It never makes sense to use foldl on lists in Haskell (2019)

Re:It never makes sense to use foldl on lists in Haskell (2019) by 08-15 on Saturday March 21, 2020 @05:11AM (#22645426)

Foldl' doesn't leak, but it sure looks like it does, if you think the folded function is strict in the accumulator argument, but isn't. For example, this attempt at computing the average of a list

  uncurry (/) . foldl' (\(s,l) x -> (s+x,l+1)) (0,0)

leaks, because during the recursion, there is no reason to scrutinize the intermediate pairs. This version is fine:

  uncurry (/) . foldl' (\(!s,!l) x -> (s+x,l+1)) (0,0)

[ Parent | on HN ]

Re:It never makes sense to use foldl on lists in Haskell (2019) by dependenttypes on Saturday March 21, 2020 @05:24AM (#22645457)

{-# LANGUAGE Strict, StrictData, BangPatterns #-} should imply the !, right?

[ Parent | on HN ]

Re:It never makes sense to use foldl on lists in Haskell (2019) by 08-15 on Saturday March 21, 2020 @05:46AM (#22645523)

I don't think it does. 'Strict' doesn't add an implicit ! to nested patterns (so (s,l) becomes !(s,l), not (!s,!l)), and 'StrictData' doesn't change anything about the tuple, which comes from a different module. (Disclaimer: I haven't actually used 'Strict', this is my interpretation of the user guide.)

[ Parent | on HN ]

Re:It never makes sense to use foldl on lists in Haskell (2019) by nimih on Sunday March 22, 2020 @12:40AM (#22652774)

Thanks, this is a helpful example :)

[ Parent | on HN ]