mapM_ for bytestring

Mario Blažević blamario at acanac.net
Fri Sep 13 04:55:13 CEST 2013


On 09/11/13 10:28, Edward Kmett wrote:
> We can't just upgrade those classes to be type family based, as this
> destroys their support for polymorphic recursion.
>
> We could write another class though.
>
> The question is if that class belongs in base or in some other package,
> as it takes is pretty far afield of anything purporting to be portable.

	Not necessarily. If you get rid of the item type, the resulting class 
is standard Haskell:

class Monoid m => FactorialMonoid m where
    foldMap :: (FactorialMonoid m, Monoid n) => (m -> n) -> m -> n

	With this approach you're folding over the atomic factors of the 
ByteString, i.e., ByteStrings containing 1 byte each, rather than the 
bytes themselves. The same interface covers strict and lazy ByteString 
and ByteString.Char8 as well.

	Yes, abstracting away the item type would have some performance impact, 
as foldMap would need to construct the singleton ByteStrings. I'm hoping 
they could be deforested away.





More information about the Libraries mailing list