()
const fakeError = { code: 'fake_error_code' }
const stub = jest.fn().mockRejectedValue(fakeError)
example.handleErrorDisplay = jest.fn()
await example.handleStatusChangeRequest(stub)
expect(example.handleErrorDisplay).toHaveBeenCalled()
expect(example.handleErrorDisplay).toHaveBeenCalledWith(fakeError)
});
});
Opinionated Disclaimer: Mocking methods of the unit under test is a smell. Consider checking for the expected effects of calling handleStatusChangeSuccess and handleErrorDisplay instead of checking to see if they were called. Then don't even expose those methods publicly unless consumers of the class need access.
Opinionated Disclaimer: Mocking methods of the unit under test is a
smell. Consider checking for the expected effects of calling
handleStatusChangeSuccess and handleErrorDisplay instead of checking
to see if they were called. Then don't even expose those methods
publicly unless consumers of the class need access.
I wholeheartedly agree with webprojohn's disclaimer. Mocks are a smell as tests should assert the behavior of the code, not its implementation. Testing the latter makes the code brittle to change.
Stepping off my soapbox... :) We're looking for a way to test an asynchronous method. I'm not sure what assertions your tests should make to verify the behavior inside handleStatusChangeSuccess() and handleErrorDisplay(err) so the example below leaves a comment where those assertions would go. The following uses Promise.resolve() and Promise.reject() to trigger the outcomes to test. I've used async/await, Jest has other async examples in their docs.
const Example = require('./example')
describe('handleStatusChangeRequest', () => {
it('should resolve successfully', async () => {
const {handleStatusChangeRequest} = new Example();
const resolvePromise = () => Promise.resolve();
await handleStatusChangeRequest(resolvePromise);
// resolution assertions here
});
it('should resolve errors', async () => {
const {handleStatusChangeRequest} = new Example();
const fakeError = new Error('eep');
const rejectPromise = () => Promise.reject(fakeError);
// if your method doesn't throw, we can remove this try/catch
// block and the fail() polyfill
try {
await example.handleStatusChangeRequest(rejectPromise);
// if we don't throw our test shouldn't get here, so we
// polyfill a fail() method since Jest doesn't give us one.
// See https://github.com/facebook/jest/issues/2129
expect(true).toBe(false);
}
catch (e) {
// rejection assertions here
}
});
});
The answer I have looks so:
**Success tests
const instance = el.find(EntryToolBar).instance()
const spy = jest.spyOn(instance, 'handleStatusChangeSuccess')
await instance.handleStatusChangeRequest(() => Promise.resolve('cool man'))
expect(spy).toHaveBeenCalledTimes(1)
**Error tests
const instance = el.find(EntryToolBar).instance()
const spy = jest.spyOn(instance, 'handleErrorDisplay')
await instance.handleStatusChangeRequest(() => Promise.reject(Error('shit')))
expect(spy).toHaveBeenCalledTimes(1)
As I stated above, the handleStatusChangeSuccess and handleError methods are test else where with some snapshots (they just set state and render out some different jsx). I feel pretty good about this. I'm using spys/mocks, but I'm testing the implementation functions elsewhere. Sufficient?
The following test is behaving oddly:
it('Should return the exchange rates for btc_ltc', function(done) {
var pair = 'btc_ltc';
shapeshift.getRate(pair)
.then(function(data){
expect(data.pair).to.equal(pair);
expect(data.rate).to.have.length(400);
done();
})
.catch(function(err){
//this should really be `.catch` for a failed request, but
//instead it looks like chai is picking this up when a test fails
done(err);
})
});
How should I properly handle a rejected promise (and test it)?
How should I properly handle a failed test (ie: expect(data.rate).to.have.length(400);?
Here is the implementation I'm testing:
var requestp = require('request-promise');
var shapeshift = module.exports = {};
var url = 'http://shapeshift.io';
shapeshift.getRate = function(pair){
return requestp({
url: url + '/rate/' + pair,
json: true
});
};
The easiest thing to do would be to use the built in promises support Mocha has in recent versions:
it('Should return the exchange rates for btc_ltc', function() { // no done
var pair = 'btc_ltc';
// note the return
return shapeshift.getRate(pair).then(function(data){
expect(data.pair).to.equal(pair);
expect(data.rate).to.have.length(400);
});// no catch, it'll figure it out since the promise is rejected
});
Or with modern Node and async/await:
it('Should return the exchange rates for btc_ltc', async () => { // no done
const pair = 'btc_ltc';
const data = await shapeshift.getRate(pair);
expect(data.pair).to.equal(pair);
expect(data.rate).to.have.length(400);
});
Since this approach is promises end to end it is easier to test and you won't have to think about the strange cases you're thinking about like the odd done() calls everywhere.
