#1965: Allow unconstrained existential contexts in newtypes -------------------------------------+------------------------------------- Reporter: guest | Owner: (none) Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 6.8.1 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 Ericson2314): Yes for @sgraf's reason I wish we used an equivalence relation on these internal notion of deep representations, and also {{{RuntimeRep}}}s. For example, given matching deep structure (not tracked in the kind/{{{RuntimeRep}}}): {{{ forall a t. exists t' t'. (t :: Kind a) ~R (t' :: Kind TupleRep [a]) ~R (t'' :: Kind SumRep [a]) }}} Maybe it's not worth the extra coercions in practice, but certainly I think its good teaching to pretend newtypes are always: {{{ newtype Foo (a :: k) :: TupleRep [k] = Foo (a :: k) }}} to explain {{{ newtype A :: TupleRep [LiftedRep] = A (Int :: LiftedRep) case undefined of A _ -> 1 }}} contrasted with pretend elaborated unpacked {{{!Int}}}: {{{ data A' :: LiftedRep = A' (!Int :: IntRep) case undefined of A' _ -> undefined }}} That the shallow representations of {{{A}}} and {{{A'}}} have different strictness, opposite what is contained inside, cleanly explains how the `case`-expression has different semantics despite the overall isomorphism of {{{A}}} and {{{A'}}}. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/1965#comment:39 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler