First, having a robust type system (whether simplified for the programmer by the presence of type inference or not) makes a large number of the types of tests performed in scripting language code unnecessary. Return types, errors raised, bounds checks, etc., are all implicit in the type signature of a function, and so do not need to be checked.

Secondly, at least in a pure language like Haskell, stateful code must be delimited via monads. Good program design dictates that you make those stateful components as small as possible, which then allows you to test the remainder of your code without resorting to mocks and stubs at all.

As a basic example, instead of passing a mocked file handle object between multiple methods, you might change your API to use a sequence of strings representing lines in the source file. At test time, you can simply pass any such sequence to your functions rather than mocking or stubbing all the potential behaviors of a "real" file.

That being said, your real question seems to be, "can I substitute X for Y, as I can in Ruby/Python/etc.?" My answer to that would be "yes, as long as there are (potentially trivial) implementations of all the required functions in the typeclass(es) of Y that we care about."

Haskell typing, at least, acts a lot more like an abstract interface than a concrete class, and a single datum is likely to be a member of many different typeclasses. However, each parameter to a function is constrained by a single typeclass, so you need not provide for every possible operation on a piece of data in order to mock it for testing.