Re: Proposal: Foldable typeclass: make foldl' and foldr' class methods

On Mon, 2011-06-20 at 22:44 +0200, Johan Tibell wrote:

On Mon, Jun 20, 2011 at 10:11 PM, Duncan Coutts wrote:

...

Mm, the '1' variants are obvious. The fold and foldMap are more interesting. I'd have to think harder about what their default definition would be, or what specialised implementations might look like (e.g. tree folds starting from the leaves and working towards the root). Are there any obvious use cases for strict monoid folds?

How about parallel folds? Here's something I'd like to support for HashMap:

parFold :: Monoid m => (v -> m) -> HashMap k v -> m sumValues = parFold Sum

The strategies stuff uses Traversable I think. To some degree I think you could do parallel evaluation with the existing Foldable operations because you can pick the monoid, so you can pick one that evaluates in parallel. Needs a bit of thought and experimentation. Duncan