2 % (c) The University of Glasgow 2006
3 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
10 tcSimplifyInfer, tcSimplifyInferCheck,
11 tcSimplifyCheck, tcSimplifyRestricted,
12 tcSimplifyRuleLhs, tcSimplifyIPs,
13 tcSimplifySuperClasses,
14 tcSimplifyTop, tcSimplifyInteractive,
15 tcSimplifyBracket, tcSimplifyCheckPat,
17 tcSimplifyDeriv, tcSimplifyDefault,
21 #include "HsVersions.h"
23 import {-# SOURCE #-} TcUnify( unifyType )
60 %************************************************************************
64 %************************************************************************
66 --------------------------------------
67 Notes on functional dependencies (a bug)
68 --------------------------------------
75 instance D a b => C a b -- Undecidable
76 -- (Not sure if it's crucial to this eg)
77 f :: C a b => a -> Bool
80 g :: C a b => a -> Bool
83 Here f typechecks, but g does not!! Reason: before doing improvement,
84 we reduce the (C a b1) constraint from the call of f to (D a b1).
86 Here is a more complicated example:
88 | > class Foo a b | a->b
90 | > class Bar a b | a->b
94 | > instance Bar Obj Obj
96 | > instance (Bar a b) => Foo a b
98 | > foo:: (Foo a b) => a -> String
101 | > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
107 | Could not deduce (Bar a b) from the context (Foo a b)
108 | arising from use of `foo' at <interactive>:1
110 | Add (Bar a b) to the expected type of an expression
111 | In the first argument of `runFoo', namely `foo'
112 | In the definition of `it': it = runFoo foo
114 | Why all of the sudden does GHC need the constraint Bar a b? The
115 | function foo didn't ask for that...
117 The trouble is that to type (runFoo foo), GHC has to solve the problem:
119 Given constraint Foo a b
120 Solve constraint Foo a b'
122 Notice that b and b' aren't the same. To solve this, just do
123 improvement and then they are the same. But GHC currently does
128 That is usually fine, but it isn't here, because it sees that Foo a b is
129 not the same as Foo a b', and so instead applies the instance decl for
130 instance Bar a b => Foo a b. And that's where the Bar constraint comes
133 The Right Thing is to improve whenever the constraint set changes at
134 all. Not hard in principle, but it'll take a bit of fiddling to do.
138 --------------------------------------
139 Notes on quantification
140 --------------------------------------
142 Suppose we are about to do a generalisation step.
146 T the type of the RHS
147 C the constraints from that RHS
149 The game is to figure out
151 Q the set of type variables over which to quantify
152 Ct the constraints we will *not* quantify over
153 Cq the constraints we will quantify over
155 So we're going to infer the type
159 and float the constraints Ct further outwards.
161 Here are the things that *must* be true:
163 (A) Q intersect fv(G) = EMPTY limits how big Q can be
164 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
166 (A) says we can't quantify over a variable that's free in the
167 environment. (B) says we must quantify over all the truly free
168 variables in T, else we won't get a sufficiently general type. We do
169 not *need* to quantify over any variable that is fixed by the free
170 vars of the environment G.
172 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
174 Example: class H x y | x->y where ...
176 fv(G) = {a} C = {H a b, H c d}
179 (A) Q intersect {a} is empty
180 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
182 So Q can be {c,d}, {b,c,d}
184 Other things being equal, however, we'd like to quantify over as few
185 variables as possible: smaller types, fewer type applications, more
186 constraints can get into Ct instead of Cq.
189 -----------------------------------------
192 fv(T) the free type vars of T
194 oclose(vs,C) The result of extending the set of tyvars vs
195 using the functional dependencies from C
197 grow(vs,C) The result of extend the set of tyvars vs
198 using all conceivable links from C.
200 E.g. vs = {a}, C = {H [a] b, K (b,Int) c, Eq e}
201 Then grow(vs,C) = {a,b,c}
203 Note that grow(vs,C) `superset` grow(vs,simplify(C))
204 That is, simplfication can only shrink the result of grow.
207 oclose is conservative one way: v `elem` oclose(vs,C) => v is definitely fixed by vs
208 grow is conservative the other way: if v might be fixed by vs => v `elem` grow(vs,C)
211 -----------------------------------------
215 Here's a good way to choose Q:
217 Q = grow( fv(T), C ) \ oclose( fv(G), C )
219 That is, quantify over all variable that that MIGHT be fixed by the
220 call site (which influences T), but which aren't DEFINITELY fixed by
221 G. This choice definitely quantifies over enough type variables,
222 albeit perhaps too many.
224 Why grow( fv(T), C ) rather than fv(T)? Consider
226 class H x y | x->y where ...
231 If we used fv(T) = {c} we'd get the type
233 forall c. H c d => c -> b
235 And then if the fn was called at several different c's, each of
236 which fixed d differently, we'd get a unification error, because
237 d isn't quantified. Solution: quantify d. So we must quantify
238 everything that might be influenced by c.
240 Why not oclose( fv(T), C )? Because we might not be able to see
241 all the functional dependencies yet:
243 class H x y | x->y where ...
244 instance H x y => Eq (T x y) where ...
249 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
250 apparent yet, and that's wrong. We must really quantify over d too.
253 There really isn't any point in quantifying over any more than
254 grow( fv(T), C ), because the call sites can't possibly influence
255 any other type variables.
259 -------------------------------------
261 -------------------------------------
263 It's very hard to be certain when a type is ambiguous. Consider
267 instance H x y => K (x,y)
269 Is this type ambiguous?
270 forall a b. (K (a,b), Eq b) => a -> a
272 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
273 now we see that a fixes b. So we can't tell about ambiguity for sure
274 without doing a full simplification. And even that isn't possible if
275 the context has some free vars that may get unified. Urgle!
277 Here's another example: is this ambiguous?
278 forall a b. Eq (T b) => a -> a
279 Not if there's an insance decl (with no context)
280 instance Eq (T b) where ...
282 You may say of this example that we should use the instance decl right
283 away, but you can't always do that:
285 class J a b where ...
286 instance J Int b where ...
288 f :: forall a b. J a b => a -> a
290 (Notice: no functional dependency in J's class decl.)
291 Here f's type is perfectly fine, provided f is only called at Int.
292 It's premature to complain when meeting f's signature, or even
293 when inferring a type for f.
297 However, we don't *need* to report ambiguity right away. It'll always
298 show up at the call site.... and eventually at main, which needs special
299 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
301 So here's the plan. We WARN about probable ambiguity if
303 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
305 (all tested before quantification).
306 That is, all the type variables in Cq must be fixed by the the variables
307 in the environment, or by the variables in the type.
309 Notice that we union before calling oclose. Here's an example:
311 class J a b c | a b -> c
315 forall b c. (J a b c) => b -> b
317 Only if we union {a} from G with {b} from T before using oclose,
318 do we see that c is fixed.
320 It's a bit vague exactly which C we should use for this oclose call. If we
321 don't fix enough variables we might complain when we shouldn't (see
322 the above nasty example). Nothing will be perfect. That's why we can
323 only issue a warning.
326 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
328 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
330 then c is a "bubble"; there's no way it can ever improve, and it's
331 certainly ambiguous. UNLESS it is a constant (sigh). And what about
336 instance H x y => K (x,y)
338 Is this type ambiguous?
339 forall a b. (K (a,b), Eq b) => a -> a
341 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
342 is a "bubble" that's a set of constraints
344 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
346 Hence another idea. To decide Q start with fv(T) and grow it
347 by transitive closure in Cq (no functional dependencies involved).
348 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
349 The definitely-ambiguous can then float out, and get smashed at top level
350 (which squashes out the constants, like Eq (T a) above)
353 --------------------------------------
354 Notes on principal types
355 --------------------------------------
360 f x = let g y = op (y::Int) in True
362 Here the principal type of f is (forall a. a->a)
363 but we'll produce the non-principal type
364 f :: forall a. C Int => a -> a
367 --------------------------------------
368 The need for forall's in constraints
369 --------------------------------------
371 [Exchange on Haskell Cafe 5/6 Dec 2000]
373 class C t where op :: t -> Bool
374 instance C [t] where op x = True
376 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
377 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
379 The definitions of p and q differ only in the order of the components in
380 the pair on their right-hand sides. And yet:
382 ghc and "Typing Haskell in Haskell" reject p, but accept q;
383 Hugs rejects q, but accepts p;
384 hbc rejects both p and q;
385 nhc98 ... (Malcolm, can you fill in the blank for us!).
387 The type signature for f forces context reduction to take place, and
388 the results of this depend on whether or not the type of y is known,
389 which in turn depends on which component of the pair the type checker
392 Solution: if y::m a, float out the constraints
393 Monad m, forall c. C (m c)
394 When m is later unified with [], we can solve both constraints.
397 --------------------------------------
398 Notes on implicit parameters
399 --------------------------------------
401 Question 1: can we "inherit" implicit parameters
402 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
407 where f is *not* a top-level binding.
408 From the RHS of f we'll get the constraint (?y::Int).
409 There are two types we might infer for f:
413 (so we get ?y from the context of f's definition), or
415 f :: (?y::Int) => Int -> Int
417 At first you might think the first was better, becuase then
418 ?y behaves like a free variable of the definition, rather than
419 having to be passed at each call site. But of course, the WHOLE
420 IDEA is that ?y should be passed at each call site (that's what
421 dynamic binding means) so we'd better infer the second.
423 BOTTOM LINE: when *inferring types* you *must* quantify
424 over implicit parameters. See the predicate isFreeWhenInferring.
427 Question 2: type signatures
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
429 BUT WATCH OUT: When you supply a type signature, we can't force you
430 to quantify over implicit parameters. For example:
434 This is perfectly reasonable. We do not want to insist on
436 (?x + 1) :: (?x::Int => Int)
438 That would be silly. Here, the definition site *is* the occurrence site,
439 so the above strictures don't apply. Hence the difference between
440 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
441 and tcSimplifyCheckBind (which does not).
443 What about when you supply a type signature for a binding?
444 Is it legal to give the following explicit, user type
445 signature to f, thus:
450 At first sight this seems reasonable, but it has the nasty property
451 that adding a type signature changes the dynamic semantics.
454 (let f x = (x::Int) + ?y
455 in (f 3, f 3 with ?y=5)) with ?y = 6
461 in (f 3, f 3 with ?y=5)) with ?y = 6
465 Indeed, simply inlining f (at the Haskell source level) would change the
468 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
469 semantics for a Haskell program without knowing its typing, so if you
470 change the typing you may change the semantics.
472 To make things consistent in all cases where we are *checking* against
473 a supplied signature (as opposed to inferring a type), we adopt the
476 a signature does not need to quantify over implicit params.
478 [This represents a (rather marginal) change of policy since GHC 5.02,
479 which *required* an explicit signature to quantify over all implicit
480 params for the reasons mentioned above.]
482 But that raises a new question. Consider
484 Given (signature) ?x::Int
485 Wanted (inferred) ?x::Int, ?y::Bool
487 Clearly we want to discharge the ?x and float the ?y out. But
488 what is the criterion that distinguishes them? Clearly it isn't
489 what free type variables they have. The Right Thing seems to be
490 to float a constraint that
491 neither mentions any of the quantified type variables
492 nor any of the quantified implicit parameters
494 See the predicate isFreeWhenChecking.
497 Question 3: monomorphism
498 ~~~~~~~~~~~~~~~~~~~~~~~~
499 There's a nasty corner case when the monomorphism restriction bites:
503 The argument above suggests that we *must* generalise
504 over the ?y parameter, to get
505 z :: (?y::Int) => Int,
506 but the monomorphism restriction says that we *must not*, giving
508 Why does the momomorphism restriction say this? Because if you have
510 let z = x + ?y in z+z
512 you might not expect the addition to be done twice --- but it will if
513 we follow the argument of Question 2 and generalise over ?y.
516 Question 4: top level
517 ~~~~~~~~~~~~~~~~~~~~~
518 At the top level, monomorhism makes no sense at all.
521 main = let ?x = 5 in print foo
525 woggle :: (?x :: Int) => Int -> Int
528 We definitely don't want (foo :: Int) with a top-level implicit parameter
529 (?x::Int) becuase there is no way to bind it.
534 (A) Always generalise over implicit parameters
535 Bindings that fall under the monomorphism restriction can't
539 * Inlining remains valid
540 * No unexpected loss of sharing
541 * But simple bindings like
543 will be rejected, unless you add an explicit type signature
544 (to avoid the monomorphism restriction)
545 z :: (?y::Int) => Int
547 This seems unacceptable
549 (B) Monomorphism restriction "wins"
550 Bindings that fall under the monomorphism restriction can't
552 Always generalise over implicit parameters *except* for bindings
553 that fall under the monomorphism restriction
556 * Inlining isn't valid in general
557 * No unexpected loss of sharing
558 * Simple bindings like
560 accepted (get value of ?y from binding site)
562 (C) Always generalise over implicit parameters
563 Bindings that fall under the monomorphism restriction can't
564 be generalised, EXCEPT for implicit parameters
566 * Inlining remains valid
567 * Unexpected loss of sharing (from the extra generalisation)
568 * Simple bindings like
570 accepted (get value of ?y from occurrence sites)
575 None of these choices seems very satisfactory. But at least we should
576 decide which we want to do.
578 It's really not clear what is the Right Thing To Do. If you see
582 would you expect the value of ?y to be got from the *occurrence sites*
583 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
584 case of function definitions, the answer is clearly the former, but
585 less so in the case of non-fucntion definitions. On the other hand,
586 if we say that we get the value of ?y from the definition site of 'z',
587 then inlining 'z' might change the semantics of the program.
589 Choice (C) really says "the monomorphism restriction doesn't apply
590 to implicit parameters". Which is fine, but remember that every
591 innocent binding 'x = ...' that mentions an implicit parameter in
592 the RHS becomes a *function* of that parameter, called at each
593 use of 'x'. Now, the chances are that there are no intervening 'with'
594 clauses that bind ?y, so a decent compiler should common up all
595 those function calls. So I think I strongly favour (C). Indeed,
596 one could make a similar argument for abolishing the monomorphism
597 restriction altogether.
599 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
603 %************************************************************************
605 \subsection{tcSimplifyInfer}
607 %************************************************************************
609 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
611 1. Compute Q = grow( fvs(T), C )
613 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
614 predicates will end up in Ct; we deal with them at the top level
616 3. Try improvement, using functional dependencies
618 4. If Step 3 did any unification, repeat from step 1
619 (Unification can change the result of 'grow'.)
621 Note: we don't reduce dictionaries in step 2. For example, if we have
622 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
623 after step 2. However note that we may therefore quantify over more
624 type variables than we absolutely have to.
626 For the guts, we need a loop, that alternates context reduction and
627 improvement with unification. E.g. Suppose we have
629 class C x y | x->y where ...
631 and tcSimplify is called with:
633 Then improvement unifies a with b, giving
636 If we need to unify anything, we rattle round the whole thing all over
643 -> TcTyVarSet -- fv(T); type vars
645 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked)
646 TcDictBinds, -- Bindings
647 [TcId]) -- Dict Ids that must be bound here (zonked)
648 -- Any free (escaping) Insts are tossed into the environment
653 tcSimplifyInfer doc tau_tvs wanted_lie
654 = inferLoop doc (varSetElems tau_tvs)
655 wanted_lie `thenM` \ (qtvs, frees, binds, irreds) ->
657 extendLIEs frees `thenM_`
658 returnM (qtvs, binds, map instToId irreds)
660 inferLoop doc tau_tvs wanteds
662 zonkTcTyVarsAndFV tau_tvs `thenM` \ tau_tvs' ->
663 mappM zonkInst wanteds `thenM` \ wanteds' ->
664 tcGetGlobalTyVars `thenM` \ gbl_tvs ->
666 preds = fdPredsOfInsts wanteds'
667 qtvs = grow preds tau_tvs' `minusVarSet` oclose preds gbl_tvs
670 | isFreeWhenInferring qtvs inst = Free
671 | isClassDict inst = Irred -- Dicts
672 | otherwise = ReduceMe NoSCs -- Lits and Methods
673 env = mkRedEnv doc try_me []
675 traceTc (text "infloop" <+> vcat [ppr tau_tvs', ppr wanteds', ppr preds,
676 ppr (grow preds tau_tvs'), ppr qtvs]) `thenM_`
678 reduceContext env wanteds' `thenM` \ (improved, frees, binds, irreds) ->
682 returnM (varSetElems qtvs, frees, binds, irreds)
684 -- If improvement did some unification, we go round again. There
685 -- are two subtleties:
686 -- a) We start again with irreds, not wanteds
687 -- Using an instance decl might have introduced a fresh type variable
688 -- which might have been unified, so we'd get an infinite loop
689 -- if we started again with wanteds! See example [LOOP]
691 -- b) It's also essential to re-process frees, because unification
692 -- might mean that a type variable that looked free isn't now.
694 -- Hence the (irreds ++ frees)
696 -- However, NOTICE that when we are done, we might have some bindings, but
697 -- the final qtvs might be empty. See [NO TYVARS] below.
699 inferLoop doc tau_tvs (irreds ++ frees) `thenM` \ (qtvs1, frees1, binds1, irreds1) ->
700 returnM (qtvs1, frees1, binds `unionBags` binds1, irreds1)
705 class If b t e r | b t e -> r
708 class Lte a b c | a b -> c where lte :: a -> b -> c
710 instance (Lte a b l,If l b a c) => Max a b c
712 Wanted: Max Z (S x) y
714 Then we'll reduce using the Max instance to:
715 (Lte Z (S x) l, If l (S x) Z y)
716 and improve by binding l->T, after which we can do some reduction
717 on both the Lte and If constraints. What we *can't* do is start again
718 with (Max Z (S x) y)!
722 class Y a b | a -> b where
725 instance Y [[a]] a where
728 k :: X a -> X a -> X a
730 g :: Num a => [X a] -> [X a]
733 h ys = ys ++ map (k (y [[0]])) xs
735 The excitement comes when simplifying the bindings for h. Initially
736 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
737 From this we get t1:=:t2, but also various bindings. We can't forget
738 the bindings (because of [LOOP]), but in fact t1 is what g is
741 The net effect of [NO TYVARS]
744 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
745 isFreeWhenInferring qtvs inst
746 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
747 && isInheritableInst inst -- And no implicit parameter involved
748 -- (see "Notes on implicit parameters")
750 {- No longer used (with implication constraints)
751 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
752 -> NameSet -- Quantified implicit parameters
754 isFreeWhenChecking qtvs ips inst
755 = isFreeWrtTyVars qtvs inst
756 && isFreeWrtIPs ips inst
759 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
760 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
764 %************************************************************************
766 \subsection{tcSimplifyCheck}
768 %************************************************************************
770 @tcSimplifyCheck@ is used when we know exactly the set of variables
771 we are going to quantify over. For example, a class or instance declaration.
774 -----------------------------------------------------------
775 -- tcSimplifyCheck is used when checking expression type signatures,
776 -- class decls, instance decls etc.
777 tcSimplifyCheck :: InstLoc
778 -> [TcTyVar] -- Quantify over these
781 -> TcM TcDictBinds -- Bindings
782 tcSimplifyCheck loc qtvs givens wanteds
783 = ASSERT( all isSkolemTyVar qtvs )
784 do { (binds, irreds) <- innerCheckLoop loc AddSCs givens wanteds
785 ; implic_bind <- bindIrreds loc [] emptyRefinement
787 ; return (binds `unionBags` implic_bind) }
789 -----------------------------------------------------------
790 -- tcSimplifyCheckPat is used for existential pattern match
791 tcSimplifyCheckPat :: InstLoc
792 -> [CoVar] -> Refinement
793 -> [TcTyVar] -- Quantify over these
796 -> TcM TcDictBinds -- Bindings
797 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
798 = ASSERT( all isSkolemTyVar qtvs )
799 do { (binds, irreds) <- innerCheckLoop loc AddSCs givens wanteds
800 ; implic_bind <- bindIrreds loc co_vars reft
802 ; return (binds `unionBags` implic_bind) }
804 -----------------------------------------------------------
805 bindIrreds :: InstLoc -> [CoVar] -> Refinement
806 -> [TcTyVar] -> [Inst] -> [Inst]
808 -- Make a binding that binds 'irreds', by generating an implication
809 -- constraint for them, *and* throwing the constraint into the LIE
810 bindIrreds loc co_vars reft qtvs givens irreds
811 = do { let givens' = filter isDict givens
812 -- The givens can include methods
814 -- If there are no 'givens', then it's safe to
815 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
816 -- See Note [Freeness and implications]
817 ; irreds' <- if null givens'
819 { let qtv_set = mkVarSet qtvs
820 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
822 ; return real_irreds }
825 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
826 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
827 -- This call does the real work
832 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
834 -> TcM ([Inst], TcDictBinds)
835 -- Make a binding that binds 'irreds', by generating an implication
836 -- constraint for them, *and* throwing the constraint into the LIE
837 -- The binding looks like
838 -- (ir1, .., irn) = f qtvs givens
839 -- where f is (evidence for) the new implication constraint
841 -- This binding must line up the 'rhs' in reduceImplication
842 makeImplicationBind loc all_tvs reft
843 givens -- Guaranteed all Dicts
845 | null irreds -- If there are no irreds, we are done
846 = return ([], emptyBag)
847 | otherwise -- Otherwise we must generate a binding
848 = do { uniq <- newUnique
849 ; span <- getSrcSpanM
850 ; let name = mkInternalName uniq (mkVarOcc "ic") (srcSpanStart span)
851 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
852 tci_tyvars = all_tvs,
854 tci_wanted = irreds, tci_loc = loc }
856 ; let n_irreds = length irreds
857 irred_ids = map instToId irreds
858 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
859 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
860 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
861 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
862 bind | n_irreds==1 = VarBind (head irred_ids) rhs
863 | otherwise = PatBind { pat_lhs = L span pat,
864 pat_rhs = unguardedGRHSs rhs,
866 bind_fvs = placeHolderNames }
867 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
868 return ([implic_inst], unitBag (L span bind)) }
870 -----------------------------------------------------------
874 [Inst]) -- Irreducible
876 topCheckLoop doc wanteds
877 = checkLoop (mkRedEnv doc try_me []) wanteds
879 try_me inst = ReduceMe AddSCs
881 -----------------------------------------------------------
882 innerCheckLoop :: InstLoc -> WantSCs
886 [Inst]) -- Irreducible
888 innerCheckLoop inst_loc want_scs givens wanteds
889 = checkLoop env wanteds
891 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
893 try_me inst | isMethodOrLit inst = ReduceMe want_scs
895 -- When checking against a given signature
896 -- we MUST be very gentle: Note [Check gently]
901 We have to very careful about not simplifying too vigorously
906 f :: Show b => T b -> b
909 Inside the pattern match, which binds (a:*, x:a), we know that
911 Hence we have a dictionary for Show [a] available; and indeed we
912 need it. We are going to build an implication contraint
913 forall a. (b~[a]) => Show [a]
914 Later, we will solve this constraint using the knowledge (Show b)
916 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
917 thing becomes insoluble. So we simplify gently (get rid of literals
918 and methods only, plus common up equal things), deferring the real
919 work until top level, when we solve the implication constraint
924 -----------------------------------------------------------
928 [Inst]) -- Irreducible
929 -- Precondition: the try_me never returns Free
930 -- givens are completely rigid
932 checkLoop env wanteds
933 = do { -- Givens are skolems, so no need to zonk them
934 wanteds' <- mappM zonkInst wanteds
936 ; (improved, _frees, binds, irreds) <- reduceContext env wanteds'
938 ; ASSERT( null _frees )
941 return (binds, irreds)
944 { (binds1, irreds1) <- checkLoop env irreds
945 ; return (binds `unionBags` binds1, irreds1) } }
950 -----------------------------------------------------------
951 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
952 -- against, but we don't know the type variables over which we are going to quantify.
953 -- This happens when we have a type signature for a mutually recursive group
956 -> TcTyVarSet -- fv(T)
959 -> TcM ([TcTyVar], -- Variables over which to quantify
960 TcDictBinds) -- Bindings
962 tcSimplifyInferCheck loc tau_tvs givens wanteds
963 = do { (binds, irreds) <- innerCheckLoop loc AddSCs givens wanteds
965 -- Figure out which type variables to quantify over
966 -- You might think it should just be the signature tyvars,
967 -- but in bizarre cases you can get extra ones
968 -- f :: forall a. Num a => a -> a
969 -- f x = fst (g (x, head [])) + 1
971 -- Here we infer g :: forall a b. a -> b -> (b,a)
972 -- We don't want g to be monomorphic in b just because
973 -- f isn't quantified over b.
974 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
975 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
976 ; gbl_tvs <- tcGetGlobalTyVars
977 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
978 -- We could close gbl_tvs, but its not necessary for
979 -- soundness, and it'll only affect which tyvars, not which
980 -- dictionaries, we quantify over
982 -- Now we are back to normal (c.f. tcSimplCheck)
983 ; implic_bind <- bindIrreds loc [] emptyRefinement
985 ; return (qtvs, binds `unionBags` implic_bind) }
989 %************************************************************************
991 tcSimplifySuperClasses
993 %************************************************************************
995 Note [SUPERCLASS-LOOP 1]
996 ~~~~~~~~~~~~~~~~~~~~~~~~
997 We have to be very, very careful when generating superclasses, lest we
998 accidentally build a loop. Here's an example:
1002 class S a => C a where { opc :: a -> a }
1003 class S b => D b where { opd :: b -> b }
1005 instance C Int where
1008 instance D Int where
1011 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1012 Simplifying, we may well get:
1013 $dfCInt = :C ds1 (opd dd)
1016 Notice that we spot that we can extract ds1 from dd.
1018 Alas! Alack! We can do the same for (instance D Int):
1020 $dfDInt = :D ds2 (opc dc)
1024 And now we've defined the superclass in terms of itself.
1026 Solution: never generate a superclass selectors at all when
1027 satisfying the superclass context of an instance declaration.
1029 Two more nasty cases are in
1034 tcSimplifySuperClasses
1039 tcSimplifySuperClasses loc givens sc_wanteds
1040 = do { (binds1, irreds) <- checkLoop env sc_wanteds
1041 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1042 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1045 env = mkRedEnv (pprInstLoc loc) try_me givens
1046 try_me inst = ReduceMe NoSCs
1047 -- Like topCheckLoop, but with NoSCs
1051 %************************************************************************
1053 \subsection{tcSimplifyRestricted}
1055 %************************************************************************
1057 tcSimplifyRestricted infers which type variables to quantify for a
1058 group of restricted bindings. This isn't trivial.
1061 We want to quantify over a to get id :: forall a. a->a
1064 We do not want to quantify over a, because there's an Eq a
1065 constraint, so we get eq :: a->a->Bool (notice no forall)
1068 RHS has type 'tau', whose free tyvars are tau_tvs
1069 RHS has constraints 'wanteds'
1072 Quantify over (tau_tvs \ ftvs(wanteds))
1073 This is bad. The constraints may contain (Monad (ST s))
1074 where we have instance Monad (ST s) where...
1075 so there's no need to be monomorphic in s!
1077 Also the constraint might be a method constraint,
1078 whose type mentions a perfectly innocent tyvar:
1079 op :: Num a => a -> b -> a
1080 Here, b is unconstrained. A good example would be
1082 We want to infer the polymorphic type
1083 foo :: forall b. b -> b
1086 Plan B (cunning, used for a long time up to and including GHC 6.2)
1087 Step 1: Simplify the constraints as much as possible (to deal
1088 with Plan A's problem). Then set
1089 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1091 Step 2: Now simplify again, treating the constraint as 'free' if
1092 it does not mention qtvs, and trying to reduce it otherwise.
1093 The reasons for this is to maximise sharing.
1095 This fails for a very subtle reason. Suppose that in the Step 2
1096 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1097 In the Step 1 this constraint might have been simplified, perhaps to
1098 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1099 This won't happen in Step 2... but that in turn might prevent some other
1100 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1101 and that in turn breaks the invariant that no constraints are quantified over.
1103 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1108 Step 1: Simplify the constraints as much as possible (to deal
1109 with Plan A's problem). Then set
1110 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1111 Return the bindings from Step 1.
1114 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1117 instance (HasBinary ty IO) => HasCodedValue ty
1119 foo :: HasCodedValue a => String -> IO a
1121 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1122 doDecodeIO codedValue view
1123 = let { act = foo "foo" } in act
1125 You might think this should work becuase the call to foo gives rise to a constraint
1126 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1127 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1128 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1130 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1134 Plan D (a variant of plan B)
1135 Step 1: Simplify the constraints as much as possible (to deal
1136 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1137 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1139 Step 2: Now simplify again, treating the constraint as 'free' if
1140 it does not mention qtvs, and trying to reduce it otherwise.
1142 The point here is that it's generally OK to have too few qtvs; that is,
1143 to make the thing more monomorphic than it could be. We don't want to
1144 do that in the common cases, but in wierd cases it's ok: the programmer
1145 can always add a signature.
1147 Too few qtvs => too many wanteds, which is what happens if you do less
1152 tcSimplifyRestricted -- Used for restricted binding groups
1153 -- i.e. ones subject to the monomorphism restriction
1156 -> [Name] -- Things bound in this group
1157 -> TcTyVarSet -- Free in the type of the RHSs
1158 -> [Inst] -- Free in the RHSs
1159 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked)
1160 TcDictBinds) -- Bindings
1161 -- tcSimpifyRestricted returns no constraints to
1162 -- quantify over; by definition there are none.
1163 -- They are all thrown back in the LIE
1165 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1166 -- Zonk everything in sight
1167 = mappM zonkInst wanteds `thenM` \ wanteds' ->
1169 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1170 -- dicts; the idea is to get rid of as many type
1171 -- variables as possible, and we don't want to stop
1172 -- at (say) Monad (ST s), because that reduces
1173 -- immediately, with no constraint on s.
1175 -- BUT do no improvement! See Plan D above
1176 -- HOWEVER, some unification may take place, if we instantiate
1177 -- a method Inst with an equality constraint
1178 let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1180 reduceContext env wanteds' `thenM` \ (_imp, _frees, _binds, constrained_dicts) ->
1182 -- Next, figure out the tyvars we will quantify over
1183 zonkTcTyVarsAndFV (varSetElems tau_tvs) `thenM` \ tau_tvs' ->
1184 tcGetGlobalTyVars `thenM` \ gbl_tvs' ->
1185 mappM zonkInst constrained_dicts `thenM` \ constrained_dicts' ->
1187 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1188 qtvs' = (tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs')
1189 `minusVarSet` constrained_tvs'
1191 traceTc (text "tcSimplifyRestricted" <+> vcat [
1192 pprInsts wanteds, pprInsts _frees, pprInsts constrained_dicts',
1194 ppr constrained_tvs', ppr tau_tvs', ppr qtvs' ]) `thenM_`
1196 -- The first step may have squashed more methods than
1197 -- necessary, so try again, this time more gently, knowing the exact
1198 -- set of type variables to quantify over.
1200 -- We quantify only over constraints that are captured by qtvs';
1201 -- these will just be a subset of non-dicts. This in contrast
1202 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1203 -- all *non-inheritable* constraints too. This implements choice
1204 -- (B) under "implicit parameter and monomorphism" above.
1206 -- Remember that we may need to do *some* simplification, to
1207 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1208 -- just to float all constraints
1210 -- At top level, we *do* squash methods becuase we want to
1211 -- expose implicit parameters to the test that follows
1213 is_nested_group = isNotTopLevel top_lvl
1214 try_me inst | isFreeWrtTyVars qtvs' inst,
1215 (is_nested_group || isDict inst) = Free
1216 | otherwise = ReduceMe AddSCs
1217 env = mkNoImproveRedEnv doc try_me
1219 reduceContext env wanteds' `thenM` \ (_imp, frees, binds, irreds) ->
1220 ASSERT( null irreds )
1222 -- See "Notes on implicit parameters, Question 4: top level"
1223 if is_nested_group then
1224 extendLIEs frees `thenM_`
1225 returnM (varSetElems qtvs', binds)
1228 (non_ips, bad_ips) = partition isClassDict frees
1230 addTopIPErrs bndrs bad_ips `thenM_`
1231 extendLIEs non_ips `thenM_`
1232 returnM (varSetElems qtvs', binds)
1236 %************************************************************************
1240 %************************************************************************
1242 On the LHS of transformation rules we only simplify methods and constants,
1243 getting dictionaries. We want to keep all of them unsimplified, to serve
1244 as the available stuff for the RHS of the rule.
1246 Example. Consider the following left-hand side of a rule
1248 f (x == y) (y > z) = ...
1250 If we typecheck this expression we get constraints
1252 d1 :: Ord a, d2 :: Eq a
1254 We do NOT want to "simplify" to the LHS
1256 forall x::a, y::a, z::a, d1::Ord a.
1257 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1261 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1262 f ((==) d2 x y) ((>) d1 y z) = ...
1264 Here is another example:
1266 fromIntegral :: (Integral a, Num b) => a -> b
1267 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1269 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1270 we *dont* want to get
1272 forall dIntegralInt.
1273 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1275 because the scsel will mess up RULE matching. Instead we want
1277 forall dIntegralInt, dNumInt.
1278 fromIntegral Int Int dIntegralInt dNumInt = id Int
1282 g (x == y) (y == z) = ..
1284 where the two dictionaries are *identical*, we do NOT WANT
1286 forall x::a, y::a, z::a, d1::Eq a
1287 f ((==) d1 x y) ((>) d1 y z) = ...
1289 because that will only match if the dict args are (visibly) equal.
1290 Instead we want to quantify over the dictionaries separately.
1292 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1293 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1294 from scratch, rather than further parameterise simpleReduceLoop etc
1297 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1298 tcSimplifyRuleLhs wanteds
1299 = go [] emptyBag wanteds
1302 = return (dicts, binds)
1303 go dicts binds (w:ws)
1305 = go (w:dicts) binds ws
1307 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1308 -- to fromInteger; this looks fragile to me
1309 ; lookup_result <- lookupSimpleInst w'
1310 ; case lookup_result of
1311 GenInst ws' rhs -> go dicts (addBind binds w rhs) (ws' ++ ws)
1312 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1316 tcSimplifyBracket is used when simplifying the constraints arising from
1317 a Template Haskell bracket [| ... |]. We want to check that there aren't
1318 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1319 Show instance), but we aren't otherwise interested in the results.
1320 Nor do we care about ambiguous dictionaries etc. We will type check
1321 this bracket again at its usage site.
1324 tcSimplifyBracket :: [Inst] -> TcM ()
1325 tcSimplifyBracket wanteds
1326 = do { topCheckLoop doc wanteds
1329 doc = text "tcSimplifyBracket"
1333 %************************************************************************
1335 \subsection{Filtering at a dynamic binding}
1337 %************************************************************************
1342 we must discharge all the ?x constraints from B. We also do an improvement
1343 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1345 Actually, the constraints from B might improve the types in ?x. For example
1347 f :: (?x::Int) => Char -> Char
1350 then the constraint (?x::Int) arising from the call to f will
1351 force the binding for ?x to be of type Int.
1354 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1357 -- We need a loop so that we do improvement, and then
1358 -- (next time round) generate a binding to connect the two
1360 -- Here the two ?x's have different types, and improvement
1361 -- makes them the same.
1363 tcSimplifyIPs given_ips wanteds
1364 = do { wanteds' <- mappM zonkInst wanteds
1365 ; given_ips' <- mappM zonkInst given_ips
1366 -- Unusually for checking, we *must* zonk the given_ips
1368 ; let env = mkRedEnv doc try_me given_ips'
1369 ; (improved, _frees, binds, irreds) <- reduceContext env wanteds'
1371 ; if not improved then
1372 ASSERT( all is_free irreds )
1373 do { extendLIEs irreds
1376 tcSimplifyIPs given_ips wanteds }
1378 doc = text "tcSimplifyIPs" <+> ppr given_ips
1379 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1380 is_free inst = isFreeWrtIPs ip_set inst
1382 -- Simplify any methods that mention the implicit parameter
1383 try_me inst | is_free inst = Irred
1384 | otherwise = ReduceMe NoSCs
1388 %************************************************************************
1390 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1392 %************************************************************************
1394 When doing a binding group, we may have @Insts@ of local functions.
1395 For example, we might have...
1397 let f x = x + 1 -- orig local function (overloaded)
1398 f.1 = f Int -- two instances of f
1403 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1404 where @f@ is in scope; those @Insts@ must certainly not be passed
1405 upwards towards the top-level. If the @Insts@ were binding-ified up
1406 there, they would have unresolvable references to @f@.
1408 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1409 For each method @Inst@ in the @init_lie@ that mentions one of the
1410 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1411 @LIE@), as well as the @HsBinds@ generated.
1414 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1415 -- Simlifies only MethodInsts, and generate only bindings of form
1417 -- We're careful not to even generate bindings of the form
1419 -- You'd think that'd be fine, but it interacts with what is
1420 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1422 bindInstsOfLocalFuns wanteds local_ids
1423 | null overloaded_ids
1425 = extendLIEs wanteds `thenM_`
1426 returnM emptyLHsBinds
1429 = do { (binds, irreds) <- checkLoop env for_me
1430 ; extendLIEs not_for_me
1434 env = mkRedEnv doc try_me []
1435 doc = text "bindInsts" <+> ppr local_ids
1436 overloaded_ids = filter is_overloaded local_ids
1437 is_overloaded id = isOverloadedTy (idType id)
1438 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1440 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1441 -- so it's worth building a set, so that
1442 -- lookup (in isMethodFor) is faster
1443 try_me inst | isMethod inst = ReduceMe NoSCs
1448 %************************************************************************
1450 \subsection{Data types for the reduction mechanism}
1452 %************************************************************************
1454 The main control over context reduction is here
1458 = RedEnv { red_doc :: SDoc -- The context
1459 , red_try_me :: Inst -> WhatToDo
1460 , red_improve :: Bool -- True <=> do improvement
1461 , red_givens :: [Inst] -- All guaranteed rigid
1463 -- but see Note [Rigidity]
1464 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1465 -- See Note [RedStack]
1469 -- The red_givens are rigid so far as cmpInst is concerned.
1470 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1471 -- let ?x = e in ...
1472 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1473 -- But that doesn't affect the comparison, which is based only on mame.
1476 -- The red_stack pair (n,insts) pair is just used for error reporting.
1477 -- 'n' is always the depth of the stack.
1478 -- The 'insts' is the stack of Insts being reduced: to produce X
1479 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1482 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1483 mkRedEnv doc try_me givens
1484 = RedEnv { red_doc = doc, red_try_me = try_me,
1485 red_givens = givens, red_stack = (0,[]),
1486 red_improve = True }
1488 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1489 -- Do not do improvement; no givens
1490 mkNoImproveRedEnv doc try_me
1491 = RedEnv { red_doc = doc, red_try_me = try_me,
1492 red_givens = [], red_stack = (0,[]),
1493 red_improve = True }
1496 = ReduceMe WantSCs -- Try to reduce this
1497 -- If there's no instance, add the inst to the
1498 -- irreductible ones, but don't produce an error
1499 -- message of any kind.
1500 -- It might be quite legitimate such as (Eq a)!
1502 | Irred -- Return as irreducible unless it can
1503 -- be reduced to a constant in one step
1505 | Free -- Return as free
1507 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1508 -- of a predicate when adding it to the avails
1509 -- The reason for this flag is entirely the super-class loop problem
1510 -- Note [SUPER-CLASS LOOP 1]
1513 %************************************************************************
1515 \subsection[reduce]{@reduce@}
1517 %************************************************************************
1521 reduceContext :: RedEnv
1523 -> TcM (ImprovementDone,
1525 TcDictBinds, -- Dictionary bindings
1526 [Inst]) -- Irreducible
1528 reduceContext env wanteds
1529 = do { traceTc (text "reduceContext" <+> (vcat [
1530 text "----------------------",
1532 text "given" <+> ppr (red_givens env),
1533 text "wanted" <+> ppr wanteds,
1534 text "----------------------"
1537 -- Build the Avail mapping from "givens"
1538 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1541 ; avails <- reduceList env wanteds init_state
1543 ; let improved = availsImproved avails
1544 ; (binds, irreds, frees) <- extractResults avails wanteds
1546 ; traceTc (text "reduceContext end" <+> (vcat [
1547 text "----------------------",
1549 text "given" <+> ppr (red_givens env),
1550 text "wanted" <+> ppr wanteds,
1552 text "avails" <+> pprAvails avails,
1553 text "frees" <+> ppr frees,
1554 text "improved =" <+> ppr improved,
1555 text "----------------------"
1558 ; return (improved, frees, binds, irreds) }
1560 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1561 tcImproveOne avails inst
1562 | not (isDict inst) = return False
1564 = do { inst_envs <- tcGetInstEnvs
1565 ; let eqns = improveOne (classInstances inst_envs)
1566 (dictPred inst, pprInstArising inst)
1567 [ (dictPred p, pprInstArising p)
1568 | p <- availsInsts avails, isDict p ]
1569 -- Avails has all the superclasses etc (good)
1570 -- It also has all the intermediates of the deduction (good)
1571 -- It does not have duplicates (good)
1572 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1573 -- so that improve will see them separate
1574 ; traceTc (text "improveOne" <+> ppr inst)
1577 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1578 -> TcM ImprovementDone
1579 unifyEqns [] = return False
1581 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1585 unify ((qtvs, pairs), what1, what2)
1586 = addErrCtxtM (mkEqnMsg what1 what2) $
1587 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1588 mapM_ (unif_pr tenv) pairs
1589 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1591 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1593 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1594 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1595 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1596 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1597 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1598 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1599 ; return (tidy_env, msg) }
1602 The main context-reduction function is @reduce@. Here's its game plan.
1605 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1606 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1607 = do { dopts <- getDOpts
1610 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1611 2 (ifPprDebug (nest 2 (pprStack stk))))
1614 ; if n >= ctxtStkDepth dopts then
1615 failWithTc (reduceDepthErr n stk)
1619 go [] state = return state
1620 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1623 -- Base case: we're done!
1624 reduce env wanted avails
1625 -- It's the same as an existing inst, or a superclass thereof
1626 | Just avail <- findAvail avails wanted
1630 = case red_try_me env wanted of {
1631 Free -> try_simple addFree -- It's free so just chuck it upstairs
1632 ; Irred -> try_simple (addIrred AddSCs) -- Assume want superclasses
1634 ; ReduceMe want_scs -> -- It should be reduced
1635 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1636 case lookup_result of
1637 NoInstance -> -- No such instance!
1638 -- Add it and its superclasses
1639 addIrred want_scs avails wanted
1641 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1643 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1644 ; avails2 <- reduceList env wanteds' avails1
1645 ; addWanted want_scs avails2 wanted rhs wanteds' }
1646 -- Temporarily do addIrred *before* the reduceList,
1647 -- which has the effect of adding the thing we are trying
1648 -- to prove to the database before trying to prove the things it
1649 -- needs. See note [RECURSIVE DICTIONARIES]
1650 -- NB: we must not do an addWanted before, because that adds the
1651 -- superclasses too, and thaat can lead to a spurious loop; see
1652 -- the examples in [SUPERCLASS-LOOP]
1653 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1657 -- First, see if the inst can be reduced to a constant in one step
1658 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1659 -- Don't bother for implication constraints, which take real work
1660 try_simple do_this_otherwise
1661 = do { res <- lookupSimpleInst wanted
1663 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1664 other -> do_this_otherwise avails wanted }
1668 Note [SUPERCLASS-LOOP 2]
1669 ~~~~~~~~~~~~~~~~~~~~~~~~
1670 But the above isn't enough. Suppose we are *given* d1:Ord a,
1671 and want to deduce (d2:C [a]) where
1673 class Ord a => C a where
1674 instance Ord [a] => C [a] where ...
1676 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1677 superclasses of C [a] to avails. But we must not overwrite the binding
1678 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1681 Here's another variant, immortalised in tcrun020
1682 class Monad m => C1 m
1683 class C1 m => C2 m x
1684 instance C2 Maybe Bool
1685 For the instance decl we need to build (C1 Maybe), and it's no good if
1686 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1687 before we search for C1 Maybe.
1689 Here's another example
1690 class Eq b => Foo a b
1691 instance Eq a => Foo [a] a
1695 we'll first deduce that it holds (via the instance decl). We must not
1696 then overwrite the Eq t constraint with a superclass selection!
1698 At first I had a gross hack, whereby I simply did not add superclass constraints
1699 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1700 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1701 I found a very obscure program (now tcrun021) in which improvement meant the
1702 simplifier got two bites a the cherry... so something seemed to be an Irred
1703 first time, but reducible next time.
1705 Now we implement the Right Solution, which is to check for loops directly
1706 when adding superclasses. It's a bit like the occurs check in unification.
1709 Note [RECURSIVE DICTIONARIES]
1710 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1712 data D r = ZeroD | SuccD (r (D r));
1714 instance (Eq (r (D r))) => Eq (D r) where
1715 ZeroD == ZeroD = True
1716 (SuccD a) == (SuccD b) = a == b
1719 equalDC :: D [] -> D [] -> Bool;
1722 We need to prove (Eq (D [])). Here's how we go:
1726 by instance decl, holds if
1730 by instance decl of Eq, holds if
1732 where d2 = dfEqList d3
1735 But now we can "tie the knot" to give
1741 and it'll even run! The trick is to put the thing we are trying to prove
1742 (in this case Eq (D []) into the database before trying to prove its
1743 contributing clauses.
1746 %************************************************************************
1748 Reducing a single constraint
1750 %************************************************************************
1753 ---------------------------------------------
1754 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1755 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1756 tci_given = extra_givens, tci_wanted = wanteds })
1757 = reduceImplication env avails reft tvs extra_givens wanteds loc
1759 reduceInst env avails other_inst
1760 = do { result <- lookupSimpleInst other_inst
1761 ; return (avails, result) }
1765 ---------------------------------------------
1766 reduceImplication :: RedEnv
1768 -> Refinement -- May refine the givens; often empty
1769 -> [TcTyVar] -- Quantified type variables; all skolems
1770 -> [Inst] -- Extra givens; all rigid
1773 -> TcM (Avails, LookupInstResult)
1776 Suppose we are simplifying the constraint
1777 forall bs. extras => wanted
1778 in the context of an overall simplification problem with givens 'givens',
1779 and refinment 'reft'.
1782 * The refinement is often empty
1784 * The 'extra givens' need not mention any of the quantified type variables
1785 e.g. forall {}. Eq a => Eq [a]
1786 forall {}. C Int => D (Tree Int)
1788 This happens when you have something like
1790 T1 :: Eq a => a -> T a
1793 f x = ...(case x of { T1 v -> v==v })...
1796 -- ToDo: should we instantiate tvs? I think it's not necessary
1798 -- ToDo: what about improvement? There may be some improvement
1799 -- exposed as a result of the simplifications done by reduceList
1800 -- which are discarded if we back off.
1801 -- This is almost certainly Wrong, but we'll fix it when dealing
1802 -- better with equality constraints
1803 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1804 = do { -- Add refined givens, and the extra givens
1805 (refined_red_givens, avails)
1806 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1807 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1808 ; avails <- foldlM addGiven avails extra_givens
1810 -- Solve the sub-problem
1811 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1812 env' = env { red_givens = refined_red_givens ++ extra_givens
1813 , red_try_me = try_me }
1815 ; traceTc (text "reduceImplication" <+> vcat
1816 [ ppr (red_givens env), ppr extra_givens, ppr reft, ppr wanteds ])
1817 ; avails <- reduceList env' wanteds avails
1819 -- Extract the binding (no frees, because try_me never says Free)
1820 ; (binds, irreds, _frees) <- extractResults avails wanteds
1822 -- We always discard the extra avails we've generated;
1823 -- but we remember if we have done any (global) improvement
1824 ; let ret_avails = updateImprovement orig_avails avails
1826 ; if isEmptyLHsBinds binds then -- No progress
1827 return (ret_avails, NoInstance)
1829 { (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
1830 -- This binding is useless if the recursive simplification
1831 -- made no progress; but currently we don't try to optimise that
1832 -- case. After all, we only try hard to reduce at top level, or
1833 -- when inferring types.
1835 ; let dict_ids = map instToId extra_givens
1836 co = mkWpTyLams tvs <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
1837 rhs = mkHsWrap co payload
1838 loc = instLocSpan inst_loc
1839 payload | isSingleton wanteds = HsVar (instToId (head wanteds))
1840 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) wanteds) Boxed
1842 -- If there are any irreds, we back off and return NoInstance
1843 ; return (ret_avails, GenInst implic_insts (L loc rhs))
1847 Note [Freeness and implications]
1848 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1849 It's hard to say when an implication constraint can be floated out. Consider
1850 forall {} Eq a => Foo [a]
1851 The (Foo [a]) doesn't mention any of the quantified variables, but it
1852 still might be partially satisfied by the (Eq a).
1854 There is a useful special case when it *is* easy to partition the
1855 constraints, namely when there are no 'givens'. Consider
1856 forall {a}. () => Bar b
1857 There are no 'givens', and so there is no reason to capture (Bar b).
1858 We can let it float out. But if there is even one constraint we
1859 must be much more careful:
1860 forall {a}. C a b => Bar (m b)
1861 because (C a b) might have a superclass (D b), from which we might
1862 deduce (Bar [b]) when m later gets instantiated to []. Ha!
1864 Here is an even more exotic example
1866 Now consider the constraint
1867 forall b. D Int b => C Int
1868 We can satisfy the (C Int) from the superclass of D, so we don't want
1869 to float the (C Int) out, even though it mentions no type variable in
1872 %************************************************************************
1874 Avails and AvailHow: the pool of evidence
1876 %************************************************************************
1880 data Avails = Avails !ImprovementDone !AvailEnv
1882 type ImprovementDone = Bool -- True <=> some unification has happened
1883 -- so some Irreds might now be reducible
1884 -- keys that are now
1886 type AvailEnv = FiniteMap Inst AvailHow
1888 = IsFree -- Used for free Insts
1889 | IsIrred -- Used for irreducible dictionaries,
1890 -- which are going to be lambda bound
1892 | Given TcId -- Used for dictionaries for which we have a binding
1893 -- e.g. those "given" in a signature
1895 | Rhs -- Used when there is a RHS
1896 (LHsExpr TcId) -- The RHS
1897 [Inst] -- Insts free in the RHS; we need these too
1899 instance Outputable Avails where
1902 pprAvails (Avails imp avails)
1903 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
1904 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
1905 | (inst,avail) <- fmToList avails ])]
1907 instance Outputable AvailHow where
1910 -------------------------
1911 pprAvail :: AvailHow -> SDoc
1912 pprAvail IsFree = text "Free"
1913 pprAvail IsIrred = text "Irred"
1914 pprAvail (Given x) = text "Given" <+> ppr x
1915 pprAvail (Rhs rhs bs) = text "Rhs" <+> ppr rhs <+> braces (ppr bs)
1917 -------------------------
1918 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
1919 extendAvailEnv env inst avail = addToFM env inst avail
1921 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
1922 findAvailEnv env wanted = lookupFM env wanted
1923 -- NB 1: the Ord instance of Inst compares by the class/type info
1924 -- *not* by unique. So
1925 -- d1::C Int == d2::C Int
1927 emptyAvails :: Avails
1928 emptyAvails = Avails False emptyFM
1930 findAvail :: Avails -> Inst -> Maybe AvailHow
1931 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
1933 elemAvails :: Inst -> Avails -> Bool
1934 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
1936 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
1938 extendAvails avails@(Avails imp env) inst avail
1939 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
1940 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
1942 availsInsts :: Avails -> [Inst]
1943 availsInsts (Avails _ avails) = keysFM avails
1945 availsImproved (Avails imp _) = imp
1947 updateImprovement :: Avails -> Avails -> Avails
1948 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
1949 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
1952 Extracting the bindings from a bunch of Avails.
1953 The bindings do *not* come back sorted in dependency order.
1954 We assume that they'll be wrapped in a big Rec, so that the
1955 dependency analyser can sort them out later
1958 extractResults :: Avails
1960 -> TcM ( TcDictBinds, -- Bindings
1961 [Inst], -- Irreducible ones
1962 [Inst]) -- Free ones
1964 extractResults (Avails _ avails) wanteds
1965 = go avails emptyBag [] [] wanteds
1967 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst] -> [Inst]
1968 -> TcM (TcDictBinds, [Inst], [Inst])
1969 go avails binds irreds frees []
1970 = returnM (binds, irreds, frees)
1972 go avails binds irreds frees (w:ws)
1973 = case findAvailEnv avails w of
1974 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
1975 go avails binds irreds frees ws
1977 Just IsFree -> go (add_free avails w) binds irreds (w:frees) ws
1978 Just IsIrred -> go (add_given avails w) binds (w:irreds) frees ws
1980 Just (Given id) -> go avails new_binds irreds frees ws
1982 new_binds | id == instToId w = binds
1983 | otherwise = addBind binds w (L (instSpan w) (HsVar id))
1984 -- The sought Id can be one of the givens, via a superclass chain
1985 -- and then we definitely don't want to generate an x=x binding!
1987 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds frees (ws' ++ ws)
1989 new_binds = addBind binds w rhs
1991 add_given avails w = extendAvailEnv avails w (Given (instToId w))
1993 add_free avails w | isMethod w = avails
1994 | otherwise = add_given avails w
1996 -- Do *not* replace Free by Given if it's a method.
1997 -- The following situation shows why this is bad:
1998 -- truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
1999 -- From an application (truncate f i) we get
2000 -- t1 = truncate at f
2002 -- If we have also have a second occurrence of truncate, we get
2003 -- t3 = truncate at f
2005 -- When simplifying with i,f free, we might still notice that
2006 -- t1=t3; but alas, the binding for t2 (which mentions t1)
2007 -- will continue to float out!
2009 addBind binds inst rhs = binds `unionBags` unitBag (L (instSpan inst)
2010 (VarBind (instToId inst) rhs))
2011 instSpan wanted = instLocSpan (instLoc wanted)
2016 -------------------------
2017 addFree :: Avails -> Inst -> TcM Avails
2018 -- When an Inst is tossed upstairs as 'free' we nevertheless add it
2019 -- to avails, so that any other equal Insts will be commoned up right
2020 -- here rather than also being tossed upstairs. This is really just
2021 -- an optimisation, and perhaps it is more trouble that it is worth,
2022 -- as the following comments show!
2024 -- NB: do *not* add superclasses. If we have
2027 -- but a is not bound here, then we *don't* want to derive
2028 -- dn from df here lest we lose sharing.
2030 addFree avails free = extendAvails avails free IsFree
2032 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2033 addWanted want_scs avails wanted rhs_expr wanteds
2034 = addAvailAndSCs want_scs avails wanted avail
2036 avail = Rhs rhs_expr wanteds
2038 addGiven :: Avails -> Inst -> TcM Avails
2039 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2040 -- Always add superclasses for 'givens'
2042 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2043 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2044 -- so the assert isn't true
2046 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2047 addRefinedGiven reft (refined_givens, avails) given
2048 | isDict given -- We sometimes have 'given' methods, but they
2049 -- are always optional, so we can drop them
2050 , Just (co, pred) <- refinePred reft (dictPred given)
2051 = do { new_given <- newDictBndr (instLoc given) pred
2052 ; let rhs = L (instSpan given) $
2053 HsWrap (WpCo co) (HsVar (instToId given))
2054 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2055 ; return (new_given:refined_givens, avails) }
2056 -- ToDo: the superclasses of the original given all exist in Avails
2057 -- so we could really just cast them, but it's more awkward to do,
2058 -- and hopefully the optimiser will spot the duplicated work
2060 = return (refined_givens, avails)
2062 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2063 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2064 addAvailAndSCs want_scs avails irred IsIrred
2066 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2067 addAvailAndSCs want_scs avails inst avail
2068 | not (isClassDict inst) = extendAvails avails inst avail
2069 | NoSCs <- want_scs = extendAvails avails inst avail
2070 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2071 ; avails' <- extendAvails avails inst avail
2072 ; addSCs is_loop avails' inst }
2074 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2075 -- Note: this compares by *type*, not by Unique
2076 deps = findAllDeps (unitVarSet (instToId inst)) avail
2077 dep_tys = map idType (varSetElems deps)
2079 findAllDeps :: IdSet -> AvailHow -> IdSet
2080 -- Find all the Insts that this one depends on
2081 -- See Note [SUPERCLASS-LOOP 2]
2082 -- Watch out, though. Since the avails may contain loops
2083 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2084 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2085 findAllDeps so_far other = so_far
2087 find_all :: IdSet -> Inst -> IdSet
2089 | kid_id `elemVarSet` so_far = so_far
2090 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2091 | otherwise = so_far'
2093 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2094 kid_id = instToId kid
2096 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2097 -- Add all the superclasses of the Inst to Avails
2098 -- The first param says "dont do this because the original thing
2099 -- depends on this one, so you'd build a loop"
2100 -- Invariant: the Inst is already in Avails.
2102 addSCs is_loop avails dict
2103 = ASSERT( isDict dict )
2104 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2105 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2107 (clas, tys) = getDictClassTys dict
2108 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2109 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2111 add_sc avails (sc_dict, sc_sel)
2112 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2113 | is_given sc_dict = return avails
2114 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2115 ; addSCs is_loop avails' sc_dict }
2117 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2118 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2120 is_given :: Inst -> Bool
2121 is_given sc_dict = case findAvail avails sc_dict of
2122 Just (Given _) -> True -- Given is cheaper than superclass selection
2126 %************************************************************************
2128 \section{tcSimplifyTop: defaulting}
2130 %************************************************************************
2133 @tcSimplifyTop@ is called once per module to simplify all the constant
2134 and ambiguous Insts.
2136 We need to be careful of one case. Suppose we have
2138 instance Num a => Num (Foo a b) where ...
2140 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2141 to (Num x), and default x to Int. But what about y??
2143 It's OK: the final zonking stage should zap y to (), which is fine.
2147 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2148 tcSimplifyTop wanteds
2149 = tc_simplify_top doc False wanteds
2151 doc = text "tcSimplifyTop"
2153 tcSimplifyInteractive wanteds
2154 = tc_simplify_top doc True wanteds
2156 doc = text "tcSimplifyInteractive"
2158 -- The TcLclEnv should be valid here, solely to improve
2159 -- error message generation for the monomorphism restriction
2160 tc_simplify_top doc interactive wanteds
2161 = do { wanteds <- mapM zonkInst wanteds
2162 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2164 ; (binds1, irreds1) <- topCheckLoop doc wanteds
2166 ; if null irreds1 then
2169 -- OK, so there are some errors
2170 { -- Use the defaulting rules to do extra unification
2171 -- NB: irreds are already zonked
2172 ; extended_default <- if interactive then return True
2173 else doptM Opt_ExtendedDefaultRules
2174 ; disambiguate extended_default irreds1 -- Does unification
2175 ; (binds2, irreds2) <- topCheckLoop doc irreds1
2177 -- Deal with implicit parameter
2178 ; let (bad_ips, non_ips) = partition isIPDict irreds2
2179 (ambigs, others) = partition isTyVarDict non_ips
2181 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2183 ; addNoInstanceErrs others
2184 ; addTopAmbigErrs ambigs
2186 ; return (binds1 `unionBags` binds2) }}
2189 If a dictionary constrains a type variable which is
2190 * not mentioned in the environment
2191 * and not mentioned in the type of the expression
2192 then it is ambiguous. No further information will arise to instantiate
2193 the type variable; nor will it be generalised and turned into an extra
2194 parameter to a function.
2196 It is an error for this to occur, except that Haskell provided for
2197 certain rules to be applied in the special case of numeric types.
2199 * at least one of its classes is a numeric class, and
2200 * all of its classes are numeric or standard
2201 then the type variable can be defaulted to the first type in the
2202 default-type list which is an instance of all the offending classes.
2204 So here is the function which does the work. It takes the ambiguous
2205 dictionaries and either resolves them (producing bindings) or
2206 complains. It works by splitting the dictionary list by type
2207 variable, and using @disambigOne@ to do the real business.
2209 @disambigOne@ assumes that its arguments dictionaries constrain all
2210 the same type variable.
2212 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2213 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2214 the most common use of defaulting is code like:
2216 _ccall_ foo `seqPrimIO` bar
2218 Since we're not using the result of @foo@, the result if (presumably)
2222 disambiguate :: Bool -> [Inst] -> TcM ()
2223 -- Just does unification to fix the default types
2224 -- The Insts are assumed to be pre-zonked
2225 disambiguate extended_defaulting insts
2226 | null defaultable_groups
2229 = do { -- Figure out what default types to use
2230 mb_defaults <- getDefaultTys
2231 ; default_tys <- case mb_defaults of
2232 Just tys -> return tys
2233 Nothing -> -- No use-supplied default;
2234 -- use [Integer, Double]
2235 do { integer_ty <- tcMetaTy integerTyConName
2236 ; checkWiredInTyCon doubleTyCon
2237 ; return [integer_ty, doubleTy] }
2238 ; mapM_ (disambigGroup default_tys) defaultable_groups }
2240 unaries :: [(Inst,Class, TcTyVar)] -- (C tv) constraints
2241 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2242 (unaries, bad_tvs) = getDefaultableDicts insts
2244 -- Group by type variable
2245 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2246 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2247 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2249 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2250 defaultable_group ds@((_,_,tv):_)
2251 = not (isSkolemTyVar tv) -- Note [Avoiding spurious errors]
2252 && not (tv `elemVarSet` bad_tvs)
2253 && defaultable_classes [c | (_,c,_) <- ds]
2254 defaultable_group [] = panic "defaultable_group"
2256 defaultable_classes clss
2257 | extended_defaulting = any isInteractiveClass clss
2258 | otherwise = all isStandardClass clss && any isNumericClass clss
2260 -- In interactive mode, or with -fextended-default-rules,
2261 -- we default Show a to Show () to avoid graututious errors on "show []"
2262 isInteractiveClass cls
2263 = isNumericClass cls
2264 || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2267 disambigGroup :: [Type] -- The default types
2268 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2269 -> TcM () -- Just does unification, to fix the default types
2271 disambigGroup default_tys dicts
2272 = try_default default_tys
2274 (_,_,tyvar) = head dicts -- Should be non-empty
2275 classes = [c | (_,c,_) <- dicts]
2277 try_default [] = return ()
2278 try_default (default_ty : default_tys)
2279 = tryTcLIE_ (try_default default_tys) $
2280 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2281 -- This may fail; then the tryTcLIE_ kicks in
2282 -- Failure here is caused by there being no type in the
2283 -- default list which can satisfy all the ambiguous classes.
2284 -- For example, if Real a is reqd, but the only type in the
2285 -- default list is Int.
2287 -- After this we can't fail
2288 ; warnDefault dicts default_ty
2289 ; unifyType default_ty (mkTyVarTy tyvar) }
2292 Note [Avoiding spurious errors]
2293 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2294 When doing the unification for defaulting, we check for skolem
2295 type variables, and simply don't default them. For example:
2296 f = (*) -- Monomorphic
2297 g :: Num a => a -> a
2299 Here, we get a complaint when checking the type signature for g,
2300 that g isn't polymorphic enough; but then we get another one when
2301 dealing with the (Num a) context arising from f's definition;
2302 we try to unify a with Int (to default it), but find that it's
2303 already been unified with the rigid variable from g's type sig
2306 %************************************************************************
2308 \subsection[simple]{@Simple@ versions}
2310 %************************************************************************
2312 Much simpler versions when there are no bindings to make!
2314 @tcSimplifyThetas@ simplifies class-type constraints formed by
2315 @deriving@ declarations and when specialising instances. We are
2316 only interested in the simplified bunch of class/type constraints.
2318 It simplifies to constraints of the form (C a b c) where
2319 a,b,c are type variables. This is required for the context of
2320 instance declarations.
2323 tcSimplifyDeriv :: InstOrigin
2326 -> ThetaType -- Wanted
2327 -> TcM ThetaType -- Needed
2329 tcSimplifyDeriv orig tc tyvars theta
2330 = tcInstTyVars tyvars `thenM` \ (tvs, _, tenv) ->
2331 -- The main loop may do unification, and that may crash if
2332 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2333 -- ToDo: what if two of them do get unified?
2334 newDictBndrsO orig (substTheta tenv theta) `thenM` \ wanteds ->
2335 topCheckLoop doc wanteds `thenM` \ (_, irreds) ->
2337 doptM Opt_GlasgowExts `thenM` \ gla_exts ->
2338 doptM Opt_AllowUndecidableInstances `thenM` \ undecidable_ok ->
2340 inst_ty = mkTyConApp tc (mkTyVarTys tvs)
2341 (ok_insts, bad_insts) = partition is_ok_inst irreds
2343 = isDict inst -- Exclude implication consraints
2344 && (isTyVarClassPred pred || (gla_exts && ok_gla_pred pred))
2346 pred = dictPred inst
2348 ok_gla_pred pred = null (checkInstTermination [inst_ty] [pred])
2349 -- See Note [Deriving context]
2351 tv_set = mkVarSet tvs
2352 simpl_theta = map dictPred ok_insts
2353 weird_preds = [pred | pred <- simpl_theta
2354 , not (tyVarsOfPred pred `subVarSet` tv_set)]
2356 -- Check for a bizarre corner case, when the derived instance decl should
2357 -- have form instance C a b => D (T a) where ...
2358 -- Note that 'b' isn't a parameter of T. This gives rise to all sorts
2359 -- of problems; in particular, it's hard to compare solutions for
2360 -- equality when finding the fixpoint. So I just rule it out for now.
2362 rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2363 -- This reverse-mapping is a Royal Pain,
2364 -- but the result should mention TyVars not TcTyVars
2366 -- In effect, the bad and wierd insts cover all of the cases that
2367 -- would make checkValidInstance fail; if it were called right after tcSimplifyDeriv
2368 -- * wierd_preds ensures unambiguous instances (checkAmbiguity in checkValidInstance)
2369 -- * ok_gla_pred ensures termination (checkInstTermination in checkValidInstance)
2370 addNoInstanceErrs bad_insts `thenM_`
2371 mapM_ (addErrTc . badDerivedPred) weird_preds `thenM_`
2372 returnM (substTheta rev_env simpl_theta)
2374 doc = ptext SLIT("deriving classes for a data type")
2377 Note [Deriving context]
2378 ~~~~~~~~~~~~~~~~~~~~~~~
2379 With -fglasgow-exts, we allow things like (C Int a) in the simplified
2380 context for a derived instance declaration, because at a use of this
2381 instance, we might know that a=Bool, and have an instance for (C Int
2384 We nevertheless insist that each predicate meets the termination
2385 conditions. If not, the deriving mechanism generates larger and larger
2386 constraints. Example:
2388 data Seq a = Cons a (Seq (Succ a)) | Nil deriving Show
2390 Note the lack of a Show instance for Succ. First we'll generate
2391 instance (Show (Succ a), Show a) => Show (Seq a)
2393 instance (Show (Succ (Succ a)), Show (Succ a), Show a) => Show (Seq a)
2394 and so on. Instead we want to complain of no instance for (Show (Succ a)).
2398 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2399 used with \tr{default} declarations. We are only interested in
2400 whether it worked or not.
2403 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2406 tcSimplifyDefault theta
2407 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2408 topCheckLoop doc wanteds `thenM` \ (_, irreds) ->
2409 addNoInstanceErrs irreds `thenM_`
2415 doc = ptext SLIT("default declaration")
2419 %************************************************************************
2421 \section{Errors and contexts}
2423 %************************************************************************
2425 ToDo: for these error messages, should we note the location as coming
2426 from the insts, or just whatever seems to be around in the monad just
2430 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2431 -> [Inst] -- The offending Insts
2433 -- Group together insts with the same origin
2434 -- We want to report them together in error messages
2436 groupErrs report_err []
2438 groupErrs report_err (inst:insts)
2439 = do_one (inst:friends) `thenM_`
2440 groupErrs report_err others
2443 -- (It may seem a bit crude to compare the error messages,
2444 -- but it makes sure that we combine just what the user sees,
2445 -- and it avoids need equality on InstLocs.)
2446 (friends, others) = partition is_friend insts
2447 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2448 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2449 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2450 -- Add location and context information derived from the Insts
2452 -- Add the "arising from..." part to a message about bunch of dicts
2453 addInstLoc :: [Inst] -> Message -> Message
2454 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2456 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2457 addTopIPErrs bndrs []
2459 addTopIPErrs bndrs ips
2460 = addErrTcM (tidy_env, mk_msg tidy_ips)
2462 (tidy_env, tidy_ips) = tidyInsts ips
2463 mk_msg ips = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2464 nest 2 (ptext SLIT("the monomorphic top-level binding(s) of")
2465 <+> pprBinders bndrs <> colon)],
2466 nest 2 (vcat (map ppr_ip ips)),
2468 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2470 topIPErrs :: [Inst] -> TcM ()
2472 = groupErrs report tidy_dicts
2474 (tidy_env, tidy_dicts) = tidyInsts dicts
2475 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2476 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2477 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2479 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2481 addNoInstanceErrs insts
2482 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2483 ; reportNoInstances tidy_env Nothing tidy_insts }
2487 -> Maybe (InstLoc, [Inst]) -- Context
2488 -- Nothing => top level
2489 -- Just (d,g) => d describes the construct
2491 -> [Inst] -- What is wanted (can include implications)
2494 reportNoInstances tidy_env mb_what insts
2495 = groupErrs (report_no_instances tidy_env mb_what) insts
2497 report_no_instances tidy_env mb_what insts
2498 = do { inst_envs <- tcGetInstEnvs
2499 ; let (implics, insts1) = partition isImplicInst insts
2500 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2501 ; traceTc (text "reportNoInstnces" <+> vcat
2502 [ppr implics, ppr insts1, ppr insts2])
2503 ; mapM_ complain_implic implics
2504 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2505 ; groupErrs complain_no_inst insts2 }
2507 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2509 complain_implic inst -- Recurse!
2510 = reportNoInstances tidy_env
2511 (Just (tci_loc inst, tci_given inst))
2514 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2515 -- Right msg => overlap message
2516 -- Left inst => no instance
2517 check_overlap inst_envs wanted
2518 | not (isClassDict wanted) = Left wanted
2520 = case lookupInstEnv inst_envs clas tys of
2521 -- The case of exactly one match and no unifiers means
2522 -- a successful lookup. That can't happen here, becuase
2523 -- dicts only end up here if they didn't match in Inst.lookupInst
2525 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2527 ([], _) -> Left wanted -- No match
2528 res -> Right (mk_overlap_msg wanted res)
2530 (clas,tys) = getDictClassTys wanted
2532 mk_overlap_msg dict (matches, unifiers)
2533 = vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2534 <+> pprPred (dictPred dict))),
2535 sep [ptext SLIT("Matching instances") <> colon,
2536 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2537 ASSERT( not (null matches) )
2538 if not (isSingleton matches)
2539 then -- Two or more matches
2541 else -- One match, plus some unifiers
2542 ASSERT( not (null unifiers) )
2543 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2544 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2545 ptext SLIT("Use -fallow-incoherent-instances to use the first choice above")])]
2547 ispecs = [ispec | (_, ispec) <- matches]
2549 mk_no_inst_err insts
2550 | null insts = empty
2552 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2553 not (isEmptyVarSet (tyVarsOfInsts insts))
2554 = vcat [ addInstLoc insts $
2555 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2556 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2557 , show_fixes (fix1 loc : fixes2) ]
2559 | otherwise -- Top level
2560 = vcat [ addInstLoc insts $
2561 ptext SLIT("No instance") <> plural insts
2562 <+> ptext SLIT("for") <+> pprDictsTheta insts
2563 , show_fixes fixes2 ]
2566 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2567 <+> ptext SLIT("to the context of"),
2568 nest 2 (ppr (instLocOrigin loc)) ]
2569 -- I'm not sure it helps to add the location
2570 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2572 fixes2 | null instance_dicts = []
2573 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2574 pprDictsTheta instance_dicts]]
2575 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2576 -- Insts for which it is worth suggesting an adding an instance declaration
2577 -- Exclude implicit parameters, and tyvar dicts
2579 show_fixes :: [SDoc] -> SDoc
2580 show_fixes [] = empty
2581 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2582 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2584 addTopAmbigErrs dicts
2585 -- Divide into groups that share a common set of ambiguous tyvars
2586 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2587 -- See Note [Avoiding spurious errors]
2588 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2590 (tidy_env, tidy_dicts) = tidyInsts dicts
2592 tvs_of :: Inst -> [TcTyVar]
2593 tvs_of d = varSetElems (tyVarsOfInst d)
2594 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2596 report :: [(Inst,[TcTyVar])] -> TcM ()
2597 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2598 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2599 setSrcSpan (instSpan inst) $
2600 -- the location of the first one will do for the err message
2601 addErrTcM (tidy_env, msg $$ mono_msg)
2603 dicts = map fst pairs
2604 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2605 pprQuotedList tvs <+> in_msg,
2606 nest 2 (pprDictsInFull dicts)]
2607 in_msg = text "in the constraint" <> plural dicts <> colon
2608 report [] = panic "addTopAmbigErrs"
2611 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2612 -- There's an error with these Insts; if they have free type variables
2613 -- it's probably caused by the monomorphism restriction.
2614 -- Try to identify the offending variable
2615 -- ASSUMPTION: the Insts are fully zonked
2616 mkMonomorphismMsg tidy_env inst_tvs
2617 = findGlobals (mkVarSet inst_tvs) tidy_env `thenM` \ (tidy_env, docs) ->
2618 returnM (tidy_env, mk_msg docs)
2620 mk_msg [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2621 -- This happens in things like
2622 -- f x = show (read "foo")
2623 -- where monomorphism doesn't play any role
2624 mk_msg docs = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2628 monomorphism_fix :: SDoc
2629 monomorphism_fix = ptext SLIT("Probable fix:") <+>
2630 (ptext SLIT("give these definition(s) an explicit type signature")
2631 $$ ptext SLIT("or use -fno-monomorphism-restriction"))
2633 warnDefault ups default_ty
2634 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2635 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2637 dicts = [d | (d,_,_) <- ups]
2640 (_, tidy_dicts) = tidyInsts dicts
2641 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2642 quotes (ppr default_ty),
2643 pprDictsInFull tidy_dicts]
2645 -- Used for the ...Thetas variants; all top level
2647 = vcat [ptext SLIT("Can't derive instances where the instance context mentions"),
2648 ptext SLIT("type variables that are not data type parameters"),
2649 nest 2 (ptext SLIT("Offending constraint:") <+> ppr pred)]
2651 reduceDepthErr n stack
2652 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2653 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2654 nest 4 (pprStack stack)]
2656 pprStack stack = vcat (map pprInstInFull stack)