Always interesting to hear different points of view on this subject. Personally I think mocks make sense to capture complex sets of interactions or otherwise difficult to reach error conditions, so I don’t think it’s a do or do-not kind of thing.
I’m assuming we’re talking about test mocks.
I’m quite partial to using mocks within hard-to-reach code. I was curious the other day when I saw the Primeagen saying test mocks are awful - I’m not that heated about it either way and I think if it will take an hour of dev time or a simple mock I would always prioritize using the mock.
Overly fine-grained “brittle” mocks are something to be wary of, I agree that you don’t want to have to change a ton of tests every time you edit the code. But the hand-wavey suggestion of “use a fake instead” with no follow up is not particularly helpful…
Mocking and dependency injection don’t seem to be mutually exclusive, if anything dependency injection can make it easier to get the component you’re testing to interface with the mock
How would you isolate your thing for testing then without mocks? Contract Driven Development is the best thing to decouple development teams from eachother so that they can continue to work on their part as the teams have agreed on the API.
And I doubt it’s different in Go.
Unless I’m missing something, the author seems to think that a “Mock” means verification of exact /fixed/fragile call sequences, but instead advocates use of (undefined) “Fake” objects or alternatively skipping those unit tests and relying on integration or other high level tests for those parts of the system…
I can’t decide if they actually believe that “Fake” != “Mock” or it’s just to drum up traffic…
