Re: darcs patch: Control.Monad: +void :: f a -> f ()

On Wed, Jan 13, 2010 at 10:48 AM, Gwern Branwen wrote:

On Wed, Jan 13, 2010 at 10:26 AM, Tyson Whitehead wrote:

...

I was wondering why forM_ cares about the return type, but then, when I checked the type signature, it seems to me like it doesn't

 forM_ :: Monad m => [a] -> (a -> m b) -> m ()

Perhaps forkIO and friends should be like this as well.  Replace them with something like the following

 forkIO . (>> return ()) :: IO a -> IO ThreadId

(presumably this would be optimized away?)

Cheers!  -Tyson

PS:  Note that I'm not saying there should be a void function, just wondering why all these functions that don't care about return type force it to be ()?

I think the idea is that the type forces the caller to explicitly deal with the return value, to avoid issues such as space leaks. Neil Mitchell has made a similar argument for >> forcing an explicit ignoring: http://neilmitchell.blogspot.com/2008/12/mapm-mapm-and-monadic-statements.ht...

I've never found this argument particularly convincing. I'd much rather generalize the types of fork and forM_ than introduce more trivial functions like void. -- Dave Menendez http://www.eyrie.org/~zednenem/