[Haskell-cafe] Type families are awesome

[John Ky]

Hi Haskell Cafe,

I'm finding that I really like type families.  For instance, the
GHC.List.lookup and Data.Map.lookup functions annoy me because their names
clash, yet their type are so similar.  With type families, I could define a
more generic lookup function like this:

import Data.Map as MAP
import GHC.List as LIST

class MapType ma where
   type Key ma
   type Item ma
   lookup :: Key ma -> ma -> Maybe (Item ma)

instance (Ord ka) => MapType (MAP.Map ka a) where
   type Key (MAP.Map ka a) = ka
   type Item (MAP.Map ka a) = a
   lookup ka ma = MAP.lookup ka ma

instance (Eq ka) => MapType [(ka, a)] where
   type Key [(ka, a)] = ka
   type Item [(ka, a)] = a
   lookup ka ma = LIST.lookup ka ma

This lookup function works on both "Map ka a" and "[(ka, a)]" types and I no
longer need to qualify my lookup function with the module name.

The downside I suppose is that lookup is no longer a function that can be
manipulated freely:

*Main> let x = lookup
*Main> let y = Fx.Data.Map.lookup

<interactive>:1:8:
    Ambiguous type variable `ma' in the constraint:
      `Fx.Data.Map.MapType ma'
        arising from a use of `Fx.Data.Map.lookup' at <interactive>:1:8-25
    Probable fix: add a type signature that fixes these type variable(s)

A shame that.  I had been hoping it would be possible to have a generic
lookup function that could be used in every way the current collection of
various lookup functions can be used.

So much nicer if 'y' above could be bound to the Fx.Data.Map.lookup with the
same type:

*Main> :t Fx.Data.Map.lookup
Fx.Data.Map.lookup :: forall ma.
                      (Fx.Data.Map.MapType ma) =>
                      Fx.Data.Map.Key ma -> ma -> Maybe (Fx.Data.Map.Item
ma)

And then have the ambiguity resolve later when 'y' is actually used.

-John
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.haskell.org/pipermail/haskell-cafe/attachments/20090121/18720ac2/attachment.htm