Conor McBride and I would like to propose some new classes for traversals of containers, based around applicative functors (the abstraction formerly known as idioms): http://www.soi.city.ac.uk/~ross/traversals/ They are proposed for the base package, and would replace FunctorM. Comments most welcome.
Ross Paterson
Conor McBride and I would like to propose some new classes for traversals of containers, based around applicative functors (the abstraction formerly known as idioms):
http://www.soi.city.ac.uk/~ross/traversals/
They are proposed for the base package, and would replace FunctorM.
Comments most welcome.
On line 83-84 or Traversable.hs: -- | Any 'Traversable' can also be made an instance of 'Functor' by -- thng 'fmap' as this function. Maybe thng should be assigning? I'm not sure which word should be there. -- Shae Matijs Erisson - http://www.ScannedInAvian.com/ - Sockmonster once said: You could switch out the unicycles for badgers, and the game would be the same.
On Fri, Nov 18, 2005 at 11:05:37AM +0000, Ross Paterson wrote:
Conor McBride and I would like to propose some new classes for traversals of containers, based around applicative functors (the abstraction formerly known as idioms):
http://www.soi.city.ac.uk/~ross/traversals/
They are proposed for the base package, and would replace FunctorM.
This is very nice (and I say this as one of proponents of FunctorM). If it's in base package, then it's going to be shipped with newer compilers, like GHC 6.6, right? Best regards Tomasz
On Fri, Nov 18, 2005 at 02:24:41PM +0100, Tomasz Zielonka wrote:
If it's in base package, then it's going to be shipped with newer compilers, like GHC 6.6, right?
That's the idea. These modules are portable (as Data.Monoid recently became portable), so they would work with all implementations.
This seems like a solid, worthwhile proposal to me as well. In fact, I will start using this in new code, regardless of its acceptance into the library. Jacques Ross Paterson wrote:
Conor McBride and I would like to propose some new classes for traversals of containers, based around applicative functors (the abstraction formerly known as idioms):
http://www.soi.city.ac.uk/~ross/traversals/
They are proposed for the base package, and would replace FunctorM.
Comments most welcome.
_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries
Hello Ross, Friday, November 18, 2005, 2:05:37 PM, you wrote: RP> http://www.soi.city.ac.uk/~ross/traversals/ RP> Comments most welcome. 1) imho, it's better to change "Prelude." to "List." here: instance Foldable [] where foldr = Prelude.foldr foldl = Prelude.foldl foldr1 = Prelude.foldr1 foldl1 = Prelude.foldl1 and in other places where you use list functions listed in Prelude 2) i can try to make Template Haskell machinery to automatically derive instances of Foldable and Traversable, using your examples for Tree as sample 3) -- If @f@ is also a 'Functor', define @('<$>') = 'fmap'@. -- If it is also a 'Monad', define @'pure' = 'return'@ and @('<*>') = 'ap'@. why you don't just define: instance Monad m => Applicative m where pure = return (<*>) = ap RP> They are proposed for the base package, and would replace FunctorM. it is a beatiful contribution. but it doesn't work without definitions for All, Any and so on :) -- Best regards, Bulat mailto:bulatz@HotPOP.com
On Sat, Nov 19, 2005 at 05:51:44PM +0300, Bulat Ziganshin wrote:
-- If @f@ is also a 'Functor', define @('<$>') = 'fmap'@. -- If it is also a 'Monad', define @'pure' = 'return'@ and @('<*>') = 'ap'@.
why you don't just define:
instance Monad m => Applicative m where pure = return (<*>) = ap
Because that instance would overlap with all others. As it is, these classes are Haskell 98. Better language support for superclasses would be useful here, but is unavailable right now.
RP> They are proposed for the base package, and would replace FunctorM.
it is a beatiful contribution. but it doesn't work without definitions for All, Any and so on :)
I'm not sure I understand this one. These types are defined in Data.Monoid in CVS, and coming soon to a Haskell implementation near you.
Hello Ross, Monday, November 21, 2005, 1:09:26 PM, you wrote:
why you don't just define:
instance Monad m => Applicative m where pure = return (<*>) = ap
RP> Because that instance would overlap with all others. As it is, these RP> classes are Haskell 98. Better language support for superclasses would RP> be useful here, but is unavailable right now. with instances for [], IO and so on? but you can just omit them, they just repeats this definition. or you think that some Monads will require alternative definitions?
it is a beatiful contribution. but it doesn't work without definitions for All, Any and so on :)
RP> I'm not sure I understand this one. These types are defined in RP> Data.Monoid in CVS, and coming soon to a Haskell implementation near you. but not in ghc 6.4.1. btw, why you not wrote about this to main Haskell list? imho this is enough important contribution to make all know about this. and you are not answered about my TH suggestion -- Best regards, Bulat mailto:bulatz@HotPOP.com
participants (5)
-
Bulat Ziganshin -
Jacques Carette -
Ross Paterson -
Shae Matijs Erisson -
Tomasz Zielonka