[GHC] #7123: Dictionary method should be applicable even in existential context

[GHC]

#7123: Dictionary method should be applicable even in existential context
-------------------------------+--------------------------------------------
  Reporter:  heisenbug         |          Owner:                  
      Type:  bug               |         Status:  closed          
  Priority:  normal            |      Milestone:                  
 Component:  Compiler          |        Version:  7.4.1           
Resolution:  invalid           |       Keywords:                  
        Os:  Unknown/Multiple  |   Architecture:  Unknown/Multiple
   Failure:  None/Unknown      |     Difficulty:  Unknown         
  Testcase:                    |      Blockedby:                  
  Blocking:                    |        Related:                  
-------------------------------+--------------------------------------------
Changes (by simonpj):

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


Comment:

 parcs is right.  See
 http://www.haskell.org/haskellwiki/Type_families#Injectivity.2C_type_inference.2C_and_ambiguity.

 It's actually impossible to call `baz`, ever!  Look at its type:
 {{{
 baz :: Foo a => Bar a -> Int
 }}}
 Suppose we have a call `(baz x)` where `x::(Int,Int)`.  How should we
 intantiate `a`?  Maybe we can instantiate `a` with `Char`, because `Bar
 Char = (Int,Int)` which matches x's type.  But suppose we also had
 {{{
 instance Foo Bool where
   type Bar Bool = (Int,Int)
   bar (i,_) = i
 }}}
 Now we could also instantiate `a` to `Bool`!

 You can fix this by using a `data` family instead; or (in the future) by
 specifing that `Bar` is injective.

 Simon

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/7123#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler