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 Note [Inheriting 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 Note [Implicit parameters and ambiguity]
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
429 What type should we infer for this?
430 f x = (show ?y, x::Int)
431 Since we must quantify over the ?y, the most plausible type is
432 f :: (Show a, ?y::a) => Int -> (String, Int)
433 But notice that the type of the RHS is (String,Int), with no type
434 varibables mentioned at all! The type of f looks ambiguous. But
435 it isn't, because at a call site we might have
436 let ?y = 5::Int in f 7
437 and all is well. In effect, implicit parameters are, well, parameters,
438 so we can take their type variables into account as part of the
439 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
442 Question 2: type signatures
443 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
444 BUT WATCH OUT: When you supply a type signature, we can't force you
445 to quantify over implicit parameters. For example:
449 This is perfectly reasonable. We do not want to insist on
451 (?x + 1) :: (?x::Int => Int)
453 That would be silly. Here, the definition site *is* the occurrence site,
454 so the above strictures don't apply. Hence the difference between
455 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
456 and tcSimplifyCheckBind (which does not).
458 What about when you supply a type signature for a binding?
459 Is it legal to give the following explicit, user type
460 signature to f, thus:
465 At first sight this seems reasonable, but it has the nasty property
466 that adding a type signature changes the dynamic semantics.
469 (let f x = (x::Int) + ?y
470 in (f 3, f 3 with ?y=5)) with ?y = 6
476 in (f 3, f 3 with ?y=5)) with ?y = 6
480 Indeed, simply inlining f (at the Haskell source level) would change the
483 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
484 semantics for a Haskell program without knowing its typing, so if you
485 change the typing you may change the semantics.
487 To make things consistent in all cases where we are *checking* against
488 a supplied signature (as opposed to inferring a type), we adopt the
491 a signature does not need to quantify over implicit params.
493 [This represents a (rather marginal) change of policy since GHC 5.02,
494 which *required* an explicit signature to quantify over all implicit
495 params for the reasons mentioned above.]
497 But that raises a new question. Consider
499 Given (signature) ?x::Int
500 Wanted (inferred) ?x::Int, ?y::Bool
502 Clearly we want to discharge the ?x and float the ?y out. But
503 what is the criterion that distinguishes them? Clearly it isn't
504 what free type variables they have. The Right Thing seems to be
505 to float a constraint that
506 neither mentions any of the quantified type variables
507 nor any of the quantified implicit parameters
509 See the predicate isFreeWhenChecking.
512 Question 3: monomorphism
513 ~~~~~~~~~~~~~~~~~~~~~~~~
514 There's a nasty corner case when the monomorphism restriction bites:
518 The argument above suggests that we *must* generalise
519 over the ?y parameter, to get
520 z :: (?y::Int) => Int,
521 but the monomorphism restriction says that we *must not*, giving
523 Why does the momomorphism restriction say this? Because if you have
525 let z = x + ?y in z+z
527 you might not expect the addition to be done twice --- but it will if
528 we follow the argument of Question 2 and generalise over ?y.
531 Question 4: top level
532 ~~~~~~~~~~~~~~~~~~~~~
533 At the top level, monomorhism makes no sense at all.
536 main = let ?x = 5 in print foo
540 woggle :: (?x :: Int) => Int -> Int
543 We definitely don't want (foo :: Int) with a top-level implicit parameter
544 (?x::Int) becuase there is no way to bind it.
549 (A) Always generalise over implicit parameters
550 Bindings that fall under the monomorphism restriction can't
554 * Inlining remains valid
555 * No unexpected loss of sharing
556 * But simple bindings like
558 will be rejected, unless you add an explicit type signature
559 (to avoid the monomorphism restriction)
560 z :: (?y::Int) => Int
562 This seems unacceptable
564 (B) Monomorphism restriction "wins"
565 Bindings that fall under the monomorphism restriction can't
567 Always generalise over implicit parameters *except* for bindings
568 that fall under the monomorphism restriction
571 * Inlining isn't valid in general
572 * No unexpected loss of sharing
573 * Simple bindings like
575 accepted (get value of ?y from binding site)
577 (C) Always generalise over implicit parameters
578 Bindings that fall under the monomorphism restriction can't
579 be generalised, EXCEPT for implicit parameters
581 * Inlining remains valid
582 * Unexpected loss of sharing (from the extra generalisation)
583 * Simple bindings like
585 accepted (get value of ?y from occurrence sites)
590 None of these choices seems very satisfactory. But at least we should
591 decide which we want to do.
593 It's really not clear what is the Right Thing To Do. If you see
597 would you expect the value of ?y to be got from the *occurrence sites*
598 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
599 case of function definitions, the answer is clearly the former, but
600 less so in the case of non-fucntion definitions. On the other hand,
601 if we say that we get the value of ?y from the definition site of 'z',
602 then inlining 'z' might change the semantics of the program.
604 Choice (C) really says "the monomorphism restriction doesn't apply
605 to implicit parameters". Which is fine, but remember that every
606 innocent binding 'x = ...' that mentions an implicit parameter in
607 the RHS becomes a *function* of that parameter, called at each
608 use of 'x'. Now, the chances are that there are no intervening 'with'
609 clauses that bind ?y, so a decent compiler should common up all
610 those function calls. So I think I strongly favour (C). Indeed,
611 one could make a similar argument for abolishing the monomorphism
612 restriction altogether.
614 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
618 %************************************************************************
620 \subsection{tcSimplifyInfer}
622 %************************************************************************
624 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
626 1. Compute Q = grow( fvs(T), C )
628 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
629 predicates will end up in Ct; we deal with them at the top level
631 3. Try improvement, using functional dependencies
633 4. If Step 3 did any unification, repeat from step 1
634 (Unification can change the result of 'grow'.)
636 Note: we don't reduce dictionaries in step 2. For example, if we have
637 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
638 after step 2. However note that we may therefore quantify over more
639 type variables than we absolutely have to.
641 For the guts, we need a loop, that alternates context reduction and
642 improvement with unification. E.g. Suppose we have
644 class C x y | x->y where ...
646 and tcSimplify is called with:
648 Then improvement unifies a with b, giving
651 If we need to unify anything, we rattle round the whole thing all over
658 -> TcTyVarSet -- fv(T); type vars
660 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked)
661 TcDictBinds, -- Bindings
662 [TcId]) -- Dict Ids that must be bound here (zonked)
663 -- Any free (escaping) Insts are tossed into the environment
668 tcSimplifyInfer doc tau_tvs wanted_lie
669 = do { let try_me inst | isDict inst = Stop -- Dicts
670 | otherwise = ReduceMe NoSCs -- Lits, Methods,
671 -- and impliciation constraints
672 -- In an effort to make the inferred types simple, we try
673 -- to squeeze out implication constraints if we can.
674 -- See Note [Squashing methods]
676 ; (binds1, irreds) <- checkLoop (mkRedEnv doc try_me []) wanted_lie
678 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
679 ; gbl_tvs <- tcGetGlobalTyVars
680 ; let preds = fdPredsOfInsts irreds
681 qtvs = grow preds tau_tvs' `minusVarSet` oclose preds gbl_tvs
682 (free, bound) = partition (isFreeWhenInferring qtvs) irreds
684 -- Remove redundant superclasses from 'bound'
685 -- The 'Stop' try_me result does not do so,
686 -- see Note [No superclasses for Stop]
687 ; let try_me inst = ReduceMe AddSCs
688 ; (binds2, irreds) <- checkLoop (mkRedEnv doc try_me []) bound
691 ; return (varSetElems qtvs, binds1 `unionBags` binds2, map instToId irreds) }
692 -- NB: when we are done, we might have some bindings, but
693 -- the final qtvs might be empty. See Note [NO TYVARS] below.
696 Note [Squashing methods]
697 ~~~~~~~~~~~~~~~~~~~~~~~~~
698 Be careful if you want to float methods more:
699 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
700 From an application (truncate f i) we get
703 If we have also have a second occurrence of truncate, we get
706 When simplifying with i,f free, we might still notice that
707 t1=t3; but alas, the binding for t2 (which mentions t1)
708 may continue to float out!
