Add deeply strict liftM #13

Activity

1. RyanGlScott commented on Aug 15, 2016

   

   RyanGlScott

   on Aug 15, 2016

   Member

   More actions

   Sanity-check: is this quite analogous to (<$!>) from base? Its definition is:

   (<$!>) :: Monad m => (a -> b) -> m a -> m b
   f <$!> m = do
     x <- m
     let z = f x
     z `seq` return z

   Whereas with (<$!!>), we have something that's closer to:

   (<$!!>) :: (Monad m, NFData b) => (a -> b) -> m a -> m b
   f <$!!> m = do
     x <- m
     f x `deepseq` return (f x)

   I'd hope that GHC is smart enough to apply CSE on the two occurrences of f x, but maybe it'd be wise to use explicit sharing here.

   👍React with 👍1Reacted by Herbert Valerio Riedel
2. treeowl commented on Aug 15, 2016

   

   treeowl

   on Aug 15, 2016

   ContributorAuthor

   More actions

   I don't understand how it could translate the way you say. Doesn't $!! take care of the sharing here? If not, that's broken.
3. treeowl commented on Aug 15, 2016

   

   treeowl

   on Aug 15, 2016

   ContributorAuthor

   More actions

   Actually, $!! shouldn't have to worry about it either. Unless GHC actually copies the expression for no reason, we should be fine.
4. RyanGlScott commented on Aug 15, 2016

   

   RyanGlScott

   on Aug 15, 2016

   Member

   More actions

   Ah, you're right, I inlined one too many definitions in my head. The fact that you're using ($!!) here explicitly takes care of that issue. Carry on :)
5. treeowl commented on Aug 15, 2016

   

   treeowl

   on Aug 15, 2016

   ContributorAuthor

   More actions

   The bigger problem is that <$!> is weird. It should really be called <!$>. <$!> should force the argument; <!$!> should force both. For deepseq, I think only <!!$> and <$!!> are really useful. But it's too late to fix this, so we should follow base for consistency.
6. RyanGlScott

   closed this as completedin #23on Nov 17, 2016