"S.D.Mechveliani" wrote:
> I do not feel aggressive against Prelude.
So you're not advocating eliminating the current min and max functions
(and others), and leaving only list-based versions?
> Concerning foldl1 min,
> there is a little spot that it tends to run in O(n) space:
> foldl1 min [1..n]
>
> While the simplest direct implementation of minimumBy tend to run
> in a constant space.
You're right; lazy folds tend to have space leaks. A strict fold should
be used instead. I think that strict versions of fold should be
exported from the prelude, along with the lazy ones. A strict foldl' is
already used internally. I would redefine minimum and maximum to use a
strict fold, the way sum and product do today (at least in hugs/GHC).
Strictness analysis might eliminate, or reduce, the need for special
strict & lazy versions of fold. I wonder; is anyone working on adding
strictness analysis to any major Haskell implementation?
Regards,
Matt Harden
