#15020: PatternSynonyms: Problems with quantified constraints / foralls -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: 8.6.1 Component: Compiler | Version: 8.5 Resolution: | Keywords: | PatternSynonyms 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): Interesting. Would you expect the following to work? {{{ data Lam2 a = Lam2 (forall mm. Monad mm => mm a) | Other newtype Forall f = Forall (forall xx. f xx) mLam2 :: Forall Lam2 -> r -> ((forall xx mm. Monad mm => mm xx) -> r) -> r mLam2 (Forall (Lam2 lam)) d k = k lam mLam2 (Forall Other) d k = d }}} Here `mLam2` is the sort of matching function that GHC builds for `PLam2`. And indeed `mLam2` is rejected because `lam` is not polymorphic enough. Is that so unreasonable? (I checked why `mkLam3` works; it's because teh pattern matching is pushed into the argument of the continuation, something like this {{{ mLam3 :: Forall Lam2 -> r -> ((forall xx mm. Monad mm => mm xx) -> r) -> r mLam3 x d k = k (case x of { Forall (Lam2 lam) -> lam }) }}} But this won't work when we want to bind multiple arguments, or when the match can fail.) I don't see an easy way to improve this. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/15020#comment:2 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler