Proposal for deepseq: Data.Sequence.Seq instance for NFData
ankarakulov at gmail.com
Tue Jul 12 17:38:51 CEST 2011
Oops.. Probably I get why toList is preferred method for making instances :)
My rnfFoldable does not work.
But this version must be correct:
> rnfFoldable :: (NFData a, Foldable f) => f a -> ()
> rnfFoldable = Foldable.foldl (\z x -> rnf x `seq` z) ()
On Tue, Jul 12, 2011 at 6:21 PM, Alexey Karakulov <ankarakulov at gmail.com>wrote:
> My proposal is qiute straightforward. The patch is attached and it's
> essence is in these lines:
> > instance NFData a => NFData (Data.Sequence.Seq a) where
> > rnf = rnf . Foldable.toList
> [Discussion part]
> I choosed toList method to keep the code style, but I'm not sure it's
> efficient to create cons'es when we could just fold:
> > rnfFoldable :: (NFData a, Foldable f) => f a -> ()
> > rnfFoldable = Foldable.foldMap rnf
> Or it's better to do this with foldl (while not so nice at is was with
> using () as monoid):
> > rnfFoldable = Foldable.foldl (const rnf) ()
> There was a proposal to add strict folds to Foldable type class, so we can
> use foldl' there.
> My proposal is to add the instance in either way.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Libraries