I wondered if something like that could work, but I wasn't sure that mdo allowed recursion in its let-bindings...
The mdo implementation in ghc does not actually...
I think there's a misunderstanding there. let expressions inside an mdo can of course be recursive; both Hugs and ghc support this: ----------------------------- import Control.Monad.Fix main = mdo x <- return y let y = 1:y z <- return y print (take 1 x, take 2 y, take 3 z) ------------------------------ will happily print ([1],[1,1],[1,1,1]); no problems there. Note that the let defined variable 'y' is used both before and after its definition; and it's also defined recursively. May be you are thinking of polymorphism problem. The restriction is this: let bound variables in an mdo cannot be use polymorphically if they are referred to before their definition. (It's also OK to use them polymorphically after their definition.) This is merely there to ensure that the translation will be a well-typed Haskell-98 expression; nothing more than that. If we choose, we can make it polymorphic everywhere, by typing mdo-expressions separately. As John pointed out, there're no technical difficulties there. The only thing we lose is that the translation may fail to type-check. This was a compromise we had to do, and we chose the light-weight view that mdo is only syntactic-sugar. -Levent.