#11028: Refactor ConDecl -------------------------------------+------------------------------------- Reporter: simonpj | Owner: alanz Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 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 alanz): The key factor is the work currently done in `rnConResult`, {{{#!hs rnConResult doc _con details (ResTyGADT ls ty) = do { (ty', fvs) <- rnLHsType doc ty ; let (arg_tys, res_ty) = splitHsFunType ty' -- We can finally split it up, -- now the renamer has dealt with fixities -- See Note [Sorting out the result type] in RdrHsSyn ; case details of InfixCon {} -> pprPanic "rnConResult" (ppr ty) -- See Note [Sorting out the result type] in RdrHsSyn RecCon {} -> do { unless (null arg_tys) (addErr (badRecResTy doc)) ; return (details, ResTyGADT ls res_ty, fvs) } PrefixCon {} -> return (PrefixCon arg_tys, ResTyGADT ls res_ty, fvs)} }}} `splitHsFunTy` on `t -> MaybeO a b` will return `([t],MaybeO a b)`, and the latter is then returned as the new `res_ty` My understanding is that we are doing this so that the ParsedSource more accurately reflects what was seen, which can simplify the API Annotations work. Perhaps we should add a `PostRn` field to the `ConDeclGADT` to capture the new `details` and `res_ty`. Alternatively, everything except the `rnLHsType` in `rnConResult` can move to the type checking. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/11028#comment:24 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler