Tom Ellis wrote:
On Wed, Jun 11, 2014 at 01:25:48AM +0200, Bertram Felgenhauer wrote:
begin cont = cont (return ()) a m cont = cont (m >> putStrLn "a") b m cont = cont (m >> putStrLn "b") c m cont = cont (m >> putStrLn "c") end m = m
main = begin a b c a b a end
This is fantastic!
Unfortunately this causes performance problems with ghc's type checker: main = begin a a a a a a a a a a a a a a a a a end takes more than a second to type-check under ghc 7.8.2, and every additional 'a' adds a factor of two. Interestingly, ghc 7.6.3 does not show this behaviour. Note that the instantiated types of 'a' grow exponentially: The first 'a' in 'begin a end' and 'begin a a end' have types M -> (M -> M) -> M and M -> (M -> (M -> M) -> M) -> (M -> M) -> M respectively, where 'M' abbreviates IO (), and each additional 'a' turns M -> r into M -> (M -> r) -> r, doubling the type's size. Cheers, Bertram P.S. a related way for causing bad type checking performance is the innocent looking main = id id id id id id id id id id id id id id id id id id id id id id (return ()) where again, the type of the first 'id' grows exponentially in the number of 'id' calls. In this case, ghc-7.6.3 performs just as badly as ghc-7.8.2.