Repository : ssh://darcs.haskell.org//srv/darcs/ghc On branch : master
http://hackage.haskell.org/trac/ghc/changeset/459fb7bcb36bea9798486b74098e800b1d55139f >--------------------------------------------------------------- commit 459fb7bcb36bea9798486b74098e800b1d55139f Author: Simon Peyton Jones <[email protected]> Date: Wed Aug 3 16:15:29 2011 +0100 isCoVarType should look at the *representation* type, rather than using isPredTy! In Core land, a PredTy and its representation type are synonymous. >--------------------------------------------------------------- compiler/types/Coercion.lhs | 7 +++++-- compiler/types/TypeRep.lhs | 26 ++++---------------------- 2 files changed, 9 insertions(+), 24 deletions(-) diff --git a/compiler/types/Coercion.lhs b/compiler/types/Coercion.lhs index db7f96f..cf458c7 100644 --- a/compiler/types/Coercion.lhs +++ b/compiler/types/Coercion.lhs @@ -99,7 +99,7 @@ import Outputable import Unique import Pair import TysPrim ( eqPredPrimTyCon ) -import PrelNames ( funTyConKey ) +import PrelNames ( funTyConKey, eqPredPrimTyConKey ) import Control.Applicative import Data.Traversable (traverse, sequenceA) import Control.Arrow (second) @@ -279,7 +279,10 @@ isCoVar :: Var -> Bool isCoVar v = isCoVarType (varType v) isCoVarType :: Type -> Bool -isCoVarType = isEqPredTy +-- Don't rely on a PredTy; look at the representation type +isCoVarType ty + | Just tc <- tyConAppTyCon_maybe ty = tc `hasKey` eqPredPrimTyConKey + | otherwise = False \end{code} diff --git a/compiler/types/TypeRep.lhs b/compiler/types/TypeRep.lhs index e0a5670..c6568e2 100644 --- a/compiler/types/TypeRep.lhs +++ b/compiler/types/TypeRep.lhs @@ -155,11 +155,11 @@ data Type | PredTy PredType -- ^ The type of evidence for a type predictate. - -- Note that a @PredTy (EqPred _ _)@ can appear only as the kind - -- of a coercion variable; never as the argument or result of a - -- 'FunTy' (unlike the 'PredType' constructors 'ClassP' or 'IParam') + -- See Note [PredTy] + -- By the time we are in Core-land, PredTys are + -- synonymous with their representation + -- (see Type.predTypeRep) - -- See Note [PredTy], and Note [Equality predicates] deriving (Data.Data, Data.Typeable) -- | The key type representing kinds in the compiler. @@ -231,24 +231,6 @@ The predicate really does turn into a real extra argument to the function. If the argument has type (PredTy p) then the predicate p is represented by evidence (a dictionary, for example, of type (predRepTy p). -Note [Equality predicates] -~~~~~~~~~~~~~~~~~~~~~~~~~~ - forall a b. (a ~ S b) => a -> b -could be represented by - ForAllTy a (ForAllTy b (FunTy (PredTy (EqPred a (S b))) ...)) -OR - ForAllTy a (ForAllTy b (ForAllTy (c::PredTy (EqPred a (S b))) ...)) - -The latter is what we do. (Unlike for class and implicit parameter -constraints, which do use FunTy.) - -Reason: - * FunTy is always a *value* function - * ForAllTy is discarded at runtime - -We often need to make a "wildcard" (c::PredTy..). We always use the same -name (wildCoVarName), since it's not mentioned. - %************************************************************************ %* * _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
