Missing dataCast1 and dataCast2 methods in Data.Data instances

[dreixel]

Having a look at Data instances in module Data.Data, Ptr a and ForeignPtr a are missing a dataCast1 = gcast1 line. And Array a b seems to be missing the dataCast2 method.

[simonpj]

Any chance of a patch? Sounds as if you know just what to do.

[dreixel]

Yes, I'll take care of this, it's trivial.

[dreixel]

Here's the fix:

diff --git a/Data/Data.hs b/Data/Data.hs
index 0a29668..cb2d894 100644
--- a/Data/Data.hs
+++ b/Data/Data.hs
@@ -1275,27 +1275,27 @@ instance (Data a, Data b, Data c, Data d, Data e, Data f
 
 ------------------------------------------------------------------------------
 
-instance Typeable a => Data (Ptr a) where
+instance (Data a, Typeable a) => Data (Ptr a) where
   toConstr _   = error "Data.Data.toConstr(Ptr)"
   gunfold _ _  = error "Data.Data.gunfold(Ptr)"
   dataTypeOf _ = mkNoRepType "GHC.Ptr.Ptr"
-
+  dataCast1 x  = gcast1 x
 
 ------------------------------------------------------------------------------
 
-instance Typeable a => Data (ForeignPtr a) where
+instance (Data a, Typeable a) => Data (ForeignPtr a) where
   toConstr _   = error "Data.Data.toConstr(ForeignPtr)"
   gunfold _ _  = error "Data.Data.gunfold(ForeignPtr)"
   dataTypeOf _ = mkNoRepType "GHC.ForeignPtr.ForeignPtr"
-
+  dataCast1 x  = gcast1 x
 
 ------------------------------------------------------------------------------
 -- The Data instance for Array preserves data abstraction at the cost of 
 -- inefficiency. We omit reflection services for the sake of data abstraction.
-instance (Typeable a, Data b, Ix a) => Data (Array a b)
+instance (Typeable a, Data a, Data b, Ix a) => Data (Array a b)
  where
   gfoldl f z a = z (listArray (bounds a)) `f` (elems a)
   toConstr _   = error "Data.Data.toConstr(Array)"
   gunfold _ _  = error "Data.Data.gunfold(Array)"
   dataTypeOf _ = mkNoRepType "Data.Array.Array"
-
+  dataCast2 x  = gcast2 x

Note however that this requires adding additional Data constraints to these instances; perhaps this is the reason why these methods were omitted in the first place. So fixing this might break client code that uses these instances. Nonetheless, they are required for the correct behavior of SYB's ext/1/2 functions.

Shall I go ahead and push the patch?

[igloo]

Replying to dreixel:

Shall I go ahead and push the patch?

Probably worth asking on libraries@, to give people the chance to object.

[igloo]

Removing patch tag until libraries@ has given an opinion.

[dreixel]

There was no opinion:  http://permalink.gmane.org/gmane.comp.lang.haskell.libraries/18099

I was going to proceed and patch.

[igloo]

Ah, great. I've applied it. Thanks!