Yes, it is very difficult. A sensible API for a standard array library is something that needs more research. FWIW, I don't know of any other language that has what I'd like to see in Haskell. C++ probably comes closest but they have it easy - they don't do fusion.
I assume you've looked at SAC? http://www.sac-home.org/ Their main research and development focus was/has been on arrays (fusion/layout/padding/tiling/slicing/data-parallelism/shape-invariance (source algorithms parameterized over array dimensionality/shape)/ whole-array ops/list-like ops/lots of surface operations reducable to a minimal set of core operations that need implementation/cache behaviour/performance/performance/performance/..). When they started out, I tried to make the point that I would have liked to have their wonderful array ideas in our otherwise wonderful language, but they preferred to follow their own way towards world domination (*). Does that sound familiar?-) Claus (*) ok, they did have a good motive: they came out of a research group that had done non-sequential functional programming in the 1980s, with all the things we see today: great speedups, shame about the sequential baseline; creating parallel threads is easy, load balancing slightly harder, but pumping (creating thread hierarchies recursively, only to see them fold into a small result, for the process to begin again) is a waste, etc.; so they decided to start from a fast sequential baseline instead of full functional language, and designed their language around arrays instead of trying to add arrays to an existing language.