diff options
author | Ryan Scott <ryan.gl.scott@gmail.com> | 2020-06-23 17:50:15 -0400 |
---|---|---|
committer | Marge Bot <ben+marge-bot@smart-cactus.org> | 2020-06-28 09:19:46 -0400 |
commit | 42f797b0ad034a92389e7081aa50ef4ab3434d01 (patch) | |
tree | 8a198d5f6b2e51fe3578586c44445d76dbe4e96d /compiler/GHC/Tc/Gen | |
parent | d8ba9e6f951a2f8c6e2429a8b2dcb035c392908f (diff) | |
download | haskell-42f797b0ad034a92389e7081aa50ef4ab3434d01.tar.gz |
Use NHsCoreTy to embed types into GND-generated code
`GeneralizedNewtypeDeriving` is in the unique situation where it must
produce an `LHsType GhcPs` from a Core `Type`. Historically, this was
done with the `typeToLHsType` function, which walked over the entire
`Type` and attempted to construct an `LHsType` with the same overall
structure. `typeToLHsType` is quite complicated, however, and has
been the subject of numerous bugs over the years (e.g., #14579).
Luckily, there is an easier way to accomplish the same thing: the
`XHsType` constructor of `HsType`. `XHsType` bundles an `NHsCoreTy`,
which allows embedding a Core `Type` directly into an `HsType`,
avoiding the need to laboriously convert from one to another (as
`typeToLHsType` did). Moreover, renaming and typechecking an
`XHsType` is simple, since one doesn't need to do anything to a
Core `Type`...
...well, almost. For the reasons described in
`Note [Typechecking NHsCoreTys]` in `GHC.Tc.Gen.HsType`, we must
apply a substitution that we build from the local `tcl_env` type
environment. But that's a relatively modest price to pay.
Now that `GeneralizedNewtypeDeriving` uses `NHsCoreTy`, the
`typeToLHsType` function no longer has any uses in GHC, so this patch
rips it out. Some additional tweaks to `hsTypeNeedsParens` were
necessary to make the new `-ddump-deriv` output correctly
parenthesized, but other than that, this patch is quite
straightforward.
This is a mostly internal refactoring, although it is likely that
`GeneralizedNewtypeDeriving`-generated code will now need fewer
language extensions in certain situations than it did before.
Diffstat (limited to 'compiler/GHC/Tc/Gen')
-rw-r--r-- | compiler/GHC/Tc/Gen/HsType.hs | 56 |
1 files changed, 54 insertions, 2 deletions
diff --git a/compiler/GHC/Tc/Gen/HsType.hs b/compiler/GHC/Tc/Gen/HsType.hs index 68d29f565e..22edf3c0b4 100644 --- a/compiler/GHC/Tc/Gen/HsType.hs +++ b/compiler/GHC/Tc/Gen/HsType.hs @@ -90,6 +90,7 @@ import GHC.Tc.Utils.TcType import GHC.Tc.Utils.Instantiate ( tcInstInvisibleTyBinders, tcInstInvisibleTyBinder ) import GHC.Core.Type import GHC.Builtin.Types.Prim +import GHC.Types.Name.Env import GHC.Types.Name.Reader( lookupLocalRdrOcc ) import GHC.Types.Var import GHC.Types.Var.Set @@ -106,6 +107,7 @@ import GHC.Types.SrcLoc import GHC.Settings.Constants ( mAX_CTUPLE_SIZE ) import GHC.Utils.Error( MsgDoc ) import GHC.Types.Unique +import GHC.Types.Unique.FM import GHC.Types.Unique.Set import GHC.Utils.Misc import GHC.Types.Unique.Supply @@ -833,8 +835,17 @@ tc_infer_hs_type mode (HsSpliceTy _ (HsSpliced _ _ (HsSplicedTy ty))) = tc_infer_hs_type mode ty tc_infer_hs_type mode (HsDocTy _ ty _) = tc_infer_lhs_type mode ty -tc_infer_hs_type _ (XHsType (NHsCoreTy ty)) - = return (ty, tcTypeKind ty) + +-- See Note [Typechecking NHsCoreTys] +tc_infer_hs_type _ (XHsType (NHsCoreTy ty)) + = do env <- getLclEnv + let subst_prs = [ (nm, tv) + | ATyVar nm tv <- nameEnvElts (tcl_env env) ] + subst = mkTvSubst + (mkInScopeSet $ mkVarSet $ map snd subst_prs) + (listToUFM $ map (liftSnd mkTyVarTy) subst_prs) + ty' = substTy subst ty + return (ty', tcTypeKind ty') tc_infer_hs_type _ (HsExplicitListTy _ _ tys) | null tys -- this is so that we can use visible kind application with '[] @@ -847,6 +858,47 @@ tc_infer_hs_type mode other_ty ; ty' <- tc_hs_type mode other_ty kv ; return (ty', kv) } +{- +Note [Typechecking NHsCoreTys] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +NHsCoreTy is an escape hatch that allows embedding Core Types in HsTypes. +As such, there's not much to be done in order to typecheck an NHsCoreTy, +since it's already been typechecked to some extent. There is one thing that +we must do, however: we must substitute the type variables from the tcl_env. +To see why, consider GeneralizedNewtypeDeriving, which is one of the main +clients of NHsCoreTy (example adapted from #14579): + + newtype T a = MkT a deriving newtype Eq + +This will produce an InstInfo GhcPs that looks roughly like this: + + instance forall a_1. Eq a_1 => Eq (T a_1) where + (==) = coerce @( a_1 -> a_1 -> Bool) -- The type within @(...) is an NHsCoreTy + @(T a_1 -> T a_1 -> Bool) -- So is this + (==) + +This is then fed into the renamer. Since all of the type variables in this +InstInfo use Exact RdrNames, the resulting InstInfo GhcRn looks basically +identical. Things get more interesting when the InstInfo is fed into the +typechecker, however. GHC will first generate fresh skolems to instantiate +the instance-bound type variables with. In the example above, we might generate +the skolem a_2 and use that to instantiate a_1, which extends the local type +environment (tcl_env) with [a_1 :-> a_2]. This gives us: + + instance forall a_2. Eq a_2 => Eq (T a_2) where ... + +To ensure that the body of this instance is well scoped, every occurrence of +the `a` type variable should refer to a_2, the new skolem. However, the +NHsCoreTys mention a_1, not a_2. Luckily, the tcl_env provides exactly the +substitution we need ([a_1 :-> a_2]) to fix up the scoping. We apply this +substitution to each NHsCoreTy and all is well: + + instance forall a_2. Eq a_2 => Eq (T a_2) where + (==) = coerce @( a_2 -> a_2 -> Bool) + @(T a_2 -> T a_2 -> Bool) + (==) +-} + ------------------------------------------ tcLHsType :: LHsType GhcRn -> TcKind -> TcM TcType tcLHsType hs_ty exp_kind |