Thank you Tillman, and Oleg, for your advices! Since ICFP contest is starting in a few hours, I will make a quick response with gratefulness and will read the full paper later.... Let me guess a few things, please tell me am I right. The share :: m a -> m (m a) is almost the thing I am looking for. I have independently (believe me!) invented an equivalent, "bind trick," for my DSL: http://nushisblogger.blogspot.jp/2012/06/builder-monad.html but am now enlighted with what it really meant. Still, `share` cannot bring the shared binding to global scope (e.g. it doesn't allow place sunMass, earthMass, marsMass in separate modules) I guess, my original question is ill-posed, since all the values in Haskell are pure, so in the following trivial example
earthCopyMass = earthMass
there is no way to distinguish two masses, thus there's no telling if Earth and EarthCopy is two reference to one planet or two distinct planets. I don't know if memo can solve this problem. I have to test. I'll try implement `memo` in your JFP paper section 4.2 Memoization; seems like it's not in explicit-sharing hackage. I'm vaguely foreseeing, that like in memoized (f2 0, f2 1, f2 0, f2 1) we need to pass around some `world` among it. That will be random generator seeds if our continuous-nondeterminism is an MonadIO when we perform Monte-Carlo simulations; or it's a virtual `world` if we make Gaussian approximation of probabilistic density functions. To Ben: Thank you for your comments anyway! But since I'm not going to use the List monad (the use of List was just for explanation,) the discreteness is not an issue here. That's my intent when I said "another story." Sorry for confusion! All the best, Takayuki