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 = do { let try_me inst | isDict inst = Stop -- Dicts
655 | otherwise = ReduceMe NoSCs -- Lits, Methods,
656 -- and impliciation constraints
657 -- In an effort to make the inferred types simple, we try
658 -- to squeeze out implication constraints if we can.
659 -- See Note [Squashing methods]
661 ; (binds1, irreds) <- checkLoop (mkRedEnv doc try_me []) wanted_lie
663 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
664 ; gbl_tvs <- tcGetGlobalTyVars
665 ; let preds = fdPredsOfInsts irreds
666 qtvs = grow preds tau_tvs' `minusVarSet` oclose preds gbl_tvs
667 (free, bound) = partition (isFreeWhenInferring qtvs) irreds
669 -- Remove redundant superclasses from 'bound'
670 -- The 'Stop' try_me result does not do so,
671 -- see Note [No superclasses for Stop]
672 ; let try_me inst = ReduceMe AddSCs
673 ; (binds2, irreds) <- checkLoop (mkRedEnv doc try_me []) bound
676 ; return (varSetElems qtvs, binds1 `unionBags` binds2, map instToId irreds) }
677 -- NB: when we are done, we might have some bindings, but
678 -- the final qtvs might be empty. See Note [NO TYVARS] below.
681 Note [Squashing methods]
682 ~~~~~~~~~~~~~~~~~~~~~~~~~
683 Be careful if you want to float methods more:
684 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
685 From an application (truncate f i) we get
688 If we have also have a second occurrence of truncate, we get
691 When simplifying with i,f free, we might still notice that
692 t1=t3; but alas, the binding for t2 (which mentions t1)
693 may continue to float out!
698 class Y a b | a -> b where
701 instance Y [[a]] a where
704 k :: X a -> X a -> X a
706 g :: Num a => [X a] -> [X a]
709 h ys = ys ++ map (k (y [[0]])) xs
711 The excitement comes when simplifying the bindings for h. Initially
712 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
713 From this we get t1:=:t2, but also various bindings. We can't forget
714 the bindings (because of [LOOP]), but in fact t1 is what g is
717 The net effect of [NO TYVARS]
720 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
721 isFreeWhenInferring qtvs inst
722 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
723 && isInheritableInst inst -- And no implicit parameter involved
724 -- (see "Notes on implicit parameters")
726 {- No longer used (with implication constraints)
727 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
728 -> NameSet -- Quantified implicit parameters
730 isFreeWhenChecking qtvs ips inst
731 = isFreeWrtTyVars qtvs inst
732 && isFreeWrtIPs ips inst
735 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
736 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
740 %************************************************************************
742 \subsection{tcSimplifyCheck}
744 %************************************************************************
746 @tcSimplifyCheck@ is used when we know exactly the set of variables
747 we are going to quantify over. For example, a class or instance declaration.
750 -----------------------------------------------------------
751 -- tcSimplifyCheck is used when checking expression type signatures,
752 -- class decls, instance decls etc.
753 tcSimplifyCheck :: InstLoc
754 -> [TcTyVar] -- Quantify over these
757 -> TcM TcDictBinds -- Bindings
758 tcSimplifyCheck loc qtvs givens wanteds
759 = ASSERT( all isSkolemTyVar qtvs )
760 do { (binds, irreds) <- innerCheckLoop loc givens wanteds
761 ; implic_bind <- bindIrreds loc [] emptyRefinement
763 ; return (binds `unionBags` implic_bind) }
765 -----------------------------------------------------------
766 -- tcSimplifyCheckPat is used for existential pattern match
767 tcSimplifyCheckPat :: InstLoc
768 -> [CoVar] -> Refinement
769 -> [TcTyVar] -- Quantify over these
772 -> TcM TcDictBinds -- Bindings
773 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
774 = ASSERT( all isSkolemTyVar qtvs )
775 do { (binds, irreds) <- innerCheckLoop loc givens wanteds
776 ; implic_bind <- bindIrreds loc co_vars reft
778 ; return (binds `unionBags` implic_bind) }
780 -----------------------------------------------------------
781 bindIrreds :: InstLoc -> [CoVar] -> Refinement
782 -> [TcTyVar] -> [Inst] -> [Inst]
784 -- Make a binding that binds 'irreds', by generating an implication
785 -- constraint for them, *and* throwing the constraint into the LIE
786 bindIrreds loc co_vars reft qtvs givens irreds
787 = do { let givens' = filter isDict givens
788 -- The givens can include methods
790 -- If there are no 'givens', then it's safe to
791 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
792 -- See Note [Freeness and implications]
793 ; irreds' <- if null givens'
795 { let qtv_set = mkVarSet qtvs
796 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
798 ; return real_irreds }
801 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
802 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
803 -- This call does the real work
808 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
810 -> TcM ([Inst], TcDictBinds)
811 -- Make a binding that binds 'irreds', by generating an implication
812 -- constraint for them, *and* throwing the constraint into the LIE
813 -- The binding looks like
814 -- (ir1, .., irn) = f qtvs givens
815 -- where f is (evidence for) the new implication constraint
817 -- This binding must line up the 'rhs' in reduceImplication
818 makeImplicationBind loc all_tvs reft
819 givens -- Guaranteed all Dicts
821 | null irreds -- If there are no irreds, we are done
822 = return ([], emptyBag)
823 | otherwise -- Otherwise we must generate a binding
824 = do { uniq <- newUnique
825 ; span <- getSrcSpanM
826 ; let name = mkInternalName uniq (mkVarOcc "ic") (srcSpanStart span)
827 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
828 tci_tyvars = all_tvs,
830 tci_wanted = irreds, tci_loc = loc }
832 ; let n_irreds = length irreds
833 irred_ids = map instToId irreds
834 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
835 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
836 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
837 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
838 bind | n_irreds==1 = VarBind (head irred_ids) rhs
839 | otherwise = PatBind { pat_lhs = L span pat,
840 pat_rhs = unguardedGRHSs rhs,
842 bind_fvs = placeHolderNames }
843 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
844 return ([implic_inst], unitBag (L span bind)) }
846 -----------------------------------------------------------
850 [Inst]) -- Irreducible
852 topCheckLoop doc wanteds
853 = checkLoop (mkRedEnv doc try_me []) wanteds
855 try_me inst = ReduceMe AddSCs
857 -----------------------------------------------------------
858 innerCheckLoop :: InstLoc
862 [Inst]) -- Irreducible
864 innerCheckLoop inst_loc givens wanteds
865 = checkLoop env wanteds
867 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
869 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
871 -- When checking against a given signature
872 -- we MUST be very gentle: Note [Check gently]
877 We have to very careful about not simplifying too vigorously
882 f :: Show b => T b -> b
885 Inside the pattern match, which binds (a:*, x:a), we know that
887 Hence we have a dictionary for Show [a] available; and indeed we
888 need it. We are going to build an implication contraint
889 forall a. (b~[a]) => Show [a]
890 Later, we will solve this constraint using the knowledge (Show b)
892 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
893 thing becomes insoluble. So we simplify gently (get rid of literals
894 and methods only, plus common up equal things), deferring the real
895 work until top level, when we solve the implication constraint
900 -----------------------------------------------------------
904 [Inst]) -- Irreducible
905 -- Precondition: the try_me never returns Free
906 -- givens are completely rigid
908 checkLoop env wanteds
909 = do { -- Givens are skolems, so no need to zonk them
910 wanteds' <- mappM zonkInst wanteds
912 ; (improved, binds, irreds) <- reduceContext env wanteds'
914 ; if not improved then
915 return (binds, irreds)
918 -- If improvement did some unification, we go round again.
919 -- We start again with irreds, not wanteds
920 -- Using an instance decl might have introduced a fresh type variable
921 -- which might have been unified, so we'd get an infinite loop
922 -- if we started again with wanteds! See Note [LOOP]
923 { (binds1, irreds1) <- checkLoop env irreds
924 ; return (binds `unionBags` binds1, irreds1) } }
929 class If b t e r | b t e -> r
932 class Lte a b c | a b -> c where lte :: a -> b -> c
934 instance (Lte a b l,If l b a c) => Max a b c
936 Wanted: Max Z (S x) y
938 Then we'll reduce using the Max instance to:
939 (Lte Z (S x) l, If l (S x) Z y)
940 and improve by binding l->T, after which we can do some reduction
941 on both the Lte and If constraints. What we *can't* do is start again
942 with (Max Z (S x) y)!
946 -----------------------------------------------------------
947 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
948 -- against, but we don't know the type variables over which we are going to quantify.
949 -- This happens when we have a type signature for a mutually recursive group
952 -> TcTyVarSet -- fv(T)
955 -> TcM ([TcTyVar], -- Variables over which to quantify
956 TcDictBinds) -- Bindings
958 tcSimplifyInferCheck loc tau_tvs givens wanteds
959 = do { (binds, irreds) <- innerCheckLoop loc givens wanteds
961 -- Figure out which type variables to quantify over
962 -- You might think it should just be the signature tyvars,
963 -- but in bizarre cases you can get extra ones
964 -- f :: forall a. Num a => a -> a
965 -- f x = fst (g (x, head [])) + 1
967 -- Here we infer g :: forall a b. a -> b -> (b,a)
968 -- We don't want g to be monomorphic in b just because
969 -- f isn't quantified over b.
970 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
971 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
972 ; gbl_tvs <- tcGetGlobalTyVars
973 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
974 -- We could close gbl_tvs, but its not necessary for
975 -- soundness, and it'll only affect which tyvars, not which
976 -- dictionaries, we quantify over
978 -- Now we are back to normal (c.f. tcSimplCheck)
979 ; implic_bind <- bindIrreds loc [] emptyRefinement
981 ; return (qtvs, binds `unionBags` implic_bind) }
985 %************************************************************************
987 tcSimplifySuperClasses
989 %************************************************************************
991 Note [SUPERCLASS-LOOP 1]
992 ~~~~~~~~~~~~~~~~~~~~~~~~
993 We have to be very, very careful when generating superclasses, lest we
994 accidentally build a loop. Here's an example:
998 class S a => C a where { opc :: a -> a }
999 class S b => D b where { opd :: b -> b }
1001 instance C Int where
1004 instance D Int where
1007 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1008 Simplifying, we may well get:
1009 $dfCInt = :C ds1 (opd dd)
1012 Notice that we spot that we can extract ds1 from dd.
1014 Alas! Alack! We can do the same for (instance D Int):
1016 $dfDInt = :D ds2 (opc dc)
1020 And now we've defined the superclass in terms of itself.
1022 Solution: never generate a superclass selectors at all when
1023 satisfying the superclass context of an instance declaration.
1025 Two more nasty cases are in
1030 tcSimplifySuperClasses
1035 tcSimplifySuperClasses loc givens sc_wanteds
1036 = do { (binds1, irreds) <- checkLoop env sc_wanteds
1037 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1038 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1041 env = mkRedEnv (pprInstLoc loc) try_me givens
1042 try_me inst = ReduceMe NoSCs
1043 -- Like topCheckLoop, but with NoSCs
1047 %************************************************************************
1049 \subsection{tcSimplifyRestricted}
1051 %************************************************************************
1053 tcSimplifyRestricted infers which type variables to quantify for a
1054 group of restricted bindings. This isn't trivial.
1057 We want to quantify over a to get id :: forall a. a->a
1060 We do not want to quantify over a, because there's an Eq a
1061 constraint, so we get eq :: a->a->Bool (notice no forall)
1064 RHS has type 'tau', whose free tyvars are tau_tvs
1065 RHS has constraints 'wanteds'
1068 Quantify over (tau_tvs \ ftvs(wanteds))
1069 This is bad. The constraints may contain (Monad (ST s))
1070 where we have instance Monad (ST s) where...
1071 so there's no need to be monomorphic in s!
1073 Also the constraint might be a method constraint,
1074 whose type mentions a perfectly innocent tyvar:
1075 op :: Num a => a -> b -> a
1076 Here, b is unconstrained. A good example would be
1078 We want to infer the polymorphic type
1079 foo :: forall b. b -> b
1082 Plan B (cunning, used for a long time up to and including GHC 6.2)
1083 Step 1: Simplify the constraints as much as possible (to deal
1084 with Plan A's problem). Then set
1085 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1087 Step 2: Now simplify again, treating the constraint as 'free' if
1088 it does not mention qtvs, and trying to reduce it otherwise.
1089 The reasons for this is to maximise sharing.
1091 This fails for a very subtle reason. Suppose that in the Step 2
1092 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1093 In the Step 1 this constraint might have been simplified, perhaps to
1094 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1095 This won't happen in Step 2... but that in turn might prevent some other
1096 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1097 and that in turn breaks the invariant that no constraints are quantified over.
1099 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1104 Step 1: Simplify the constraints as much as possible (to deal
1105 with Plan A's problem). Then set
1106 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1107 Return the bindings from Step 1.
1110 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1113 instance (HasBinary ty IO) => HasCodedValue ty
1115 foo :: HasCodedValue a => String -> IO a
1117 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1118 doDecodeIO codedValue view
1119 = let { act = foo "foo" } in act
1121 You might think this should work becuase the call to foo gives rise to a constraint
1122 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1123 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1124 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1126 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1130 Plan D (a variant of plan B)
1131 Step 1: Simplify the constraints as much as possible (to deal
1132 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1133 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1135 Step 2: Now simplify again, treating the constraint as 'free' if
1136 it does not mention qtvs, and trying to reduce it otherwise.
1138 The point here is that it's generally OK to have too few qtvs; that is,
1139 to make the thing more monomorphic than it could be. We don't want to
1140 do that in the common cases, but in wierd cases it's ok: the programmer
1141 can always add a signature.
1143 Too few qtvs => too many wanteds, which is what happens if you do less
1148 tcSimplifyRestricted -- Used for restricted binding groups
1149 -- i.e. ones subject to the monomorphism restriction
1152 -> [Name] -- Things bound in this group
1153 -> TcTyVarSet -- Free in the type of the RHSs
1154 -> [Inst] -- Free in the RHSs
1155 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked)
1156 TcDictBinds) -- Bindings
1157 -- tcSimpifyRestricted returns no constraints to
1158 -- quantify over; by definition there are none.
1159 -- They are all thrown back in the LIE
1161 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1162 -- Zonk everything in sight
1163 = mappM zonkInst wanteds `thenM` \ wanteds' ->
1165 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1166 -- dicts; the idea is to get rid of as many type
1167 -- variables as possible, and we don't want to stop
1168 -- at (say) Monad (ST s), because that reduces
1169 -- immediately, with no constraint on s.
1171 -- BUT do no improvement! See Plan D above
1172 -- HOWEVER, some unification may take place, if we instantiate
1173 -- a method Inst with an equality constraint
1174 let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1176 reduceContext env wanteds' `thenM` \ (_imp, _binds, constrained_dicts) ->
1178 -- Next, figure out the tyvars we will quantify over
1179 zonkTcTyVarsAndFV (varSetElems tau_tvs) `thenM` \ tau_tvs' ->
1180 tcGetGlobalTyVars `thenM` \ gbl_tvs' ->
1181 mappM zonkInst constrained_dicts `thenM` \ constrained_dicts' ->
1183 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1184 qtvs = (tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs')
1185 `minusVarSet` constrained_tvs'
1187 traceTc (text "tcSimplifyRestricted" <+> vcat [
1188 pprInsts wanteds, pprInsts constrained_dicts',
1190 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ]) `thenM_`
1192 -- The first step may have squashed more methods than
1193 -- necessary, so try again, this time more gently, knowing the exact
1194 -- set of type variables to quantify over.
1196 -- We quantify only over constraints that are captured by qtvs;
1197 -- these will just be a subset of non-dicts. This in contrast
1198 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1199 -- all *non-inheritable* constraints too. This implements choice
1200 -- (B) under "implicit parameter and monomorphism" above.
1202 -- Remember that we may need to do *some* simplification, to
1203 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1204 -- just to float all constraints
1206 -- At top level, we *do* squash methods becuase we want to
1207 -- expose implicit parameters to the test that follows
1209 is_nested_group = isNotTopLevel top_lvl
1210 try_me inst | isFreeWrtTyVars qtvs inst,
1211 (is_nested_group || isDict inst) = Stop
1212 | otherwise = ReduceMe AddSCs
1213 env = mkNoImproveRedEnv doc try_me
1215 reduceContext env wanteds' `thenM` \ (_imp, binds, irreds) ->
1216 ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1218 -- See "Notes on implicit parameters, Question 4: top level"
1219 if is_nested_group then
1220 extendLIEs irreds `thenM_`
1221 returnM (varSetElems qtvs, binds)
1224 (non_ips, bad_ips) = partition isClassDict irreds
1226 addTopIPErrs bndrs bad_ips `thenM_`
1227 extendLIEs non_ips `thenM_`
1228 returnM (varSetElems qtvs, binds)
1232 %************************************************************************
1236 %************************************************************************
1238 On the LHS of transformation rules we only simplify methods and constants,
1239 getting dictionaries. We want to keep all of them unsimplified, to serve
1240 as the available stuff for the RHS of the rule.
1242 Example. Consider the following left-hand side of a rule
1244 f (x == y) (y > z) = ...
1246 If we typecheck this expression we get constraints
1248 d1 :: Ord a, d2 :: Eq a
1250 We do NOT want to "simplify" to the LHS
1252 forall x::a, y::a, z::a, d1::Ord a.
1253 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1257 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1258 f ((==) d2 x y) ((>) d1 y z) = ...
1260 Here is another example:
1262 fromIntegral :: (Integral a, Num b) => a -> b
1263 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1265 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1266 we *dont* want to get
1268 forall dIntegralInt.
1269 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1271 because the scsel will mess up RULE matching. Instead we want
1273 forall dIntegralInt, dNumInt.
1274 fromIntegral Int Int dIntegralInt dNumInt = id Int
1278 g (x == y) (y == z) = ..
1280 where the two dictionaries are *identical*, we do NOT WANT
1282 forall x::a, y::a, z::a, d1::Eq a
1283 f ((==) d1 x y) ((>) d1 y z) = ...
1285 because that will only match if the dict args are (visibly) equal.
1286 Instead we want to quantify over the dictionaries separately.
1288 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1289 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1290 from scratch, rather than further parameterise simpleReduceLoop etc
1293 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1294 tcSimplifyRuleLhs wanteds
1295 = go [] emptyBag wanteds
1298 = return (dicts, binds)
1299 go dicts binds (w:ws)
1301 = go (w:dicts) binds ws
1303 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1304 -- to fromInteger; this looks fragile to me
1305 ; lookup_result <- lookupSimpleInst w'
1306 ; case lookup_result of
1307 GenInst ws' rhs -> go dicts (addBind binds w rhs) (ws' ++ ws)
1308 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1312 tcSimplifyBracket is used when simplifying the constraints arising from
1313 a Template Haskell bracket [| ... |]. We want to check that there aren't
1314 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1315 Show instance), but we aren't otherwise interested in the results.
1316 Nor do we care about ambiguous dictionaries etc. We will type check
1317 this bracket again at its usage site.
1320 tcSimplifyBracket :: [Inst] -> TcM ()
1321 tcSimplifyBracket wanteds
1322 = do { topCheckLoop doc wanteds
1325 doc = text "tcSimplifyBracket"
1329 %************************************************************************
1331 \subsection{Filtering at a dynamic binding}
1333 %************************************************************************
1338 we must discharge all the ?x constraints from B. We also do an improvement
1339 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1341 Actually, the constraints from B might improve the types in ?x. For example
1343 f :: (?x::Int) => Char -> Char
1346 then the constraint (?x::Int) arising from the call to f will
1347 force the binding for ?x to be of type Int.
1350 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1353 -- We need a loop so that we do improvement, and then
1354 -- (next time round) generate a binding to connect the two
1356 -- Here the two ?x's have different types, and improvement
1357 -- makes them the same.
1359 tcSimplifyIPs given_ips wanteds
1360 = do { wanteds' <- mappM zonkInst wanteds
1361 ; given_ips' <- mappM zonkInst given_ips
1362 -- Unusually for checking, we *must* zonk the given_ips
1364 ; let env = mkRedEnv doc try_me given_ips'
1365 ; (improved, binds, irreds) <- reduceContext env wanteds'
1367 ; if not improved then
1368 ASSERT( all is_free irreds )
1369 do { extendLIEs irreds
1372 tcSimplifyIPs given_ips wanteds }
1374 doc = text "tcSimplifyIPs" <+> ppr given_ips
1375 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1376 is_free inst = isFreeWrtIPs ip_set inst
1378 -- Simplify any methods that mention the implicit parameter
1379 try_me inst | is_free inst = Stop
1380 | otherwise = ReduceMe NoSCs
1384 %************************************************************************
1386 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1388 %************************************************************************
1390 When doing a binding group, we may have @Insts@ of local functions.
1391 For example, we might have...
1393 let f x = x + 1 -- orig local function (overloaded)
1394 f.1 = f Int -- two instances of f
1399 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1400 where @f@ is in scope; those @Insts@ must certainly not be passed
1401 upwards towards the top-level. If the @Insts@ were binding-ified up
1402 there, they would have unresolvable references to @f@.
1404 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1405 For each method @Inst@ in the @init_lie@ that mentions one of the
1406 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1407 @LIE@), as well as the @HsBinds@ generated.
1410 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1411 -- Simlifies only MethodInsts, and generate only bindings of form
1413 -- We're careful not to even generate bindings of the form
1415 -- You'd think that'd be fine, but it interacts with what is
1416 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1418 bindInstsOfLocalFuns wanteds local_ids
1419 | null overloaded_ids
1421 = extendLIEs wanteds `thenM_`
1422 returnM emptyLHsBinds
1425 = do { (binds, irreds) <- checkLoop env for_me
1426 ; extendLIEs not_for_me
1430 env = mkRedEnv doc try_me []
1431 doc = text "bindInsts" <+> ppr local_ids
1432 overloaded_ids = filter is_overloaded local_ids
1433 is_overloaded id = isOverloadedTy (idType id)
1434 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1436 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1437 -- so it's worth building a set, so that
1438 -- lookup (in isMethodFor) is faster
1439 try_me inst | isMethod inst = ReduceMe NoSCs
1444 %************************************************************************
1446 \subsection{Data types for the reduction mechanism}
1448 %************************************************************************
1450 The main control over context reduction is here
1454 = RedEnv { red_doc :: SDoc -- The context
1455 , red_try_me :: Inst -> WhatToDo
1456 , red_improve :: Bool -- True <=> do improvement
1457 , red_givens :: [Inst] -- All guaranteed rigid
1459 -- but see Note [Rigidity]
1460 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1461 -- See Note [RedStack]
1465 -- The red_givens are rigid so far as cmpInst is concerned.
1466 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1467 -- let ?x = e in ...
1468 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1469 -- But that doesn't affect the comparison, which is based only on mame.
1472 -- The red_stack pair (n,insts) pair is just used for error reporting.
1473 -- 'n' is always the depth of the stack.
1474 -- The 'insts' is the stack of Insts being reduced: to produce X
1475 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1478 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1479 mkRedEnv doc try_me givens
1480 = RedEnv { red_doc = doc, red_try_me = try_me,
1481 red_givens = givens, red_stack = (0,[]),
1482 red_improve = True }
1484 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1485 -- Do not do improvement; no givens
1486 mkNoImproveRedEnv doc try_me
1487 = RedEnv { red_doc = doc, red_try_me = try_me,
1488 red_givens = [], red_stack = (0,[]),
1489 red_improve = True }
1492 = ReduceMe WantSCs -- Try to reduce this
1493 -- If there's no instance, add the inst to the
1494 -- irreductible ones, but don't produce an error
1495 -- message of any kind.
1496 -- It might be quite legitimate such as (Eq a)!
1498 | Stop -- Return as irreducible unless it can
1499 -- be reduced to a constant in one step
1500 -- Do not add superclasses; see
1502 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1503 -- of a predicate when adding it to the avails
1504 -- The reason for this flag is entirely the super-class loop problem
1505 -- Note [SUPER-CLASS LOOP 1]
1508 %************************************************************************
1510 \subsection[reduce]{@reduce@}
1512 %************************************************************************
1516 reduceContext :: RedEnv
1518 -> TcM (ImprovementDone,
1519 TcDictBinds, -- Dictionary bindings
1520 [Inst]) -- Irreducible
1522 reduceContext env wanteds
1523 = do { traceTc (text "reduceContext" <+> (vcat [
1524 text "----------------------",
1526 text "given" <+> ppr (red_givens env),
1527 text "wanted" <+> ppr wanteds,
1528 text "----------------------"
1531 -- Build the Avail mapping from "givens"
1532 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1535 ; avails <- reduceList env wanteds init_state
1537 ; let improved = availsImproved avails
1538 ; (binds, irreds) <- extractResults avails wanteds
1540 ; traceTc (text "reduceContext end" <+> (vcat [
1541 text "----------------------",
1543 text "given" <+> ppr (red_givens env),
1544 text "wanted" <+> ppr wanteds,
1546 text "avails" <+> pprAvails avails,
1547 text "improved =" <+> ppr improved,
1548 text "----------------------"
1551 ; return (improved, binds, irreds) }
1553 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1554 tcImproveOne avails inst
1555 | not (isDict inst) = return False
1557 = do { inst_envs <- tcGetInstEnvs
1558 ; let eqns = improveOne (classInstances inst_envs)
1559 (dictPred inst, pprInstArising inst)
1560 [ (dictPred p, pprInstArising p)
1561 | p <- availsInsts avails, isDict p ]
1562 -- Avails has all the superclasses etc (good)
1563 -- It also has all the intermediates of the deduction (good)
1564 -- It does not have duplicates (good)
1565 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1566 -- so that improve will see them separate
1567 ; traceTc (text "improveOne" <+> ppr inst)
1570 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1571 -> TcM ImprovementDone
1572 unifyEqns [] = return False
1574 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1578 unify ((qtvs, pairs), what1, what2)
1579 = addErrCtxtM (mkEqnMsg what1 what2) $
1580 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1581 mapM_ (unif_pr tenv) pairs
1582 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1584 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1586 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1587 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1588 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1589 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1590 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1591 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1592 ; return (tidy_env, msg) }
1595 The main context-reduction function is @reduce@. Here's its game plan.
1598 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1599 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1600 = do { dopts <- getDOpts
1603 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1604 2 (ifPprDebug (nest 2 (pprStack stk))))
1607 ; if n >= ctxtStkDepth dopts then
1608 failWithTc (reduceDepthErr n stk)
1612 go [] state = return state
1613 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1616 -- Base case: we're done!
1617 reduce env wanted avails
1618 -- It's the same as an existing inst, or a superclass thereof
1619 | Just avail <- findAvail avails wanted
1623 = case red_try_me env wanted of {
1624 ; Stop -> try_simple (addIrred NoSCs) -- See Note [No superclasses for Stop]
1626 ; ReduceMe want_scs -> -- It should be reduced
1627 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1628 case lookup_result of
1629 NoInstance -> -- No such instance!
1630 -- Add it and its superclasses
1631 addIrred want_scs avails wanted
1633 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1635 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1636 ; avails2 <- reduceList env wanteds' avails1
1637 ; addWanted want_scs avails2 wanted rhs wanteds' }
1638 -- Temporarily do addIrred *before* the reduceList,
1639 -- which has the effect of adding the thing we are trying
1640 -- to prove to the database before trying to prove the things it
1641 -- needs. See note [RECURSIVE DICTIONARIES]
1642 -- NB: we must not do an addWanted before, because that adds the
1643 -- superclasses too, and thaat can lead to a spurious loop; see
1644 -- the examples in [SUPERCLASS-LOOP]
1645 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1649 -- First, see if the inst can be reduced to a constant in one step
1650 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1651 -- Don't bother for implication constraints, which take real work
1652 try_simple do_this_otherwise
1653 = do { res <- lookupSimpleInst wanted
1655 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1656 other -> do_this_otherwise avails wanted }
1660 Note [SUPERCLASS-LOOP 2]
1661 ~~~~~~~~~~~~~~~~~~~~~~~~
1662 But the above isn't enough. Suppose we are *given* d1:Ord a,
1663 and want to deduce (d2:C [a]) where
1665 class Ord a => C a where
1666 instance Ord [a] => C [a] where ...
1668 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1669 superclasses of C [a] to avails. But we must not overwrite the binding
1670 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1673 Here's another variant, immortalised in tcrun020
1674 class Monad m => C1 m
1675 class C1 m => C2 m x
1676 instance C2 Maybe Bool
1677 For the instance decl we need to build (C1 Maybe), and it's no good if
1678 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1679 before we search for C1 Maybe.
1681 Here's another example
1682 class Eq b => Foo a b
1683 instance Eq a => Foo [a] a
1687 we'll first deduce that it holds (via the instance decl). We must not
1688 then overwrite the Eq t constraint with a superclass selection!
1690 At first I had a gross hack, whereby I simply did not add superclass constraints
1691 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1692 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1693 I found a very obscure program (now tcrun021) in which improvement meant the
1694 simplifier got two bites a the cherry... so something seemed to be an Stop
1695 first time, but reducible next time.
1697 Now we implement the Right Solution, which is to check for loops directly
1698 when adding superclasses. It's a bit like the occurs check in unification.
1701 Note [RECURSIVE DICTIONARIES]
1702 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1704 data D r = ZeroD | SuccD (r (D r));
1706 instance (Eq (r (D r))) => Eq (D r) where
1707 ZeroD == ZeroD = True
1708 (SuccD a) == (SuccD b) = a == b
1711 equalDC :: D [] -> D [] -> Bool;
1714 We need to prove (Eq (D [])). Here's how we go:
1718 by instance decl, holds if
1722 by instance decl of Eq, holds if
1724 where d2 = dfEqList d3
1727 But now we can "tie the knot" to give
1733 and it'll even run! The trick is to put the thing we are trying to prove
1734 (in this case Eq (D []) into the database before trying to prove its
1735 contributing clauses.
1738 %************************************************************************
1740 Reducing a single constraint
1742 %************************************************************************
1745 ---------------------------------------------
1746 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1747 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1748 tci_given = extra_givens, tci_wanted = wanteds })
1749 = reduceImplication env avails reft tvs extra_givens wanteds loc
1751 reduceInst env avails other_inst
1752 = do { result <- lookupSimpleInst other_inst
1753 ; return (avails, result) }
1757 ---------------------------------------------
1758 reduceImplication :: RedEnv
1760 -> Refinement -- May refine the givens; often empty
1761 -> [TcTyVar] -- Quantified type variables; all skolems
1762 -> [Inst] -- Extra givens; all rigid
1765 -> TcM (Avails, LookupInstResult)
1768 Suppose we are simplifying the constraint
1769 forall bs. extras => wanted
1770 in the context of an overall simplification problem with givens 'givens',
1771 and refinment 'reft'.
1774 * The refinement is often empty
1776 * The 'extra givens' need not mention any of the quantified type variables
1777 e.g. forall {}. Eq a => Eq [a]
1778 forall {}. C Int => D (Tree Int)
1780 This happens when you have something like
1782 T1 :: Eq a => a -> T a
1785 f x = ...(case x of { T1 v -> v==v })...
1788 -- ToDo: should we instantiate tvs? I think it's not necessary
1790 -- ToDo: what about improvement? There may be some improvement
1791 -- exposed as a result of the simplifications done by reduceList
1792 -- which are discarded if we back off.
1793 -- This is almost certainly Wrong, but we'll fix it when dealing
1794 -- better with equality constraints
1795 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1796 = do { -- Add refined givens, and the extra givens
1797 (refined_red_givens, avails)
1798 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1799 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1800 ; avails <- foldlM addGiven avails extra_givens
1802 -- Solve the sub-problem
1803 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1804 env' = env { red_givens = refined_red_givens ++ extra_givens
1805 , red_try_me = try_me }
1807 ; traceTc (text "reduceImplication" <+> vcat
1808 [ ppr (red_givens env), ppr extra_givens, ppr reft, ppr wanteds ])
1809 ; avails <- reduceList env' wanteds avails
1811 -- Extract the binding (no frees, because try_me never says Free)
1812 ; (binds, irreds) <- extractResults avails wanteds
1814 -- We always discard the extra avails we've generated;
1815 -- but we remember if we have done any (global) improvement
1816 ; let ret_avails = updateImprovement orig_avails avails
1818 ; if isEmptyLHsBinds binds then -- No progress
1819 return (ret_avails, NoInstance)
1821 { (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
1822 -- This binding is useless if the recursive simplification
1823 -- made no progress; but currently we don't try to optimise that
1824 -- case. After all, we only try hard to reduce at top level, or
1825 -- when inferring types.
1827 ; let dict_ids = map instToId extra_givens
1828 co = mkWpTyLams tvs <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
1829 rhs = mkHsWrap co payload
1830 loc = instLocSpan inst_loc
1831 payload | isSingleton wanteds = HsVar (instToId (head wanteds))
1832 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) wanteds) Boxed
1834 -- If there are any irreds, we back off and return NoInstance
1835 ; return (ret_avails, GenInst implic_insts (L loc rhs))
1839 Note [Freeness and implications]
1840 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1841 It's hard to say when an implication constraint can be floated out. Consider
1842 forall {} Eq a => Foo [a]
1843 The (Foo [a]) doesn't mention any of the quantified variables, but it
1844 still might be partially satisfied by the (Eq a).
1846 There is a useful special case when it *is* easy to partition the
1847 constraints, namely when there are no 'givens'. Consider
1848 forall {a}. () => Bar b
1849 There are no 'givens', and so there is no reason to capture (Bar b).
1850 We can let it float out. But if there is even one constraint we
1851 must be much more careful:
1852 forall {a}. C a b => Bar (m b)
1853 because (C a b) might have a superclass (D b), from which we might
1854 deduce (Bar [b]) when m later gets instantiated to []. Ha!
1856 Here is an even more exotic example
1858 Now consider the constraint
1859 forall b. D Int b => C Int
1860 We can satisfy the (C Int) from the superclass of D, so we don't want
1861 to float the (C Int) out, even though it mentions no type variable in
1864 %************************************************************************
1866 Avails and AvailHow: the pool of evidence
1868 %************************************************************************
1872 data Avails = Avails !ImprovementDone !AvailEnv
1874 type ImprovementDone = Bool -- True <=> some unification has happened
1875 -- so some Irreds might now be reducible
1876 -- keys that are now
1878 type AvailEnv = FiniteMap Inst AvailHow
1880 = IsIrred -- Used for irreducible dictionaries,
1881 -- which are going to be lambda bound
1883 | Given TcId -- Used for dictionaries for which we have a binding
1884 -- e.g. those "given" in a signature
1886 | Rhs -- Used when there is a RHS
1887 (LHsExpr TcId) -- The RHS
1888 [Inst] -- Insts free in the RHS; we need these too
1890 instance Outputable Avails where
1893 pprAvails (Avails imp avails)
1894 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
1895 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
1896 | (inst,avail) <- fmToList avails ])]
1898 instance Outputable AvailHow where
1901 -------------------------
1902 pprAvail :: AvailHow -> SDoc
1903 pprAvail IsIrred = text "Irred"
1904 pprAvail (Given x) = text "Given" <+> ppr x
1905 pprAvail (Rhs rhs bs) = text "Rhs" <+> ppr rhs <+> braces (ppr bs)
1907 -------------------------
1908 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
1909 extendAvailEnv env inst avail = addToFM env inst avail
1911 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
1912 findAvailEnv env wanted = lookupFM env wanted
1913 -- NB 1: the Ord instance of Inst compares by the class/type info
1914 -- *not* by unique. So
1915 -- d1::C Int == d2::C Int
1917 emptyAvails :: Avails
1918 emptyAvails = Avails False emptyFM
1920 findAvail :: Avails -> Inst -> Maybe AvailHow
1921 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
1923 elemAvails :: Inst -> Avails -> Bool
1924 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
1926 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
1928 extendAvails avails@(Avails imp env) inst avail
1929 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
1930 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
1932 availsInsts :: Avails -> [Inst]
1933 availsInsts (Avails _ avails) = keysFM avails
1935 availsImproved (Avails imp _) = imp
1937 updateImprovement :: Avails -> Avails -> Avails
1938 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
1939 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
1942 Extracting the bindings from a bunch of Avails.
1943 The bindings do *not* come back sorted in dependency order.
1944 We assume that they'll be wrapped in a big Rec, so that the
1945 dependency analyser can sort them out later
1948 extractResults :: Avails
1950 -> TcM ( TcDictBinds, -- Bindings
1951 [Inst]) -- Irreducible ones
1953 extractResults (Avails _ avails) wanteds
1954 = go avails emptyBag [] wanteds
1956 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
1957 -> TcM (TcDictBinds, [Inst])
1958 go avails binds irreds []
1959 = returnM (binds, irreds)
1961 go avails binds irreds (w:ws)
1962 = case findAvailEnv avails w of
1963 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
1964 go avails binds irreds ws
1966 Just IsIrred -> go (add_given avails w) binds (w:irreds) ws
1970 -> go avails binds irreds ws
1971 -- The sought Id can be one of the givens, via a superclass chain
1972 -- and then we definitely don't want to generate an x=x binding!
1975 -> go avails (addBind binds w (nlHsVar id)) irreds ws
1977 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds (ws' ++ ws)
1979 new_binds = addBind binds w rhs
1984 add_given avails w = extendAvailEnv avails w (Given (instToId w))
1986 addBind binds inst rhs = binds `unionBags` unitBag (L (instSpan inst)
1987 (VarBind (instToId inst) rhs))
1991 Note [No superclasses for Stop]
1992 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1993 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
1994 add it to avails, so that any other equal Insts will be commoned up
1995 right here. However, we do *not* add superclasses. If we have
1998 but a is not bound here, then we *don't* want to derive dn from df
1999 here lest we lose sharing.
2002 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2003 addWanted want_scs avails wanted rhs_expr wanteds
2004 = addAvailAndSCs want_scs avails wanted avail
2006 avail = Rhs rhs_expr wanteds
2008 addGiven :: Avails -> Inst -> TcM Avails
2009 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2010 -- Always add superclasses for 'givens'
2012 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2013 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2014 -- so the assert isn't true
2016 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2017 addRefinedGiven reft (refined_givens, avails) given
2018 | isDict given -- We sometimes have 'given' methods, but they
2019 -- are always optional, so we can drop them
2020 , Just (co, pred) <- refinePred reft (dictPred given)
2021 = do { new_given <- newDictBndr (instLoc given) pred
2022 ; let rhs = L (instSpan given) $
2023 HsWrap (WpCo co) (HsVar (instToId given))
2024 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2025 ; return (new_given:refined_givens, avails) }
2026 -- ToDo: the superclasses of the original given all exist in Avails
2027 -- so we could really just cast them, but it's more awkward to do,
2028 -- and hopefully the optimiser will spot the duplicated work
2030 = return (refined_givens, avails)
2032 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2033 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2034 addAvailAndSCs want_scs avails irred IsIrred
2036 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2037 addAvailAndSCs want_scs avails inst avail
2038 | not (isClassDict inst) = extendAvails avails inst avail
2039 | NoSCs <- want_scs = extendAvails avails inst avail
2040 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2041 ; avails' <- extendAvails avails inst avail
2042 ; addSCs is_loop avails' inst }
2044 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2045 -- Note: this compares by *type*, not by Unique
2046 deps = findAllDeps (unitVarSet (instToId inst)) avail
2047 dep_tys = map idType (varSetElems deps)
2049 findAllDeps :: IdSet -> AvailHow -> IdSet
2050 -- Find all the Insts that this one depends on
2051 -- See Note [SUPERCLASS-LOOP 2]
2052 -- Watch out, though. Since the avails may contain loops
2053 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2054 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2055 findAllDeps so_far other = so_far
2057 find_all :: IdSet -> Inst -> IdSet
2059 | kid_id `elemVarSet` so_far = so_far
2060 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2061 | otherwise = so_far'
2063 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2064 kid_id = instToId kid
2066 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2067 -- Add all the superclasses of the Inst to Avails
2068 -- The first param says "dont do this because the original thing
2069 -- depends on this one, so you'd build a loop"
2070 -- Invariant: the Inst is already in Avails.
2072 addSCs is_loop avails dict
2073 = ASSERT( isDict dict )
2074 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2075 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2077 (clas, tys) = getDictClassTys dict
2078 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2079 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2081 add_sc avails (sc_dict, sc_sel)
2082 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2083 | is_given sc_dict = return avails
2084 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2085 ; addSCs is_loop avails' sc_dict }
2087 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2088 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2090 is_given :: Inst -> Bool
2091 is_given sc_dict = case findAvail avails sc_dict of
2092 Just (Given _) -> True -- Given is cheaper than superclass selection
2096 %************************************************************************
2098 \section{tcSimplifyTop: defaulting}
2100 %************************************************************************
2103 @tcSimplifyTop@ is called once per module to simplify all the constant
2104 and ambiguous Insts.
2106 We need to be careful of one case. Suppose we have
2108 instance Num a => Num (Foo a b) where ...
2110 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2111 to (Num x), and default x to Int. But what about y??
2113 It's OK: the final zonking stage should zap y to (), which is fine.
2117 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2118 tcSimplifyTop wanteds
2119 = tc_simplify_top doc False wanteds
2121 doc = text "tcSimplifyTop"
2123 tcSimplifyInteractive wanteds
2124 = tc_simplify_top doc True wanteds
2126 doc = text "tcSimplifyInteractive"
2128 -- The TcLclEnv should be valid here, solely to improve
2129 -- error message generation for the monomorphism restriction
2130 tc_simplify_top doc interactive wanteds
2131 = do { wanteds <- mapM zonkInst wanteds
2132 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2134 ; (binds1, irreds1) <- topCheckLoop doc wanteds
2136 ; if null irreds1 then
2139 -- OK, so there are some errors
2140 { -- Use the defaulting rules to do extra unification
2141 -- NB: irreds are already zonked
2142 ; extended_default <- if interactive then return True
2143 else doptM Opt_ExtendedDefaultRules
2144 ; disambiguate extended_default irreds1 -- Does unification
2145 ; (binds2, irreds2) <- topCheckLoop doc irreds1
2147 -- Deal with implicit parameter
2148 ; let (bad_ips, non_ips) = partition isIPDict irreds2
2149 (ambigs, others) = partition isTyVarDict non_ips
2151 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2153 ; addNoInstanceErrs others
2154 ; addTopAmbigErrs ambigs
2156 ; return (binds1 `unionBags` binds2) }}
2159 If a dictionary constrains a type variable which is
2160 * not mentioned in the environment
2161 * and not mentioned in the type of the expression
2162 then it is ambiguous. No further information will arise to instantiate
2163 the type variable; nor will it be generalised and turned into an extra
2164 parameter to a function.
2166 It is an error for this to occur, except that Haskell provided for
2167 certain rules to be applied in the special case of numeric types.
2169 * at least one of its classes is a numeric class, and
2170 * all of its classes are numeric or standard
2171 then the type variable can be defaulted to the first type in the
2172 default-type list which is an instance of all the offending classes.
2174 So here is the function which does the work. It takes the ambiguous
2175 dictionaries and either resolves them (producing bindings) or
2176 complains. It works by splitting the dictionary list by type
2177 variable, and using @disambigOne@ to do the real business.
2179 @disambigOne@ assumes that its arguments dictionaries constrain all
2180 the same type variable.
2182 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2183 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2184 the most common use of defaulting is code like:
2186 _ccall_ foo `seqPrimIO` bar
2188 Since we're not using the result of @foo@, the result if (presumably)
2192 disambiguate :: Bool -> [Inst] -> TcM ()
2193 -- Just does unification to fix the default types
2194 -- The Insts are assumed to be pre-zonked
2195 disambiguate extended_defaulting insts
2196 | null defaultable_groups
2199 = do { -- Figure out what default types to use
2200 mb_defaults <- getDefaultTys
2201 ; default_tys <- case mb_defaults of
2202 Just tys -> return tys
2203 Nothing -> -- No use-supplied default;
2204 -- use [Integer, Double]
2205 do { integer_ty <- tcMetaTy integerTyConName
2206 ; checkWiredInTyCon doubleTyCon
2207 ; return [integer_ty, doubleTy] }
2208 ; mapM_ (disambigGroup default_tys) defaultable_groups }
2210 unaries :: [(Inst,Class, TcTyVar)] -- (C tv) constraints
2211 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2212 (unaries, bad_tvs) = getDefaultableDicts insts
2214 -- Group by type variable
2215 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2216 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2217 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2219 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2220 defaultable_group ds@((_,_,tv):_)
2221 = not (isSkolemTyVar tv) -- Note [Avoiding spurious errors]
2222 && not (tv `elemVarSet` bad_tvs)
2223 && defaultable_classes [c | (_,c,_) <- ds]
2224 defaultable_group [] = panic "defaultable_group"
2226 defaultable_classes clss
2227 | extended_defaulting = any isInteractiveClass clss
2228 | otherwise = all isStandardClass clss && any isNumericClass clss
2230 -- In interactive mode, or with -fextended-default-rules,
2231 -- we default Show a to Show () to avoid graututious errors on "show []"
2232 isInteractiveClass cls
2233 = isNumericClass cls
2234 || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2237 disambigGroup :: [Type] -- The default types
2238 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2239 -> TcM () -- Just does unification, to fix the default types
2241 disambigGroup default_tys dicts
2242 = try_default default_tys
2244 (_,_,tyvar) = head dicts -- Should be non-empty
2245 classes = [c | (_,c,_) <- dicts]
2247 try_default [] = return ()
2248 try_default (default_ty : default_tys)
2249 = tryTcLIE_ (try_default default_tys) $
2250 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2251 -- This may fail; then the tryTcLIE_ kicks in
2252 -- Failure here is caused by there being no type in the
2253 -- default list which can satisfy all the ambiguous classes.
2254 -- For example, if Real a is reqd, but the only type in the
2255 -- default list is Int.
2257 -- After this we can't fail
2258 ; warnDefault dicts default_ty
2259 ; unifyType default_ty (mkTyVarTy tyvar) }
2262 Note [Avoiding spurious errors]
2263 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2264 When doing the unification for defaulting, we check for skolem
2265 type variables, and simply don't default them. For example:
2266 f = (*) -- Monomorphic
2267 g :: Num a => a -> a
2269 Here, we get a complaint when checking the type signature for g,
2270 that g isn't polymorphic enough; but then we get another one when
2271 dealing with the (Num a) context arising from f's definition;
2272 we try to unify a with Int (to default it), but find that it's
2273 already been unified with the rigid variable from g's type sig
2276 %************************************************************************
2278 \subsection[simple]{@Simple@ versions}
2280 %************************************************************************
2282 Much simpler versions when there are no bindings to make!
2284 @tcSimplifyThetas@ simplifies class-type constraints formed by
2285 @deriving@ declarations and when specialising instances. We are
2286 only interested in the simplified bunch of class/type constraints.
2288 It simplifies to constraints of the form (C a b c) where
2289 a,b,c are type variables. This is required for the context of
2290 instance declarations.
2293 tcSimplifyDeriv :: InstOrigin
2296 -> ThetaType -- Wanted
2297 -> TcM ThetaType -- Needed
2299 tcSimplifyDeriv orig tc tyvars theta
2300 = tcInstTyVars tyvars `thenM` \ (tvs, _, tenv) ->
2301 -- The main loop may do unification, and that may crash if
2302 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2303 -- ToDo: what if two of them do get unified?
2304 newDictBndrsO orig (substTheta tenv theta) `thenM` \ wanteds ->
2305 topCheckLoop doc wanteds `thenM` \ (_, irreds) ->
2307 doptM Opt_GlasgowExts `thenM` \ gla_exts ->
2308 doptM Opt_AllowUndecidableInstances `thenM` \ undecidable_ok ->
2310 inst_ty = mkTyConApp tc (mkTyVarTys tvs)
2311 (ok_insts, bad_insts) = partition is_ok_inst irreds
2313 = isDict inst -- Exclude implication consraints
2314 && (isTyVarClassPred pred || (gla_exts && ok_gla_pred pred))
2316 pred = dictPred inst
2318 ok_gla_pred pred = null (checkInstTermination [inst_ty] [pred])
2319 -- See Note [Deriving context]
2321 tv_set = mkVarSet tvs
2322 simpl_theta = map dictPred ok_insts
2323 weird_preds = [pred | pred <- simpl_theta
2324 , not (tyVarsOfPred pred `subVarSet` tv_set)]
2326 -- Check for a bizarre corner case, when the derived instance decl should
2327 -- have form instance C a b => D (T a) where ...
2328 -- Note that 'b' isn't a parameter of T. This gives rise to all sorts
2329 -- of problems; in particular, it's hard to compare solutions for
2330 -- equality when finding the fixpoint. So I just rule it out for now.
2332 rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2333 -- This reverse-mapping is a Royal Pain,
2334 -- but the result should mention TyVars not TcTyVars
2336 -- In effect, the bad and wierd insts cover all of the cases that
2337 -- would make checkValidInstance fail; if it were called right after tcSimplifyDeriv
2338 -- * wierd_preds ensures unambiguous instances (checkAmbiguity in checkValidInstance)
2339 -- * ok_gla_pred ensures termination (checkInstTermination in checkValidInstance)
2340 addNoInstanceErrs bad_insts `thenM_`
2341 mapM_ (addErrTc . badDerivedPred) weird_preds `thenM_`
2342 returnM (substTheta rev_env simpl_theta)
2344 doc = ptext SLIT("deriving classes for a data type")
2347 Note [Deriving context]
2348 ~~~~~~~~~~~~~~~~~~~~~~~
2349 With -fglasgow-exts, we allow things like (C Int a) in the simplified
2350 context for a derived instance declaration, because at a use of this
2351 instance, we might know that a=Bool, and have an instance for (C Int
2354 We nevertheless insist that each predicate meets the termination
2355 conditions. If not, the deriving mechanism generates larger and larger
2356 constraints. Example:
2358 data Seq a = Cons a (Seq (Succ a)) | Nil deriving Show
2360 Note the lack of a Show instance for Succ. First we'll generate
2361 instance (Show (Succ a), Show a) => Show (Seq a)
2363 instance (Show (Succ (Succ a)), Show (Succ a), Show a) => Show (Seq a)
2364 and so on. Instead we want to complain of no instance for (Show (Succ a)).
2368 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2369 used with \tr{default} declarations. We are only interested in
2370 whether it worked or not.
2373 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2376 tcSimplifyDefault theta
2377 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2378 topCheckLoop doc wanteds `thenM` \ (_, irreds) ->
2379 addNoInstanceErrs irreds `thenM_`
2385 doc = ptext SLIT("default declaration")
2389 %************************************************************************
2391 \section{Errors and contexts}
2393 %************************************************************************
2395 ToDo: for these error messages, should we note the location as coming
2396 from the insts, or just whatever seems to be around in the monad just
2400 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2401 -> [Inst] -- The offending Insts
2403 -- Group together insts with the same origin
2404 -- We want to report them together in error messages
2406 groupErrs report_err []
2408 groupErrs report_err (inst:insts)
2409 = do_one (inst:friends) `thenM_`
2410 groupErrs report_err others
2413 -- (It may seem a bit crude to compare the error messages,
2414 -- but it makes sure that we combine just what the user sees,
2415 -- and it avoids need equality on InstLocs.)
2416 (friends, others) = partition is_friend insts
2417 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2418 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2419 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2420 -- Add location and context information derived from the Insts
2422 -- Add the "arising from..." part to a message about bunch of dicts
2423 addInstLoc :: [Inst] -> Message -> Message
2424 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2426 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2427 addTopIPErrs bndrs []
2429 addTopIPErrs bndrs ips
2430 = addErrTcM (tidy_env, mk_msg tidy_ips)
2432 (tidy_env, tidy_ips) = tidyInsts ips
2433 mk_msg ips = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2434 nest 2 (ptext SLIT("the monomorphic top-level binding(s) of")
2435 <+> pprBinders bndrs <> colon)],
2436 nest 2 (vcat (map ppr_ip ips)),
2438 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2440 topIPErrs :: [Inst] -> TcM ()
2442 = groupErrs report tidy_dicts
2444 (tidy_env, tidy_dicts) = tidyInsts dicts
2445 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2446 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2447 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2449 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2451 addNoInstanceErrs insts
2452 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2453 ; reportNoInstances tidy_env Nothing tidy_insts }
2457 -> Maybe (InstLoc, [Inst]) -- Context
2458 -- Nothing => top level
2459 -- Just (d,g) => d describes the construct
2461 -> [Inst] -- What is wanted (can include implications)
2464 reportNoInstances tidy_env mb_what insts
2465 = groupErrs (report_no_instances tidy_env mb_what) insts
2467 report_no_instances tidy_env mb_what insts
2468 = do { inst_envs <- tcGetInstEnvs
2469 ; let (implics, insts1) = partition isImplicInst insts
2470 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2471 ; traceTc (text "reportNoInstnces" <+> vcat
2472 [ppr implics, ppr insts1, ppr insts2])
2473 ; mapM_ complain_implic implics
2474 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2475 ; groupErrs complain_no_inst insts2 }
2477 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2479 complain_implic inst -- Recurse!
2480 = reportNoInstances tidy_env
2481 (Just (tci_loc inst, tci_given inst))
2484 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2485 -- Right msg => overlap message
2486 -- Left inst => no instance
2487 check_overlap inst_envs wanted
2488 | not (isClassDict wanted) = Left wanted
2490 = case lookupInstEnv inst_envs clas tys of
2491 -- The case of exactly one match and no unifiers means
2492 -- a successful lookup. That can't happen here, becuase
2493 -- dicts only end up here if they didn't match in Inst.lookupInst
2495 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2497 ([], _) -> Left wanted -- No match
2498 res -> Right (mk_overlap_msg wanted res)
2500 (clas,tys) = getDictClassTys wanted
2502 mk_overlap_msg dict (matches, unifiers)
2503 = vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2504 <+> pprPred (dictPred dict))),
2505 sep [ptext SLIT("Matching instances") <> colon,
2506 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2507 ASSERT( not (null matches) )
2508 if not (isSingleton matches)
2509 then -- Two or more matches
2511 else -- One match, plus some unifiers
2512 ASSERT( not (null unifiers) )
2513 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2514 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2515 ptext SLIT("Use -fallow-incoherent-instances to use the first choice above")])]
2517 ispecs = [ispec | (_, ispec) <- matches]
2519 mk_no_inst_err insts
2520 | null insts = empty
2522 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2523 not (isEmptyVarSet (tyVarsOfInsts insts))
2524 = vcat [ addInstLoc insts $
2525 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2526 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2527 , show_fixes (fix1 loc : fixes2) ]
2529 | otherwise -- Top level
2530 = vcat [ addInstLoc insts $
2531 ptext SLIT("No instance") <> plural insts
2532 <+> ptext SLIT("for") <+> pprDictsTheta insts
2533 , show_fixes fixes2 ]
2536 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2537 <+> ptext SLIT("to the context of"),
2538 nest 2 (ppr (instLocOrigin loc)) ]
2539 -- I'm not sure it helps to add the location
2540 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2542 fixes2 | null instance_dicts = []
2543 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2544 pprDictsTheta instance_dicts]]
2545 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2546 -- Insts for which it is worth suggesting an adding an instance declaration
2547 -- Exclude implicit parameters, and tyvar dicts
2549 show_fixes :: [SDoc] -> SDoc
2550 show_fixes [] = empty
2551 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2552 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2554 addTopAmbigErrs dicts
2555 -- Divide into groups that share a common set of ambiguous tyvars
2556 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2557 -- See Note [Avoiding spurious errors]
2558 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2560 (tidy_env, tidy_dicts) = tidyInsts dicts
2562 tvs_of :: Inst -> [TcTyVar]
2563 tvs_of d = varSetElems (tyVarsOfInst d)
2564 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2566 report :: [(Inst,[TcTyVar])] -> TcM ()
2567 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2568 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2569 setSrcSpan (instSpan inst) $
2570 -- the location of the first one will do for the err message
2571 addErrTcM (tidy_env, msg $$ mono_msg)
2573 dicts = map fst pairs
2574 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2575 pprQuotedList tvs <+> in_msg,
2576 nest 2 (pprDictsInFull dicts)]
2577 in_msg = text "in the constraint" <> plural dicts <> colon
2578 report [] = panic "addTopAmbigErrs"
2581 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2582 -- There's an error with these Insts; if they have free type variables
2583 -- it's probably caused by the monomorphism restriction.
2584 -- Try to identify the offending variable
2585 -- ASSUMPTION: the Insts are fully zonked
2586 mkMonomorphismMsg tidy_env inst_tvs
2587 = findGlobals (mkVarSet inst_tvs) tidy_env `thenM` \ (tidy_env, docs) ->
2588 returnM (tidy_env, mk_msg docs)
2590 mk_msg [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2591 -- This happens in things like
2592 -- f x = show (read "foo")
2593 -- where monomorphism doesn't play any role
2594 mk_msg docs = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2598 monomorphism_fix :: SDoc
2599 monomorphism_fix = ptext SLIT("Probable fix:") <+>
2600 (ptext SLIT("give these definition(s) an explicit type signature")
2601 $$ ptext SLIT("or use -fno-monomorphism-restriction"))
2603 warnDefault ups default_ty
2604 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2605 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2607 dicts = [d | (d,_,_) <- ups]
2610 (_, tidy_dicts) = tidyInsts dicts
2611 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2612 quotes (ppr default_ty),
2613 pprDictsInFull tidy_dicts]
2615 -- Used for the ...Thetas variants; all top level
2617 = vcat [ptext SLIT("Can't derive instances where the instance context mentions"),
2618 ptext SLIT("type variables that are not data type parameters"),
2619 nest 2 (ptext SLIT("Offending constraint:") <+> ppr pred)]
2621 reduceDepthErr n stack
2622 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2623 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2624 nest 4 (pprStack stack)]
2626 pprStack stack = vcat (map pprInstInFull stack)