On 2006-08-25 at 19:09PDT Ashley Yakeley wrote:
Jon Fairbairn wrote:
There has been discussion in the past about whether Monad should be defined as
class Functor m => Monad m where ...
It's more complicated now that we have Ross Patterson's "Applicative".
http://haskell.org/ghc/dist/current/docs/libraries/base/Control-Applicative....
FSVO "complicated"... it looks like a Good Thing to me, although I don't like the names much. [It would be nice if there were a bit more commentary in the Haddock -- referring to a paper is fine, but in day to day programming a scan through several modules to find what one wants is hampered by having to step out to read the papers.]
The correct decl I think is this:
class Applicative m => Monad m where -- remove "return" ...
and changing the names of the Applicative functions:
class Functor f => Applicative f where return :: a -> f a ap :: f (a -> b) -> f a -> f b
Better, I think, and though I can see a case for using “unit” instead of return (as it's more neutral; “return (+1) <*> [1..10]” looks a bit odd) using “return” would be less of a shock to the system.
This would mean moving Applicative into the Prelude.
Good, though I'm still very much in favour of moving as much out of the Prelude as possible.
I think "joining up the classes" is a good idea,
Definitely -- as is slicing them into finer layers (of which this is also an example). -- Jón Fairbairn Jon.Fairbairn at cl.cam.ac.uk