713 class Y a b | a -> b where
716 instance Y [[a]] a where
719 k :: X a -> X a -> X a
721 g :: Num a => [X a] -> [X a]
724 h ys = ys ++ map (k (y [[0]])) xs
726 The excitement comes when simplifying the bindings for h. Initially
727 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
728 From this we get t1:=:t2, but also various bindings. We can't forget
729 the bindings (because of [LOOP]), but in fact t1 is what g is
732 The net effect of [NO TYVARS]
735 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
736 isFreeWhenInferring qtvs inst
737 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
738 && isInheritableInst inst -- and no implicit parameter involved
739 -- see Note [Inheriting implicit parameters]
741 {- No longer used (with implication constraints)
742 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
743 -> NameSet -- Quantified implicit parameters
745 isFreeWhenChecking qtvs ips inst
746 = isFreeWrtTyVars qtvs inst
747 && isFreeWrtIPs ips inst
750 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
751 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
755 %************************************************************************
757 \subsection{tcSimplifyCheck}
759 %************************************************************************
761 @tcSimplifyCheck@ is used when we know exactly the set of variables
762 we are going to quantify over. For example, a class or instance declaration.
765 -----------------------------------------------------------
766 -- tcSimplifyCheck is used when checking expression type signatures,
767 -- class decls, instance decls etc.
768 tcSimplifyCheck :: InstLoc
769 -> [TcTyVar] -- Quantify over these
772 -> TcM TcDictBinds -- Bindings
773 tcSimplifyCheck loc qtvs givens wanteds
774 = ASSERT( all isSkolemTyVar qtvs )
775 do { (binds, irreds) <- innerCheckLoop loc givens wanteds
776 ; implic_bind <- bindIrreds loc [] emptyRefinement
778 ; return (binds `unionBags` implic_bind) }
780 -----------------------------------------------------------
781 -- tcSimplifyCheckPat is used for existential pattern match
782 tcSimplifyCheckPat :: InstLoc
783 -> [CoVar] -> Refinement
784 -> [TcTyVar] -- Quantify over these
787 -> TcM TcDictBinds -- Bindings
788 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
789 = ASSERT( all isSkolemTyVar qtvs )
790 do { (binds, irreds) <- innerCheckLoop loc givens wanteds
791 ; implic_bind <- bindIrreds loc co_vars reft
793 ; return (binds `unionBags` implic_bind) }
795 -----------------------------------------------------------
796 bindIrreds :: InstLoc -> [CoVar] -> Refinement
797 -> [TcTyVar] -> [Inst] -> [Inst]
799 -- Make a binding that binds 'irreds', by generating an implication
800 -- constraint for them, *and* throwing the constraint into the LIE
801 bindIrreds loc co_vars reft qtvs givens irreds
802 = do { let givens' = filter isDict givens
803 -- The givens can include methods
805 -- If there are no 'givens', then it's safe to
806 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
807 -- See Note [Freeness and implications]
808 ; irreds' <- if null givens'
810 { let qtv_set = mkVarSet qtvs
811 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
813 ; return real_irreds }
816 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
817 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
818 -- This call does the real work
823 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
825 -> TcM ([Inst], TcDictBinds)
826 -- Make a binding that binds 'irreds', by generating an implication
827 -- constraint for them, *and* throwing the constraint into the LIE
828 -- The binding looks like
829 -- (ir1, .., irn) = f qtvs givens
830 -- where f is (evidence for) the new implication constraint
832 -- This binding must line up the 'rhs' in reduceImplication
833 makeImplicationBind loc all_tvs reft
834 givens -- Guaranteed all Dicts
836 | null irreds -- If there are no irreds, we are done
837 = return ([], emptyBag)
838 | otherwise -- Otherwise we must generate a binding
839 = do { uniq <- newUnique
840 ; span <- getSrcSpanM
841 ; let name = mkInternalName uniq (mkVarOcc "ic") (srcSpanStart span)
842 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
843 tci_tyvars = all_tvs,
845 tci_wanted = irreds, tci_loc = loc }
847 ; let n_irreds = length irreds
848 irred_ids = map instToId irreds
849 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
850 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
851 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
852 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
853 bind | n_irreds==1 = VarBind (head irred_ids) rhs
854 | otherwise = PatBind { pat_lhs = L span pat,
855 pat_rhs = unguardedGRHSs rhs,
857 bind_fvs = placeHolderNames }
858 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
859 return ([implic_inst], unitBag (L span bind)) }
861 -----------------------------------------------------------
865 [Inst]) -- Irreducible
867 topCheckLoop doc wanteds
868 = checkLoop (mkRedEnv doc try_me []) wanteds
870 try_me inst = ReduceMe AddSCs
872 -----------------------------------------------------------
873 innerCheckLoop :: InstLoc
877 [Inst]) -- Irreducible
879 innerCheckLoop inst_loc givens wanteds
880 = checkLoop env wanteds
882 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
884 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
886 -- When checking against a given signature
887 -- we MUST be very gentle: Note [Check gently]
892 We have to very careful about not simplifying too vigorously
897 f :: Show b => T b -> b
900 Inside the pattern match, which binds (a:*, x:a), we know that
902 Hence we have a dictionary for Show [a] available; and indeed we
903 need it. We are going to build an implication contraint
904 forall a. (b~[a]) => Show [a]
905 Later, we will solve this constraint using the knowledge (Show b)
907 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
908 thing becomes insoluble. So we simplify gently (get rid of literals
909 and methods only, plus common up equal things), deferring the real
910 work until top level, when we solve the implication constraint
915 -----------------------------------------------------------
919 [Inst]) -- Irreducible
920 -- Precondition: the try_me never returns Free
921 -- givens are completely rigid
923 checkLoop env wanteds
924 = do { -- Givens are skolems, so no need to zonk them
925 wanteds' <- mappM zonkInst wanteds
927 ; (improved, binds, irreds) <- reduceContext env wanteds'
929 ; if not improved then
930 return (binds, irreds)
933 -- If improvement did some unification, we go round again.
934 -- We start again with irreds, not wanteds
935 -- Using an instance decl might have introduced a fresh type variable
936 -- which might have been unified, so we'd get an infinite loop
937 -- if we started again with wanteds! See Note [LOOP]
938 { (binds1, irreds1) <- checkLoop env irreds
939 ; return (binds `unionBags` binds1, irreds1) } }
944 class If b t e r | b t e -> r
947 class Lte a b c | a b -> c where lte :: a -> b -> c
949 instance (Lte a b l,If l b a c) => Max a b c
951 Wanted: Max Z (S x) y
953 Then we'll reduce using the Max instance to:
954 (Lte Z (S x) l, If l (S x) Z y)
955 and improve by binding l->T, after which we can do some reduction
956 on both the Lte and If constraints. What we *can't* do is start again
957 with (Max Z (S x) y)!
961 -----------------------------------------------------------
962 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
963 -- against, but we don't know the type variables over which we are going to quantify.
964 -- This happens when we have a type signature for a mutually recursive group
967 -> TcTyVarSet -- fv(T)
970 -> TcM ([TcTyVar], -- Variables over which to quantify
971 TcDictBinds) -- Bindings
973 tcSimplifyInferCheck loc tau_tvs givens wanteds
974 = do { (binds, irreds) <- innerCheckLoop loc givens wanteds
976 -- Figure out which type variables to quantify over
977 -- You might think it should just be the signature tyvars,
978 -- but in bizarre cases you can get extra ones
979 -- f :: forall a. Num a => a -> a
980 -- f x = fst (g (x, head [])) + 1
982 -- Here we infer g :: forall a b. a -> b -> (b,a)
983 -- We don't want g to be monomorphic in b just because
984 -- f isn't quantified over b.
985 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
986 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
987 ; gbl_tvs <- tcGetGlobalTyVars
988 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
989 -- We could close gbl_tvs, but its not necessary for
990 -- soundness, and it'll only affect which tyvars, not which
991 -- dictionaries, we quantify over
993 -- Now we are back to normal (c.f. tcSimplCheck)
994 ; implic_bind <- bindIrreds loc [] emptyRefinement
996 ; return (qtvs, binds `unionBags` implic_bind) }
1000 %************************************************************************
1002 tcSimplifySuperClasses
1004 %************************************************************************
1006 Note [SUPERCLASS-LOOP 1]
1007 ~~~~~~~~~~~~~~~~~~~~~~~~
1008 We have to be very, very careful when generating superclasses, lest we
1009 accidentally build a loop. Here's an example:
1013 class S a => C a where { opc :: a -> a }
1014 class S b => D b where { opd :: b -> b }
1016 instance C Int where
1019 instance D Int where
1022 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1023 Simplifying, we may well get:
1024 $dfCInt = :C ds1 (opd dd)
1027 Notice that we spot that we can extract ds1 from dd.
1029 Alas! Alack! We can do the same for (instance D Int):
1031 $dfDInt = :D ds2 (opc dc)
1035 And now we've defined the superclass in terms of itself.
1037 Solution: never generate a superclass selectors at all when
1038 satisfying the superclass context of an instance declaration.
1040 Two more nasty cases are in
1045 tcSimplifySuperClasses
1050 tcSimplifySuperClasses loc givens sc_wanteds
1051 = do { (binds1, irreds) <- checkLoop env sc_wanteds
1052 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1053 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1056 env = mkRedEnv (pprInstLoc loc) try_me givens
1057 try_me inst = ReduceMe NoSCs
1058 -- Like topCheckLoop, but with NoSCs
1062 %************************************************************************
1064 \subsection{tcSimplifyRestricted}
1066 %************************************************************************
1068 tcSimplifyRestricted infers which type variables to quantify for a
1069 group of restricted bindings. This isn't trivial.
1072 We want to quantify over a to get id :: forall a. a->a
1075 We do not want to quantify over a, because there's an Eq a
1076 constraint, so we get eq :: a->a->Bool (notice no forall)
1079 RHS has type 'tau', whose free tyvars are tau_tvs
1080 RHS has constraints 'wanteds'
1083 Quantify over (tau_tvs \ ftvs(wanteds))
1084 This is bad. The constraints may contain (Monad (ST s))
1085 where we have instance Monad (ST s) where...
1086 so there's no need to be monomorphic in s!
1088 Also the constraint might be a method constraint,
1089 whose type mentions a perfectly innocent tyvar:
1090 op :: Num a => a -> b -> a
1091 Here, b is unconstrained. A good example would be
1093 We want to infer the polymorphic type
1094 foo :: forall b. b -> b
1097 Plan B (cunning, used for a long time up to and including GHC 6.2)
1098 Step 1: Simplify the constraints as much as possible (to deal
1099 with Plan A's problem). Then set
1100 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1102 Step 2: Now simplify again, treating the constraint as 'free' if
1103 it does not mention qtvs, and trying to reduce it otherwise.
1104 The reasons for this is to maximise sharing.
1106 This fails for a very subtle reason. Suppose that in the Step 2
1107 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1108 In the Step 1 this constraint might have been simplified, perhaps to
1109 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1110 This won't happen in Step 2... but that in turn might prevent some other
1111 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1112 and that in turn breaks the invariant that no constraints are quantified over.
1114 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1119 Step 1: Simplify the constraints as much as possible (to deal
1120 with Plan A's problem). Then set
1121 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1122 Return the bindings from Step 1.
1125 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1128 instance (HasBinary ty IO) => HasCodedValue ty
1130 foo :: HasCodedValue a => String -> IO a
1132 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1133 doDecodeIO codedValue view
1134 = let { act = foo "foo" } in act
1136 You might think this should work becuase the call to foo gives rise to a constraint
1137 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1138 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1139 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1141 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1145 Plan D (a variant of plan B)
1146 Step 1: Simplify the constraints as much as possible (to deal
1147 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1148 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1150 Step 2: Now simplify again, treating the constraint as 'free' if
1151 it does not mention qtvs, and trying to reduce it otherwise.
1153 The point here is that it's generally OK to have too few qtvs; that is,
1154 to make the thing more monomorphic than it could be. We don't want to
1155 do that in the common cases, but in wierd cases it's ok: the programmer
1156 can always add a signature.
1158 Too few qtvs => too many wanteds, which is what happens if you do less
1163 tcSimplifyRestricted -- Used for restricted binding groups
1164 -- i.e. ones subject to the monomorphism restriction
1167 -> [Name] -- Things bound in this group
1168 -> TcTyVarSet -- Free in the type of the RHSs
1169 -> [Inst] -- Free in the RHSs
1170 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked)
1171 TcDictBinds) -- Bindings
1172 -- tcSimpifyRestricted returns no constraints to
1173 -- quantify over; by definition there are none.
1174 -- They are all thrown back in the LIE
1176 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1177 -- Zonk everything in sight
1178 = mappM zonkInst wanteds `thenM` \ wanteds' ->
1180 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1181 -- dicts; the idea is to get rid of as many type
1182 -- variables as possible, and we don't want to stop
1183 -- at (say) Monad (ST s), because that reduces
1184 -- immediately, with no constraint on s.
1186 -- BUT do no improvement! See Plan D above
1187 -- HOWEVER, some unification may take place, if we instantiate
1188 -- a method Inst with an equality constraint
1189 let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1191 reduceContext env wanteds' `thenM` \ (_imp, _binds, constrained_dicts) ->
1193 -- Next, figure out the tyvars we will quantify over
1194 zonkTcTyVarsAndFV (varSetElems tau_tvs) `thenM` \ tau_tvs' ->
1195 tcGetGlobalTyVars `thenM` \ gbl_tvs' ->
1196 mappM zonkInst constrained_dicts `thenM` \ constrained_dicts' ->
1198 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1199 qtvs = (tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs')
1200 `minusVarSet` constrained_tvs'
1202 traceTc (text "tcSimplifyRestricted" <+> vcat [
1203 pprInsts wanteds, pprInsts constrained_dicts',
1205 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ]) `thenM_`
1207 -- The first step may have squashed more methods than
1208 -- necessary, so try again, this time more gently, knowing the exact
1209 -- set of type variables to quantify over.
1211 -- We quantify only over constraints that are captured by qtvs;
1212 -- these will just be a subset of non-dicts. This in contrast
1213 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1214 -- all *non-inheritable* constraints too. This implements choice
1215 -- (B) under "implicit parameter and monomorphism" above.
1217 -- Remember that we may need to do *some* simplification, to
1218 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1219 -- just to float all constraints
1221 -- At top level, we *do* squash methods becuase we want to
1222 -- expose implicit parameters to the test that follows
1224 is_nested_group = isNotTopLevel top_lvl
1225 try_me inst | isFreeWrtTyVars qtvs inst,
1226 (is_nested_group || isDict inst) = Stop
1227 | otherwise = ReduceMe AddSCs
1228 env = mkNoImproveRedEnv doc try_me
1230 reduceContext env wanteds' `thenM` \ (_imp, binds, irreds) ->
1231 ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1233 -- See "Notes on implicit parameters, Question 4: top level"
1234 if is_nested_group then
1235 extendLIEs irreds `thenM_`
1236 returnM (varSetElems qtvs, binds)
1239 (bad_ips, non_ips) = partition isIPDict irreds
1241 addTopIPErrs bndrs bad_ips `thenM_`
1242 extendLIEs non_ips `thenM_`
1243 returnM (varSetElems qtvs, binds)
1247 %************************************************************************
1251 %************************************************************************
1253 On the LHS of transformation rules we only simplify methods and constants,
1254 getting dictionaries. We want to keep all of them unsimplified, to serve
1255 as the available stuff for the RHS of the rule.
1257 Example. Consider the following left-hand side of a rule
1259 f (x == y) (y > z) = ...
1261 If we typecheck this expression we get constraints
1263 d1 :: Ord a, d2 :: Eq a
1265 We do NOT want to "simplify" to the LHS
1267 forall x::a, y::a, z::a, d1::Ord a.
1268 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1272 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1273 f ((==) d2 x y) ((>) d1 y z) = ...
1275 Here is another example:
1277 fromIntegral :: (Integral a, Num b) => a -> b
1278 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1280 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1281 we *dont* want to get
1283 forall dIntegralInt.
1284 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1286 because the scsel will mess up RULE matching. Instead we want
1288 forall dIntegralInt, dNumInt.
1289 fromIntegral Int Int dIntegralInt dNumInt = id Int
1293 g (x == y) (y == z) = ..
1295 where the two dictionaries are *identical*, we do NOT WANT
1297 forall x::a, y::a, z::a, d1::Eq a
1298 f ((==) d1 x y) ((>) d1 y z) = ...
1300 because that will only match if the dict args are (visibly) equal.
1301 Instead we want to quantify over the dictionaries separately.
1303 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1304 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1305 from scratch, rather than further parameterise simpleReduceLoop etc
1308 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1309 tcSimplifyRuleLhs wanteds
1310 = go [] emptyBag wanteds
1313 = return (dicts, binds)
1314 go dicts binds (w:ws)
1316 = go (w:dicts) binds ws
1318 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1319 -- to fromInteger; this looks fragile to me
1320 ; lookup_result <- lookupSimpleInst w'
1321 ; case lookup_result of
1322 GenInst ws' rhs -> go dicts (addBind binds w rhs) (ws' ++ ws)
1323 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1327 tcSimplifyBracket is used when simplifying the constraints arising from
1328 a Template Haskell bracket [| ... |]. We want to check that there aren't
1329 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1330 Show instance), but we aren't otherwise interested in the results.
1331 Nor do we care about ambiguous dictionaries etc. We will type check
1332 this bracket again at its usage site.
1335 tcSimplifyBracket :: [Inst] -> TcM ()
1336 tcSimplifyBracket wanteds
1337 = do { topCheckLoop doc wanteds
1340 doc = text "tcSimplifyBracket"
1344 %************************************************************************
1346 \subsection{Filtering at a dynamic binding}
1348 %************************************************************************
1353 we must discharge all the ?x constraints from B. We also do an improvement
1354 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1356 Actually, the constraints from B might improve the types in ?x. For example
1358 f :: (?x::Int) => Char -> Char
1361 then the constraint (?x::Int) arising from the call to f will
1362 force the binding for ?x to be of type Int.
1365 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1368 -- We need a loop so that we do improvement, and then
1369 -- (next time round) generate a binding to connect the two
1371 -- Here the two ?x's have different types, and improvement
1372 -- makes them the same.
1374 tcSimplifyIPs given_ips wanteds
1375 = do { wanteds' <- mappM zonkInst wanteds
1376 ; given_ips' <- mappM zonkInst given_ips
1377 -- Unusually for checking, we *must* zonk the given_ips
1379 ; let env = mkRedEnv doc try_me given_ips'
1380 ; (improved, binds, irreds) <- reduceContext env wanteds'
1382 ; if not improved then
1383 ASSERT( all is_free irreds )
1384 do { extendLIEs irreds
1387 tcSimplifyIPs given_ips wanteds }
1389 doc = text "tcSimplifyIPs" <+> ppr given_ips
1390 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1391 is_free inst = isFreeWrtIPs ip_set inst
1393 -- Simplify any methods that mention the implicit parameter
1394 try_me inst | is_free inst = Stop
1395 | otherwise = ReduceMe NoSCs
1399 %************************************************************************
1401 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1403 %************************************************************************
1405 When doing a binding group, we may have @Insts@ of local functions.
1406 For example, we might have...
1408 let f x = x + 1 -- orig local function (overloaded)
1409 f.1 = f Int -- two instances of f
1414 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1415 where @f@ is in scope; those @Insts@ must certainly not be passed
1416 upwards towards the top-level. If the @Insts@ were binding-ified up
1417 there, they would have unresolvable references to @f@.
1419 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1420 For each method @Inst@ in the @init_lie@ that mentions one of the
1421 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1422 @LIE@), as well as the @HsBinds@ generated.
1425 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1426 -- Simlifies only MethodInsts, and generate only bindings of form
1428 -- We're careful not to even generate bindings of the form
1430 -- You'd think that'd be fine, but it interacts with what is
1431 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1433 bindInstsOfLocalFuns wanteds local_ids
1434 | null overloaded_ids
1436 = extendLIEs wanteds `thenM_`
1437 returnM emptyLHsBinds
1440 = do { (binds, irreds) <- checkLoop env for_me
1441 ; extendLIEs not_for_me
1445 env = mkRedEnv doc try_me []
1446 doc = text "bindInsts" <+> ppr local_ids
1447 overloaded_ids = filter is_overloaded local_ids
1448 is_overloaded id = isOverloadedTy (idType id)
1449 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1451 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1452 -- so it's worth building a set, so that
1453 -- lookup (in isMethodFor) is faster
1454 try_me inst | isMethod inst = ReduceMe NoSCs
1459 %************************************************************************
1461 \subsection{Data types for the reduction mechanism}
1463 %************************************************************************
1465 The main control over context reduction is here
1469 = RedEnv { red_doc :: SDoc -- The context
1470 , red_try_me :: Inst -> WhatToDo
1471 , red_improve :: Bool -- True <=> do improvement
1472 , red_givens :: [Inst] -- All guaranteed rigid
1474 -- but see Note [Rigidity]
1475 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1476 -- See Note [RedStack]
1480 -- The red_givens are rigid so far as cmpInst is concerned.
1481 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1482 -- let ?x = e in ...
1483 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1484 -- But that doesn't affect the comparison, which is based only on mame.
1487 -- The red_stack pair (n,insts) pair is just used for error reporting.
1488 -- 'n' is always the depth of the stack.
1489 -- The 'insts' is the stack of Insts being reduced: to produce X
1490 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1493 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1494 mkRedEnv doc try_me givens
1495 = RedEnv { red_doc = doc, red_try_me = try_me,
1496 red_givens = givens, red_stack = (0,[]),
1497 red_improve = True }
1499 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1500 -- Do not do improvement; no givens
1501 mkNoImproveRedEnv doc try_me
1502 = RedEnv { red_doc = doc, red_try_me = try_me,
1503 red_givens = [], red_stack = (0,[]),
1504 red_improve = True }
1507 = ReduceMe WantSCs -- Try to reduce this
1508 -- If there's no instance, add the inst to the
1509 -- irreductible ones, but don't produce an error
1510 -- message of any kind.
1511 -- It might be quite legitimate such as (Eq a)!
1513 | Stop -- Return as irreducible unless it can
1514 -- be reduced to a constant in one step
1515 -- Do not add superclasses; see
1517 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1518 -- of a predicate when adding it to the avails
1519 -- The reason for this flag is entirely the super-class loop problem
1520 -- Note [SUPER-CLASS LOOP 1]
1523 %************************************************************************
1525 \subsection[reduce]{@reduce@}
1527 %************************************************************************
1531 reduceContext :: RedEnv
1533 -> TcM (ImprovementDone,
1534 TcDictBinds, -- Dictionary bindings
1535 [Inst]) -- Irreducible
1537 reduceContext env wanteds
1538 = do { traceTc (text "reduceContext" <+> (vcat [
1539 text "----------------------",
1541 text "given" <+> ppr (red_givens env),
1542 text "wanted" <+> ppr wanteds,
1543 text "----------------------"
1546 -- Build the Avail mapping from "givens"
1547 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1550 ; avails <- reduceList env wanteds init_state
1552 ; let improved = availsImproved avails
1553 ; (binds, irreds) <- extractResults avails wanteds
1555 ; traceTc (text "reduceContext end" <+> (vcat [
1556 text "----------------------",
1558 text "given" <+> ppr (red_givens env),
1559 text "wanted" <+> ppr wanteds,
1561 text "avails" <+> pprAvails avails,
1562 text "improved =" <+> ppr improved,
1563 text "----------------------"
1566 ; return (improved, binds, irreds) }
1568 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1569 tcImproveOne avails inst
1570 | not (isDict inst) = return False
1572 = do { inst_envs <- tcGetInstEnvs
1573 ; let eqns = improveOne (classInstances inst_envs)
1574 (dictPred inst, pprInstArising inst)
1575 [ (dictPred p, pprInstArising p)
1576 | p <- availsInsts avails, isDict p ]
1577 -- Avails has all the superclasses etc (good)
1578 -- It also has all the intermediates of the deduction (good)
1579 -- It does not have duplicates (good)
1580 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1581 -- so that improve will see them separate
1582 ; traceTc (text "improveOne" <+> ppr inst)
1585 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1586 -> TcM ImprovementDone
1587 unifyEqns [] = return False
1589 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1593 unify ((qtvs, pairs), what1, what2)
1594 = addErrCtxtM (mkEqnMsg what1 what2) $
1595 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1596 mapM_ (unif_pr tenv) pairs
1597 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1599 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1601 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1602 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1603 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1604 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1605 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1606 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1607 ; return (tidy_env, msg) }
1610 The main context-reduction function is @reduce@. Here's its game plan.
1613 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1614 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1615 = do { dopts <- getDOpts
1618 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1619 2 (ifPprDebug (nest 2 (pprStack stk))))
1622 ; if n >= ctxtStkDepth dopts then
1623 failWithTc (reduceDepthErr n stk)
1627 go [] state = return state
1628 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1631 -- Base case: we're done!
1632 reduce env wanted avails
1633 -- It's the same as an existing inst, or a superclass thereof
1634 | Just avail <- findAvail avails wanted
1638 = case red_try_me env wanted of {
1639 ; Stop -> try_simple (addIrred NoSCs) -- See Note [No superclasses for Stop]
1641 ; ReduceMe want_scs -> -- It should be reduced
1642 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1643 case lookup_result of
1644 NoInstance -> -- No such instance!
1645 -- Add it and its superclasses
1646 addIrred want_scs avails wanted
1648 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1650 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1651 ; avails2 <- reduceList env wanteds' avails1
1652 ; addWanted want_scs avails2 wanted rhs wanteds' }
1653 -- Temporarily do addIrred *before* the reduceList,
1654 -- which has the effect of adding the thing we are trying
1655 -- to prove to the database before trying to prove the things it
1656 -- needs. See note [RECURSIVE DICTIONARIES]
1657 -- NB: we must not do an addWanted before, because that adds the
1658 -- superclasses too, and thaat can lead to a spurious loop; see
1659 -- the examples in [SUPERCLASS-LOOP]
1660 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1664 -- First, see if the inst can be reduced to a constant in one step
1665 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1666 -- Don't bother for implication constraints, which take real work
1667 try_simple do_this_otherwise
1668 = do { res <- lookupSimpleInst wanted
1670 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1671 other -> do_this_otherwise avails wanted }
1675 Note [SUPERCLASS-LOOP 2]
1676 ~~~~~~~~~~~~~~~~~~~~~~~~
1677 But the above isn't enough. Suppose we are *given* d1:Ord a,
1678 and want to deduce (d2:C [a]) where
1680 class Ord a => C a where
1681 instance Ord [a] => C [a] where ...
1683 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1684 superclasses of C [a] to avails. But we must not overwrite the binding
1685 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1688 Here's another variant, immortalised in tcrun020
1689 class Monad m => C1 m
1690 class C1 m => C2 m x
1691 instance C2 Maybe Bool
1692 For the instance decl we need to build (C1 Maybe), and it's no good if
1693 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1694 before we search for C1 Maybe.
1696 Here's another example
1697 class Eq b => Foo a b
1698 instance Eq a => Foo [a] a
1702 we'll first deduce that it holds (via the instance decl). We must not
1703 then overwrite the Eq t constraint with a superclass selection!
1705 At first I had a gross hack, whereby I simply did not add superclass constraints
1706 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1707 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1708 I found a very obscure program (now tcrun021) in which improvement meant the
1709 simplifier got two bites a the cherry... so something seemed to be an Stop
1710 first time, but reducible next time.
1712 Now we implement the Right Solution, which is to check for loops directly
1713 when adding superclasses. It's a bit like the occurs check in unification.
1716 Note [RECURSIVE DICTIONARIES]
1717 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1719 data D r = ZeroD | SuccD (r (D r));
1721 instance (Eq (r (D r))) => Eq (D r) where
1722 ZeroD == ZeroD = True
1723 (SuccD a) == (SuccD b) = a == b
1726 equalDC :: D [] -> D [] -> Bool;
1729 We need to prove (Eq (D [])). Here's how we go:
1733 by instance decl, holds if
1737 by instance decl of Eq, holds if
1739 where d2 = dfEqList d3
1742 But now we can "tie the knot" to give
1748 and it'll even run! The trick is to put the thing we are trying to prove
1749 (in this case Eq (D []) into the database before trying to prove its
1750 contributing clauses.
1753 %************************************************************************
1755 Reducing a single constraint
1757 %************************************************************************
1760 ---------------------------------------------
1761 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1762 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1763 tci_given = extra_givens, tci_wanted = wanteds })
1764 = reduceImplication env avails reft tvs extra_givens wanteds loc
1766 reduceInst env avails other_inst
1767 = do { result <- lookupSimpleInst other_inst
1768 ; return (avails, result) }
1772 ---------------------------------------------
1773 reduceImplication :: RedEnv
1775 -> Refinement -- May refine the givens; often empty
1776 -> [TcTyVar] -- Quantified type variables; all skolems
1777 -> [Inst] -- Extra givens; all rigid
1780 -> TcM (Avails, LookupInstResult)
1783 Suppose we are simplifying the constraint
1784 forall bs. extras => wanted
1785 in the context of an overall simplification problem with givens 'givens',
1786 and refinment 'reft'.
1789 * The refinement is often empty
1791 * The 'extra givens' need not mention any of the quantified type variables
1792 e.g. forall {}. Eq a => Eq [a]
1793 forall {}. C Int => D (Tree Int)
1795 This happens when you have something like
1797 T1 :: Eq a => a -> T a
1800 f x = ...(case x of { T1 v -> v==v })...
1803 -- ToDo: should we instantiate tvs? I think it's not necessary
1805 -- ToDo: what about improvement? There may be some improvement
1806 -- exposed as a result of the simplifications done by reduceList
1807 -- which are discarded if we back off.
1808 -- This is almost certainly Wrong, but we'll fix it when dealing
1809 -- better with equality constraints
1810 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1811 = do { -- Add refined givens, and the extra givens
1812 (refined_red_givens, avails)
1813 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1814 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1815 ; avails <- foldlM addGiven avails extra_givens
1817 -- Solve the sub-problem
1818 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1819 env' = env { red_givens = refined_red_givens ++ extra_givens
1820 , red_try_me = try_me }
1822 ; traceTc (text "reduceImplication" <+> vcat
1823 [ ppr (red_givens env), ppr extra_givens, ppr reft, ppr wanteds ])
1824 ; avails <- reduceList env' wanteds avails
1826 -- Extract the binding (no frees, because try_me never says Free)
1827 ; (binds, irreds) <- extractResults avails wanteds
1829 -- We always discard the extra avails we've generated;
1830 -- but we remember if we have done any (global) improvement
1831 ; let ret_avails = updateImprovement orig_avails avails
1833 ; if isEmptyLHsBinds binds then -- No progress
1834 return (ret_avails, NoInstance)
1836 { (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
1837 -- This binding is useless if the recursive simplification
1838 -- made no progress; but currently we don't try to optimise that
1839 -- case. After all, we only try hard to reduce at top level, or
1840 -- when inferring types.
1842 ; let dict_ids = map instToId extra_givens
1843 co = mkWpTyLams tvs <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
1844 rhs = mkHsWrap co payload
1845 loc = instLocSpan inst_loc
1846 payload | isSingleton wanteds = HsVar (instToId (head wanteds))
1847 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) wanteds) Boxed
1849 -- If there are any irreds, we back off and return NoInstance
1850 ; return (ret_avails, GenInst implic_insts (L loc rhs))
1854 Note [Freeness and implications]
1855 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1856 It's hard to say when an implication constraint can be floated out. Consider
1857 forall {} Eq a => Foo [a]
1858 The (Foo [a]) doesn't mention any of the quantified variables, but it
1859 still might be partially satisfied by the (Eq a).
1861 There is a useful special case when it *is* easy to partition the
1862 constraints, namely when there are no 'givens'. Consider
1863 forall {a}. () => Bar b
1864 There are no 'givens', and so there is no reason to capture (Bar b).
1865 We can let it float out. But if there is even one constraint we
1866 must be much more careful:
1867 forall {a}. C a b => Bar (m b)
1868 because (C a b) might have a superclass (D b), from which we might
1869 deduce (Bar [b]) when m later gets instantiated to []. Ha!
1871 Here is an even more exotic example
1873 Now consider the constraint
1874 forall b. D Int b => C Int
1875 We can satisfy the (C Int) from the superclass of D, so we don't want
1876 to float the (C Int) out, even though it mentions no type variable in
1879 %************************************************************************
1881 Avails and AvailHow: the pool of evidence
1883 %************************************************************************
1887 data Avails = Avails !ImprovementDone !AvailEnv
1889 type ImprovementDone = Bool -- True <=> some unification has happened
1890 -- so some Irreds might now be reducible
1891 -- keys that are now
1893 type AvailEnv = FiniteMap Inst AvailHow
1895 = IsIrred -- Used for irreducible dictionaries,
1896 -- which are going to be lambda bound
1898 | Given TcId -- Used for dictionaries for which we have a binding
1899 -- e.g. those "given" in a signature
1901 | Rhs -- Used when there is a RHS
1902 (LHsExpr TcId) -- The RHS
1903 [Inst] -- Insts free in the RHS; we need these too
1905 instance Outputable Avails where
1908 pprAvails (Avails imp avails)
1909 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
1910 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
1911 | (inst,avail) <- fmToList avails ])]
1913 instance Outputable AvailHow where
1916 -------------------------
1917 pprAvail :: AvailHow -> SDoc
1918 pprAvail IsIrred = text "Irred"
1919 pprAvail (Given x) = text "Given" <+> ppr x
1920 pprAvail (Rhs rhs bs) = text "Rhs" <+> ppr rhs <+> braces (ppr bs)
1922 -------------------------
1923 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
1924 extendAvailEnv env inst avail = addToFM env inst avail
1926 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
1927 findAvailEnv env wanted = lookupFM env wanted
1928 -- NB 1: the Ord instance of Inst compares by the class/type info
1929 -- *not* by unique. So
1930 -- d1::C Int == d2::C Int
1932 emptyAvails :: Avails
1933 emptyAvails = Avails False emptyFM
1935 findAvail :: Avails -> Inst -> Maybe AvailHow
1936 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
1938 elemAvails :: Inst -> Avails -> Bool
1939 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
1941 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
1943 extendAvails avails@(Avails imp env) inst avail
1944 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
1945 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
1947 availsInsts :: Avails -> [Inst]
1948 availsInsts (Avails _ avails) = keysFM avails
1950 availsImproved (Avails imp _) = imp
1952 updateImprovement :: Avails -> Avails -> Avails
1953 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
1954 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
1957 Extracting the bindings from a bunch of Avails.
1958 The bindings do *not* come back sorted in dependency order.
1959 We assume that they'll be wrapped in a big Rec, so that the
1960 dependency analyser can sort them out later
1963 extractResults :: Avails
1965 -> TcM ( TcDictBinds, -- Bindings
1966 [Inst]) -- Irreducible ones
1968 extractResults (Avails _ avails) wanteds
1969 = go avails emptyBag [] wanteds
1971 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
1972 -> TcM (TcDictBinds, [Inst])
1973 go avails binds irreds []
1974 = returnM (binds, irreds)
1976 go avails binds irreds (w:ws)
1977 = case findAvailEnv avails w of
1978 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
1979 go avails binds irreds ws
1981 Just IsIrred -> go (add_given avails w) binds (w:irreds) ws
1985 -> go avails binds irreds ws
1986 -- The sought Id can be one of the givens, via a superclass chain
1987 -- and then we definitely don't want to generate an x=x binding!
1990 -> go avails (addBind binds w (nlHsVar id)) irreds ws
1992 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds (ws' ++ ws)
1994 new_binds = addBind binds w rhs
1996 add_given avails w = extendAvailEnv avails w (Given (instToId w))
1998 addBind binds inst rhs = binds `unionBags` unitBag (L (instSpan inst)
1999 (VarBind (instToId inst) rhs))
2003 Note [No superclasses for Stop]
2004 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2005 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2006 add it to avails, so that any other equal Insts will be commoned up
2007 right here. However, we do *not* add superclasses. If we have
2010 but a is not bound here, then we *don't* want to derive dn from df
2011 here lest we lose sharing.
2014 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2015 addWanted want_scs avails wanted rhs_expr wanteds
2016 = addAvailAndSCs want_scs avails wanted avail
2018 avail = Rhs rhs_expr wanteds
2020 addGiven :: Avails -> Inst -> TcM Avails
2021 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2022 -- Always add superclasses for 'givens'
2024 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2025 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2026 -- so the assert isn't true
2028 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2029 addRefinedGiven reft (refined_givens, avails) given
2030 | isDict given -- We sometimes have 'given' methods, but they
2031 -- are always optional, so we can drop them
2032 , Just (co, pred) <- refinePred reft (dictPred given)
2033 = do { new_given <- newDictBndr (instLoc given) pred
2034 ; let rhs = L (instSpan given) $
2035 HsWrap (WpCo co) (HsVar (instToId given))
2036 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2037 ; return (new_given:refined_givens, avails) }
2038 -- ToDo: the superclasses of the original given all exist in Avails
2039 -- so we could really just cast them, but it's more awkward to do,
2040 -- and hopefully the optimiser will spot the duplicated work
2042 = return (refined_givens, avails)
2044 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2045 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2046 addAvailAndSCs want_scs avails irred IsIrred
2048 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2049 addAvailAndSCs want_scs avails inst avail
2050 | not (isClassDict inst) = extendAvails avails inst avail
2051 | NoSCs <- want_scs = extendAvails avails inst avail
2052 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2053 ; avails' <- extendAvails avails inst avail
2054 ; addSCs is_loop avails' inst }
2056 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2057 -- Note: this compares by *type*, not by Unique
2058 deps = findAllDeps (unitVarSet (instToId inst)) avail
2059 dep_tys = map idType (varSetElems deps)
2061 findAllDeps :: IdSet -> AvailHow -> IdSet
2062 -- Find all the Insts that this one depends on
2063 -- See Note [SUPERCLASS-LOOP 2]
2064 -- Watch out, though. Since the avails may contain loops
2065 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2066 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2067 findAllDeps so_far other = so_far
2069 find_all :: IdSet -> Inst -> IdSet
2071 | kid_id `elemVarSet` so_far = so_far
2072 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2073 | otherwise = so_far'
2075 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2076 kid_id = instToId kid
2078 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2079 -- Add all the superclasses of the Inst to Avails
2080 -- The first param says "dont do this because the original thing
2081 -- depends on this one, so you'd build a loop"
2082 -- Invariant: the Inst is already in Avails.
2084 addSCs is_loop avails dict
2085 = ASSERT( isDict dict )
2086 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2087 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2089 (clas, tys) = getDictClassTys dict
2090 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2091 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2093 add_sc avails (sc_dict, sc_sel)
2094 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2095 | is_given sc_dict = return avails
2096 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2097 ; addSCs is_loop avails' sc_dict }
2099 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2100 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2102 is_given :: Inst -> Bool
2103 is_given sc_dict = case findAvail avails sc_dict of
2104 Just (Given _) -> True -- Given is cheaper than superclass selection
2108 %************************************************************************
2110 \section{tcSimplifyTop: defaulting}
2112 %************************************************************************
2115 @tcSimplifyTop@ is called once per module to simplify all the constant
2116 and ambiguous Insts.
2118 We need to be careful of one case. Suppose we have
2120 instance Num a => Num (Foo a b) where ...
2122 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2123 to (Num x), and default x to Int. But what about y??
2125 It's OK: the final zonking stage should zap y to (), which is fine.
2129 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2130 tcSimplifyTop wanteds
2131 = tc_simplify_top doc False wanteds
2133 doc = text "tcSimplifyTop"
2135 tcSimplifyInteractive wanteds
2136 = tc_simplify_top doc True wanteds
2138 doc = text "tcSimplifyInteractive"
2140 -- The TcLclEnv should be valid here, solely to improve
2141 -- error message generation for the monomorphism restriction
2142 tc_simplify_top doc interactive wanteds
2143 = do { wanteds <- mapM zonkInst wanteds
2144 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2146 ; (binds1, irreds1) <- topCheckLoop doc wanteds
2148 ; if null irreds1 then
2151 -- OK, so there are some errors
2152 { -- Use the defaulting rules to do extra unification
2153 -- NB: irreds are already zonked
2154 ; extended_default <- if interactive then return True
2155 else doptM Opt_ExtendedDefaultRules
2156 ; disambiguate extended_default irreds1 -- Does unification
2157 ; (binds2, irreds2) <- topCheckLoop doc irreds1
2159 -- Deal with implicit parameter
2160 ; let (bad_ips, non_ips) = partition isIPDict irreds2
2161 (ambigs, others) = partition isTyVarDict non_ips
2163 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2165 ; addNoInstanceErrs others
2166 ; addTopAmbigErrs ambigs
2168 ; return (binds1 `unionBags` binds2) }}
2171 If a dictionary constrains a type variable which is
2172 * not mentioned in the environment
2173 * and not mentioned in the type of the expression
2174 then it is ambiguous. No further information will arise to instantiate
2175 the type variable; nor will it be generalised and turned into an extra
2176 parameter to a function.
2178 It is an error for this to occur, except that Haskell provided for
2179 certain rules to be applied in the special case of numeric types.
2181 * at least one of its classes is a numeric class, and
2182 * all of its classes are numeric or standard
2183 then the type variable can be defaulted to the first type in the
2184 default-type list which is an instance of all the offending classes.
2186 So here is the function which does the work. It takes the ambiguous
2187 dictionaries and either resolves them (producing bindings) or
2188 complains. It works by splitting the dictionary list by type
2189 variable, and using @disambigOne@ to do the real business.
2191 @disambigOne@ assumes that its arguments dictionaries constrain all
2192 the same type variable.
2194 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2195 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2196 the most common use of defaulting is code like:
2198 _ccall_ foo `seqPrimIO` bar
2200 Since we're not using the result of @foo@, the result if (presumably)
2204 disambiguate :: Bool -> [Inst] -> TcM ()
2205 -- Just does unification to fix the default types
2206 -- The Insts are assumed to be pre-zonked
2207 disambiguate extended_defaulting insts
2208 | null defaultable_groups
2211 = do { -- Figure out what default types to use
2212 mb_defaults <- getDefaultTys
2213 ; default_tys <- case mb_defaults of
2214 Just tys -> return tys
2215 Nothing -> -- No use-supplied default;
2216 -- use [Integer, Double]
2217 do { integer_ty <- tcMetaTy integerTyConName
2218 ; checkWiredInTyCon doubleTyCon
2219 ; return [integer_ty, doubleTy] }
2220 ; mapM_ (disambigGroup default_tys) defaultable_groups }
2222 unaries :: [(Inst,Class, TcTyVar)] -- (C tv) constraints
2223 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2224 (unaries, bad_tvs) = getDefaultableDicts insts
2226 -- Group by type variable
2227 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2228 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2229 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2231 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2232 defaultable_group ds@((_,_,tv):_)
2233 = not (isSkolemTyVar tv) -- Note [Avoiding spurious errors]
2234 && not (tv `elemVarSet` bad_tvs)
2235 && defaultable_classes [c | (_,c,_) <- ds]
2236 defaultable_group [] = panic "defaultable_group"
2238 defaultable_classes clss
2239 | extended_defaulting = any isInteractiveClass clss
2240 | otherwise = all isStandardClass clss && any isNumericClass clss
2242 -- In interactive mode, or with -fextended-default-rules,
2243 -- we default Show a to Show () to avoid graututious errors on "show []"
2244 isInteractiveClass cls
2245 = isNumericClass cls
2246 || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2249 disambigGroup :: [Type] -- The default types
2250 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2251 -> TcM () -- Just does unification, to fix the default types
2253 disambigGroup default_tys dicts
2254 = try_default default_tys
2256 (_,_,tyvar) = head dicts -- Should be non-empty
2257 classes = [c | (_,c,_) <- dicts]
2259 try_default [] = return ()
2260 try_default (default_ty : default_tys)
2261 = tryTcLIE_ (try_default default_tys) $
2262 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2263 -- This may fail; then the tryTcLIE_ kicks in
2264 -- Failure here is caused by there being no type in the
2265 -- default list which can satisfy all the ambiguous classes.
2266 -- For example, if Real a is reqd, but the only type in the
2267 -- default list is Int.
2269 -- After this we can't fail
2270 ; warnDefault dicts default_ty
2271 ; unifyType default_ty (mkTyVarTy tyvar) }
2274 Note [Avoiding spurious errors]
2275 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2276 When doing the unification for defaulting, we check for skolem
2277 type variables, and simply don't default them. For example:
2278 f = (*) -- Monomorphic
2279 g :: Num a => a -> a
2281 Here, we get a complaint when checking the type signature for g,
2282 that g isn't polymorphic enough; but then we get another one when
2283 dealing with the (Num a) context arising from f's definition;
2284 we try to unify a with Int (to default it), but find that it's
2285 already been unified with the rigid variable from g's type sig
2288 %************************************************************************
2290 \subsection[simple]{@Simple@ versions}
2292 %************************************************************************
2294 Much simpler versions when there are no bindings to make!
2296 @tcSimplifyThetas@ simplifies class-type constraints formed by
2297 @deriving@ declarations and when specialising instances. We are
2298 only interested in the simplified bunch of class/type constraints.
2300 It simplifies to constraints of the form (C a b c) where
2301 a,b,c are type variables. This is required for the context of
2302 instance declarations.
2305 tcSimplifyDeriv :: InstOrigin
2308 -> ThetaType -- Wanted
2309 -> TcM ThetaType -- Needed
2311 tcSimplifyDeriv orig tc tyvars theta
2312 = tcInstTyVars tyvars `thenM` \ (tvs, _, tenv) ->
2313 -- The main loop may do unification, and that may crash if
2314 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2315 -- ToDo: what if two of them do get unified?
2316 newDictBndrsO orig (substTheta tenv theta) `thenM` \ wanteds ->
2317 topCheckLoop doc wanteds `thenM` \ (_, irreds) ->
2319 doptM Opt_GlasgowExts `thenM` \ gla_exts ->
2320 doptM Opt_AllowUndecidableInstances `thenM` \ undecidable_ok ->
2322 inst_ty = mkTyConApp tc (mkTyVarTys tvs)
2323 (ok_insts, bad_insts) = partition is_ok_inst irreds
2325 = isDict inst -- Exclude implication consraints
2326 && (isTyVarClassPred pred || (gla_exts && ok_gla_pred pred))
2328 pred = dictPred inst
2330 ok_gla_pred pred = null (checkInstTermination [inst_ty] [pred])
2331 -- See Note [Deriving context]
2333 tv_set = mkVarSet tvs
2334 simpl_theta = map dictPred ok_insts
2335 weird_preds = [pred | pred <- simpl_theta
2336 , not (tyVarsOfPred pred `subVarSet` tv_set)]
2338 -- Check for a bizarre corner case, when the derived instance decl should
2339 -- have form instance C a b => D (T a) where ...
2340 -- Note that 'b' isn't a parameter of T. This gives rise to all sorts
2341 -- of problems; in particular, it's hard to compare solutions for
2342 -- equality when finding the fixpoint. So I just rule it out for now.
2344 rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2345 -- This reverse-mapping is a Royal Pain,
2346 -- but the result should mention TyVars not TcTyVars
2348 -- In effect, the bad and wierd insts cover all of the cases that
2349 -- would make checkValidInstance fail; if it were called right after tcSimplifyDeriv
2350 -- * wierd_preds ensures unambiguous instances (checkAmbiguity in checkValidInstance)
2351 -- * ok_gla_pred ensures termination (checkInstTermination in checkValidInstance)
2352 addNoInstanceErrs bad_insts `thenM_`
2353 mapM_ (addErrTc . badDerivedPred) weird_preds `thenM_`
2354 returnM (substTheta rev_env simpl_theta)
2356 doc = ptext SLIT("deriving classes for a data type")
2359 Note [Deriving context]
2360 ~~~~~~~~~~~~~~~~~~~~~~~
2361 With -fglasgow-exts, we allow things like (C Int a) in the simplified
2362 context for a derived instance declaration, because at a use of this
2363 instance, we might know that a=Bool, and have an instance for (C Int
2366 We nevertheless insist that each predicate meets the termination
2367 conditions. If not, the deriving mechanism generates larger and larger
2368 constraints. Example:
2370 data Seq a = Cons a (Seq (Succ a)) | Nil deriving Show
2372 Note the lack of a Show instance for Succ. First we'll generate
2373 instance (Show (Succ a), Show a) => Show (Seq a)
2375 instance (Show (Succ (Succ a)), Show (Succ a), Show a) => Show (Seq a)
2376 and so on. Instead we want to complain of no instance for (Show (Succ a)).
2380 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2381 used with \tr{default} declarations. We are only interested in
2382 whether it worked or not.
2385 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2388 tcSimplifyDefault theta
2389 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2390 topCheckLoop doc wanteds `thenM` \ (_, irreds) ->
2391 addNoInstanceErrs irreds `thenM_`
2397 doc = ptext SLIT("default declaration")
2401 %************************************************************************
2403 \section{Errors and contexts}
2405 %************************************************************************
2407 ToDo: for these error messages, should we note the location as coming
2408 from the insts, or just whatever seems to be around in the monad just
2412 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2413 -> [Inst] -- The offending Insts
2415 -- Group together insts with the same origin
2416 -- We want to report them together in error messages
2418 groupErrs report_err []
2420 groupErrs report_err (inst:insts)
2421 = do_one (inst:friends) `thenM_`
2422 groupErrs report_err others
2425 -- (It may seem a bit crude to compare the error messages,
2426 -- but it makes sure that we combine just what the user sees,
2427 -- and it avoids need equality on InstLocs.)
2428 (friends, others) = partition is_friend insts
2429 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2430 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2431 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2432 -- Add location and context information derived from the Insts
2434 -- Add the "arising from..." part to a message about bunch of dicts
2435 addInstLoc :: [Inst] -> Message -> Message
2436 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2438 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2439 addTopIPErrs bndrs []
2441 addTopIPErrs bndrs ips
2442 = addErrTcM (tidy_env, mk_msg tidy_ips)
2444 (tidy_env, tidy_ips) = tidyInsts ips
2445 mk_msg ips = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2446 nest 2 (ptext SLIT("the monomorphic top-level binding")
2447 <> plural bndrs <+> ptext SLIT("of")
2448 <+> pprBinders bndrs <> colon)],
2449 nest 2 (vcat (map ppr_ip ips)),
2451 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2453 topIPErrs :: [Inst] -> TcM ()
2455 = groupErrs report tidy_dicts
2457 (tidy_env, tidy_dicts) = tidyInsts dicts
2458 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2459 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2460 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2462 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2464 addNoInstanceErrs insts
2465 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2466 ; reportNoInstances tidy_env Nothing tidy_insts }
2470 -> Maybe (InstLoc, [Inst]) -- Context
2471 -- Nothing => top level
2472 -- Just (d,g) => d describes the construct
2474 -> [Inst] -- What is wanted (can include implications)
2477 reportNoInstances tidy_env mb_what insts
2478 = groupErrs (report_no_instances tidy_env mb_what) insts
2480 report_no_instances tidy_env mb_what insts
2481 = do { inst_envs <- tcGetInstEnvs
2482 ; let (implics, insts1) = partition isImplicInst insts
2483 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2484 ; traceTc (text "reportNoInstnces" <+> vcat
2485 [ppr implics, ppr insts1, ppr insts2])
2486 ; mapM_ complain_implic implics
2487 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2488 ; groupErrs complain_no_inst insts2 }
2490 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2492 complain_implic inst -- Recurse!
2493 = reportNoInstances tidy_env
2494 (Just (tci_loc inst, tci_given inst))
2497 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2498 -- Right msg => overlap message
2499 -- Left inst => no instance
2500 check_overlap inst_envs wanted
2501 | not (isClassDict wanted) = Left wanted
2503 = case lookupInstEnv inst_envs clas tys of
2504 -- The case of exactly one match and no unifiers means
2505 -- a successful lookup. That can't happen here, becuase
2506 -- dicts only end up here if they didn't match in Inst.lookupInst
2508 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2510 ([], _) -> Left wanted -- No match
2511 res -> Right (mk_overlap_msg wanted res)
2513 (clas,tys) = getDictClassTys wanted
2515 mk_overlap_msg dict (matches, unifiers)
2516 = vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2517 <+> pprPred (dictPred dict))),
2518 sep [ptext SLIT("Matching instances") <> colon,
2519 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2520 ASSERT( not (null matches) )
2521 if not (isSingleton matches)
2522 then -- Two or more matches
2524 else -- One match, plus some unifiers
2525 ASSERT( not (null unifiers) )
2526 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2527 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2528 ptext SLIT("Use -fallow-incoherent-instances to use the first choice above")])]
2530 ispecs = [ispec | (_, ispec) <- matches]
2532 mk_no_inst_err insts
2533 | null insts = empty
2535 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2536 not (isEmptyVarSet (tyVarsOfInsts insts))
2537 = vcat [ addInstLoc insts $
2538 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2539 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2540 , show_fixes (fix1 loc : fixes2) ]
2542 | otherwise -- Top level
2543 = vcat [ addInstLoc insts $
2544 ptext SLIT("No instance") <> plural insts
2545 <+> ptext SLIT("for") <+> pprDictsTheta insts
2546 , show_fixes fixes2 ]
2549 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2550 <+> ptext SLIT("to the context of"),
2551 nest 2 (ppr (instLocOrigin loc)) ]
2552 -- I'm not sure it helps to add the location
2553 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2555 fixes2 | null instance_dicts = []
2556 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2557 pprDictsTheta instance_dicts]]
2558 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2559 -- Insts for which it is worth suggesting an adding an instance declaration
2560 -- Exclude implicit parameters, and tyvar dicts
2562 show_fixes :: [SDoc] -> SDoc
2563 show_fixes [] = empty
2564 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2565 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2567 addTopAmbigErrs dicts
2568 -- Divide into groups that share a common set of ambiguous tyvars
2569 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2570 -- See Note [Avoiding spurious errors]
2571 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2573 (tidy_env, tidy_dicts) = tidyInsts dicts
2575 tvs_of :: Inst -> [TcTyVar]
2576 tvs_of d = varSetElems (tyVarsOfInst d)
2577 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2579 report :: [(Inst,[TcTyVar])] -> TcM ()
2580 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2581 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2582 setSrcSpan (instSpan inst) $
2583 -- the location of the first one will do for the err message
2584 addErrTcM (tidy_env, msg $$ mono_msg)
2586 dicts = map fst pairs
2587 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2588 pprQuotedList tvs <+> in_msg,
2589 nest 2 (pprDictsInFull dicts)]
2590 in_msg = text "in the constraint" <> plural dicts <> colon
2591 report [] = panic "addTopAmbigErrs"
2594 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2595 -- There's an error with these Insts; if they have free type variables
2596 -- it's probably caused by the monomorphism restriction.
2597 -- Try to identify the offending variable
2598 -- ASSUMPTION: the Insts are fully zonked
2599 mkMonomorphismMsg tidy_env inst_tvs
2600 = findGlobals (mkVarSet inst_tvs) tidy_env `thenM` \ (tidy_env, docs) ->
2601 returnM (tidy_env, mk_msg docs)
2603 mk_msg [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2604 -- This happens in things like
2605 -- f x = show (read "foo")
2606 -- where monomorphism doesn't play any role
2607 mk_msg docs = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2611 monomorphism_fix :: SDoc
2612 monomorphism_fix = ptext SLIT("Probable fix:") <+>
2613 (ptext SLIT("give these definition(s) an explicit type signature")
2614 $$ ptext SLIT("or use -fno-monomorphism-restriction"))
2616 warnDefault ups default_ty
2617 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2618 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2620 dicts = [d | (d,_,_) <- ups]
2623 (_, tidy_dicts) = tidyInsts dicts
2624 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2625 quotes (ppr default_ty),
2626 pprDictsInFull tidy_dicts]
2628 -- Used for the ...Thetas variants; all top level
2630 = vcat [ptext SLIT("Can't derive instances where the instance context mentions"),
2631 ptext SLIT("type variables that are not data type parameters"),
2632 nest 2 (ptext SLIT("Offending constraint:") <+> ppr pred)]
2634 reduceDepthErr n stack
2635 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2636 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2637 nest 4 (pprStack stack)]
2639 pprStack stack = vcat (map pprInstInFull stack)