- Testing is not a guarantee of being free from bugs, and it’s not the point of testing. The point of testing is in reducing risks by verifying that main and important requirements are met.
- The more meaningful tests you have, the lower the risks are, but the higher the maintenance cost.
- Different testing covers different sets of requirements and thus risks.
- White-box tests verify the functional requirements (what the code should and should not do).
- Black-box testing may be used for verifying the functional requirements or validating the non-functional requirements.
- Property-based is more suitable for CPU-bound and pure code, but less suitable for IO-bound code.
- Regression tests protect from unexpected breakages. And sometimes we write tests to find bugs.
- If we’re shipping an application framework, a good idea will be to provide a testing framework so that our clients could test their scenarios.
- Establishing a proper testing environment may be expensive or even impossible if the architecture isn’t testable.
- Three-layered Free monadic architecture has a very good testability compared to all other architectural approaches.
- Mocking is important. Achieving good (meaning useful) test coverage without mocks doesn’t seem possible for most of the applications which are IO-bound by nature.
- Mocking is only possible with good architecture that follows all the design principles such as SOLID or ‘low coupling, high cohesion’.
- Free monadic architectures achieve very good testability by decoupling interfaces and implementation.
- Automatic white-box testing is good for golden tests although it’s a bit clumsy in setup and maintenance.
