I vaguely remember someone (maybe Duncan Coutts?) saying that this was a commonly held misconception, and that GHC did indeed GC CAFs when optimization is enabled. If I am remembering incorrectly, does anyone have a reference to a ticket outlining this non-GC'ed CAF behavior?
Thanks, Dan On Thu, Apr 16, 2009 at 4:57 AM, Eugene Kirpichov <ekirpic...@gmail.com> wrote: > The parameterless version is a top-level definition and won't get > garbage-collected, IIRC. > So, if you evaluate primes!!10000000, you'll end up with a > 10000000-element list hanging in memory forever. > If you evaluate (primes' ()) !! 10000000, you won't. > > 2009/4/16 Niemeijer, R.A. <r.a.niemei...@tue.nl>: >> Heinrich Apfelmus wrote: >>> +1 except that exporting the potentially infinite list of primes is >>> problematic in that it may become a memory leak. >>> >>> I'd suggest to export two versions >>> >>> primes :: [Integer] >>> primes' :: () -> [Integer] >>> >>> for casual (i.e. throwaway program to solve a Project Euler problem) and >>> for memory aware use respectively. >> >> I'm afraid I don't quite follow. Would you mind explaining what the first >> parameter is for and how it would solve the memory leak? >> _______________________________________________ >> Haskell-Cafe mailing list >> Haskell-Cafe@haskell.org >> http://www.haskell.org/mailman/listinfo/haskell-cafe >> > > > > -- > Eugene Kirpichov > Web IR developer, market.yandex.ru > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe@haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe > _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
