Commit e1ff2b49 authored by Simon Peyton Jones's avatar Simon Peyton Jones

Fix partial sigs and pattern bindings interaction

It turns out that GHC 8.0 would accept entirely bogus
programs like

   f2 :: (True, _) -> Char
   Just f2 = Just (\x->x)

(which is now partial-sigs/should_fail/PatBind3)

This also fixes Trac #9478, test `PatBind2`.
parent cc75a5d9
......@@ -158,29 +158,43 @@ lookupPragEnv prag_fn n = lookupNameEnv prag_fn n `orElse` []
* *
********************************************************************* -}
tcPatBndr :: PatEnv -> Name -> ExpSigmaType -> TcM (TcCoercionN, TcId)
tcPatBndr :: PatEnv -> Name -> ExpSigmaType -> TcM (HsWrapper, TcId)
-- (coi, xp) = tcPatBndr penv x pat_ty
-- Then coi : pat_ty ~ typeof(xp)
tcPatBndr (PE { pe_ctxt = LetPat lookup_sig no_gen}) bndr_name pat_ty
tcPatBndr (PE { pe_ctxt = LetPat lookup_sig no_gen
, pe_orig = orig }) bndr_name pat_ty
-- See Note [Typing patterns in pattern bindings]
| LetGblBndr prags <- no_gen
, Just (TcIdSig sig) <- lookup_sig bndr_name
, Just (TcIdSig sig) <- mb_sig
, Just poly_id <- completeIdSigPolyId_maybe sig
= do { bndr_id <- addInlinePrags poly_id (lookupPragEnv prags bndr_name)
; traceTc "tcPatBndr(gbl,sig)" (ppr bndr_id $$ ppr (idType bndr_id))
; co <- unifyPatType bndr_id (idType bndr_id) pat_ty
; return (co, bndr_id) }
; return (mkWpCastN co, bndr_id) }
-- See Note [Partial signatures for pattern bindings]
| LetLclBndr <- no_gen
, Just (TcIdSig sig) <- mb_sig
= do { mono_name <- newLocalName bndr_name
; (subst, _) <- newMetaSigTyVars (map snd (sig_skols sig))
; let tau = substTy subst (sig_tau sig)
mono_id = mkLocalId mono_name tau
; wrap <- tcSubTypeET orig pat_ty tau
; traceTc "tcPatBndr(lsl,sig)" (ppr mono_id $$ ppr tau $$ ppr pat_ty)
; return (wrap, mono_id) }
| otherwise
= do { pat_ty <- expTypeToType pat_ty
; bndr_id <- newNoSigLetBndr no_gen bndr_name pat_ty
; traceTc "tcPatBndr(no-sig)" (ppr bndr_id $$ ppr (idType bndr_id))
; return (mkTcNomReflCo pat_ty, bndr_id) }
; return (idHsWrapper, bndr_id) }
mb_sig = lookup_sig bndr_name
tcPatBndr (PE { pe_ctxt = _lam_or_proc }) bndr_name pat_ty
= do { pat_ty <- expTypeToType pat_ty
; return (mkTcNomReflCo pat_ty, mkLocalId bndr_name pat_ty) }
; return (idHsWrapper, mkLocalId bndr_name pat_ty) }
-- whether or not there is a sig is irrelevant, as this
-- is local
......@@ -226,7 +240,63 @@ addInlinePrags poly_id prags
pp_inl (L loc prag) = ppr prag <+> parens (ppr loc)
{- Note [Partial signatures for pattern bindings]
Consider a function binding and a pattern binding, both
with a partial type signature
f1 :: (True, _) -> Char
f1 = \x -> x
f2 :: (True, _) -> Char
Just f2 = Just (\x->x)
Obviously, both should be rejected. That happens naturally for the
function binding, f1, because we typecheck the RHS with "expected"
type '(True, apha) -> Char', which correctly fails.
But what of the pattern binding for f2? We infer the type of the
pattern, and check tha the RHS has that type. So we must feed in the
type of f2 when inferring the type of the pattern! We do this right
here, in tcPatBndr, for a LetLclBndr. The signature already has fresh
unification variables for the wildcards (if any).
Extra notes
* For /complete/ type signatures, we could im principle ignore all this
and just infer the most general type for f2, and check (in
TcBinds.mkExport) whether it has the claimed type.
But not so for /partial/ signatures; to get the wildcard unification
variables into the game we really must inject them here. If we don't
we never get /any/ value assigned to the wildcards; and programs that
are bogus, like f2, are accepted.
Moreover, by feeding in the expected type we do less fruitless
creation of unification variables, and improve error messages.
* We need to take care with the skolems. Consider
data T a = MkT a a
f :: forall a. a->a
g :: forall b. b->b
MkT f g = MkT (\x->x) (\y->y)
Here we'll infer a type from the pattern of 'T a', but if we feed in
the signature types for f and g, we'll end up unifying 'a' and 'b'.
So we instantiate the skolems with SigTvs; hence newMetaSigTyVars.
All we are doing here is getting the "shapes" right. In tcExport
we'll check that the Id really does have the claimed type, with
the claimed polymorphism.
* We need to do a subsumption, not equality, check. If
data T = MkT (forall a. a->a)
f :: forall b. [b]->[b]
MkT f = blah
Since 'blah' returns a value of type T, its payload is a polymorphic
function of type (forall a. a->a). And that's enough to bind the
less-polymorphic function 'f', but we need some impedence matching
to witness the instantiation.
Note [Typing patterns in pattern bindings]
Suppose we are typing a pattern binding
......@@ -331,10 +401,10 @@ tc_pat :: PatEnv
a) -- Result of thing inside
tc_pat penv (VarPat (L l name)) pat_ty thing_inside
= do { (co, id) <- tcPatBndr penv name pat_ty
= do { (wrap, id) <- tcPatBndr penv name pat_ty
; res <- tcExtendIdEnv1 name id thing_inside
; pat_ty <- readExpType pat_ty
; return (mkHsWrapPatCo co (VarPat (L l id)) pat_ty, res) }
; return (mkHsWrapPat wrap (VarPat (L l id)) pat_ty, res) }
tc_pat penv (ParPat pat) pat_ty thing_inside
= do { (pat', res) <- tc_lpat pat pat_ty penv thing_inside
......@@ -370,7 +440,7 @@ tc_pat _ (WildPat _) pat_ty thing_inside
; return (WildPat pat_ty, res) }
tc_pat penv (AsPat (L nm_loc name) pat) pat_ty thing_inside
= do { (co, bndr_id) <- setSrcSpan nm_loc (tcPatBndr penv name pat_ty)
= do { (wrap, bndr_id) <- setSrcSpan nm_loc (tcPatBndr penv name pat_ty)
; (pat', res) <- tcExtendIdEnv1 name bndr_id $
tc_lpat pat (mkCheckExpType $ idType bndr_id)
penv thing_inside
......@@ -382,7 +452,7 @@ tc_pat penv (AsPat (L nm_loc name) pat) pat_ty thing_inside
-- If you fix it, don't forget the bindInstsOfPatIds!
; pat_ty <- readExpType pat_ty
; return (mkHsWrapPatCo co (AsPat (L nm_loc bndr_id) pat') pat_ty, res) }
; return (mkHsWrapPat wrap (AsPat (L nm_loc bndr_id) pat') pat_ty, res) }
tc_pat penv (ViewPat expr pat _) overall_pat_ty thing_inside
= do {
......@@ -579,12 +649,12 @@ tc_pat penv (NPlusKPat (L nm_loc name) (L loc lit) _ ge minus _) pat_ty thing_in
<- tcSyntaxOpGen orig minus [synKnownType pat_ty, SynRho] SynAny $
\ [lit2_ty, var_ty] ->
do { lit2' <- newOverloadedLit lit (mkCheckExpType lit2_ty)
; (co, bndr_id) <- setSrcSpan nm_loc $
; (wrap, bndr_id) <- setSrcSpan nm_loc $
tcPatBndr penv name (mkCheckExpType var_ty)
-- co :: var_ty ~ idType bndr_id
-- minus_wrap is applicable to minus'
; return (lit2', mkWpCastN co, bndr_id) }
; return (lit2', wrap, bndr_id) }
-- The Report says that n+k patterns must be in Integral
-- but it's silly to insist on this in the RebindableSyntax case
......@@ -36,7 +36,7 @@ test('NamedWildcardsAsTyVars', normal, compile, [''])
test('ParensAroundContext', normal, compile, ['-ddump-types -fno-warn-partial-type-signatures'])
test('PatBind', normal, compile, ['-ddump-types -fno-warn-partial-type-signatures'])
# Bug
test('PatBind2', expect_broken(9478), compile, ['-ddump-types -fno-warn-partial-type-signatures'])
test('PatBind2', normal, compile, ['-ddump-types -fno-warn-partial-type-signatures'])
test('PatternSig', normal, compile, ['-ddump-types -fno-warn-partial-type-signatures'])
test('Recursive', normal, compile, ['-ddump-types -fno-warn-partial-type-signatures'])
test('ScopedNamedWildcards', normal, compile, ['-ddump-types -fno-warn-partial-type-signatures'])
{-# LANGUAGE PartialTypeSignatures #-}
module PatBind3 where
-- Oddly GHC 8.0 accepted this, but it should obvoiusly fail!
foo :: (Bool, _) -> Char
Just foo = Just id
PatBind3.hs:6:12: error:
• Couldn't match type ‘(Bool, t)’ with ‘Char’
Expected type: Maybe ((Bool, t) -> Char)
Actual type: Maybe ((Bool, t) -> (Bool, t))
• In the expression: Just id
In a pattern binding: Just foo = Just id
• Relevant bindings include
foo :: (Bool, t) -> Char (bound at PatBind3.hs:6:6)
......@@ -60,3 +60,4 @@ test('T10045', normal, compile_fail, [''])
test('T10999', normal, compile_fail, [''])
test('T11122', normal, compile, [''])
test('T11976', normal, compile_fail, [''])
test('PatBind3', normal, compile_fail, [''])
Markdown is supported
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment