Indeed. I did this with one program: I wrote some number crunching code in Haskell, and exposed some functions to C via the FFI. Then I wrote the user interface in Cocoa/Objective-C. The primary thing that requires some though is how to pass data between C and Haskell. You usually end up marshaling/unmarshaling some structs:
Is the gain rom Haskell worth the impedance mismatch of cross communication?
I guess it depends on how much you love Haskell ;), and the usage scenario. I think using Haskell's FFI in this direction is especially attractive if you prefer to write most code in Haskell, but need some functionality that does not map naturally to a pure functional language (such as GUIs in Cocoa). If you want to mix a lot of imperative and functional code seamlessly, F#/C# or Scala/Java are probably better options.
Thanks what's the size like? The last time I built a Haskell application it was quite large.
4969KB with the runtime and package dependencies linked in the dynamic library (Darwin, 64-bit). Since I do not have dynamic libraries of its dependencies installed, I cannot quickly try a completely dynamic version...
interestingly Wp7 is by far the easiest mobile to get a good functional language running on.
I purchased a Windows Phone last week (iOS user here) to see how it works, and I am surprised how good the integration is.
http://www.haskell.org/haskellwiki/FFICookBook#Working_with_...