Hey everyone! I've been struggling to maintain stable unit tests while making changes to my code. It seems like whenever I tweak something, the tests tend to break, even if the inputs and outputs are unchanged and the code still functions correctly. I've heard that the key is to focus on testing the behavior of the units rather than their implementation details. But I find that isolating my units with test doubles or mocks often leads to a certain level of dependency on the implementation itself. Any advice on how to approach this issue? Thanks!
1 Answer
Many fragile tests arise from over-mocking and focusing too much on internal behavior instead of final outcomes. If your tests break even though the inputs and outputs are unchanged, it's a sign they're tightly coupled to the underlying implementation. Aim to mock only true external interactions, like network or database calls, and keep your assertions centered on what the function returns or the observable effects. Refactoring your code into smaller, pure functions can also help make behavior-based testing simpler and more robust.
I totally see what you're saying! I often create mocks just to explore different paths in the code, which leads to upholding certain calls in the mocks. Are there strategies to handle this while keeping the tests durable?