How does TDD compare with Functional Programming Languages?

How does TDD compare with Functional Programming Languages like F# and Erlang?

I haven't actually worked directly with a functional programming language yet, but from what I've seen of it, you have two sides of an equation and they have to balance like in algebra or accounting; this seems somewhat reminiscent of TDD where you define your expected outputs as Assert statements (one side of the equation) and the rest of the functionality goes into a class decoupled from the test (the other side of the equation), except that functional programming IMHO seems a bit cleaner.

Do the two actually have similarit开发者_如何学Goies, or am I just overthinking this a bit?

---

Software Design v Development Methodology

They're orthogonal.

TDD is an approach to developing software which focuses on ensuring correctness by developing tests against specifications before the production code is written. Functional programming is a paradigm for designing and implementing software.

---

I think TDD and functional programming (FP) are different in that TDD is a methodology and FP is programming paradigm.

I would say that FP help when practicing TDD as FP encourages you to make things deterministic when possible. Deterministic functions are much easier to test than non-deterministic ones.

---

Chris is correct in saying that they are orthogonal. However, there are some aspects of functional programming that make testing of functional programs a lot easier.

• Functional programs are composed from functions and guarantee that the function will behave the same in all contexts. This means that when you test a function in unit test, you know that it will always work this way. You don't have to test whether it works if you plug it into some other environment.

• Functions take arguments and return results and that's the only thing they do. This means that you can usually avoid mocking and similar tricks, because you don't need to verify whether a function does some call on some object. You only need to verify that it returns the expected result for given arguments.

• Finally, there are some nice automatic tools for testing functional programs. For F#, we have FsCheck (which is based on QuickCheck known from Haskell). These benefit from various properties of functional programs.

So, they both have different purpose and are essentially a different thing, but there are some nice relations (perhaps like a tea and a teapot :-) they are completely different things, but work very well together!)

---

You're correct that when writing a functional program you might use equational reasoning to derive the definition of a function. However, that reasoning typically doesn't exist in some reified form (such as tests), so it is not generally the case that a function is proven correct in a way that is machine- or human-checkable. It is certainly possible to use TDD with functional languages (e.g. using any .NET compatible TDD library with F#) to verify that functions have been derived correctly, but there are also other testing strategies which might be more unique to functional languages, such as using QuickCheck for randomized specification checking.

---

I think the similar feel between the two stems from the fact that, with both, functions are supposed to be deterministic. FP functions shouldn't have side effects and side effects in test functions for object orientated code should be removed by injecting stubs.