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,
18 bindInstsOfLocalFuns, bindIrreds,
21 #include "HsVersions.h"
23 import {-# SOURCE #-} TcUnify( unifyType )
59 %************************************************************************
63 %************************************************************************
65 --------------------------------------
66 Notes on functional dependencies (a bug)
67 --------------------------------------
74 instance D a b => C a b -- Undecidable
75 -- (Not sure if it's crucial to this eg)
76 f :: C a b => a -> Bool
79 g :: C a b => a -> Bool
82 Here f typechecks, but g does not!! Reason: before doing improvement,
83 we reduce the (C a b1) constraint from the call of f to (D a b1).
85 Here is a more complicated example:
87 | > class Foo a b | a->b
89 | > class Bar a b | a->b
93 | > instance Bar Obj Obj
95 | > instance (Bar a b) => Foo a b
97 | > foo:: (Foo a b) => a -> String
100 | > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
106 | Could not deduce (Bar a b) from the context (Foo a b)
107 | arising from use of `foo' at <interactive>:1
109 | Add (Bar a b) to the expected type of an expression
110 | In the first argument of `runFoo', namely `foo'
111 | In the definition of `it': it = runFoo foo
113 | Why all of the sudden does GHC need the constraint Bar a b? The
114 | function foo didn't ask for that...
116 The trouble is that to type (runFoo foo), GHC has to solve the problem:
118 Given constraint Foo a b
119 Solve constraint Foo a b'
121 Notice that b and b' aren't the same. To solve this, just do
122 improvement and then they are the same. But GHC currently does
127 That is usually fine, but it isn't here, because it sees that Foo a b is
128 not the same as Foo a b', and so instead applies the instance decl for
129 instance Bar a b => Foo a b. And that's where the Bar constraint comes
132 The Right Thing is to improve whenever the constraint set changes at
133 all. Not hard in principle, but it'll take a bit of fiddling to do.
137 --------------------------------------
138 Notes on quantification
139 --------------------------------------
141 Suppose we are about to do a generalisation step.
145 T the type of the RHS
146 C the constraints from that RHS
148 The game is to figure out
150 Q the set of type variables over which to quantify
151 Ct the constraints we will *not* quantify over
152 Cq the constraints we will quantify over
154 So we're going to infer the type
158 and float the constraints Ct further outwards.
160 Here are the things that *must* be true:
162 (A) Q intersect fv(G) = EMPTY limits how big Q can be
163 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
165 (A) says we can't quantify over a variable that's free in the
166 environment. (B) says we must quantify over all the truly free
167 variables in T, else we won't get a sufficiently general type. We do
168 not *need* to quantify over any variable that is fixed by the free
169 vars of the environment G.
171 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
173 Example: class H x y | x->y where ...
175 fv(G) = {a} C = {H a b, H c d}
178 (A) Q intersect {a} is empty
179 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
181 So Q can be {c,d}, {b,c,d}
183 Other things being equal, however, we'd like to quantify over as few
184 variables as possible: smaller types, fewer type applications, more
185 constraints can get into Ct instead of Cq.
188 -----------------------------------------
191 fv(T) the free type vars of T
193 oclose(vs,C) The result of extending the set of tyvars vs
194 using the functional dependencies from C
196 grow(vs,C) The result of extend the set of tyvars vs
197 using all conceivable links from C.
199 E.g. vs = {a}, C = {H [a] b, K (b,Int) c, Eq e}
200 Then grow(vs,C) = {a,b,c}
202 Note that grow(vs,C) `superset` grow(vs,simplify(C))
203 That is, simplfication can only shrink the result of grow.
206 oclose is conservative one way: v `elem` oclose(vs,C) => v is definitely fixed by vs
207 grow is conservative the other way: if v might be fixed by vs => v `elem` grow(vs,C)
210 -----------------------------------------
214 Here's a good way to choose Q:
216 Q = grow( fv(T), C ) \ oclose( fv(G), C )
218 That is, quantify over all variable that that MIGHT be fixed by the
219 call site (which influences T), but which aren't DEFINITELY fixed by
220 G. This choice definitely quantifies over enough type variables,
221 albeit perhaps too many.
223 Why grow( fv(T), C ) rather than fv(T)? Consider
225 class H x y | x->y where ...
230 If we used fv(T) = {c} we'd get the type
232 forall c. H c d => c -> b
234 And then if the fn was called at several different c's, each of
235 which fixed d differently, we'd get a unification error, because
236 d isn't quantified. Solution: quantify d. So we must quantify
237 everything that might be influenced by c.
239 Why not oclose( fv(T), C )? Because we might not be able to see
240 all the functional dependencies yet:
242 class H x y | x->y where ...
243 instance H x y => Eq (T x y) where ...
248 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
249 apparent yet, and that's wrong. We must really quantify over d too.
252 There really isn't any point in quantifying over any more than
253 grow( fv(T), C ), because the call sites can't possibly influence
254 any other type variables.
258 -------------------------------------
260 -------------------------------------
262 It's very hard to be certain when a type is ambiguous. Consider
266 instance H x y => K (x,y)
268 Is this type ambiguous?
269 forall a b. (K (a,b), Eq b) => a -> a
271 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
272 now we see that a fixes b. So we can't tell about ambiguity for sure
273 without doing a full simplification. And even that isn't possible if
274 the context has some free vars that may get unified. Urgle!
276 Here's another example: is this ambiguous?
277 forall a b. Eq (T b) => a -> a
278 Not if there's an insance decl (with no context)
279 instance Eq (T b) where ...
281 You may say of this example that we should use the instance decl right
282 away, but you can't always do that:
284 class J a b where ...
285 instance J Int b where ...
287 f :: forall a b. J a b => a -> a
289 (Notice: no functional dependency in J's class decl.)
290 Here f's type is perfectly fine, provided f is only called at Int.
291 It's premature to complain when meeting f's signature, or even
292 when inferring a type for f.
296 However, we don't *need* to report ambiguity right away. It'll always
297 show up at the call site.... and eventually at main, which needs special
298 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
300 So here's the plan. We WARN about probable ambiguity if
302 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
304 (all tested before quantification).
305 That is, all the type variables in Cq must be fixed by the the variables
306 in the environment, or by the variables in the type.
308 Notice that we union before calling oclose. Here's an example:
310 class J a b c | a b -> c
314 forall b c. (J a b c) => b -> b
316 Only if we union {a} from G with {b} from T before using oclose,
317 do we see that c is fixed.
319 It's a bit vague exactly which C we should use for this oclose call. If we
320 don't fix enough variables we might complain when we shouldn't (see
321 the above nasty example). Nothing will be perfect. That's why we can
322 only issue a warning.
325 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
327 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
329 then c is a "bubble"; there's no way it can ever improve, and it's
330 certainly ambiguous. UNLESS it is a constant (sigh). And what about
335 instance H x y => K (x,y)
337 Is this type ambiguous?
338 forall a b. (K (a,b), Eq b) => a -> a
340 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
341 is a "bubble" that's a set of constraints
343 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
345 Hence another idea. To decide Q start with fv(T) and grow it
346 by transitive closure in Cq (no functional dependencies involved).
347 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
348 The definitely-ambiguous can then float out, and get smashed at top level
349 (which squashes out the constants, like Eq (T a) above)
352 --------------------------------------
353 Notes on principal types
354 --------------------------------------
359 f x = let g y = op (y::Int) in True
361 Here the principal type of f is (forall a. a->a)
362 but we'll produce the non-principal type
363 f :: forall a. C Int => a -> a
366 --------------------------------------
367 The need for forall's in constraints
368 --------------------------------------
370 [Exchange on Haskell Cafe 5/6 Dec 2000]
372 class C t where op :: t -> Bool
373 instance C [t] where op x = True
375 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
376 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
378 The definitions of p and q differ only in the order of the components in
379 the pair on their right-hand sides. And yet:
381 ghc and "Typing Haskell in Haskell" reject p, but accept q;
382 Hugs rejects q, but accepts p;
383 hbc rejects both p and q;
384 nhc98 ... (Malcolm, can you fill in the blank for us!).
386 The type signature for f forces context reduction to take place, and
387 the results of this depend on whether or not the type of y is known,
388 which in turn depends on which component of the pair the type checker
391 Solution: if y::m a, float out the constraints
392 Monad m, forall c. C (m c)
393 When m is later unified with [], we can solve both constraints.
396 --------------------------------------
397 Notes on implicit parameters
398 --------------------------------------
400 Note [Inheriting implicit parameters]
401 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
406 where f is *not* a top-level binding.
407 From the RHS of f we'll get the constraint (?y::Int).
408 There are two types we might infer for f:
412 (so we get ?y from the context of f's definition), or
414 f :: (?y::Int) => Int -> Int
416 At first you might think the first was better, becuase then
417 ?y behaves like a free variable of the definition, rather than
418 having to be passed at each call site. But of course, the WHOLE
419 IDEA is that ?y should be passed at each call site (that's what
420 dynamic binding means) so we'd better infer the second.
422 BOTTOM LINE: when *inferring types* you *must* quantify
423 over implicit parameters. See the predicate isFreeWhenInferring.
426 Note [Implicit parameters and ambiguity]
427 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
428 What type should we infer for this?
429 f x = (show ?y, x::Int)
430 Since we must quantify over the ?y, the most plausible type is
431 f :: (Show a, ?y::a) => Int -> (String, Int)
432 But notice that the type of the RHS is (String,Int), with no type
433 varibables mentioned at all! The type of f looks ambiguous. But
434 it isn't, because at a call site we might have
435 let ?y = 5::Int in f 7
436 and all is well. In effect, implicit parameters are, well, parameters,
437 so we can take their type variables into account as part of the
438 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
441 Question 2: type signatures
442 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
443 BUT WATCH OUT: When you supply a type signature, we can't force you
444 to quantify over implicit parameters. For example:
448 This is perfectly reasonable. We do not want to insist on
450 (?x + 1) :: (?x::Int => Int)
452 That would be silly. Here, the definition site *is* the occurrence site,
453 so the above strictures don't apply. Hence the difference between
454 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
455 and tcSimplifyCheckBind (which does not).
457 What about when you supply a type signature for a binding?
458 Is it legal to give the following explicit, user type
459 signature to f, thus:
464 At first sight this seems reasonable, but it has the nasty property
465 that adding a type signature changes the dynamic semantics.
468 (let f x = (x::Int) + ?y
469 in (f 3, f 3 with ?y=5)) with ?y = 6
475 in (f 3, f 3 with ?y=5)) with ?y = 6
479 Indeed, simply inlining f (at the Haskell source level) would change the
482 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
483 semantics for a Haskell program without knowing its typing, so if you
484 change the typing you may change the semantics.
486 To make things consistent in all cases where we are *checking* against
487 a supplied signature (as opposed to inferring a type), we adopt the
490 a signature does not need to quantify over implicit params.
492 [This represents a (rather marginal) change of policy since GHC 5.02,
493 which *required* an explicit signature to quantify over all implicit
494 params for the reasons mentioned above.]
496 But that raises a new question. Consider
498 Given (signature) ?x::Int
499 Wanted (inferred) ?x::Int, ?y::Bool
501 Clearly we want to discharge the ?x and float the ?y out. But
502 what is the criterion that distinguishes them? Clearly it isn't
503 what free type variables they have. The Right Thing seems to be
504 to float a constraint that
505 neither mentions any of the quantified type variables
506 nor any of the quantified implicit parameters
508 See the predicate isFreeWhenChecking.
511 Question 3: monomorphism
512 ~~~~~~~~~~~~~~~~~~~~~~~~
513 There's a nasty corner case when the monomorphism restriction bites:
517 The argument above suggests that we *must* generalise
518 over the ?y parameter, to get
519 z :: (?y::Int) => Int,
520 but the monomorphism restriction says that we *must not*, giving
522 Why does the momomorphism restriction say this? Because if you have
524 let z = x + ?y in z+z
526 you might not expect the addition to be done twice --- but it will if
527 we follow the argument of Question 2 and generalise over ?y.
530 Question 4: top level
531 ~~~~~~~~~~~~~~~~~~~~~
532 At the top level, monomorhism makes no sense at all.
535 main = let ?x = 5 in print foo
539 woggle :: (?x :: Int) => Int -> Int
542 We definitely don't want (foo :: Int) with a top-level implicit parameter
543 (?x::Int) becuase there is no way to bind it.
548 (A) Always generalise over implicit parameters
549 Bindings that fall under the monomorphism restriction can't
553 * Inlining remains valid
554 * No unexpected loss of sharing
555 * But simple bindings like
557 will be rejected, unless you add an explicit type signature
558 (to avoid the monomorphism restriction)
559 z :: (?y::Int) => Int
561 This seems unacceptable
563 (B) Monomorphism restriction "wins"
564 Bindings that fall under the monomorphism restriction can't
566 Always generalise over implicit parameters *except* for bindings
567 that fall under the monomorphism restriction
570 * Inlining isn't valid in general
571 * No unexpected loss of sharing
572 * Simple bindings like
574 accepted (get value of ?y from binding site)
576 (C) Always generalise over implicit parameters
577 Bindings that fall under the monomorphism restriction can't
578 be generalised, EXCEPT for implicit parameters
580 * Inlining remains valid
581 * Unexpected loss of sharing (from the extra generalisation)
582 * Simple bindings like
584 accepted (get value of ?y from occurrence sites)
589 None of these choices seems very satisfactory. But at least we should
590 decide which we want to do.
592 It's really not clear what is the Right Thing To Do. If you see
596 would you expect the value of ?y to be got from the *occurrence sites*
597 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
598 case of function definitions, the answer is clearly the former, but
599 less so in the case of non-fucntion definitions. On the other hand,
600 if we say that we get the value of ?y from the definition site of 'z',
601 then inlining 'z' might change the semantics of the program.
603 Choice (C) really says "the monomorphism restriction doesn't apply
604 to implicit parameters". Which is fine, but remember that every
605 innocent binding 'x = ...' that mentions an implicit parameter in
606 the RHS becomes a *function* of that parameter, called at each
607 use of 'x'. Now, the chances are that there are no intervening 'with'
608 clauses that bind ?y, so a decent compiler should common up all
609 those function calls. So I think I strongly favour (C). Indeed,
610 one could make a similar argument for abolishing the monomorphism
611 restriction altogether.
613 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
617 %************************************************************************
619 \subsection{tcSimplifyInfer}
621 %************************************************************************
623 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
625 1. Compute Q = grow( fvs(T), C )
627 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
628 predicates will end up in Ct; we deal with them at the top level
630 3. Try improvement, using functional dependencies
632 4. If Step 3 did any unification, repeat from step 1
633 (Unification can change the result of 'grow'.)
635 Note: we don't reduce dictionaries in step 2. For example, if we have
636 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
637 after step 2. However note that we may therefore quantify over more
638 type variables than we absolutely have to.
640 For the guts, we need a loop, that alternates context reduction and
641 improvement with unification. E.g. Suppose we have
643 class C x y | x->y where ...
645 and tcSimplify is called with:
647 Then improvement unifies a with b, giving
650 If we need to unify anything, we rattle round the whole thing all over
657 -> TcTyVarSet -- fv(T); type vars
659 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
660 [Inst], -- Dict Ids that must be bound here (zonked)
661 TcDictBinds) -- Bindings
662 -- Any free (escaping) Insts are tossed into the environment
667 tcSimplifyInfer doc tau_tvs wanted
668 = do { tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
669 ; wanted' <- mappM zonkInst wanted -- Zonk before deciding quantified tyvars
670 ; gbl_tvs <- tcGetGlobalTyVars
671 ; let preds = fdPredsOfInsts wanted'
672 qtvs = grow preds tau_tvs' `minusVarSet` oclose preds gbl_tvs
673 (free, bound) = partition (isFreeWhenInferring qtvs) wanted'
674 ; traceTc (text "infer" <+> (ppr preds $$ ppr (grow preds tau_tvs') $$ ppr gbl_tvs $$ ppr (oclose preds gbl_tvs) $$ ppr free $$ ppr bound))
677 -- To make types simple, reduce as much as possible
678 ; let try_me inst = ReduceMe AddSCs
679 ; (irreds, binds) <- checkLoop (mkRedEnv doc try_me []) bound
681 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
683 -- We can't abstract over implications
684 ; let (dicts, implics) = partition isDict irreds
685 ; loc <- getInstLoc (ImplicOrigin doc)
686 ; implic_bind <- bindIrreds loc qtvs' dicts implics
688 ; return (qtvs', dicts, binds `unionBags` implic_bind) }
689 -- NB: when we are done, we might have some bindings, but
690 -- the final qtvs might be empty. See Note [NO TYVARS] below.
694 -----------------------------------------------------------
695 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
696 -- against, but we don't know the type variables over which we are going to quantify.
697 -- This happens when we have a type signature for a mutually recursive group
700 -> TcTyVarSet -- fv(T)
703 -> TcM ([TyVar], -- Fully zonked, and quantified
704 TcDictBinds) -- Bindings
706 tcSimplifyInferCheck loc tau_tvs givens wanteds
707 = do { (irreds, binds) <- innerCheckLoop loc givens wanteds
709 -- Figure out which type variables to quantify over
710 -- You might think it should just be the signature tyvars,
711 -- but in bizarre cases you can get extra ones
712 -- f :: forall a. Num a => a -> a
713 -- f x = fst (g (x, head [])) + 1
715 -- Here we infer g :: forall a b. a -> b -> (b,a)
716 -- We don't want g to be monomorphic in b just because
717 -- f isn't quantified over b.
718 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
719 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
720 ; gbl_tvs <- tcGetGlobalTyVars
721 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
722 -- We could close gbl_tvs, but its not necessary for
723 -- soundness, and it'll only affect which tyvars, not which
724 -- dictionaries, we quantify over
726 ; qtvs' <- zonkQuantifiedTyVars qtvs
728 -- Now we are back to normal (c.f. tcSimplCheck)
729 ; implic_bind <- bindIrreds loc qtvs' givens irreds
731 ; return (qtvs', binds `unionBags` implic_bind) }
734 Note [Squashing methods]
735 ~~~~~~~~~~~~~~~~~~~~~~~~~
736 Be careful if you want to float methods more:
737 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
738 From an application (truncate f i) we get
741 If we have also have a second occurrence of truncate, we get
744 When simplifying with i,f free, we might still notice that
745 t1=t3; but alas, the binding for t2 (which mentions t1)
746 may continue to float out!
751 class Y a b | a -> b where
754 instance Y [[a]] a where
757 k :: X a -> X a -> X a
759 g :: Num a => [X a] -> [X a]
762 h ys = ys ++ map (k (y [[0]])) xs
764 The excitement comes when simplifying the bindings for h. Initially
765 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
766 From this we get t1:=:t2, but also various bindings. We can't forget
767 the bindings (because of [LOOP]), but in fact t1 is what g is
770 The net effect of [NO TYVARS]
773 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
774 isFreeWhenInferring qtvs inst
775 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
776 && isInheritableInst inst -- and no implicit parameter involved
777 -- see Note [Inheriting implicit parameters]
779 {- No longer used (with implication constraints)
780 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
781 -> NameSet -- Quantified implicit parameters
783 isFreeWhenChecking qtvs ips inst
784 = isFreeWrtTyVars qtvs inst
785 && isFreeWrtIPs ips inst
788 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
789 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
793 %************************************************************************
795 \subsection{tcSimplifyCheck}
797 %************************************************************************
799 @tcSimplifyCheck@ is used when we know exactly the set of variables
800 we are going to quantify over. For example, a class or instance declaration.
803 -----------------------------------------------------------
804 -- tcSimplifyCheck is used when checking expression type signatures,
805 -- class decls, instance decls etc.
806 tcSimplifyCheck :: InstLoc
807 -> [TcTyVar] -- Quantify over these
810 -> TcM TcDictBinds -- Bindings
811 tcSimplifyCheck loc qtvs givens wanteds
812 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
813 do { (irreds, binds) <- innerCheckLoop loc givens wanteds
814 ; implic_bind <- bindIrreds loc qtvs givens irreds
815 ; return (binds `unionBags` implic_bind) }
817 -----------------------------------------------------------
818 -- tcSimplifyCheckPat is used for existential pattern match
819 tcSimplifyCheckPat :: InstLoc
820 -> [CoVar] -> Refinement
821 -> [TcTyVar] -- Quantify over these
824 -> TcM TcDictBinds -- Bindings
825 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
826 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
827 do { (irreds, binds) <- innerCheckLoop loc givens wanteds
828 ; implic_bind <- bindIrredsR loc qtvs co_vars reft
830 ; return (binds `unionBags` implic_bind) }
832 -----------------------------------------------------------
833 bindIrreds :: InstLoc -> [TcTyVar]
836 bindIrreds loc qtvs givens irreds
837 = bindIrredsR loc qtvs [] emptyRefinement givens irreds
839 bindIrredsR :: InstLoc -> [TcTyVar] -> [CoVar]
840 -> Refinement -> [Inst] -> [Inst]
842 -- Make a binding that binds 'irreds', by generating an implication
843 -- constraint for them, *and* throwing the constraint into the LIE
844 bindIrredsR loc qtvs co_vars reft givens irreds
848 = do { let givens' = filter isDict givens
849 -- The givens can include methods
850 -- See Note [Pruning the givens in an implication constraint]
852 -- If there are no 'givens', then it's safe to
853 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
854 -- See Note [Freeness and implications]
855 ; irreds' <- if null givens'
857 { let qtv_set = mkVarSet qtvs
858 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
860 ; return real_irreds }
863 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
864 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
865 -- This call does the real work
870 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
872 -> TcM ([Inst], TcDictBinds)
873 -- Make a binding that binds 'irreds', by generating an implication
874 -- constraint for them, *and* throwing the constraint into the LIE
875 -- The binding looks like
876 -- (ir1, .., irn) = f qtvs givens
877 -- where f is (evidence for) the new implication constraint
879 -- This binding must line up the 'rhs' in reduceImplication
880 makeImplicationBind loc all_tvs reft
881 givens -- Guaranteed all Dicts
883 | null irreds -- If there are no irreds, we are done
884 = return ([], emptyBag)
885 | otherwise -- Otherwise we must generate a binding
886 = do { uniq <- newUnique
887 ; span <- getSrcSpanM
888 ; let name = mkInternalName uniq (mkVarOcc "ic") (srcSpanStart span)
889 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
890 tci_tyvars = all_tvs,
892 tci_wanted = irreds, tci_loc = loc }
894 ; let n_irreds = length irreds
895 irred_ids = map instToId irreds
896 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
897 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
898 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
899 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
900 bind | n_irreds==1 = VarBind (head irred_ids) rhs
901 | otherwise = PatBind { pat_lhs = L span pat,
902 pat_rhs = unguardedGRHSs rhs,
904 bind_fvs = placeHolderNames }
905 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
906 return ([implic_inst], unitBag (L span bind)) }
908 -----------------------------------------------------------
911 -> TcM ([Inst], TcDictBinds)
913 topCheckLoop doc wanteds
914 = checkLoop (mkRedEnv doc try_me []) wanteds
916 try_me inst = ReduceMe AddSCs
918 -----------------------------------------------------------
919 innerCheckLoop :: InstLoc
922 -> TcM ([Inst], TcDictBinds)
924 innerCheckLoop inst_loc givens wanteds
925 = checkLoop env wanteds
927 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
929 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
931 -- When checking against a given signature
932 -- we MUST be very gentle: Note [Check gently]
937 We have to very careful about not simplifying too vigorously
942 f :: Show b => T b -> b
945 Inside the pattern match, which binds (a:*, x:a), we know that
947 Hence we have a dictionary for Show [a] available; and indeed we
948 need it. We are going to build an implication contraint
949 forall a. (b~[a]) => Show [a]
950 Later, we will solve this constraint using the knowledge (Show b)
952 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
953 thing becomes insoluble. So we simplify gently (get rid of literals
954 and methods only, plus common up equal things), deferring the real
955 work until top level, when we solve the implication constraint
960 -----------------------------------------------------------
963 -> TcM ([Inst], TcDictBinds)
964 -- Precondition: givens are completely rigid
966 checkLoop env wanteds
967 = do { -- Givens are skolems, so no need to zonk them
968 wanteds' <- mappM zonkInst wanteds
970 ; (improved, binds, irreds) <- reduceContext env wanteds'
972 ; if not improved then
973 return (irreds, binds)
976 -- If improvement did some unification, we go round again.
977 -- We start again with irreds, not wanteds
978 -- Using an instance decl might have introduced a fresh type variable
979 -- which might have been unified, so we'd get an infinite loop
980 -- if we started again with wanteds! See Note [LOOP]
981 { (irreds1, binds1) <- checkLoop env irreds
982 ; return (irreds1, binds `unionBags` binds1) } }
987 class If b t e r | b t e -> r
990 class Lte a b c | a b -> c where lte :: a -> b -> c
992 instance (Lte a b l,If l b a c) => Max a b c
994 Wanted: Max Z (S x) y
996 Then we'll reduce using the Max instance to:
997 (Lte Z (S x) l, If l (S x) Z y)
998 and improve by binding l->T, after which we can do some reduction
999 on both the Lte and If constraints. What we *can't* do is start again
1000 with (Max Z (S x) y)!
1004 %************************************************************************
1006 tcSimplifySuperClasses
1008 %************************************************************************
1010 Note [SUPERCLASS-LOOP 1]
1011 ~~~~~~~~~~~~~~~~~~~~~~~~
1012 We have to be very, very careful when generating superclasses, lest we
1013 accidentally build a loop. Here's an example:
1017 class S a => C a where { opc :: a -> a }
1018 class S b => D b where { opd :: b -> b }
1020 instance C Int where
1023 instance D Int where
1026 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1027 Simplifying, we may well get:
1028 $dfCInt = :C ds1 (opd dd)
1031 Notice that we spot that we can extract ds1 from dd.
1033 Alas! Alack! We can do the same for (instance D Int):
1035 $dfDInt = :D ds2 (opc dc)
1039 And now we've defined the superclass in terms of itself.
1041 Solution: never generate a superclass selectors at all when
1042 satisfying the superclass context of an instance declaration.
1044 Two more nasty cases are in
1049 tcSimplifySuperClasses
1054 tcSimplifySuperClasses loc givens sc_wanteds
1055 = do { (irreds, binds1) <- checkLoop env sc_wanteds
1056 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1057 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1060 env = mkRedEnv (pprInstLoc loc) try_me givens
1061 try_me inst = ReduceMe NoSCs
1062 -- Like topCheckLoop, but with NoSCs
1066 %************************************************************************
1068 \subsection{tcSimplifyRestricted}
1070 %************************************************************************
1072 tcSimplifyRestricted infers which type variables to quantify for a
1073 group of restricted bindings. This isn't trivial.
1076 We want to quantify over a to get id :: forall a. a->a
1079 We do not want to quantify over a, because there's an Eq a
1080 constraint, so we get eq :: a->a->Bool (notice no forall)
1083 RHS has type 'tau', whose free tyvars are tau_tvs
1084 RHS has constraints 'wanteds'
1087 Quantify over (tau_tvs \ ftvs(wanteds))
1088 This is bad. The constraints may contain (Monad (ST s))
1089 where we have instance Monad (ST s) where...
1090 so there's no need to be monomorphic in s!
1092 Also the constraint might be a method constraint,
1093 whose type mentions a perfectly innocent tyvar:
1094 op :: Num a => a -> b -> a
1095 Here, b is unconstrained. A good example would be
1097 We want to infer the polymorphic type
1098 foo :: forall b. b -> b
1101 Plan B (cunning, used for a long time up to and including GHC 6.2)
1102 Step 1: Simplify the constraints as much as possible (to deal
1103 with Plan A's problem). Then set
1104 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1106 Step 2: Now simplify again, treating the constraint as 'free' if
1107 it does not mention qtvs, and trying to reduce it otherwise.
1108 The reasons for this is to maximise sharing.
1110 This fails for a very subtle reason. Suppose that in the Step 2
1111 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1112 In the Step 1 this constraint might have been simplified, perhaps to
1113 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1114 This won't happen in Step 2... but that in turn might prevent some other
1115 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1116 and that in turn breaks the invariant that no constraints are quantified over.
1118 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1123 Step 1: Simplify the constraints as much as possible (to deal
1124 with Plan A's problem). Then set
1125 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1126 Return the bindings from Step 1.
1129 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1132 instance (HasBinary ty IO) => HasCodedValue ty
1134 foo :: HasCodedValue a => String -> IO a
1136 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1137 doDecodeIO codedValue view
1138 = let { act = foo "foo" } in act
1140 You might think this should work becuase the call to foo gives rise to a constraint
1141 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1142 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1143 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1145 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1149 Plan D (a variant of plan B)
1150 Step 1: Simplify the constraints as much as possible (to deal
1151 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1152 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1154 Step 2: Now simplify again, treating the constraint as 'free' if
1155 it does not mention qtvs, and trying to reduce it otherwise.
1157 The point here is that it's generally OK to have too few qtvs; that is,
1158 to make the thing more monomorphic than it could be. We don't want to
1159 do that in the common cases, but in wierd cases it's ok: the programmer
1160 can always add a signature.
1162 Too few qtvs => too many wanteds, which is what happens if you do less
1167 tcSimplifyRestricted -- Used for restricted binding groups
1168 -- i.e. ones subject to the monomorphism restriction
1171 -> [Name] -- Things bound in this group
1172 -> TcTyVarSet -- Free in the type of the RHSs
1173 -> [Inst] -- Free in the RHSs
1174 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1175 TcDictBinds) -- Bindings
1176 -- tcSimpifyRestricted returns no constraints to
1177 -- quantify over; by definition there are none.
1178 -- They are all thrown back in the LIE
1180 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1181 -- Zonk everything in sight
1182 = do { wanteds' <- mappM zonkInst wanteds
1184 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1185 -- dicts; the idea is to get rid of as many type
1186 -- variables as possible, and we don't want to stop
1187 -- at (say) Monad (ST s), because that reduces
1188 -- immediately, with no constraint on s.
1190 -- BUT do no improvement! See Plan D above
1191 -- HOWEVER, some unification may take place, if we instantiate
1192 -- a method Inst with an equality constraint
1193 ; let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1194 ; (_imp, _binds, constrained_dicts) <- reduceContext env wanteds'
1196 -- Next, figure out the tyvars we will quantify over
1197 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1198 ; gbl_tvs' <- tcGetGlobalTyVars
1199 ; constrained_dicts' <- mappM zonkInst constrained_dicts
1201 ; let constrained_tvs' = tyVarsOfInsts constrained_dicts'
1202 qtvs = (tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs')
1203 `minusVarSet` constrained_tvs'
1204 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1205 pprInsts wanteds, pprInsts constrained_dicts',
1207 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1209 -- The first step may have squashed more methods than
1210 -- necessary, so try again, this time more gently, knowing the exact
1211 -- set of type variables to quantify over.
1213 -- We quantify only over constraints that are captured by qtvs;
1214 -- these will just be a subset of non-dicts. This in contrast
1215 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1216 -- all *non-inheritable* constraints too. This implements choice
1217 -- (B) under "implicit parameter and monomorphism" above.
1219 -- Remember that we may need to do *some* simplification, to
1220 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1221 -- just to float all constraints
1223 -- At top level, we *do* squash methods becuase we want to
1224 -- expose implicit parameters to the test that follows
1225 ; let is_nested_group = isNotTopLevel top_lvl
1226 try_me inst | isFreeWrtTyVars qtvs inst,
1227 (is_nested_group || isDict inst) = Stop
1228 | otherwise = ReduceMe AddSCs
1229 env = mkNoImproveRedEnv doc try_me
1230 ; (_imp, binds, irreds) <- reduceContext env wanteds'
1232 -- See "Notes on implicit parameters, Question 4: top level"
1233 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1234 if is_nested_group then
1236 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1237 ; addTopIPErrs bndrs bad_ips
1238 ; extendLIEs non_ips }
1240 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1241 ; return (qtvs', binds) }
1245 %************************************************************************
1249 %************************************************************************
1251 On the LHS of transformation rules we only simplify methods and constants,
1252 getting dictionaries. We want to keep all of them unsimplified, to serve
1253 as the available stuff for the RHS of the rule.
1255 Example. Consider the following left-hand side of a rule
1257 f (x == y) (y > z) = ...
1259 If we typecheck this expression we get constraints
1261 d1 :: Ord a, d2 :: Eq a
1263 We do NOT want to "simplify" to the LHS
1265 forall x::a, y::a, z::a, d1::Ord a.
1266 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1270 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1271 f ((==) d2 x y) ((>) d1 y z) = ...
1273 Here is another example:
1275 fromIntegral :: (Integral a, Num b) => a -> b
1276 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1278 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1279 we *dont* want to get
1281 forall dIntegralInt.
1282 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1284 because the scsel will mess up RULE matching. Instead we want
1286 forall dIntegralInt, dNumInt.
1287 fromIntegral Int Int dIntegralInt dNumInt = id Int
1291 g (x == y) (y == z) = ..
1293 where the two dictionaries are *identical*, we do NOT WANT
1295 forall x::a, y::a, z::a, d1::Eq a
1296 f ((==) d1 x y) ((>) d1 y z) = ...
1298 because that will only match if the dict args are (visibly) equal.
1299 Instead we want to quantify over the dictionaries separately.
1301 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1302 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1303 from scratch, rather than further parameterise simpleReduceLoop etc
1306 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1307 tcSimplifyRuleLhs wanteds
1308 = go [] emptyBag wanteds
1311 = return (dicts, binds)
1312 go dicts binds (w:ws)
1314 = go (w:dicts) binds ws
1316 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1317 -- to fromInteger; this looks fragile to me
1318 ; lookup_result <- lookupSimpleInst w'
1319 ; case lookup_result of
1320 GenInst ws' rhs -> go dicts (addBind binds w rhs) (ws' ++ ws)
1321 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1325 tcSimplifyBracket is used when simplifying the constraints arising from
1326 a Template Haskell bracket [| ... |]. We want to check that there aren't
1327 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1328 Show instance), but we aren't otherwise interested in the results.
1329 Nor do we care about ambiguous dictionaries etc. We will type check
1330 this bracket again at its usage site.
1333 tcSimplifyBracket :: [Inst] -> TcM ()
1334 tcSimplifyBracket wanteds
1335 = do { topCheckLoop doc wanteds
1338 doc = text "tcSimplifyBracket"
1342 %************************************************************************
1344 \subsection{Filtering at a dynamic binding}
1346 %************************************************************************
1351 we must discharge all the ?x constraints from B. We also do an improvement
1352 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1354 Actually, the constraints from B might improve the types in ?x. For example
1356 f :: (?x::Int) => Char -> Char
1359 then the constraint (?x::Int) arising from the call to f will
1360 force the binding for ?x to be of type Int.
1363 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1366 -- We need a loop so that we do improvement, and then
1367 -- (next time round) generate a binding to connect the two
1369 -- Here the two ?x's have different types, and improvement
1370 -- makes them the same.
1372 tcSimplifyIPs given_ips wanteds
1373 = do { wanteds' <- mappM zonkInst wanteds
1374 ; given_ips' <- mappM zonkInst given_ips
1375 -- Unusually for checking, we *must* zonk the given_ips
1377 ; let env = mkRedEnv doc try_me given_ips'
1378 ; (improved, binds, irreds) <- reduceContext env wanteds'
1380 ; if not improved then
1381 ASSERT( all is_free irreds )
1382 do { extendLIEs irreds
1385 tcSimplifyIPs given_ips wanteds }
1387 doc = text "tcSimplifyIPs" <+> ppr given_ips
1388 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1389 is_free inst = isFreeWrtIPs ip_set inst
1391 -- Simplify any methods that mention the implicit parameter
1392 try_me inst | is_free inst = Stop
1393 | otherwise = ReduceMe NoSCs
1397 %************************************************************************
1399 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1401 %************************************************************************
1403 When doing a binding group, we may have @Insts@ of local functions.
1404 For example, we might have...
1406 let f x = x + 1 -- orig local function (overloaded)
1407 f.1 = f Int -- two instances of f
1412 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1413 where @f@ is in scope; those @Insts@ must certainly not be passed
1414 upwards towards the top-level. If the @Insts@ were binding-ified up
1415 there, they would have unresolvable references to @f@.
1417 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1418 For each method @Inst@ in the @init_lie@ that mentions one of the
1419 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1420 @LIE@), as well as the @HsBinds@ generated.
1423 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1424 -- Simlifies only MethodInsts, and generate only bindings of form
1426 -- We're careful not to even generate bindings of the form
1428 -- You'd think that'd be fine, but it interacts with what is
1429 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1431 bindInstsOfLocalFuns wanteds local_ids
1432 | null overloaded_ids
1434 = extendLIEs wanteds `thenM_`
1435 returnM emptyLHsBinds
1438 = do { (irreds, binds) <- checkLoop env for_me
1439 ; extendLIEs not_for_me
1443 env = mkRedEnv doc try_me []
1444 doc = text "bindInsts" <+> ppr local_ids
1445 overloaded_ids = filter is_overloaded local_ids
1446 is_overloaded id = isOverloadedTy (idType id)
1447 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1449 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1450 -- so it's worth building a set, so that
1451 -- lookup (in isMethodFor) is faster
1452 try_me inst | isMethod inst = ReduceMe NoSCs
1457 %************************************************************************
1459 \subsection{Data types for the reduction mechanism}
1461 %************************************************************************
1463 The main control over context reduction is here
1467 = RedEnv { red_doc :: SDoc -- The context
1468 , red_try_me :: Inst -> WhatToDo
1469 , red_improve :: Bool -- True <=> do improvement
1470 , red_givens :: [Inst] -- All guaranteed rigid
1472 -- but see Note [Rigidity]
1473 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1474 -- See Note [RedStack]
1478 -- The red_givens are rigid so far as cmpInst is concerned.
1479 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1480 -- let ?x = e in ...
1481 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1482 -- But that doesn't affect the comparison, which is based only on mame.
1485 -- The red_stack pair (n,insts) pair is just used for error reporting.
1486 -- 'n' is always the depth of the stack.
1487 -- The 'insts' is the stack of Insts being reduced: to produce X
1488 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1491 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1492 mkRedEnv doc try_me givens
1493 = RedEnv { red_doc = doc, red_try_me = try_me,
1494 red_givens = givens, red_stack = (0,[]),
1495 red_improve = True }
1497 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1498 -- Do not do improvement; no givens
1499 mkNoImproveRedEnv doc try_me
1500 = RedEnv { red_doc = doc, red_try_me = try_me,
1501 red_givens = [], red_stack = (0,[]),
1502 red_improve = True }
1505 = ReduceMe WantSCs -- Try to reduce this
1506 -- If there's no instance, add the inst to the
1507 -- irreductible ones, but don't produce an error
1508 -- message of any kind.
1509 -- It might be quite legitimate such as (Eq a)!
1511 | Stop -- Return as irreducible unless it can
1512 -- be reduced to a constant in one step
1513 -- Do not add superclasses; see
1515 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1516 -- of a predicate when adding it to the avails
1517 -- The reason for this flag is entirely the super-class loop problem
1518 -- Note [SUPER-CLASS LOOP 1]
1521 %************************************************************************
1523 \subsection[reduce]{@reduce@}
1525 %************************************************************************
1529 reduceContext :: RedEnv
1531 -> TcM (ImprovementDone,
1532 TcDictBinds, -- Dictionary bindings
1533 [Inst]) -- Irreducible
1535 reduceContext env wanteds
1536 = do { traceTc (text "reduceContext" <+> (vcat [
1537 text "----------------------",
1539 text "given" <+> ppr (red_givens env),
1540 text "wanted" <+> ppr wanteds,
1541 text "----------------------"
1544 -- Build the Avail mapping from "givens"
1545 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1548 ; avails <- reduceList env wanteds init_state
1550 ; let improved = availsImproved avails
1551 ; (binds, irreds) <- extractResults avails wanteds
1553 ; traceTc (text "reduceContext end" <+> (vcat [
1554 text "----------------------",
1556 text "given" <+> ppr (red_givens env),
1557 text "wanted" <+> ppr wanteds,
1559 text "avails" <+> pprAvails avails,
1560 text "improved =" <+> ppr improved,
1561 text "----------------------"
1564 ; return (improved, binds, irreds) }
1566 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1567 tcImproveOne avails inst
1568 | not (isDict inst) = return False
1570 = do { inst_envs <- tcGetInstEnvs
1571 ; let eqns = improveOne (classInstances inst_envs)
1572 (dictPred inst, pprInstArising inst)
1573 [ (dictPred p, pprInstArising p)
1574 | p <- availsInsts avails, isDict p ]
1575 -- Avails has all the superclasses etc (good)
1576 -- It also has all the intermediates of the deduction (good)
1577 -- It does not have duplicates (good)
1578 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1579 -- so that improve will see them separate
1580 ; traceTc (text "improveOne" <+> ppr inst)
1583 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1584 -> TcM ImprovementDone
1585 unifyEqns [] = return False
1587 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1591 unify ((qtvs, pairs), what1, what2)
1592 = addErrCtxtM (mkEqnMsg what1 what2) $
1593 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1594 mapM_ (unif_pr tenv) pairs
1595 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1597 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1599 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1600 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1601 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1602 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1603 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1604 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1605 ; return (tidy_env, msg) }
1608 The main context-reduction function is @reduce@. Here's its game plan.
1611 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1612 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1613 = do { dopts <- getDOpts
1616 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1617 2 (ifPprDebug (nest 2 (pprStack stk))))
1620 ; if n >= ctxtStkDepth dopts then
1621 failWithTc (reduceDepthErr n stk)
1625 go [] state = return state
1626 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1629 -- Base case: we're done!
1630 reduce env wanted avails
1631 -- It's the same as an existing inst, or a superclass thereof
1632 | Just avail <- findAvail avails wanted
1636 = case red_try_me env wanted of {
1637 ; Stop -> try_simple (addIrred NoSCs) -- See Note [No superclasses for Stop]
1639 ; ReduceMe want_scs -> -- It should be reduced
1640 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1641 case lookup_result of
1642 NoInstance -> -- No such instance!
1643 -- Add it and its superclasses
1644 addIrred want_scs avails wanted
1646 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1648 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1649 ; avails2 <- reduceList env wanteds' avails1
1650 ; addWanted want_scs avails2 wanted rhs wanteds' }
1651 -- Temporarily do addIrred *before* the reduceList,
1652 -- which has the effect of adding the thing we are trying
1653 -- to prove to the database before trying to prove the things it
1654 -- needs. See note [RECURSIVE DICTIONARIES]
1655 -- NB: we must not do an addWanted before, because that adds the
1656 -- superclasses too, and thaat can lead to a spurious loop; see
1657 -- the examples in [SUPERCLASS-LOOP]
1658 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1662 -- First, see if the inst can be reduced to a constant in one step
1663 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1664 -- Don't bother for implication constraints, which take real work
1665 try_simple do_this_otherwise
1666 = do { res <- lookupSimpleInst wanted
1668 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1669 other -> do_this_otherwise avails wanted }
1673 Note [SUPERCLASS-LOOP 2]
1674 ~~~~~~~~~~~~~~~~~~~~~~~~
1675 But the above isn't enough. Suppose we are *given* d1:Ord a,
1676 and want to deduce (d2:C [a]) where
1678 class Ord a => C a where
1679 instance Ord [a] => C [a] where ...
1681 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1682 superclasses of C [a] to avails. But we must not overwrite the binding
1683 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1686 Here's another variant, immortalised in tcrun020
1687 class Monad m => C1 m
1688 class C1 m => C2 m x
1689 instance C2 Maybe Bool
1690 For the instance decl we need to build (C1 Maybe), and it's no good if
1691 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1692 before we search for C1 Maybe.
1694 Here's another example
1695 class Eq b => Foo a b
1696 instance Eq a => Foo [a] a
1700 we'll first deduce that it holds (via the instance decl). We must not
1701 then overwrite the Eq t constraint with a superclass selection!
1703 At first I had a gross hack, whereby I simply did not add superclass constraints
1704 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1705 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1706 I found a very obscure program (now tcrun021) in which improvement meant the
1707 simplifier got two bites a the cherry... so something seemed to be an Stop
1708 first time, but reducible next time.
1710 Now we implement the Right Solution, which is to check for loops directly
1711 when adding superclasses. It's a bit like the occurs check in unification.
1714 Note [RECURSIVE DICTIONARIES]
1715 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1717 data D r = ZeroD | SuccD (r (D r));
1719 instance (Eq (r (D r))) => Eq (D r) where
1720 ZeroD == ZeroD = True
1721 (SuccD a) == (SuccD b) = a == b
1724 equalDC :: D [] -> D [] -> Bool;
1727 We need to prove (Eq (D [])). Here's how we go:
1731 by instance decl, holds if
1735 by instance decl of Eq, holds if
1737 where d2 = dfEqList d3
1740 But now we can "tie the knot" to give
1746 and it'll even run! The trick is to put the thing we are trying to prove
1747 (in this case Eq (D []) into the database before trying to prove its
1748 contributing clauses.
1751 %************************************************************************
1753 Reducing a single constraint
1755 %************************************************************************
1758 ---------------------------------------------
1759 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1760 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1761 tci_given = extra_givens, tci_wanted = wanteds })
1762 = reduceImplication env avails reft tvs extra_givens wanteds loc
1764 reduceInst env avails other_inst
1765 = do { result <- lookupSimpleInst other_inst
1766 ; return (avails, result) }
1770 ---------------------------------------------
1771 reduceImplication :: RedEnv
1773 -> Refinement -- May refine the givens; often empty
1774 -> [TcTyVar] -- Quantified type variables; all skolems
1775 -> [Inst] -- Extra givens; all rigid
1778 -> TcM (Avails, LookupInstResult)
1781 Suppose we are simplifying the constraint
1782 forall bs. extras => wanted
1783 in the context of an overall simplification problem with givens 'givens',
1784 and refinment 'reft'.
1787 * The refinement is often empty
1789 * The 'extra givens' need not mention any of the quantified type variables
1790 e.g. forall {}. Eq a => Eq [a]
1791 forall {}. C Int => D (Tree Int)
1793 This happens when you have something like
1795 T1 :: Eq a => a -> T a
1798 f x = ...(case x of { T1 v -> v==v })...
1801 -- ToDo: should we instantiate tvs? I think it's not necessary
1803 -- ToDo: what about improvement? There may be some improvement
1804 -- exposed as a result of the simplifications done by reduceList
1805 -- which are discarded if we back off.
1806 -- This is almost certainly Wrong, but we'll fix it when dealing
1807 -- better with equality constraints
1808 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1809 = do { -- Add refined givens, and the extra givens
1810 (refined_red_givens, avails)
1811 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1812 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1813 ; avails <- foldlM addGiven avails extra_givens
1815 -- Solve the sub-problem
1816 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1817 env' = env { red_givens = refined_red_givens ++ extra_givens
1818 , red_try_me = try_me }
1820 ; traceTc (text "reduceImplication" <+> vcat
1822 ppr (red_givens env), ppr extra_givens,
1823 ppr reft, ppr wanteds, ppr avails ])
1824 ; avails <- reduceList env' wanteds avails
1826 -- Extract the binding
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 Note [Pruning the givens in an implication constraint]
1880 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1881 Suppose we are about to form the implication constraint
1882 forall tvs. Eq a => Ord b
1883 The (Eq a) cannot contribute to the (Ord b), because it has no access to
1884 the type variable 'b'. So we could filter out the (Eq a) from the givens.
1886 Doing so would be a bit tidier, but all the implication constraints get
1887 simplified away by the optimiser, so it's no great win. So I don't take
1888 advantage of that at the moment.
1890 If you do, BE CAREFUL of wobbly type variables.
1893 %************************************************************************
1895 Avails and AvailHow: the pool of evidence
1897 %************************************************************************
1901 data Avails = Avails !ImprovementDone !AvailEnv
1903 type ImprovementDone = Bool -- True <=> some unification has happened
1904 -- so some Irreds might now be reducible
1905 -- keys that are now
1907 type AvailEnv = FiniteMap Inst AvailHow
1909 = IsIrred -- Used for irreducible dictionaries,
1910 -- which are going to be lambda bound
1912 | Given TcId -- Used for dictionaries for which we have a binding
1913 -- e.g. those "given" in a signature
1915 | Rhs -- Used when there is a RHS
1916 (LHsExpr TcId) -- The RHS
1917 [Inst] -- Insts free in the RHS; we need these too
1919 instance Outputable Avails where
1922 pprAvails (Avails imp avails)
1923 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
1924 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
1925 | (inst,avail) <- fmToList avails ])]
1927 instance Outputable AvailHow where
1930 -------------------------
1931 pprAvail :: AvailHow -> SDoc
1932 pprAvail IsIrred = text "Irred"
1933 pprAvail (Given x) = text "Given" <+> ppr x
1934 pprAvail (Rhs rhs bs) = text "Rhs" <+> ppr rhs <+> braces (ppr bs)
1936 -------------------------
1937 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
1938 extendAvailEnv env inst avail = addToFM env inst avail
1940 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
1941 findAvailEnv env wanted = lookupFM env wanted
1942 -- NB 1: the Ord instance of Inst compares by the class/type info
1943 -- *not* by unique. So
1944 -- d1::C Int == d2::C Int
1946 emptyAvails :: Avails
1947 emptyAvails = Avails False emptyFM
1949 findAvail :: Avails -> Inst -> Maybe AvailHow
1950 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
1952 elemAvails :: Inst -> Avails -> Bool
1953 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
1955 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
1957 extendAvails avails@(Avails imp env) inst avail
1958 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
1959 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
1961 availsInsts :: Avails -> [Inst]
1962 availsInsts (Avails _ avails) = keysFM avails
1964 availsImproved (Avails imp _) = imp
1966 updateImprovement :: Avails -> Avails -> Avails
1967 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
1968 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
1971 Extracting the bindings from a bunch of Avails.
1972 The bindings do *not* come back sorted in dependency order.
1973 We assume that they'll be wrapped in a big Rec, so that the
1974 dependency analyser can sort them out later
1977 extractResults :: Avails
1979 -> TcM ( TcDictBinds, -- Bindings
1980 [Inst]) -- Irreducible ones
1982 extractResults (Avails _ avails) wanteds
1983 = go avails emptyBag [] wanteds
1985 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
1986 -> TcM (TcDictBinds, [Inst])
1987 go avails binds irreds []
1988 = returnM (binds, irreds)
1990 go avails binds irreds (w:ws)
1991 = case findAvailEnv avails w of
1992 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
1993 go avails binds irreds ws
1995 Just IsIrred -> go (add_given avails w) binds (w:irreds) ws
1999 -> go avails binds irreds ws
2000 -- The sought Id can be one of the givens, via a superclass chain
2001 -- and then we definitely don't want to generate an x=x binding!
2004 -> go avails (addBind binds w (nlHsVar id)) irreds ws
2006 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds (ws' ++ ws)
2008 new_binds = addBind binds w rhs
2010 add_given avails w = extendAvailEnv avails w (Given (instToId w))
2012 addBind binds inst rhs = binds `unionBags` unitBag (L (instSpan inst)
2013 (VarBind (instToId inst) rhs))
2017 Note [No superclasses for Stop]
2018 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2019 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2020 add it to avails, so that any other equal Insts will be commoned up
2021 right here. However, we do *not* add superclasses. If we have
2024 but a is not bound here, then we *don't* want to derive dn from df
2025 here lest we lose sharing.
2028 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2029 addWanted want_scs avails wanted rhs_expr wanteds
2030 = addAvailAndSCs want_scs avails wanted avail
2032 avail = Rhs rhs_expr wanteds
2034 addGiven :: Avails -> Inst -> TcM Avails
2035 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2036 -- Always add superclasses for 'givens'
2038 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2039 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2040 -- so the assert isn't true
2042 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2043 addRefinedGiven reft (refined_givens, avails) given
2044 | isDict given -- We sometimes have 'given' methods, but they
2045 -- are always optional, so we can drop them
2046 , let pred = dictPred given
2047 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2048 , Just (co, pred) <- refinePred reft pred
2049 = do { new_given <- newDictBndr (instLoc given) pred
2050 ; let rhs = L (instSpan given) $
2051 HsWrap (WpCo co) (HsVar (instToId given))
2052 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2053 ; return (new_given:refined_givens, avails) }
2054 -- ToDo: the superclasses of the original given all exist in Avails
2055 -- so we could really just cast them, but it's more awkward to do,
2056 -- and hopefully the optimiser will spot the duplicated work
2058 = return (refined_givens, avails)
2061 Note [ImplicInst rigidity]
2062 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2064 C :: forall ab. (Eq a, Ord b) => b -> T a
2066 ...(case x of C v -> <body>)...
2068 From the case (where x::T ty) we'll get an implication constraint
2069 forall b. (Eq ty, Ord b) => <body-constraints>
2070 Now suppose <body-constraints> itself has an implication constraint
2072 forall c. <reft> => <payload>
2073 Then, we can certainly apply the refinement <reft> to the Ord b, becuase it is
2074 existential, but we probably should not apply it to the (Eq ty) because it may
2075 be wobbly. Hence the isRigidInst
2077 @Insts@ are ordered by their class/type info, rather than by their
2078 unique. This allows the context-reduction mechanism to use standard finite
2079 maps to do their stuff. It's horrible that this code is here, rather
2080 than with the Avails handling stuff in TcSimplify
2083 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2084 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2085 addAvailAndSCs want_scs avails irred IsIrred
2087 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2088 addAvailAndSCs want_scs avails inst avail
2089 | not (isClassDict inst) = extendAvails avails inst avail
2090 | NoSCs <- want_scs = extendAvails avails inst avail
2091 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2092 ; avails' <- extendAvails avails inst avail
2093 ; addSCs is_loop avails' inst }
2095 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2096 -- Note: this compares by *type*, not by Unique
2097 deps = findAllDeps (unitVarSet (instToId inst)) avail
2098 dep_tys = map idType (varSetElems deps)
2100 findAllDeps :: IdSet -> AvailHow -> IdSet
2101 -- Find all the Insts that this one depends on
2102 -- See Note [SUPERCLASS-LOOP 2]
2103 -- Watch out, though. Since the avails may contain loops
2104 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2105 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2106 findAllDeps so_far other = so_far
2108 find_all :: IdSet -> Inst -> IdSet
2110 | kid_id `elemVarSet` so_far = so_far
2111 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2112 | otherwise = so_far'
2114 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2115 kid_id = instToId kid
2117 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2118 -- Add all the superclasses of the Inst to Avails
2119 -- The first param says "dont do this because the original thing
2120 -- depends on this one, so you'd build a loop"
2121 -- Invariant: the Inst is already in Avails.
2123 addSCs is_loop avails dict
2124 = ASSERT( isDict dict )
2125 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2126 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2128 (clas, tys) = getDictClassTys dict
2129 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2130 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2132 add_sc avails (sc_dict, sc_sel)
2133 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2134 | is_given sc_dict = return avails
2135 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2136 ; addSCs is_loop avails' sc_dict }
2138 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2139 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2141 is_given :: Inst -> Bool
2142 is_given sc_dict = case findAvail avails sc_dict of
2143 Just (Given _) -> True -- Given is cheaper than superclass selection
2147 %************************************************************************
2149 \section{tcSimplifyTop: defaulting}
2151 %************************************************************************
2154 @tcSimplifyTop@ is called once per module to simplify all the constant
2155 and ambiguous Insts.
2157 We need to be careful of one case. Suppose we have
2159 instance Num a => Num (Foo a b) where ...
2161 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2162 to (Num x), and default x to Int. But what about y??
2164 It's OK: the final zonking stage should zap y to (), which is fine.
2168 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2169 tcSimplifyTop wanteds
2170 = tc_simplify_top doc False wanteds
2172 doc = text "tcSimplifyTop"
2174 tcSimplifyInteractive wanteds
2175 = tc_simplify_top doc True wanteds
2177 doc = text "tcSimplifyInteractive"
2179 -- The TcLclEnv should be valid here, solely to improve
2180 -- error message generation for the monomorphism restriction
2181 tc_simplify_top doc interactive wanteds
2182 = do { wanteds <- mapM zonkInst wanteds
2183 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2185 ; (irreds1, binds1) <- topCheckLoop doc wanteds
2187 ; if null irreds1 then
2190 -- OK, so there are some errors
2191 { -- Use the defaulting rules to do extra unification
2192 -- NB: irreds are already zonked
2193 ; extended_default <- if interactive then return True
2194 else doptM Opt_ExtendedDefaultRules
2195 ; disambiguate extended_default irreds1 -- Does unification
2196 ; (irreds2, binds2) <- topCheckLoop doc irreds1
2198 -- Deal with implicit parameter
2199 ; let (bad_ips, non_ips) = partition isIPDict irreds2
2200 (ambigs, others) = partition isTyVarDict non_ips
2202 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2204 ; addNoInstanceErrs others
2205 ; addTopAmbigErrs ambigs
2207 ; return (binds1 `unionBags` binds2) }}
2210 If a dictionary constrains a type variable which is
2211 * not mentioned in the environment
2212 * and not mentioned in the type of the expression
2213 then it is ambiguous. No further information will arise to instantiate
2214 the type variable; nor will it be generalised and turned into an extra
2215 parameter to a function.
2217 It is an error for this to occur, except that Haskell provided for
2218 certain rules to be applied in the special case of numeric types.
2220 * at least one of its classes is a numeric class, and
2221 * all of its classes are numeric or standard
2222 then the type variable can be defaulted to the first type in the
2223 default-type list which is an instance of all the offending classes.
2225 So here is the function which does the work. It takes the ambiguous
2226 dictionaries and either resolves them (producing bindings) or
2227 complains. It works by splitting the dictionary list by type
2228 variable, and using @disambigOne@ to do the real business.
2230 @disambigOne@ assumes that its arguments dictionaries constrain all
2231 the same type variable.
2233 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2234 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2235 the most common use of defaulting is code like:
2237 _ccall_ foo `seqPrimIO` bar
2239 Since we're not using the result of @foo@, the result if (presumably)
2243 disambiguate :: Bool -> [Inst] -> TcM ()
2244 -- Just does unification to fix the default types
2245 -- The Insts are assumed to be pre-zonked
2246 disambiguate extended_defaulting insts
2247 | null defaultable_groups
2248 = do { traceTc (text "disambigutate" <+> vcat [ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2251 = do { -- Figure out what default types to use
2252 mb_defaults <- getDefaultTys
2253 ; default_tys <- case mb_defaults of
2254 Just tys -> return tys
2255 Nothing -> -- No use-supplied default;
2256 -- use [Integer, Double]
2257 do { integer_ty <- tcMetaTy integerTyConName
2258 ; checkWiredInTyCon doubleTyCon
2259 ; return [integer_ty, doubleTy] }
2260 ; traceTc (text "disambigutate" <+> vcat [ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2261 ; mapM_ (disambigGroup default_tys) defaultable_groups }
2263 unaries :: [(Inst,Class, TcTyVar)] -- (C tv) constraints
2264 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2265 (unaries, bad_tvs) = getDefaultableDicts insts
2267 -- Group by type variable
2268 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2269 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2270 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2272 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2273 defaultable_group ds@((_,_,tv):_)
2274 = not (isImmutableTyVar tv) -- Note [Avoiding spurious errors]
2275 && not (tv `elemVarSet` bad_tvs)
2276 && defaultable_classes [c | (_,c,_) <- ds]
2277 defaultable_group [] = panic "defaultable_group"
2279 defaultable_classes clss
2280 | extended_defaulting = any isInteractiveClass clss
2281 | otherwise = all isStandardClass clss && any isNumericClass clss
2283 -- In interactive mode, or with -fextended-default-rules,
2284 -- we default Show a to Show () to avoid graututious errors on "show []"
2285 isInteractiveClass cls
2286 = isNumericClass cls
2287 || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2290 disambigGroup :: [Type] -- The default types
2291 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2292 -> TcM () -- Just does unification, to fix the default types
2294 disambigGroup default_tys dicts
2295 = try_default default_tys
2297 (_,_,tyvar) = head dicts -- Should be non-empty
2298 classes = [c | (_,c,_) <- dicts]
2300 try_default [] = return ()
2301 try_default (default_ty : default_tys)
2302 = tryTcLIE_ (try_default default_tys) $
2303 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2304 -- This may fail; then the tryTcLIE_ kicks in
2305 -- Failure here is caused by there being no type in the
2306 -- default list which can satisfy all the ambiguous classes.
2307 -- For example, if Real a is reqd, but the only type in the
2308 -- default list is Int.
2310 -- After this we can't fail
2311 ; warnDefault dicts default_ty
2312 ; unifyType default_ty (mkTyVarTy tyvar) }
2315 Note [Avoiding spurious errors]
2316 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2317 When doing the unification for defaulting, we check for skolem
2318 type variables, and simply don't default them. For example:
2319 f = (*) -- Monomorphic
2320 g :: Num a => a -> a
2322 Here, we get a complaint when checking the type signature for g,
2323 that g isn't polymorphic enough; but then we get another one when
2324 dealing with the (Num a) context arising from f's definition;
2325 we try to unify a with Int (to default it), but find that it's
2326 already been unified with the rigid variable from g's type sig
2329 %************************************************************************
2331 \subsection[simple]{@Simple@ versions}
2333 %************************************************************************
2335 Much simpler versions when there are no bindings to make!
2337 @tcSimplifyThetas@ simplifies class-type constraints formed by
2338 @deriving@ declarations and when specialising instances. We are
2339 only interested in the simplified bunch of class/type constraints.
2341 It simplifies to constraints of the form (C a b c) where
2342 a,b,c are type variables. This is required for the context of
2343 instance declarations.
2346 tcSimplifyDeriv :: InstOrigin
2348 -> ThetaType -- Wanted
2349 -> TcM ThetaType -- Needed
2350 -- Given instance (wanted) => C inst_ty
2351 -- Simplify 'wanted' as much as possible
2352 -- The inst_ty is needed only for the termination check
2354 tcSimplifyDeriv orig tyvars theta
2355 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2356 -- The main loop may do unification, and that may crash if
2357 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2358 -- ToDo: what if two of them do get unified?
2359 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2360 ; (irreds, _) <- topCheckLoop doc wanteds
2362 -- Insist that the context of a derived instance declaration
2363 -- consists of constraints of form (C a b), where a,b are
2365 -- NB: the caller will further check the tv_dicts for
2366 -- legal instance-declaration form
2367 ; let (tv_dicts, non_tv_dicts) = partition isTyVarDict irreds
2368 ; addNoInstanceErrs non_tv_dicts
2370 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2371 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
2372 -- This reverse-mapping is a pain, but the result
2373 -- should mention the original TyVars not TcTyVars
2375 ; return simpl_theta }
2377 doc = ptext SLIT("deriving classes for a data type")
2380 Note [Deriving context]
2381 ~~~~~~~~~~~~~~~~~~~~~~~
2382 With -fglasgow-exts, we allow things like (C Int a) in the simplified
2383 context for a derived instance declaration, because at a use of this
2384 instance, we might know that a=Bool, and have an instance for (C Int
2387 We nevertheless insist that each predicate meets the termination
2388 conditions. If not, the deriving mechanism generates larger and larger
2389 constraints. Example:
2391 data Seq a = Cons a (Seq (Succ a)) | Nil deriving Show
2393 Note the lack of a Show instance for Succ. First we'll generate
2394 instance (Show (Succ a), Show a) => Show (Seq a)
2396 instance (Show (Succ (Succ a)), Show (Succ a), Show a) => Show (Seq a)
2397 and so on. Instead we want to complain of no instance for (Show (Succ a)).
2401 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2402 used with \tr{default} declarations. We are only interested in
2403 whether it worked or not.
2406 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2409 tcSimplifyDefault theta
2410 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2411 topCheckLoop doc wanteds `thenM` \ (irreds, _) ->
2412 addNoInstanceErrs irreds `thenM_`
2418 doc = ptext SLIT("default declaration")
2422 %************************************************************************
2424 \section{Errors and contexts}
2426 %************************************************************************
2428 ToDo: for these error messages, should we note the location as coming
2429 from the insts, or just whatever seems to be around in the monad just
2433 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2434 -> [Inst] -- The offending Insts
2436 -- Group together insts with the same origin
2437 -- We want to report them together in error messages
2439 groupErrs report_err []
2441 groupErrs report_err (inst:insts)
2442 = do_one (inst:friends) `thenM_`
2443 groupErrs report_err others
2446 -- (It may seem a bit crude to compare the error messages,
2447 -- but it makes sure that we combine just what the user sees,
2448 -- and it avoids need equality on InstLocs.)
2449 (friends, others) = partition is_friend insts
2450 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2451 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2452 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2453 -- Add location and context information derived from the Insts
2455 -- Add the "arising from..." part to a message about bunch of dicts
2456 addInstLoc :: [Inst] -> Message -> Message
2457 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2459 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2460 addTopIPErrs bndrs []
2462 addTopIPErrs bndrs ips
2463 = addErrTcM (tidy_env, mk_msg tidy_ips)
2465 (tidy_env, tidy_ips) = tidyInsts ips
2466 mk_msg ips = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2467 nest 2 (ptext SLIT("the monomorphic top-level binding")
2468 <> plural bndrs <+> ptext SLIT("of")
2469 <+> pprBinders bndrs <> colon)],
2470 nest 2 (vcat (map ppr_ip ips)),
2472 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2474 topIPErrs :: [Inst] -> TcM ()
2476 = groupErrs report tidy_dicts
2478 (tidy_env, tidy_dicts) = tidyInsts dicts
2479 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2480 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2481 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2483 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2485 addNoInstanceErrs insts
2486 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2487 ; reportNoInstances tidy_env Nothing tidy_insts }
2491 -> Maybe (InstLoc, [Inst]) -- Context
2492 -- Nothing => top level
2493 -- Just (d,g) => d describes the construct
2495 -> [Inst] -- What is wanted (can include implications)
2498 reportNoInstances tidy_env mb_what insts
2499 = groupErrs (report_no_instances tidy_env mb_what) insts
2501 report_no_instances tidy_env mb_what insts
2502 = do { inst_envs <- tcGetInstEnvs
2503 ; let (implics, insts1) = partition isImplicInst insts
2504 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2505 ; traceTc (text "reportNoInstnces" <+> vcat
2506 [ppr implics, ppr insts1, ppr insts2])
2507 ; mapM_ complain_implic implics
2508 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2509 ; groupErrs complain_no_inst insts2 }
2511 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2513 complain_implic inst -- Recurse!
2514 = reportNoInstances tidy_env
2515 (Just (tci_loc inst, tci_given inst))
2518 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2519 -- Right msg => overlap message
2520 -- Left inst => no instance
2521 check_overlap inst_envs wanted
2522 | not (isClassDict wanted) = Left wanted
2524 = case lookupInstEnv inst_envs clas tys of
2525 -- The case of exactly one match and no unifiers means
2526 -- a successful lookup. That can't happen here, becuase
2527 -- dicts only end up here if they didn't match in Inst.lookupInst
2529 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2531 ([], _) -> Left wanted -- No match
2532 res -> Right (mk_overlap_msg wanted res)
2534 (clas,tys) = getDictClassTys wanted
2536 mk_overlap_msg dict (matches, unifiers)
2537 = vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2538 <+> pprPred (dictPred dict))),
2539 sep [ptext SLIT("Matching instances") <> colon,
2540 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2541 ASSERT( not (null matches) )
2542 if not (isSingleton matches)
2543 then -- Two or more matches
2545 else -- One match, plus some unifiers
2546 ASSERT( not (null unifiers) )
2547 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2548 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2549 ptext SLIT("Use -fallow-incoherent-instances to use the first choice above")])]
2551 ispecs = [ispec | (_, ispec) <- matches]
2553 mk_no_inst_err insts
2554 | null insts = empty
2556 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2557 not (isEmptyVarSet (tyVarsOfInsts insts))
2558 = vcat [ addInstLoc insts $
2559 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2560 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2561 , show_fixes (fix1 loc : fixes2) ]
2563 | otherwise -- Top level
2564 = vcat [ addInstLoc insts $
2565 ptext SLIT("No instance") <> plural insts
2566 <+> ptext SLIT("for") <+> pprDictsTheta insts
2567 , show_fixes fixes2 ]
2570 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2571 <+> ptext SLIT("to the context of"),
2572 nest 2 (ppr (instLocOrigin loc)) ]
2573 -- I'm not sure it helps to add the location
2574 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2576 fixes2 | null instance_dicts = []
2577 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2578 pprDictsTheta instance_dicts]]
2579 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2580 -- Insts for which it is worth suggesting an adding an instance declaration
2581 -- Exclude implicit parameters, and tyvar dicts
2583 show_fixes :: [SDoc] -> SDoc
2584 show_fixes [] = empty
2585 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2586 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2588 addTopAmbigErrs dicts
2589 -- Divide into groups that share a common set of ambiguous tyvars
2590 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2591 -- See Note [Avoiding spurious errors]
2592 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2594 (tidy_env, tidy_dicts) = tidyInsts dicts
2596 tvs_of :: Inst -> [TcTyVar]
2597 tvs_of d = varSetElems (tyVarsOfInst d)
2598 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2600 report :: [(Inst,[TcTyVar])] -> TcM ()
2601 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2602 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2603 setSrcSpan (instSpan inst) $
2604 -- the location of the first one will do for the err message
2605 addErrTcM (tidy_env, msg $$ mono_msg)
2607 dicts = map fst pairs
2608 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2609 pprQuotedList tvs <+> in_msg,
2610 nest 2 (pprDictsInFull dicts)]
2611 in_msg = text "in the constraint" <> plural dicts <> colon
2612 report [] = panic "addTopAmbigErrs"
2615 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2616 -- There's an error with these Insts; if they have free type variables
2617 -- it's probably caused by the monomorphism restriction.
2618 -- Try to identify the offending variable
2619 -- ASSUMPTION: the Insts are fully zonked
2620 mkMonomorphismMsg tidy_env inst_tvs
2621 = findGlobals (mkVarSet inst_tvs) tidy_env `thenM` \ (tidy_env, docs) ->
2622 returnM (tidy_env, mk_msg docs)
2624 mk_msg [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2625 -- This happens in things like
2626 -- f x = show (read "foo")
2627 -- where monomorphism doesn't play any role
2628 mk_msg docs = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2632 monomorphism_fix :: SDoc
2633 monomorphism_fix = ptext SLIT("Probable fix:") <+>
2634 (ptext SLIT("give these definition(s) an explicit type signature")
2635 $$ ptext SLIT("or use -fno-monomorphism-restriction"))
2637 warnDefault ups default_ty
2638 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2639 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2641 dicts = [d | (d,_,_) <- ups]
2644 (_, tidy_dicts) = tidyInsts dicts
2645 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2646 quotes (ppr default_ty),
2647 pprDictsInFull tidy_dicts]
2649 reduceDepthErr n stack
2650 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2651 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2652 nest 4 (pprStack stack)]
2654 pprStack stack = vcat (map pprInstInFull stack)