But I think the reasons behind lazy pattern usage in the mplus implementation are similar to those behind lazy pattern usage in the (>>=) implementation.
I find the explanation of "Lazy patterns are useful in contexts where infinite data structures are being defined recursively" is easier to understand. Actually all the lazy pattern examples I have seen so far belong to this category. Here I want to correct a statement I made previously about: mplus m1 m2 = do ... ~(a,s') <- lift (mplus m1' m2') In one of my emails to the thread I mentioned m1/m2 and m1'/m2' could be different Monads so mplus works differently (so that this is not exactly a recusrive data structure), but even so the lazy pattern works. The problem is when m1/m2 and m1'/m2' are of the same Monad type, then the problem of recursive definition comes up and this is the place lazy patterns are particularly useful. Thanks, Fan