On Tue, 2008-01-22 at 01:41 +0200, Yitzchak Gale wrote:
I wrote:
Generating an infinite list from a random generator "burns up" the generator, making it unusable for any further calculations.
Jake McArthur wrote:
That's what the split function is for. ^_^
Yes, that is a nice approach. I have been avoiding it due to the following comment in the docs for System.Random:
"This is very useful in functional programs... but very little work has been done on statistically robust implementations of split ([System.Random#Burton, System.Random#Hellekalek] are the only examples we know of)."
And my own experience has been that cases where I need split tend to be in a state monad anyway, where there isn't any real advantage to split.
That said, looking around briefly, I came up with this paper by L'Ecuyer et al that does seem to describe a decent random generator with properties of split worked out:
http://citeseer.ist.psu.edu/493863.html
L'Ecuyer's implementations in C, C++ and Java are here:
http://www.iro.umontreal.ca/~lecuyer/myftp/streams00/
If we had something like that in Haskell, I might use split more often.
According to the documentation http://www.haskell.org/ghc/docs/latest/html/libraries/random/System-Random.h... That -is- what we have.