I am unconvinded that this is any more wrong than using a for loop in an imperative language. Remember that the lists are lazy, so it’s misleading to say “convert the array to a list” since what happens most of the time is that elements are taken out of the array and passed to the processing function and then thrown away before the next element is processed.
I don't think it's misleading to say "convert to a list". Without the presence of some fusion-esque optimization, the entire list is created and destroyed. It's nice if it can be done incrementally, but that's still a lot of garbage. And for a concrete example, consider concatenating a bunch of arrays. Unless it gets deforested, you have to copy each element a couple of times, since there is intermediate list concatenation. And then this kind of API is incompatible with unboxing... if you later decide to unbox the data type then I don't think you could ever get the intermediate list approach to work, since it's boxed by nature.
I’m not familiar with Repa or NumPy, but what can they do that cannot be done with judicious use of ixmap, which is a very powerful mechanism.
I spent hours messing with ixmap before getting something that seemed to work right. Maybe I just have more trouble than most with it, but though it may be powerful it's certainly not easy to use! After spending too much time on it I learned to just convert to a list and back.