This is an advantage Mocha has over other libraries like Jasmine at the moment. You might also want to check Chai As Promised which would make it even easier (no .then) but personally I prefer the clarity and simplicity of the current version
As already pointed out here, the newer versions of Mocha are already Promise-aware. But since the OP asked specifically about Chai, it's only fair to point out the chai-as-promised package which provides a clean syntax for testing promises:
using chai-as-promised
Here's how you can use chai-as-promised to test both resolve and reject cases for a Promise:
var chai = require('chai');
var expect = chai.expect;
var chaiAsPromised = require("chai-as-promised");
chai.use(chaiAsPromised);
...
it('resolves as promised', function() {
return expect(Promise.resolve('woof')).to.eventually.equal('woof');
});
it('rejects as promised', function() {
return expect(Promise.reject('caw')).to.be.rejectedWith('caw');
});
without chai-as-promised
To make it really clear as to what's getting tested, here's the same example coded without chai-as-promised:
it('resolves as promised', function() {
return Promise.resolve("woof")
.then(function(m) { expect(m).to.equal('woof'); })
.catch(function(m) { throw new Error('was not supposed to fail'); })
;
});
it('rejects as promised', function() {
return Promise.reject("caw")
.then(function(m) { throw new Error('was not supposed to succeed'); })
.catch(function(m) { expect(m).to.equal('caw'); })
;
});
Here's my take:
using async/await
not needing extra chai modules
avoiding the catch issue, #TheCrazyProgrammer pointed out above
A delayed promise function, that fails, if given a delay of 0:
const timeoutPromise = (time) => {
return new Promise((resolve, reject) => {
if (time === 0)
reject({ 'message': 'invalid time 0' })
setTimeout(() => resolve('done', time))
})
}
// ↓ ↓ ↓
it('promise selftest', async () => {
// positive test
let r = await timeoutPromise(500)
assert.equal(r, 'done')
// negative test
try {
await timeoutPromise(0)
// a failing assert here is a bad idea, since it would lead into the catch clause…
} catch (err) {
// optional, check for specific error (or error.type, error. message to contain …)
assert.deepEqual(err, { 'message': 'invalid time 0' })
return // this is important
}
assert.isOk(false, 'timeOut must throw')
log('last')
})
Positive test is rather simple. Unexpected failure (simulate by 500→0) will fail the test automatically, as rejected promise escalates.
Negative test uses the try-catch-idea. However: 'complaining' about an undesired pass happens only after the catch clause (that way, it does not end up in the catch() clause, triggering further but misleading errors.
For this strategy to work, one must return the test from the catch clause. If you want't to test anything else, use another it()-block.
Thre is a better solution. Just return the error with done in a catch block.
// ...
it('fail', (done) => {
// any async call that will return a Promise
ajaxJson({})
.then((req) => {
expect(1).to.equal(11); //this will throw a error
done(); //this will resove the test if there is no error
}).catch((e) => {
done(e); //this will catch the thrown error
});
});
this test will fail with following message: AssertionError: expected 1 to equal 11
As part of my redux action, it makes several sequential api requests. The apiCall method returns a Promise with some value, and that value is used by a subsequent apiCall to make another request, and so on. I'm using Jest to test these api calls.
const myAPI = {
apiCall(param: any): Promise<any> {
return new Promise((resolve, reject) => {
resolve('result');
});
},
};
const serialAPIRequests = () => {
myAPI.apiCall('first_param')
.then((result) => {
console.log(result);
return myAPI.apiCall(result);
})
.then((result) => {
console.log(result);
return myAPI.apiCall(result);
})
.then((result) => {
console.log(result);
return Promise.resolve(result);
});
};
I am trying to write a test to ensure apiCall has been called the correct number of times and with the right parameters.
describe.only('my test', () => {
it('should test api stuff', () => {
myAPI.apiCall = jest.fn()
.mockReturnValueOnce(Promise.resolve('result1'))
.mockReturnValueOnce(Promise.resolve('result2'))
.mockReturnValueOnce(Promise.resolve('result3'));
serialAPIRequests();
expect(myAPI.apiCall).toHaveBeenCalledTimes(3);
});
});
What happens is that Jest report Expected mock function to have been called three times, but it was called one time.
Jest test result shows that
● Console
console.log
result1
console.log
result2
console.log
result3
● my test › should test api stuff
expect(jest.fn()).toHaveBeenCalledTimes(3)
Expected mock function to have been called three times, but it was called one time.
The fact that console.log showed different values means the mocked return was properly passed through the mock function and it was called 3 times.
What could be causing this and how do I properly test this function?
Use async/await to test async code. Read more here: https://facebook.github.io/jest/docs/en/tutorial-async.html
describe.only('my test', () => {
it('should test api stuff', async () => {
myAPI.apiCall = jest.fn()
.mockReturnValueOnce(Promise.resolve('result1'))
.mockReturnValueOnce(Promise.resolve('result2'))
.mockReturnValueOnce(Promise.resolve('result3'));
await serialAPIRequests();
expect(myAPI.apiCall).toHaveBeenCalledTimes(3);
});
});
Promises are async so by the time you do you check the mock was actually called once.
You could do this instead. Wait for all calls to be done and return a promise to indicate the test is async.
return serialAPIRequests().then(() => {
expect(myAPI.apiCall).toHaveBeenCalledTimes(3);
})