[faut_reflechir]

I appreciate the polemical framing -- not many people are comfortable taking the "wrong" side of this issue publicly and it's easy to preach purity when you have a nine-to-five.

I think, though, that even if you grant that speed is more important than all else, and even if you are willing to accept a 200% interest rate on tech debt to get your releases out, you still want unit tests from day one.

Done right, writing unit testing in parallel with your code speeds things up rather than slowing them down. Personally, I find this is the case for even tiny projects like a Project Euler problem unless they can be done in under about an hour.

[pclmulqdq]

Personally, I have found that unit tests speed you up a lot, while integration tests often slow you down. In other words, if you're writing a "mock_[X]" class/function, it's probably slower than just testing with X.

[alexgrover]

Typically people refer to those by the opposite naming convention - unit tests just test one thing and mock any dependencies, while integration tests integrate multiple things and generally involve less mocking