[GHC] #13731: DeriveFunctor and friends don't understand type families

[GHC]

#13731: DeriveFunctor and friends don't understand type families
-------------------------------------+-------------------------------------
        Reporter:  spacekitteh       |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.1-rc2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Consider the instance we want:
 {{{
 instance Functor (Test ListExtension) where
   fmap f (Extend x) = Extend (fmap f x)
 }}}
 > Am I missing something?

 From the inner `fmap` we get `[W] Functor (ExtensionType ListExtension)`.
 If `ExtensionType` has arity 2, that would be an un-saturated type family.
 But if it has arity 1 it is saturated, and reduces to `[]`, so all is
 well.

 > ... the field of `Extend` will have type `ExtensionType  ListExtension
 a`. This is a fully saturated type family application

 Yes, it's saturated in the field, but the use of `fmap` requires us to
 decompose the type application, we it must be decomposable.  It's only
 decomposable if `ExtensionType` has arity 1.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13731#comment:7>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler