Jeff Martin

@NicolasRinaudo@functional.cafe @danicheg@indieweb.social Mocks should only be mocking behaviour not data. It’s unnecessary and undermines the value of mocking as a technique which exposes behaviours. The presence of multiple mocks should be taken as a signal that the code is encapsulating more behaviours than it should and should be refactored to reduce the number of required mocks / implementation concerns.

@NicolasRinaudo@functional.cafe @danicheg@indieweb.social Which is entirely unnecessary. Either the value has some effect on the outcome of the procedure in which case we assert something about the relationship between the value and the output. Or it doesn’t and it’s existence is redundant. Or it’s used in a side effect and that’s the code which should have been mocked or better factored out. We’ve ignored the signal that having this type of mock is telling us about the design of the code.

@NicolasRinaudo@functional.cafe @danicheg@indieweb.social you shouldn’t it’s because people are lazy and don’t understand side effects. It’s not a mocking problem, it’s a mocking done badly problem. The point of mocking is to say “how can I refactor my code so that this mock is redundant” rather than how can I use a mock to paper over poor design. I do think that newer mocking libraries have made this problem worse by enabling bad practice as a feature, but I don’t think that the problem is mocks rather than their misuse.

@danicheg@indieweb.social @NicolasRinaudo@functional.cafe I don’t think that’s a helpful position. It’s semantic slight of hand. What’s the definition of a stub vs a fake vs a mock. Why is stubbing (whatever that is) data better than mocking data? Use random or example data rather than some partial magic implementation of it. If your data model is so complex that you can’t easily define it for testing I’d question its monolithic structure.