[GHC] #7485: Tuple constraints not properly kinded

[GHC]

#7485: Tuple constraints not properly kinded
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                   Owner:
            Type:  bug               |                  Status:  closed
        Priority:  normal            |               Milestone:
       Component:  Compiler          |                 Version:  7.7
      Resolution:  invalid           |                Keywords:
Operating System:  Unknown/Multiple  |  ConstraintKinds
 Type of failure:  None/Unknown      |            Architecture:
      Blocked By:                    |  Unknown/Multiple
 Related Tickets:                    |               Test Case:
                                     |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by goldfire):

 Replying to [comment:3 crockeea]:
 > Is @simonpj 's comment
 > > `(,)` has to have a particular kind
 > still true with advent of `-XPolyKinds`?

 Yes. The problem is that `(,)` isn't really poly-kinded. It's not, for
 example, `k -> k -> k`. It's either `* -> * -> *` or `Constraint ->
 Constraint -> Constraint`, and there's no way of saying this in normal
 Haskell. So, we have the hack Simon describes above.

 One possible way forward is to follow the NoSubKinds route: introduce a
 new type `ValueKind` parameterized by elements of `data StarOrConstraint =
 Star | Ct`. Then `*` would be a synonym for `ValueKind Star` and
 `Constraint` would be `ValueKind Ct`. We would then have `(,) :: forall (x
 :: StarOrConstraint). ValueKind x -> ValueKind x -> ValueKind x`. This
 would be very hard to implement before merging my DependentHaskell branch
 though, because it requires kind synonyms and other funny business.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/7485#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler