#14728: Is (GeneralizedNewtypeDeriving + associated type classes) completely bogus? -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: (none) Type: bug | Status: closed Priority: normal | Milestone: 8.4.1 Component: Compiler (Type | Version: 8.2.2 checker) | Keywords: deriving, Resolution: fixed | TypeFamilies Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D4402 Wiki Page: | -------------------------------------+------------------------------------- Comment (by RyanGlScott): I completely take back what I said in comment:15 about this being unblocked—that couldn't be further from the truth! In fact, after talking with goldfire and kcsongor about this, we've come to the realization that all of the ideas in comment:3 are completely untenable at present. The issue is that we're trying to construct the type `(x |> g)`, where `x` is a type and `g` is a coercion. However, in order for this to kind-check, `g` must be nominally roled. This is never the case in GND, as we always use newtype axioms, which are by definition representationally roled! This pretty much makes this idea dead in the water, at least until we figure out a way to have representational casts in kinds (which is likely a ways away). -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14728#comment:16 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler