5. What is a unit?
● Smallest bit of code you can test?
● Talking to the actual resource may be OK if it’s stable and fast
● Classic versus mockist styles (Martin Fowler)
● Solitary versus sociable tests (Jay Fields)
● White box versus black box testing
● What’s important is the contract
http://martinfowler.com/bliki/UnitTest.html
8. You don’t know unit testing until you’ve unit tested
There’s a first time for every developer. Some are more lucky than others because
they ramp up in an environment that already embraces unit testing.
“But can already write flawless code when I’m in the zone.”
True. Because you’re actually running unit tests, without realizing, in your mind
when you’re in the zone.
Try taking a 3 week break and see what happens to those ephemeral unit tests.
Turn those tests into unit test code so that they’re repeatable and unforgettable.
10. Good unit tests
● Are functionally correct. They don’t just exercise code for the sake of
exercising code.
● Don’t depend on subsequent tests -- every test runs in its own clean
environment, failure of a test doesn’t bring the entire test suite down
● Run fast. You need to be able to run all of your tests as quickly and as
frequently as possible. Otherwise, they lose value.
● Are actually run. Automatically. So that you don’t forget to run them.
● Add new unit tests for newly discovered [and fixed] issues.
12. Good code
● Good code is more unit testable
● It all comes down to good architecture and design
● Planning for unit tests facilitates good code
● Good encapsulation: interfaces with small surfaces, well-defined contracts,
non-leaky abstractions
● Keep interdependencies low
14. Why and what are you unit testing?
● Misguided reasons: processes, meeting performance numbers
● Testing just for testing: glue code that doesn’t have any logic, ineffective tests
that don’t actually test the functionality
● Testing legacy code that is actually un-unit-testable
Be pragmatic. Don’t waste effort. Sometimes unit testing is not the answer (try
end-to-end instead).
16. Benefits of unit testing
Benefits beyond finding bugs:
● Better code
● Safety net for refactoring
● Documentation of functionality (especially when in BDD style)
● Prevents code from becoming an untestable entangled mass
18. Be test-ready on day one
● Even if you’re not planning to add test yet
● Even if there’s no code worth testing yet
● Prime your environment for future unit tests
● Especially, CI environment setup can be time consuming
● You never know when that moment will come when you have some critical
code that needs unit testing
Do this. Please.
19. Sidenote: At a bare minimum...
Even you have no time or energy to write unit tests as you go, prepare a manual
test plan, and someone in your team execute them (manually) prior to releases.
Bonus: share the effort as a team.
Basic smoke tests, checking for end-to-end sanity and regression.
Do this. Please.
30. Test correctness
● Should not be just exercising code
● Should be functionally correct
● Subject to peer review?
I don’t know of any solutions to ensure test correctness.
35. Timeout test
test/testTimeout.js:
it('should call the callback after the delay', (done) => {
const start = Date.now();
timeout.set(() => {
const elapsed = Date.now() - start;
expect(elapsed).to.equal(100);
done();
}, 100);
});
36. Run
npm test
timeout
.set()
when called with a callback and a delay
1) should call the callback after the delay
Uncaught AssertionError: expected 105 to equal 100
+ expected - actual
-105
+100
38. Write deterministic tests that run fast
● Don’t rely on chance
● A less than 100% pass rate is not acceptable
● Don’t waste time with arbitrary delays
● Use the right tools for the [right] job
41. Use a spy and a fake timer
test/testTimeout.js:
const sinon = require('sinon');
describe('timeout', () => {
let clock = null;
beforeEach(() => {
clock = sinon.useFakeTimers();
});
afterEach(() => {
clock.restore();
});
42. Use a spy and a fake timer (continued)
describe('.set()', () => {
describe('when called with a callback and a delay', () => {
it('should call the callback after the delay', () => {
const callback = sinon.spy();
timeout.set(callback, 100);
clock.tick(100);
expect(callback).to.have.been.called;
});
});
});
43. Run
npm test -- --grep timeout
timeout
.set()
when called with a callback and a delay
✓ should call the callback after the delay
100% pass rate.
58. When account not found
db.query.withArgs('account', { id: -1 }).rejects(
new Error('Account not found')
);
return accountService.findById(-1)
.catch((error) => {
expect(error).to.deep.equal(
new Error('Account not found')
);
});
60. Run
npm test -- --grep account
AccountService
.findById()
when called for an existing account
✓ should return a promise resolved with the account
when called for a non-existent account
✓ should return a promise rejected with an error
61. Need the positive case to fail the test
return accountService.findById(-1)
.catch((error) => {
expect(error).to.deep.equal(
new Error('Account not found')
);
})
.then(() => {
throw new Error('Should not have been resolved');
});
62. Run
npm test -- --grep account
AccountService
.findById()
when called for an existing account
✓ should return a promise resolved with the account
when called for a non-existent account
1) should return a promise rejected with an error
66. Updated negative test
return expect(accountService.findById(-1))
.to.eventually.be.rejectedWith(Error, 'Account not found');
67. Run
npm test -- --grep account
AccountService
.findById()
when called for an existing account
✓ should return a promise resolved with the account
when called for a non-existent account
1) should return a promise rejected with an error
AssertionError:
expected promise to be rejected with 'Error'
but it was fulfilled with { id: -1, name: 'Negative One' }