[GHC] #14869: Documentation for isLiftedTypeKind is incorrect

[GHC]

#14869: Documentation for isLiftedTypeKind is incorrect
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler (Type    |              Version:  8.2.2
  checker)                           |
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Documentation     |  Unknown/Multiple
  bug                                |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 Amusingly, the `Constraint` counterpart, `isLiftedTypeKind`, actually
 appears to do
 [http://git.haskell.org/ghc.git/blob/df2c3b3364834d2fd038192c89348fc50a2e0475:/compiler/types/Kind.hs#l68
 the right thing]:

 {{{#!hs
 isConstraintKind :: Kind -> Bool
 isConstraintKindCon :: TyCon -> Bool

 isConstraintKindCon   tc = tyConUnique tc == constraintKindTyConKey

 isConstraintKind (TyConApp tc _) = isConstraintKindCon tc
 isConstraintKind _               = False
 }}}

 This actually checks the unique of the tycon, and per `Note [Kind
 Constraint and kind *]`, the uniques for `Type` and `Constraint`
 [http://git.haskell.org/ghc.git/blob/df2c3b3364834d2fd038192c89348fc50a2e0475:/compiler/types/Kind.hs#l43
 are different]. Thus:

 {{{
 $ inplace/bin/ghc-stage2 --interactive -package ghc
 GHCi, version 8.5.20180221: http://www.haskell.org/ghc/  :? for help
 Loaded GHCi configuration from /home/rgscott/.ghci
 λ> :m + Kind TysWiredIn
 λ> isConstraintKind liftedTypeKind
 False
 λ> isConstraintKind constraintKind
 True
 }}}

 Thus, to fix the aforementioned Template Haskell bug, all we'd have to do
 is make this change:

 {{{#!diff
 diff --git a/compiler/typecheck/TcSplice.hs
 b/compiler/typecheck/TcSplice.hs
 index 45e18e6..b052e8e 100644
 --- a/compiler/typecheck/TcSplice.hs
 +++ b/compiler/typecheck/TcSplice.hs
 @@ -1707,8 +1707,8 @@ reifyFamilyInstance is_poly_tvs inst@(FamInst {
 fi_flavor = flavor
  ------------------------------
  reifyType :: TyCoRep.Type -> TcM TH.Type
  -- Monadic only because of failure
 -reifyType ty                | isLiftedTypeKind ty = return TH.StarT
 -                            | isConstraintKind ty = return TH.ConstraintT
 +reifyType ty                | isConstraintKind ty = return TH.ConstraintT
 +                            | isLiftedTypeKind ty = return TH.StarT
  reifyType ty@(ForAllTy {})  = reify_for_all ty
  reifyType (LitTy t)         = do { r <- reifyTyLit t; return (TH.LitT r)
 }
  reifyType (TyVarTy tv)      = return (TH.VarT (reifyName tv))
 }}}

 -----

 I propose that we:

 1. Change the documentation for `isLiftedTypeKind` to reflect the fact
 that it does not distinguish between `Constraint` and `Type`.
 2. Make the above change to `TcSplice` to fix the bug.

 Does that sound reasonable?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14869#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler