[GHC] #4083: Generalized newtype deriving fails for newtypes that are type family instances

[GHC]

#4083: Generalized newtype deriving fails for newtypes that are type family
instances
--------------------------------------+-------------------------------------
  Reporter:  RANDY Ruler of Zexernet  |          Owner:                  
      Type:  bug                      |         Status:  closed          
  Priority:  normal                   |      Milestone:                  
 Component:  Compiler                 |        Version:  6.12.1          
Resolution:  invalid                  |       Keywords:                  
Difficulty:                           |             Os:  Unknown/Multiple
  Testcase:                           |   Architecture:  Unknown/Multiple
   Failure:  None/Unknown             |  
--------------------------------------+-------------------------------------
Changes (by simonpj):

  * status:  new => closed
  * resolution:  => invalid


Comment:

 Thanks.  It's definitely a bug that you get an 'impossible' error,
 fortunately fixed in 6.12.3:
 {{{
 T4083.hs:6:1:
     Can't make a derived instance of `C (F [a])'
       (even with cunning newtype deriving):
       the class has associated types
     In the stand-alone deriving instance for
       `(C (Maybe a)) => C (F [a])'
 }}}
 And indeed, generalised newtype deriving doesn't support classes with
 associated types.  Remember it's just a mechanism for saving you typing.
 So suppose you had
 {{{
 class C a where
   data T a
   op :: a -> T a

 instance C Int where
   data T Int = TI Bool
   op n = TI (n>0)

 newtype Age = MkAge Int deriving( C )
 }}}
 What instance declaration would GHC generate?  Something like
 {{{
 instance C Age where
   data T Age = ???
   op (Age n) = ???
 }}}
 The trouble is that each instance generates fresh data constructors.

 Anyway at the moment it's just not supported, and you now get a civilised
 message to that effect.

 I'll close as  'invalid', but I really mean 'not supported'.  You could
 open a feature request, but I'm quite unsure what the design would look
 like.  (It might be just possible for associated type *synonyms* but I'm
 not certain.)

 Simon

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/4083#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler