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,
25 #include "HsVersions.h"
27 import {-# SOURCE #-} TcUnify( unifyType )
31 import TcHsSyn ( hsLPatType )
39 import DsUtils -- Big-tuple functions
69 %************************************************************************
73 %************************************************************************
75 --------------------------------------
76 Notes on functional dependencies (a bug)
77 --------------------------------------
84 instance D a b => C a b -- Undecidable
85 -- (Not sure if it's crucial to this eg)
86 f :: C a b => a -> Bool
89 g :: C a b => a -> Bool
92 Here f typechecks, but g does not!! Reason: before doing improvement,
93 we reduce the (C a b1) constraint from the call of f to (D a b1).
95 Here is a more complicated example:
98 > class Foo a b | a->b
100 > class Bar a b | a->b
104 > instance Bar Obj Obj
106 > instance (Bar a b) => Foo a b
108 > foo:: (Foo a b) => a -> String
111 > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
117 Could not deduce (Bar a b) from the context (Foo a b)
118 arising from use of `foo' at <interactive>:1
120 Add (Bar a b) to the expected type of an expression
121 In the first argument of `runFoo', namely `foo'
122 In the definition of `it': it = runFoo foo
124 Why all of the sudden does GHC need the constraint Bar a b? The
125 function foo didn't ask for that...
128 The trouble is that to type (runFoo foo), GHC has to solve the problem:
130 Given constraint Foo a b
131 Solve constraint Foo a b'
133 Notice that b and b' aren't the same. To solve this, just do
134 improvement and then they are the same. But GHC currently does
139 That is usually fine, but it isn't here, because it sees that Foo a b is
140 not the same as Foo a b', and so instead applies the instance decl for
141 instance Bar a b => Foo a b. And that's where the Bar constraint comes
144 The Right Thing is to improve whenever the constraint set changes at
145 all. Not hard in principle, but it'll take a bit of fiddling to do.
147 Note [Choosing which variables to quantify]
148 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
149 Suppose we are about to do a generalisation step. We have in our hand
152 T the type of the RHS
153 C the constraints from that RHS
155 The game is to figure out
157 Q the set of type variables over which to quantify
158 Ct the constraints we will *not* quantify over
159 Cq the constraints we will quantify over
161 So we're going to infer the type
165 and float the constraints Ct further outwards.
167 Here are the things that *must* be true:
169 (A) Q intersect fv(G) = EMPTY limits how big Q can be
170 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
172 (A) says we can't quantify over a variable that's free in the environment.
173 (B) says we must quantify over all the truly free variables in T, else
174 we won't get a sufficiently general type.
176 We do not *need* to quantify over any variable that is fixed by the
177 free vars of the environment G.
179 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
181 Example: class H x y | x->y where ...
183 fv(G) = {a} C = {H a b, H c d}
186 (A) Q intersect {a} is empty
187 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
189 So Q can be {c,d}, {b,c,d}
191 In particular, it's perfectly OK to quantify over more type variables
192 than strictly necessary; there is no need to quantify over 'b', since
193 it is determined by 'a' which is free in the envt, but it's perfectly
194 OK to do so. However we must not quantify over 'a' itself.
196 Other things being equal, however, we'd like to quantify over as few
197 variables as possible: smaller types, fewer type applications, more
198 constraints can get into Ct instead of Cq. Here's a good way to
201 Q = grow( fv(T), C ) \ oclose( fv(G), C )
203 That is, quantify over all variable that that MIGHT be fixed by the
204 call site (which influences T), but which aren't DEFINITELY fixed by
205 G. This choice definitely quantifies over enough type variables,
206 albeit perhaps too many.
208 Why grow( fv(T), C ) rather than fv(T)? Consider
210 class H x y | x->y where ...
215 If we used fv(T) = {c} we'd get the type
217 forall c. H c d => c -> b
219 And then if the fn was called at several different c's, each of
220 which fixed d differently, we'd get a unification error, because
221 d isn't quantified. Solution: quantify d. So we must quantify
222 everything that might be influenced by c.
224 Why not oclose( fv(T), C )? Because we might not be able to see
225 all the functional dependencies yet:
227 class H x y | x->y where ...
228 instance H x y => Eq (T x y) where ...
233 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
234 apparent yet, and that's wrong. We must really quantify over d too.
236 There really isn't any point in quantifying over any more than
237 grow( fv(T), C ), because the call sites can't possibly influence
238 any other type variables.
242 -------------------------------------
244 -------------------------------------
246 It's very hard to be certain when a type is ambiguous. Consider
250 instance H x y => K (x,y)
252 Is this type ambiguous?
253 forall a b. (K (a,b), Eq b) => a -> a
255 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
256 now we see that a fixes b. So we can't tell about ambiguity for sure
257 without doing a full simplification. And even that isn't possible if
258 the context has some free vars that may get unified. Urgle!
260 Here's another example: is this ambiguous?
261 forall a b. Eq (T b) => a -> a
262 Not if there's an insance decl (with no context)
263 instance Eq (T b) where ...
265 You may say of this example that we should use the instance decl right
266 away, but you can't always do that:
268 class J a b where ...
269 instance J Int b where ...
271 f :: forall a b. J a b => a -> a
273 (Notice: no functional dependency in J's class decl.)
274 Here f's type is perfectly fine, provided f is only called at Int.
275 It's premature to complain when meeting f's signature, or even
276 when inferring a type for f.
280 However, we don't *need* to report ambiguity right away. It'll always
281 show up at the call site.... and eventually at main, which needs special
282 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
284 So here's the plan. We WARN about probable ambiguity if
286 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
288 (all tested before quantification).
289 That is, all the type variables in Cq must be fixed by the the variables
290 in the environment, or by the variables in the type.
292 Notice that we union before calling oclose. Here's an example:
294 class J a b c | a b -> c
298 forall b c. (J a b c) => b -> b
300 Only if we union {a} from G with {b} from T before using oclose,
301 do we see that c is fixed.
303 It's a bit vague exactly which C we should use for this oclose call. If we
304 don't fix enough variables we might complain when we shouldn't (see
305 the above nasty example). Nothing will be perfect. That's why we can
306 only issue a warning.
309 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
311 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
313 then c is a "bubble"; there's no way it can ever improve, and it's
314 certainly ambiguous. UNLESS it is a constant (sigh). And what about
319 instance H x y => K (x,y)
321 Is this type ambiguous?
322 forall a b. (K (a,b), Eq b) => a -> a
324 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
325 is a "bubble" that's a set of constraints
327 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
329 Hence another idea. To decide Q start with fv(T) and grow it
330 by transitive closure in Cq (no functional dependencies involved).
331 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
332 The definitely-ambiguous can then float out, and get smashed at top level
333 (which squashes out the constants, like Eq (T a) above)
336 --------------------------------------
337 Notes on principal types
338 --------------------------------------
343 f x = let g y = op (y::Int) in True
345 Here the principal type of f is (forall a. a->a)
346 but we'll produce the non-principal type
347 f :: forall a. C Int => a -> a
350 --------------------------------------
351 The need for forall's in constraints
352 --------------------------------------
354 [Exchange on Haskell Cafe 5/6 Dec 2000]
356 class C t where op :: t -> Bool
357 instance C [t] where op x = True
359 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
360 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
362 The definitions of p and q differ only in the order of the components in
363 the pair on their right-hand sides. And yet:
365 ghc and "Typing Haskell in Haskell" reject p, but accept q;
366 Hugs rejects q, but accepts p;
367 hbc rejects both p and q;
368 nhc98 ... (Malcolm, can you fill in the blank for us!).
370 The type signature for f forces context reduction to take place, and
371 the results of this depend on whether or not the type of y is known,
372 which in turn depends on which component of the pair the type checker
375 Solution: if y::m a, float out the constraints
376 Monad m, forall c. C (m c)
377 When m is later unified with [], we can solve both constraints.
380 --------------------------------------
381 Notes on implicit parameters
382 --------------------------------------
384 Note [Inheriting implicit parameters]
385 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
390 where f is *not* a top-level binding.
391 From the RHS of f we'll get the constraint (?y::Int).
392 There are two types we might infer for f:
396 (so we get ?y from the context of f's definition), or
398 f :: (?y::Int) => Int -> Int
400 At first you might think the first was better, becuase then
401 ?y behaves like a free variable of the definition, rather than
402 having to be passed at each call site. But of course, the WHOLE
403 IDEA is that ?y should be passed at each call site (that's what
404 dynamic binding means) so we'd better infer the second.
406 BOTTOM LINE: when *inferring types* you *must* quantify
407 over implicit parameters. See the predicate isFreeWhenInferring.
410 Note [Implicit parameters and ambiguity]
411 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
412 Only a *class* predicate can give rise to ambiguity
413 An *implicit parameter* cannot. For example:
414 foo :: (?x :: [a]) => Int
416 is fine. The call site will suppply a particular 'x'
418 Furthermore, the type variables fixed by an implicit parameter
419 propagate to the others. E.g.
420 foo :: (Show a, ?x::[a]) => Int
422 The type of foo looks ambiguous. But it isn't, because at a call site
424 let ?x = 5::Int in foo
425 and all is well. In effect, implicit parameters are, well, parameters,
426 so we can take their type variables into account as part of the
427 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
430 Question 2: type signatures
431 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
432 BUT WATCH OUT: When you supply a type signature, we can't force you
433 to quantify over implicit parameters. For example:
437 This is perfectly reasonable. We do not want to insist on
439 (?x + 1) :: (?x::Int => Int)
441 That would be silly. Here, the definition site *is* the occurrence site,
442 so the above strictures don't apply. Hence the difference between
443 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
444 and tcSimplifyCheckBind (which does not).
446 What about when you supply a type signature for a binding?
447 Is it legal to give the following explicit, user type
448 signature to f, thus:
453 At first sight this seems reasonable, but it has the nasty property
454 that adding a type signature changes the dynamic semantics.
457 (let f x = (x::Int) + ?y
458 in (f 3, f 3 with ?y=5)) with ?y = 6
464 in (f 3, f 3 with ?y=5)) with ?y = 6
468 Indeed, simply inlining f (at the Haskell source level) would change the
471 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
472 semantics for a Haskell program without knowing its typing, so if you
473 change the typing you may change the semantics.
475 To make things consistent in all cases where we are *checking* against
476 a supplied signature (as opposed to inferring a type), we adopt the
479 a signature does not need to quantify over implicit params.
481 [This represents a (rather marginal) change of policy since GHC 5.02,
482 which *required* an explicit signature to quantify over all implicit
483 params for the reasons mentioned above.]
485 But that raises a new question. Consider
487 Given (signature) ?x::Int
488 Wanted (inferred) ?x::Int, ?y::Bool
490 Clearly we want to discharge the ?x and float the ?y out. But
491 what is the criterion that distinguishes them? Clearly it isn't
492 what free type variables they have. The Right Thing seems to be
493 to float a constraint that
494 neither mentions any of the quantified type variables
495 nor any of the quantified implicit parameters
497 See the predicate isFreeWhenChecking.
500 Question 3: monomorphism
501 ~~~~~~~~~~~~~~~~~~~~~~~~
502 There's a nasty corner case when the monomorphism restriction bites:
506 The argument above suggests that we *must* generalise
507 over the ?y parameter, to get
508 z :: (?y::Int) => Int,
509 but the monomorphism restriction says that we *must not*, giving
511 Why does the momomorphism restriction say this? Because if you have
513 let z = x + ?y in z+z
515 you might not expect the addition to be done twice --- but it will if
516 we follow the argument of Question 2 and generalise over ?y.
519 Question 4: top level
520 ~~~~~~~~~~~~~~~~~~~~~
521 At the top level, monomorhism makes no sense at all.
524 main = let ?x = 5 in print foo
528 woggle :: (?x :: Int) => Int -> Int
531 We definitely don't want (foo :: Int) with a top-level implicit parameter
532 (?x::Int) becuase there is no way to bind it.
537 (A) Always generalise over implicit parameters
538 Bindings that fall under the monomorphism restriction can't
542 * Inlining remains valid
543 * No unexpected loss of sharing
544 * But simple bindings like
546 will be rejected, unless you add an explicit type signature
547 (to avoid the monomorphism restriction)
548 z :: (?y::Int) => Int
550 This seems unacceptable
552 (B) Monomorphism restriction "wins"
553 Bindings that fall under the monomorphism restriction can't
555 Always generalise over implicit parameters *except* for bindings
556 that fall under the monomorphism restriction
559 * Inlining isn't valid in general
560 * No unexpected loss of sharing
561 * Simple bindings like
563 accepted (get value of ?y from binding site)
565 (C) Always generalise over implicit parameters
566 Bindings that fall under the monomorphism restriction can't
567 be generalised, EXCEPT for implicit parameters
569 * Inlining remains valid
570 * Unexpected loss of sharing (from the extra generalisation)
571 * Simple bindings like
573 accepted (get value of ?y from occurrence sites)
578 None of these choices seems very satisfactory. But at least we should
579 decide which we want to do.
581 It's really not clear what is the Right Thing To Do. If you see
585 would you expect the value of ?y to be got from the *occurrence sites*
586 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
587 case of function definitions, the answer is clearly the former, but
588 less so in the case of non-fucntion definitions. On the other hand,
589 if we say that we get the value of ?y from the definition site of 'z',
590 then inlining 'z' might change the semantics of the program.
592 Choice (C) really says "the monomorphism restriction doesn't apply
593 to implicit parameters". Which is fine, but remember that every
594 innocent binding 'x = ...' that mentions an implicit parameter in
595 the RHS becomes a *function* of that parameter, called at each
596 use of 'x'. Now, the chances are that there are no intervening 'with'
597 clauses that bind ?y, so a decent compiler should common up all
598 those function calls. So I think I strongly favour (C). Indeed,
599 one could make a similar argument for abolishing the monomorphism
600 restriction altogether.
602 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
606 %************************************************************************
608 \subsection{tcSimplifyInfer}
610 %************************************************************************
612 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
614 1. Compute Q = grow( fvs(T), C )
616 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
617 predicates will end up in Ct; we deal with them at the top level
619 3. Try improvement, using functional dependencies
621 4. If Step 3 did any unification, repeat from step 1
622 (Unification can change the result of 'grow'.)
624 Note: we don't reduce dictionaries in step 2. For example, if we have
625 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
626 after step 2. However note that we may therefore quantify over more
627 type variables than we absolutely have to.
629 For the guts, we need a loop, that alternates context reduction and
630 improvement with unification. E.g. Suppose we have
632 class C x y | x->y where ...
634 and tcSimplify is called with:
636 Then improvement unifies a with b, giving
639 If we need to unify anything, we rattle round the whole thing all over
646 -> TcTyVarSet -- fv(T); type vars
648 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
649 [Inst], -- Dict Ids that must be bound here (zonked)
650 TcDictBinds) -- Bindings
651 -- Any free (escaping) Insts are tossed into the environment
656 tcSimplifyInfer doc tau_tvs wanted
657 = do { tau_tvs1 <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
658 ; wanted' <- mapM zonkInst wanted -- Zonk before deciding quantified tyvars
659 ; gbl_tvs <- tcGetGlobalTyVars
660 ; let preds1 = fdPredsOfInsts wanted'
661 gbl_tvs1 = oclose preds1 gbl_tvs
662 qtvs = grow preds1 tau_tvs1 `minusVarSet` gbl_tvs1
663 -- See Note [Choosing which variables to quantify]
665 -- To maximise sharing, remove from consideration any
666 -- constraints that don't mention qtvs at all
667 ; let (free, bound) = partition (isFreeWhenInferring qtvs) wanted'
670 -- To make types simple, reduce as much as possible
671 ; traceTc (text "infer" <+> (ppr preds1 $$ ppr (grow preds1 tau_tvs1) $$ ppr gbl_tvs $$
672 ppr gbl_tvs1 $$ ppr free $$ ppr bound))
673 ; (irreds1, binds1) <- tryHardCheckLoop doc bound
675 -- Note [Inference and implication constraints]
676 ; let want_dict d = tyVarsOfInst d `intersectsVarSet` qtvs
677 ; (irreds2, binds2) <- approximateImplications doc want_dict irreds1
679 -- Now work out all over again which type variables to quantify,
680 -- exactly in the same way as before, but starting from irreds2. Why?
681 -- a) By now improvment may have taken place, and we must *not*
682 -- quantify over any variable free in the environment
683 -- tc137 (function h inside g) is an example
685 -- b) Do not quantify over constraints that *now* do not
686 -- mention quantified type variables, because they are
687 -- simply ambiguous (or might be bound further out). Example:
688 -- f :: Eq b => a -> (a, b)
690 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
691 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
692 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
693 -- constraint (Eq beta), which we dump back into the free set
694 -- See test tcfail181
696 -- c) irreds may contain type variables not previously mentioned,
697 -- e.g. instance D a x => Foo [a]
699 -- Then after simplifying we'll get (D a x), and x is fresh
700 -- We must quantify over x else it'll be totally unbound
701 ; tau_tvs2 <- zonkTcTyVarsAndFV (varSetElems tau_tvs1)
702 ; gbl_tvs2 <- zonkTcTyVarsAndFV (varSetElems gbl_tvs1)
703 -- Note that we start from gbl_tvs1
704 -- We use tcGetGlobalTyVars, then oclose wrt preds2, because
705 -- we've already put some of the original preds1 into frees
706 -- E.g. wanteds = C a b (where a->b)
709 -- Then b is fixed by gbl_tvs, so (C a b) will be in free, and
710 -- irreds2 will be empty. But we don't want to generalise over b!
711 ; let preds2 = fdPredsOfInsts irreds2 -- irreds2 is zonked
712 qtvs = grow preds2 tau_tvs2 `minusVarSet` oclose preds2 gbl_tvs2
713 ---------------------------------------------------
714 -- BUG WARNING: there's a nasty bug lurking here
715 -- fdPredsOfInsts may return preds that mention variables quantified in
716 -- one of the implication constraints in irreds2; and that is clearly wrong:
717 -- we might quantify over too many variables through accidental capture
718 ---------------------------------------------------
720 ; let (free, irreds3) = partition (isFreeWhenInferring qtvs) irreds2
723 -- Turn the quantified meta-type variables into real type variables
724 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems qtvs)
726 -- We can't abstract over any remaining unsolved
727 -- implications so instead just float them outwards. Ugh.
728 ; let (q_dicts0, implics) = partition isAbstractableInst irreds3
729 ; loc <- getInstLoc (ImplicOrigin doc)
730 ; implic_bind <- bindIrreds loc qtvs2 q_dicts0 implics
732 -- Prepare equality instances for quantification
733 ; let (q_eqs0,q_dicts) = partition isEqInst q_dicts0
734 ; q_eqs <- mapM finalizeEqInst q_eqs0
736 ; return (qtvs2, q_eqs ++ q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
737 -- NB: when we are done, we might have some bindings, but
738 -- the final qtvs might be empty. See Note [NO TYVARS] below.
740 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
741 -- Note [Inference and implication constraints]
742 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
743 -- - fetching any dicts inside them that are free
744 -- - using those dicts as cruder constraints, to solve the implications
745 -- - returning the extra ones too
747 approximateImplications doc want_dict irreds
749 = return (irreds, emptyBag)
751 = do { extra_dicts' <- mapM cloneDict extra_dicts
752 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
753 -- By adding extra_dicts', we make them
754 -- available to solve the implication constraints
756 extra_dicts = get_dicts (filter isImplicInst irreds)
758 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
759 -- Find the wanted constraints in implication constraints that satisfy
760 -- want_dict, and are not bound by forall's in the constraint itself
761 get_dicts ds = concatMap get_dict ds
763 get_dict d@(Dict {}) | want_dict d = [d]
765 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
766 = [ d | let tv_set = mkVarSet tvs
767 , d <- get_dicts wanteds
768 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
769 get_dict i@(EqInst {}) | want_dict i = [i]
771 get_dict other = pprPanic "approximateImplications" (ppr other)
774 Note [Inference and implication constraints]
775 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
776 Suppose we have a wanted implication constraint (perhaps arising from
777 a nested pattern match) like
779 and we are now trying to quantify over 'a' when inferring the type for
780 a function. In principle it's possible that there might be an instance
781 instance (C a, E a) => D [a]
782 so the context (E a) would suffice. The Right Thing is to abstract over
783 the implication constraint, but we don't do that (a) because it'll be
784 surprising to programmers and (b) because we don't have the machinery to deal
785 with 'given' implications.
787 So our best approximation is to make (D [a]) part of the inferred
788 context, so we can use that to discharge the implication. Hence
789 the strange function get_dicts in approximateImplications.
791 The common cases are more clear-cut, when we have things like
793 Here, abstracting over (C b) is not an approximation at all -- but see
794 Note [Freeness and implications].
796 See Trac #1430 and test tc228.
800 -----------------------------------------------------------
801 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
802 -- against, but we don't know the type variables over which we are going to quantify.
803 -- This happens when we have a type signature for a mutually recursive group
806 -> TcTyVarSet -- fv(T)
809 -> TcM ([TyVar], -- Fully zonked, and quantified
810 TcDictBinds) -- Bindings
812 tcSimplifyInferCheck loc tau_tvs givens wanteds
813 = do { traceTc (text "tcSimplifyInferCheck <-" <+> ppr wanteds)
814 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
816 -- Figure out which type variables to quantify over
817 -- You might think it should just be the signature tyvars,
818 -- but in bizarre cases you can get extra ones
819 -- f :: forall a. Num a => a -> a
820 -- f x = fst (g (x, head [])) + 1
822 -- Here we infer g :: forall a b. a -> b -> (b,a)
823 -- We don't want g to be monomorphic in b just because
824 -- f isn't quantified over b.
825 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
826 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
827 ; gbl_tvs <- tcGetGlobalTyVars
828 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
829 -- We could close gbl_tvs, but its not necessary for
830 -- soundness, and it'll only affect which tyvars, not which
831 -- dictionaries, we quantify over
833 ; qtvs' <- zonkQuantifiedTyVars qtvs
835 -- Now we are back to normal (c.f. tcSimplCheck)
836 ; implic_bind <- bindIrreds loc qtvs' givens irreds
838 ; traceTc (text "tcSimplifyInferCheck ->" <+> ppr (implic_bind))
839 ; return (qtvs', binds `unionBags` implic_bind) }
842 Note [Squashing methods]
843 ~~~~~~~~~~~~~~~~~~~~~~~~~
844 Be careful if you want to float methods more:
845 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
846 From an application (truncate f i) we get
849 If we have also have a second occurrence of truncate, we get
852 When simplifying with i,f free, we might still notice that
853 t1=t3; but alas, the binding for t2 (which mentions t1)
854 may continue to float out!
859 class Y a b | a -> b where
862 instance Y [[a]] a where
865 k :: X a -> X a -> X a
867 g :: Num a => [X a] -> [X a]
870 h ys = ys ++ map (k (y [[0]])) xs
872 The excitement comes when simplifying the bindings for h. Initially
873 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
874 From this we get t1~t2, but also various bindings. We can't forget
875 the bindings (because of [LOOP]), but in fact t1 is what g is
878 The net effect of [NO TYVARS]
881 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
882 isFreeWhenInferring qtvs inst
883 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
884 && isInheritableInst inst -- and no implicit parameter involved
885 -- see Note [Inheriting implicit parameters]
887 {- No longer used (with implication constraints)
888 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
889 -> NameSet -- Quantified implicit parameters
891 isFreeWhenChecking qtvs ips inst
892 = isFreeWrtTyVars qtvs inst
893 && isFreeWrtIPs ips inst
896 isFreeWrtTyVars :: VarSet -> Inst -> Bool
897 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
898 isFreeWrtIPs :: NameSet -> Inst -> Bool
899 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
903 %************************************************************************
905 \subsection{tcSimplifyCheck}
907 %************************************************************************
909 @tcSimplifyCheck@ is used when we know exactly the set of variables
910 we are going to quantify over. For example, a class or instance declaration.
913 -----------------------------------------------------------
914 -- tcSimplifyCheck is used when checking expression type signatures,
915 -- class decls, instance decls etc.
916 tcSimplifyCheck :: InstLoc
917 -> [TcTyVar] -- Quantify over these
920 -> TcM TcDictBinds -- Bindings
921 tcSimplifyCheck loc qtvs givens wanteds
922 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
923 do { traceTc (text "tcSimplifyCheck")
924 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
925 ; implic_bind <- bindIrreds loc qtvs givens irreds
926 ; return (binds `unionBags` implic_bind) }
928 -----------------------------------------------------------
929 -- tcSimplifyCheckPat is used for existential pattern match
930 tcSimplifyCheckPat :: InstLoc
931 -> [TcTyVar] -- Quantify over these
934 -> TcM TcDictBinds -- Bindings
935 tcSimplifyCheckPat loc qtvs givens wanteds
936 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
937 do { traceTc (text "tcSimplifyCheckPat")
938 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
939 ; implic_bind <- bindIrredsR loc qtvs givens irreds
940 ; return (binds `unionBags` implic_bind) }
942 -----------------------------------------------------------
943 bindIrreds :: InstLoc -> [TcTyVar]
946 bindIrreds loc qtvs givens irreds
947 = bindIrredsR loc qtvs givens irreds
949 bindIrredsR :: InstLoc -> [TcTyVar] -> [Inst] -> [Inst] -> TcM TcDictBinds
950 -- Make a binding that binds 'irreds', by generating an implication
951 -- constraint for them, *and* throwing the constraint into the LIE
952 bindIrredsR loc qtvs givens irreds
956 = do { let givens' = filter isAbstractableInst givens
957 -- The givens can (redundantly) include methods
958 -- We want to retain both EqInsts and Dicts
959 -- There should be no implicadtion constraints
960 -- See Note [Pruning the givens in an implication constraint]
962 -- If there are no 'givens', then it's safe to
963 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
964 -- See Note [Freeness and implications]
965 ; irreds' <- if null givens'
967 { let qtv_set = mkVarSet qtvs
968 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
970 ; return real_irreds }
973 ; (implics, bind) <- makeImplicationBind loc qtvs givens' irreds'
974 -- This call does the real work
975 -- If irreds' is empty, it does something sensible
980 makeImplicationBind :: InstLoc -> [TcTyVar]
982 -> TcM ([Inst], TcDictBinds)
983 -- Make a binding that binds 'irreds', by generating an implication
984 -- constraint for them.
986 -- The binding looks like
987 -- (ir1, .., irn) = f qtvs givens
988 -- where f is (evidence for) the new implication constraint
989 -- f :: forall qtvs. givens => (ir1, .., irn)
990 -- qtvs includes coercion variables
992 -- This binding must line up the 'rhs' in reduceImplication
993 makeImplicationBind loc all_tvs
994 givens -- Guaranteed all Dicts or EqInsts
996 | null irreds -- If there are no irreds, we are done
997 = return ([], emptyBag)
998 | otherwise -- Otherwise we must generate a binding
999 = do { uniq <- newUnique
1000 ; span <- getSrcSpanM
1001 ; let (eq_givens, dict_givens) = partition isEqInst givens
1003 -- extract equality binders
1004 eq_cotvs = map eqInstType eq_givens
1006 -- make the implication constraint instance
1007 name = mkInternalName uniq (mkVarOcc "ic") span
1008 implic_inst = ImplicInst { tci_name = name,
1009 tci_tyvars = all_tvs,
1010 tci_given = eq_givens ++ dict_givens,
1011 -- same order as binders
1012 tci_wanted = irreds,
1015 -- create binders for the irreducible dictionaries
1016 dict_irreds = filter (not . isEqInst) irreds
1017 dict_irred_ids = map instToId dict_irreds
1018 lpat = mkBigLHsPatTup (map (L span . VarPat) dict_irred_ids)
1020 -- create the binding
1021 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
1022 co = mkWpApps (map instToId dict_givens)
1023 <.> mkWpTyApps eq_cotvs
1024 <.> mkWpTyApps (mkTyVarTys all_tvs)
1025 bind | [dict_irred_id] <- dict_irred_ids
1026 = mkVarBind dict_irred_id rhs
1029 PatBind { pat_lhs = lpat
1030 , pat_rhs = unguardedGRHSs rhs
1031 , pat_rhs_ty = hsLPatType lpat
1032 , bind_fvs = placeHolderNames
1035 ; traceTc $ text "makeImplicationBind" <+> ppr implic_inst
1036 ; return ([implic_inst], unitBag bind)
1039 -----------------------------------------------------------
1040 tryHardCheckLoop :: SDoc
1042 -> TcM ([Inst], TcDictBinds)
1044 tryHardCheckLoop doc wanteds
1045 = do { (irreds,binds) <- checkLoop (mkInferRedEnv doc try_me) wanteds
1046 ; return (irreds,binds)
1050 -- Here's the try-hard bit
1052 -----------------------------------------------------------
1053 gentleCheckLoop :: InstLoc
1056 -> TcM ([Inst], TcDictBinds)
1058 gentleCheckLoop inst_loc givens wanteds
1059 = do { (irreds,binds) <- checkLoop env wanteds
1060 ; return (irreds,binds)
1063 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1065 try_me inst | isMethodOrLit inst = ReduceMe
1067 -- When checking against a given signature
1068 -- we MUST be very gentle: Note [Check gently]
1070 gentleInferLoop :: SDoc -> [Inst]
1071 -> TcM ([Inst], TcDictBinds)
1072 gentleInferLoop doc wanteds
1073 = do { (irreds, binds) <- checkLoop env wanteds
1074 ; return (irreds, binds) }
1076 env = mkInferRedEnv doc try_me
1077 try_me inst | isMethodOrLit inst = ReduceMe
1082 ~~~~~~~~~~~~~~~~~~~~
1083 We have to very careful about not simplifying too vigorously
1088 f :: Show b => T b -> b
1089 f (MkT x) = show [x]
1091 Inside the pattern match, which binds (a:*, x:a), we know that
1093 Hence we have a dictionary for Show [a] available; and indeed we
1094 need it. We are going to build an implication contraint
1095 forall a. (b~[a]) => Show [a]
1096 Later, we will solve this constraint using the knowledge (Show b)
1098 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1099 thing becomes insoluble. So we simplify gently (get rid of literals
1100 and methods only, plus common up equal things), deferring the real
1101 work until top level, when we solve the implication constraint
1102 with tryHardCheckLooop.
1106 -----------------------------------------------------------
1109 -> TcM ([Inst], TcDictBinds)
1110 -- Precondition: givens are completely rigid
1111 -- Postcondition: returned Insts are zonked
1113 checkLoop env wanteds
1115 where go env wanteds
1116 = do { -- We do need to zonk the givens; cf Note [Zonking RedEnv]
1117 ; env' <- zonkRedEnv env
1118 ; wanteds' <- zonkInsts wanteds
1120 ; (improved, binds, irreds) <- reduceContext env' wanteds'
1122 ; if null irreds || not improved then
1123 return (irreds, binds)
1126 -- If improvement did some unification, we go round again.
1127 -- We start again with irreds, not wanteds
1128 -- Using an instance decl might have introduced a fresh type
1129 -- variable which might have been unified, so we'd get an
1130 -- infinite loop if we started again with wanteds!
1132 { (irreds1, binds1) <- go env' irreds
1133 ; return (irreds1, binds `unionBags` binds1) } }
1136 Note [Zonking RedEnv]
1137 ~~~~~~~~~~~~~~~~~~~~~
1138 It might appear as if the givens in RedEnv are always rigid, but that is not
1139 necessarily the case for programs involving higher-rank types that have class
1140 contexts constraining the higher-rank variables. An example from tc237 in the
1143 class Modular s a | s -> a
1145 wim :: forall a w. Integral a
1146 => a -> (forall s. Modular s a => M s w) -> w
1147 wim i k = error "urk"
1149 test5 :: (Modular s a, Integral a) => M s a
1152 test4 = wim 4 test4'
1154 Notice how the variable 'a' of (Modular s a) in the rank-2 type of wim is
1155 quantified further outside. When type checking test4, we have to check
1156 whether the signature of test5 is an instance of
1158 (forall s. Modular s a => M s w)
1160 Consequently, we will get (Modular s t_a), where t_a is a TauTv into the
1163 Given the FD of Modular in this example, class improvement will instantiate
1164 t_a to 'a', where 'a' is the skolem from test5's signatures (due to the
1165 Modular s a predicate in that signature). If we don't zonk (Modular s t_a) in
1166 the givens, we will get into a loop as improveOne uses the unification engine
1167 Unify.tcUnifyTys, which doesn't know about mutable type variables.
1172 class If b t e r | b t e -> r
1175 class Lte a b c | a b -> c where lte :: a -> b -> c
1177 instance (Lte a b l,If l b a c) => Max a b c
1179 Wanted: Max Z (S x) y
1181 Then we'll reduce using the Max instance to:
1182 (Lte Z (S x) l, If l (S x) Z y)
1183 and improve by binding l->T, after which we can do some reduction
1184 on both the Lte and If constraints. What we *can't* do is start again
1185 with (Max Z (S x) y)!
1189 %************************************************************************
1191 tcSimplifySuperClasses
1193 %************************************************************************
1195 Note [SUPERCLASS-LOOP 1]
1196 ~~~~~~~~~~~~~~~~~~~~~~~~
1197 We have to be very, very careful when generating superclasses, lest we
1198 accidentally build a loop. Here's an example:
1202 class S a => C a where { opc :: a -> a }
1203 class S b => D b where { opd :: b -> b }
1205 instance C Int where
1208 instance D Int where
1211 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1212 Simplifying, we may well get:
1213 $dfCInt = :C ds1 (opd dd)
1216 Notice that we spot that we can extract ds1 from dd.
1218 Alas! Alack! We can do the same for (instance D Int):
1220 $dfDInt = :D ds2 (opc dc)
1224 And now we've defined the superclass in terms of itself.
1225 Two more nasty cases are in
1230 - Satisfy the superclass context *all by itself*
1231 (tcSimplifySuperClasses)
1232 - And do so completely; i.e. no left-over constraints
1233 to mix with the constraints arising from method declarations
1236 Note [Recursive instances and superclases]
1237 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1238 Consider this code, which arises in the context of "Scrap Your
1239 Boilerplate with Class".
1243 instance Sat (ctx Char) => Data ctx Char
1244 instance (Sat (ctx [a]), Data ctx a) => Data ctx [a]
1246 class Data Maybe a => Foo a
1248 instance Foo t => Sat (Maybe t)
1250 instance Data Maybe a => Foo a
1251 instance Foo a => Foo [a]
1254 In the instance for Foo [a], when generating evidence for the superclasses
1255 (ie in tcSimplifySuperClasses) we need a superclass (Data Maybe [a]).
1256 Using the instance for Data, we therefore need
1257 (Sat (Maybe [a], Data Maybe a)
1258 But we are given (Foo a), and hence its superclass (Data Maybe a).
1259 So that leaves (Sat (Maybe [a])). Using the instance for Sat means
1260 we need (Foo [a]). And that is the very dictionary we are bulding
1261 an instance for! So we must put that in the "givens". So in this
1263 Given: Foo a, Foo [a]
1264 Watend: Data Maybe [a]
1266 BUT we must *not not not* put the *superclasses* of (Foo [a]) in
1267 the givens, which is what 'addGiven' would normally do. Why? Because
1268 (Data Maybe [a]) is the superclass, so we'd "satisfy" the wanted
1269 by selecting a superclass from Foo [a], which simply makes a loop.
1271 On the other hand we *must* put the superclasses of (Foo a) in
1272 the givens, as you can see from the derivation described above.
1274 Conclusion: in the very special case of tcSimplifySuperClasses
1275 we have one 'given' (namely the "this" dictionary) whose superclasses
1276 must not be added to 'givens' by addGiven. That is the *whole* reason
1277 for the red_given_scs field in RedEnv, and the function argument to
1281 tcSimplifySuperClasses
1283 -> Inst -- The dict whose superclasses
1284 -- are being figured out
1288 tcSimplifySuperClasses loc this givens sc_wanteds
1289 = do { traceTc (text "tcSimplifySuperClasses")
1290 ; (irreds,binds1) <- checkLoop env sc_wanteds
1291 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1292 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1295 env = RedEnv { red_doc = pprInstLoc loc,
1296 red_try_me = try_me,
1297 red_givens = this:givens,
1298 red_given_scs = add_scs,
1300 red_improve = False } -- No unification vars
1301 add_scs g | g==this = NoSCs
1302 | otherwise = AddSCs
1304 try_me _ = ReduceMe -- Try hard, so we completely solve the superclass
1305 -- constraints right here. See Note [SUPERCLASS-LOOP 1]
1309 %************************************************************************
1311 \subsection{tcSimplifyRestricted}
1313 %************************************************************************
1315 tcSimplifyRestricted infers which type variables to quantify for a
1316 group of restricted bindings. This isn't trivial.
1319 We want to quantify over a to get id :: forall a. a->a
1322 We do not want to quantify over a, because there's an Eq a
1323 constraint, so we get eq :: a->a->Bool (notice no forall)
1326 RHS has type 'tau', whose free tyvars are tau_tvs
1327 RHS has constraints 'wanteds'
1330 Quantify over (tau_tvs \ ftvs(wanteds))
1331 This is bad. The constraints may contain (Monad (ST s))
1332 where we have instance Monad (ST s) where...
1333 so there's no need to be monomorphic in s!
1335 Also the constraint might be a method constraint,
1336 whose type mentions a perfectly innocent tyvar:
1337 op :: Num a => a -> b -> a
1338 Here, b is unconstrained. A good example would be
1340 We want to infer the polymorphic type
1341 foo :: forall b. b -> b
1344 Plan B (cunning, used for a long time up to and including GHC 6.2)
1345 Step 1: Simplify the constraints as much as possible (to deal
1346 with Plan A's problem). Then set
1347 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1349 Step 2: Now simplify again, treating the constraint as 'free' if
1350 it does not mention qtvs, and trying to reduce it otherwise.
1351 The reasons for this is to maximise sharing.
1353 This fails for a very subtle reason. Suppose that in the Step 2
1354 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1355 In the Step 1 this constraint might have been simplified, perhaps to
1356 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1357 This won't happen in Step 2... but that in turn might prevent some other
1358 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1359 and that in turn breaks the invariant that no constraints are quantified over.
1361 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1366 Step 1: Simplify the constraints as much as possible (to deal
1367 with Plan A's problem). Then set
1368 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1369 Return the bindings from Step 1.
1372 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1375 instance (HasBinary ty IO) => HasCodedValue ty
1377 foo :: HasCodedValue a => String -> IO a
1379 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1380 doDecodeIO codedValue view
1381 = let { act = foo "foo" } in act
1383 You might think this should work becuase the call to foo gives rise to a constraint
1384 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1385 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1386 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1388 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1392 Plan D (a variant of plan B)
1393 Step 1: Simplify the constraints as much as possible (to deal
1394 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1395 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1397 Step 2: Now simplify again, treating the constraint as 'free' if
1398 it does not mention qtvs, and trying to reduce it otherwise.
1400 The point here is that it's generally OK to have too few qtvs; that is,
1401 to make the thing more monomorphic than it could be. We don't want to
1402 do that in the common cases, but in wierd cases it's ok: the programmer
1403 can always add a signature.
1405 Too few qtvs => too many wanteds, which is what happens if you do less
1410 tcSimplifyRestricted -- Used for restricted binding groups
1411 -- i.e. ones subject to the monomorphism restriction
1414 -> [Name] -- Things bound in this group
1415 -> TcTyVarSet -- Free in the type of the RHSs
1416 -> [Inst] -- Free in the RHSs
1417 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1418 TcDictBinds) -- Bindings
1419 -- tcSimpifyRestricted returns no constraints to
1420 -- quantify over; by definition there are none.
1421 -- They are all thrown back in the LIE
1423 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1424 -- Zonk everything in sight
1425 = do { traceTc (text "tcSimplifyRestricted")
1426 ; wanteds_z <- zonkInsts wanteds
1428 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1429 -- dicts; the idea is to get rid of as many type
1430 -- variables as possible, and we don't want to stop
1431 -- at (say) Monad (ST s), because that reduces
1432 -- immediately, with no constraint on s.
1434 -- BUT do no improvement! See Plan D above
1435 -- HOWEVER, some unification may take place, if we instantiate
1436 -- a method Inst with an equality constraint
1437 ; let env = mkNoImproveRedEnv doc (\_ -> ReduceMe)
1438 ; (_imp, _binds, constrained_dicts) <- reduceContext env wanteds_z
1440 -- Next, figure out the tyvars we will quantify over
1441 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1442 ; gbl_tvs' <- tcGetGlobalTyVars
1443 ; constrained_dicts' <- zonkInsts constrained_dicts
1445 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1446 -- As in tcSimplifyInfer
1448 -- Do not quantify over constrained type variables:
1449 -- this is the monomorphism restriction
1450 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1451 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1452 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1455 ; warn_mono <- doptM Opt_WarnMonomorphism
1456 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1457 (vcat[ ptext (sLit "the Monomorphism Restriction applies to the binding")
1458 <> plural bndrs <+> ptext (sLit "for") <+> pp_bndrs,
1459 ptext (sLit "Consider giving a type signature for") <+> pp_bndrs])
1461 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1462 pprInsts wanteds, pprInsts constrained_dicts',
1464 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1466 -- Zonk wanteds again! The first call to reduceContext may have
1467 -- instantiated some variables.
1468 -- FIXME: If red_improve would work, we could propagate that into
1469 -- the equality solver, too, to prevent instantating any
1471 ; wanteds_zz <- zonkInsts wanteds_z
1473 -- The first step may have squashed more methods than
1474 -- necessary, so try again, this time more gently, knowing the exact
1475 -- set of type variables to quantify over.
1477 -- We quantify only over constraints that are captured by qtvs;
1478 -- these will just be a subset of non-dicts. This in contrast
1479 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1480 -- all *non-inheritable* constraints too. This implements choice
1481 -- (B) under "implicit parameter and monomorphism" above.
1483 -- Remember that we may need to do *some* simplification, to
1484 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1485 -- just to float all constraints
1487 -- At top level, we *do* squash methods becuase we want to
1488 -- expose implicit parameters to the test that follows
1489 ; let is_nested_group = isNotTopLevel top_lvl
1490 try_me inst | isFreeWrtTyVars qtvs inst,
1491 (is_nested_group || isDict inst) = Stop
1492 | otherwise = ReduceMe
1493 env = mkNoImproveRedEnv doc try_me
1494 ; (_imp, binds, irreds) <- reduceContext env wanteds_zz
1496 -- See "Notes on implicit parameters, Question 4: top level"
1497 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1498 if is_nested_group then
1500 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1501 ; addTopIPErrs bndrs bad_ips
1502 ; extendLIEs non_ips }
1504 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1505 ; return (qtvs', binds) }
1509 %************************************************************************
1513 %************************************************************************
1515 On the LHS of transformation rules we only simplify methods and constants,
1516 getting dictionaries. We want to keep all of them unsimplified, to serve
1517 as the available stuff for the RHS of the rule.
1519 Example. Consider the following left-hand side of a rule
1521 f (x == y) (y > z) = ...
1523 If we typecheck this expression we get constraints
1525 d1 :: Ord a, d2 :: Eq a
1527 We do NOT want to "simplify" to the LHS
1529 forall x::a, y::a, z::a, d1::Ord a.
1530 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1534 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1535 f ((==) d2 x y) ((>) d1 y z) = ...
1537 Here is another example:
1539 fromIntegral :: (Integral a, Num b) => a -> b
1540 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1542 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1543 we *dont* want to get
1545 forall dIntegralInt.
1546 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1548 because the scsel will mess up RULE matching. Instead we want
1550 forall dIntegralInt, dNumInt.
1551 fromIntegral Int Int dIntegralInt dNumInt = id Int
1555 g (x == y) (y == z) = ..
1557 where the two dictionaries are *identical*, we do NOT WANT
1559 forall x::a, y::a, z::a, d1::Eq a
1560 f ((==) d1 x y) ((>) d1 y z) = ...
1562 because that will only match if the dict args are (visibly) equal.
1563 Instead we want to quantify over the dictionaries separately.
1565 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1566 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1567 from scratch, rather than further parameterise simpleReduceLoop etc.
1568 Simpler, maybe, but alas not simple (see Trac #2494)
1570 * Type errors may give rise to an (unsatisfiable) equality constraint
1572 * Applications of a higher-rank function on the LHS may give
1573 rise to an implication constraint, esp if there are unsatisfiable
1574 equality constraints inside.
1577 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1578 tcSimplifyRuleLhs wanteds
1579 = do { wanteds' <- zonkInsts wanteds
1580 ; (irreds, binds) <- go [] emptyBag wanteds'
1581 ; let (dicts, bad_irreds) = partition isDict irreds
1582 ; traceTc (text "tcSimplifyrulelhs" <+> pprInsts bad_irreds)
1583 ; addNoInstanceErrs (nub bad_irreds)
1584 -- The nub removes duplicates, which has
1585 -- not happened otherwise (see notes above)
1586 ; return (dicts, binds) }
1588 go :: [Inst] -> TcDictBinds -> [Inst] -> TcM ([Inst], TcDictBinds)
1590 = return (irreds, binds)
1591 go irreds binds (w:ws)
1593 = go (w:irreds) binds ws
1594 | isImplicInst w -- Have a go at reducing the implication
1595 = do { (binds1, irreds1) <- reduceImplication red_env w
1596 ; let (bad_irreds, ok_irreds) = partition isImplicInst irreds1
1597 ; go (bad_irreds ++ irreds)
1598 (binds `unionBags` binds1)
1601 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1602 -- to fromInteger; this looks fragile to me
1603 ; lookup_result <- lookupSimpleInst w'
1604 ; case lookup_result of
1605 NoInstance -> go (w:irreds) binds ws
1606 GenInst ws' rhs -> go irreds binds' (ws' ++ ws)
1608 binds' = addInstToDictBind binds w rhs
1611 -- Sigh: we need to reduce inside implications
1612 red_env = mkInferRedEnv doc try_me
1613 doc = ptext (sLit "Implication constraint in RULE lhs")
1614 try_me inst | isMethodOrLit inst = ReduceMe
1615 | otherwise = Stop -- Be gentle
1618 tcSimplifyBracket is used when simplifying the constraints arising from
1619 a Template Haskell bracket [| ... |]. We want to check that there aren't
1620 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1621 Show instance), but we aren't otherwise interested in the results.
1622 Nor do we care about ambiguous dictionaries etc. We will type check
1623 this bracket again at its usage site.
1626 tcSimplifyBracket :: [Inst] -> TcM ()
1627 tcSimplifyBracket wanteds
1628 = do { tryHardCheckLoop doc wanteds
1631 doc = text "tcSimplifyBracket"
1635 %************************************************************************
1637 \subsection{Filtering at a dynamic binding}
1639 %************************************************************************
1644 we must discharge all the ?x constraints from B. We also do an improvement
1645 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1647 Actually, the constraints from B might improve the types in ?x. For example
1649 f :: (?x::Int) => Char -> Char
1652 then the constraint (?x::Int) arising from the call to f will
1653 force the binding for ?x to be of type Int.
1656 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1659 -- We need a loop so that we do improvement, and then
1660 -- (next time round) generate a binding to connect the two
1662 -- Here the two ?x's have different types, and improvement
1663 -- makes them the same.
1665 tcSimplifyIPs given_ips wanteds
1666 = do { wanteds' <- zonkInsts wanteds
1667 ; given_ips' <- zonkInsts given_ips
1668 -- Unusually for checking, we *must* zonk the given_ips
1670 ; let env = mkRedEnv doc try_me given_ips'
1671 ; (improved, binds, irreds) <- reduceContext env wanteds'
1673 ; if null irreds || not improved then
1674 ASSERT( all is_free irreds )
1675 do { extendLIEs irreds
1678 -- If improvement did some unification, we go round again.
1679 -- We start again with irreds, not wanteds
1680 -- Using an instance decl might have introduced a fresh type
1681 -- variable which might have been unified, so we'd get an
1682 -- infinite loop if we started again with wanteds!
1684 { binds1 <- tcSimplifyIPs given_ips' irreds
1685 ; return $ binds `unionBags` binds1
1688 doc = text "tcSimplifyIPs" <+> ppr given_ips
1689 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1690 is_free inst = isFreeWrtIPs ip_set inst
1692 -- Simplify any methods that mention the implicit parameter
1693 try_me inst | is_free inst = Stop
1694 | otherwise = ReduceMe
1698 %************************************************************************
1700 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1702 %************************************************************************
1704 When doing a binding group, we may have @Insts@ of local functions.
1705 For example, we might have...
1707 let f x = x + 1 -- orig local function (overloaded)
1708 f.1 = f Int -- two instances of f
1713 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1714 where @f@ is in scope; those @Insts@ must certainly not be passed
1715 upwards towards the top-level. If the @Insts@ were binding-ified up
1716 there, they would have unresolvable references to @f@.
1718 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1719 For each method @Inst@ in the @init_lie@ that mentions one of the
1720 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1721 @LIE@), as well as the @HsBinds@ generated.
1724 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1725 -- Simlifies only MethodInsts, and generate only bindings of form
1727 -- We're careful not to even generate bindings of the form
1729 -- You'd think that'd be fine, but it interacts with what is
1730 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1732 bindInstsOfLocalFuns wanteds local_ids
1733 | null overloaded_ids = do
1736 return emptyLHsBinds
1739 = do { (irreds, binds) <- gentleInferLoop doc for_me
1740 ; extendLIEs not_for_me
1744 doc = text "bindInsts" <+> ppr local_ids
1745 overloaded_ids = filter is_overloaded local_ids
1746 is_overloaded id = isOverloadedTy (idType id)
1747 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1749 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1750 -- so it's worth building a set, so that
1751 -- lookup (in isMethodFor) is faster
1755 %************************************************************************
1757 \subsection{Data types for the reduction mechanism}
1759 %************************************************************************
1761 The main control over context reduction is here
1765 = RedEnv { red_doc :: SDoc -- The context
1766 , red_try_me :: Inst -> WhatToDo
1767 , red_improve :: Bool -- True <=> do improvement
1768 , red_givens :: [Inst] -- All guaranteed rigid
1769 -- Always dicts & equalities
1770 -- but see Note [Rigidity]
1772 , red_given_scs :: Inst -> WantSCs -- See Note [Recursive instances and superclases]
1774 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1775 -- See Note [RedStack]
1779 -- The red_givens are rigid so far as cmpInst is concerned.
1780 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1781 -- let ?x = e in ...
1782 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1783 -- But that doesn't affect the comparison, which is based only on mame.
1786 -- The red_stack pair (n,insts) pair is just used for error reporting.
1787 -- 'n' is always the depth of the stack.
1788 -- The 'insts' is the stack of Insts being reduced: to produce X
1789 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1792 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1793 mkRedEnv doc try_me givens
1794 = RedEnv { red_doc = doc, red_try_me = try_me,
1795 red_givens = givens,
1796 red_given_scs = const AddSCs,
1798 red_improve = True }
1800 mkInferRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1802 mkInferRedEnv doc try_me
1803 = RedEnv { red_doc = doc, red_try_me = try_me,
1805 red_given_scs = const AddSCs,
1807 red_improve = True }
1809 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1810 -- Do not do improvement; no givens
1811 mkNoImproveRedEnv doc try_me
1812 = RedEnv { red_doc = doc, red_try_me = try_me,
1814 red_given_scs = const AddSCs,
1816 red_improve = True }
1819 = ReduceMe -- Try to reduce this
1820 -- If there's no instance, add the inst to the
1821 -- irreductible ones, but don't produce an error
1822 -- message of any kind.
1823 -- It might be quite legitimate such as (Eq a)!
1825 | Stop -- Return as irreducible unless it can
1826 -- be reduced to a constant in one step
1827 -- Do not add superclasses; see
1829 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1830 -- of a predicate when adding it to the avails
1831 -- The reason for this flag is entirely the super-class loop problem
1832 -- Note [SUPER-CLASS LOOP 1]
1834 zonkRedEnv :: RedEnv -> TcM RedEnv
1836 = do { givens' <- mapM zonkInst (red_givens env)
1837 ; return $ env {red_givens = givens'}
1842 %************************************************************************
1844 \subsection[reduce]{@reduce@}
1846 %************************************************************************
1848 Note [Ancestor Equalities]
1849 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1850 During context reduction, we add to the wanted equalities also those
1851 equalities that (transitively) occur in superclass contexts of wanted
1852 class constraints. Consider the following code
1854 class a ~ Int => C a
1857 If (C a) is wanted, we want to add (a ~ Int), which will be discharged by
1858 substituting Int for a. Hence, we ultimately want (C Int), which we
1859 discharge with the explicit instance.
1862 reduceContext :: RedEnv
1864 -> TcM (ImprovementDone,
1865 TcDictBinds, -- Dictionary bindings
1866 [Inst]) -- Irreducible
1868 reduceContext env wanteds0
1869 = do { traceTc (text "reduceContext" <+> (vcat [
1870 text "----------------------",
1872 text "given" <+> ppr (red_givens env),
1873 text "wanted" <+> ppr wanteds0,
1874 text "----------------------"
1877 -- We want to add as wanted equalities those that (transitively)
1878 -- occur in superclass contexts of wanted class constraints.
1879 -- See Note [Ancestor Equalities]
1880 ; ancestor_eqs <- ancestorEqualities wanteds0
1881 ; traceTc $ text "reduceContext: ancestor eqs" <+> ppr ancestor_eqs
1883 -- Normalise and solve all equality constraints as far as possible
1884 -- and normalise all dictionary constraints wrt to the reduced
1885 -- equalities. The returned wanted constraints include the
1886 -- irreducible wanted equalities.
1887 ; let wanteds = wanteds0 ++ ancestor_eqs
1888 givens = red_givens env
1892 eq_improved) <- tcReduceEqs givens wanteds
1893 ; traceTc $ text "reduceContext: tcReduceEqs result" <+> vcat
1894 [ppr givens', ppr wanteds', ppr normalise_binds]
1896 -- Build the Avail mapping from "given_dicts"
1897 ; (init_state, _) <- getLIE $ do
1898 { init_state <- foldlM (addGiven (red_given_scs env))
1903 -- Solve the *wanted* *dictionary* constraints (not implications)
1904 -- This may expose some further equational constraints in the course
1905 -- of improvement due to functional dependencies if any of the
1906 -- involved unifications gets deferred.
1907 ; let (wanted_implics, wanted_dicts) = partition isImplicInst wanteds'
1908 ; (avails, extra_eqs) <- getLIE (reduceList env wanted_dicts init_state)
1909 -- The getLIE is reqd because reduceList does improvement
1910 -- (via extendAvails) which may in turn do unification
1913 dict_irreds) <- extractResults avails wanted_dicts
1914 ; traceTc $ text "reduceContext: extractResults" <+> vcat
1915 [ppr avails, ppr wanted_dicts, ppr dict_binds]
1917 -- Solve the wanted *implications*. In doing so, we can provide
1918 -- as "given" all the dicts that were originally given,
1919 -- *or* for which we now have bindings,
1920 -- *or* which are now irreds
1921 -- NB: Equality irreds need to be converted, as the recursive
1922 -- invocation of the solver will still treat them as wanteds
1924 ; let implic_env = env { red_givens
1925 = givens ++ bound_dicts ++
1926 map wantedToLocalEqInst dict_irreds }
1927 ; (implic_binds_s, implic_irreds_s)
1928 <- mapAndUnzipM (reduceImplication implic_env) wanted_implics
1929 ; let implic_binds = unionManyBags implic_binds_s
1930 implic_irreds = concat implic_irreds_s
1932 -- Collect all irreducible instances, and determine whether we should
1933 -- go round again. We do so in either of two cases:
1934 -- (1) If dictionary reduction or equality solving led to
1935 -- improvement (i.e., instantiated type variables).
1936 -- (2) If we reduced dictionaries (i.e., got dictionary bindings),
1937 -- they may have exposed further opportunities to normalise
1938 -- family applications. See Note [Dictionary Improvement]
1940 -- NB: We do *not* go around for new extra_eqs. Morally, we should,
1941 -- but we can't without risking non-termination (see #2688). By
1942 -- not going around, we miss some legal programs mixing FDs and
1943 -- TFs, but we never claimed to support such programs in the
1944 -- current implementation anyway.
1946 ; let all_irreds = dict_irreds ++ implic_irreds ++ extra_eqs
1947 avails_improved = availsImproved avails
1948 improvedFlexible = avails_improved || eq_improved
1949 reduced_dicts = not (isEmptyBag dict_binds)
1950 improved = improvedFlexible || reduced_dicts
1952 improvedHint = (if avails_improved then " [AVAILS]" else "") ++
1953 (if eq_improved then " [EQ]" else "")
1955 ; traceTc (text "reduceContext end" <+> (vcat [
1956 text "----------------------",
1958 text "given" <+> ppr givens,
1959 text "wanted" <+> ppr wanteds0,
1961 text "avails" <+> pprAvails avails,
1962 text "improved =" <+> ppr improved <+> text improvedHint,
1963 text "(all) irreds = " <+> ppr all_irreds,
1964 text "dict-binds = " <+> ppr dict_binds,
1965 text "implic-binds = " <+> ppr implic_binds,
1966 text "----------------------"
1970 normalise_binds `unionBags` dict_binds
1971 `unionBags` implic_binds,
1975 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1976 tcImproveOne avails inst
1977 | not (isDict inst) = return False
1979 = do { inst_envs <- tcGetInstEnvs
1980 ; let eqns = improveOne (classInstances inst_envs)
1981 (dictPred inst, pprInstArising inst)
1982 [ (dictPred p, pprInstArising p)
1983 | p <- availsInsts avails, isDict p ]
1984 -- Avails has all the superclasses etc (good)
1985 -- It also has all the intermediates of the deduction (good)
1986 -- It does not have duplicates (good)
1987 -- NB that (?x::t1) and (?x::t2) will be held separately in
1988 -- avails so that improve will see them separate
1989 ; traceTc (text "improveOne" <+> ppr inst)
1992 unifyEqns :: [(Equation, (PredType, SDoc), (PredType, SDoc))]
1993 -> TcM ImprovementDone
1994 unifyEqns [] = return False
1996 = do { traceTc (ptext (sLit "Improve:") <+> vcat (map pprEquationDoc eqns))
1997 ; improved <- mapM unify eqns
1998 ; return $ or improved
2001 unify ((qtvs, pairs), what1, what2)
2002 = addErrCtxtM (mkEqnMsg what1 what2) $
2003 do { let freeTyVars = unionVarSets (map tvs_pr pairs)
2005 ; (_, _, tenv) <- tcInstTyVars (varSetElems qtvs)
2006 ; mapM_ (unif_pr tenv) pairs
2007 ; anyM isFilledMetaTyVar $ varSetElems freeTyVars
2010 unif_pr tenv (ty1, ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
2012 tvs_pr (ty1, ty2) = tyVarsOfType ty1 `unionVarSet` tyVarsOfType ty2
2014 pprEquationDoc :: (Equation, (PredType, SDoc), (PredType, SDoc)) -> SDoc
2015 pprEquationDoc (eqn, (p1, _), (p2, _))
2016 = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
2018 mkEqnMsg :: (TcPredType, SDoc) -> (TcPredType, SDoc) -> TidyEnv
2019 -> TcM (TidyEnv, SDoc)
2020 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
2021 = do { pred1' <- zonkTcPredType pred1
2022 ; pred2' <- zonkTcPredType pred2
2023 ; let { pred1'' = tidyPred tidy_env pred1'
2024 ; pred2'' = tidyPred tidy_env pred2' }
2025 ; let msg = vcat [ptext (sLit "When using functional dependencies to combine"),
2026 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
2027 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
2028 ; return (tidy_env, msg) }
2031 Note [Dictionary Improvement]
2032 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2033 In reduceContext, we first reduce equalities and then class constraints.
2034 However, the letter may expose further opportunities for the former. Hence,
2035 we need to go around again if dictionary reduction produced any dictionary
2036 bindings. The following example demonstrated the point:
2038 data EX _x _y (p :: * -> *)
2043 class Base (Def p) => Prop p where
2047 instance Prop () where
2050 instance (Base (Def (p ANY))) => Base (EX _x _y p)
2051 instance (Prop (p ANY)) => Prop (EX _x _y p) where
2052 type Def (EX _x _y p) = EX _x _y p
2055 instance Prop (FOO x) where
2056 type Def (FOO x) = ()
2059 instance Prop BAR where
2060 type Def BAR = EX () () FOO
2062 During checking the last instance declaration, we need to check the superclass
2063 cosntraint Base (Def BAR), which family normalisation reduced to
2064 Base (EX () () FOO). Chasing the instance for Base (EX _x _y p), gives us
2065 Base (Def (FOO ANY)), which again requires family normalisation of Def to
2066 Base () before we can finish.
2069 The main context-reduction function is @reduce@. Here's its game plan.
2072 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
2073 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
2074 = do { traceTc (text "reduceList " <+> (ppr wanteds $$ ppr state))
2076 ; when (debugIsOn && (n > 8)) $ do
2077 debugDumpTcRn (hang (ptext (sLit "Interesting! Context reduction stack depth") <+> int n)
2078 2 (ifPprDebug (nest 2 (pprStack stk))))
2079 ; if n >= ctxtStkDepth dopts then
2080 failWithTc (reduceDepthErr n stk)
2084 go [] state = return state
2085 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
2088 -- Base case: we're done!
2089 reduce :: RedEnv -> Inst -> Avails -> TcM Avails
2090 reduce env wanted avails
2092 -- We don't reduce equalities here (and they must not end up as irreds
2097 -- It's the same as an existing inst, or a superclass thereof
2098 | Just _ <- findAvail avails wanted
2099 = do { traceTc (text "reduce: found " <+> ppr wanted)
2104 = do { traceTc (text "reduce" <+> ppr wanted $$ ppr avails)
2105 ; case red_try_me env wanted of {
2106 Stop -> try_simple (addIrred NoSCs);
2107 -- See Note [No superclasses for Stop]
2109 ReduceMe -> do -- It should be reduced
2110 { (avails, lookup_result) <- reduceInst env avails wanted
2111 ; case lookup_result of
2112 NoInstance -> addIrred AddSCs avails wanted
2113 -- Add it and its superclasses
2115 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
2117 GenInst wanteds' rhs
2118 -> do { avails1 <- addIrred NoSCs avails wanted
2119 ; avails2 <- reduceList env wanteds' avails1
2120 ; addWanted AddSCs avails2 wanted rhs wanteds' } }
2121 -- Temporarily do addIrred *before* the reduceList,
2122 -- which has the effect of adding the thing we are trying
2123 -- to prove to the database before trying to prove the things it
2124 -- needs. See note [RECURSIVE DICTIONARIES]
2125 -- NB: we must not do an addWanted before, because that adds the
2126 -- superclasses too, and that can lead to a spurious loop; see
2127 -- the examples in [SUPERCLASS-LOOP]
2128 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
2131 -- First, see if the inst can be reduced to a constant in one step
2132 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
2133 -- Don't bother for implication constraints, which take real work
2134 try_simple do_this_otherwise
2135 = do { res <- lookupSimpleInst wanted
2137 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
2138 _ -> do_this_otherwise avails wanted }
2142 Note [RECURSIVE DICTIONARIES]
2143 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2145 data D r = ZeroD | SuccD (r (D r));
2147 instance (Eq (r (D r))) => Eq (D r) where
2148 ZeroD == ZeroD = True
2149 (SuccD a) == (SuccD b) = a == b
2152 equalDC :: D [] -> D [] -> Bool;
2155 We need to prove (Eq (D [])). Here's how we go:
2159 by instance decl, holds if
2163 by instance decl of Eq, holds if
2165 where d2 = dfEqList d3
2168 But now we can "tie the knot" to give
2174 and it'll even run! The trick is to put the thing we are trying to prove
2175 (in this case Eq (D []) into the database before trying to prove its
2176 contributing clauses.
2178 Note [SUPERCLASS-LOOP 2]
2179 ~~~~~~~~~~~~~~~~~~~~~~~~
2180 We need to be careful when adding "the constaint we are trying to prove".
2181 Suppose we are *given* d1:Ord a, and want to deduce (d2:C [a]) where
2183 class Ord a => C a where
2184 instance Ord [a] => C [a] where ...
2186 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
2187 superclasses of C [a] to avails. But we must not overwrite the binding
2188 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
2191 Here's another variant, immortalised in tcrun020
2192 class Monad m => C1 m
2193 class C1 m => C2 m x
2194 instance C2 Maybe Bool
2195 For the instance decl we need to build (C1 Maybe), and it's no good if
2196 we run around and add (C2 Maybe Bool) and its superclasses to the avails
2197 before we search for C1 Maybe.
2199 Here's another example
2200 class Eq b => Foo a b
2201 instance Eq a => Foo [a] a
2205 we'll first deduce that it holds (via the instance decl). We must not
2206 then overwrite the Eq t constraint with a superclass selection!
2208 At first I had a gross hack, whereby I simply did not add superclass constraints
2209 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
2210 becuase it lost legitimate superclass sharing, and it still didn't do the job:
2211 I found a very obscure program (now tcrun021) in which improvement meant the
2212 simplifier got two bites a the cherry... so something seemed to be an Stop
2213 first time, but reducible next time.
2215 Now we implement the Right Solution, which is to check for loops directly
2216 when adding superclasses. It's a bit like the occurs check in unification.
2220 %************************************************************************
2222 Reducing a single constraint
2224 %************************************************************************
2227 ---------------------------------------------
2228 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
2229 reduceInst _ avails other_inst
2230 = do { result <- lookupSimpleInst other_inst
2231 ; return (avails, result) }
2234 Note [Equational Constraints in Implication Constraints]
2235 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2237 An implication constraint is of the form
2239 where Given and Wanted may contain both equational and dictionary
2240 constraints. The delay and reduction of these two kinds of constraints
2243 -) In the generated code, wanted Dictionary constraints are wrapped up in an
2244 implication constraint that is created at the code site where the wanted
2245 dictionaries can be reduced via a let-binding. This let-bound implication
2246 constraint is deconstructed at the use-site of the wanted dictionaries.
2248 -) While the reduction of equational constraints is also delayed, the delay
2249 is not manifest in the generated code. The required evidence is generated
2250 in the code directly at the use-site. There is no let-binding and deconstruction
2251 necessary. The main disadvantage is that we cannot exploit sharing as the
2252 same evidence may be generated at multiple use-sites. However, this disadvantage
2253 is limited because it only concerns coercions which are erased.
2255 The different treatment is motivated by the different in representation. Dictionary
2256 constraints require manifest runtime dictionaries, while equations require coercions
2260 ---------------------------------------------
2261 reduceImplication :: RedEnv
2263 -> TcM (TcDictBinds, [Inst])
2266 Suppose we are simplifying the constraint
2267 forall bs. extras => wanted
2268 in the context of an overall simplification problem with givens 'givens'.
2271 * The 'givens' need not mention any of the quantified type variables
2272 e.g. forall {}. Eq a => Eq [a]
2273 forall {}. C Int => D (Tree Int)
2275 This happens when you have something like
2277 T1 :: Eq a => a -> T a
2280 f x = ...(case x of { T1 v -> v==v })...
2283 -- ToDo: should we instantiate tvs? I think it's not necessary
2285 -- Note on coercion variables:
2287 -- The extra given coercion variables are bound at two different
2290 -- -) in the creation context of the implication constraint
2291 -- the solved equational constraints use these binders
2293 -- -) at the solving site of the implication constraint
2294 -- the solved dictionaries use these binders;
2295 -- these binders are generated by reduceImplication
2297 -- Note [Binders for equalities]
2298 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2299 -- To reuse the binders of local/given equalities in the binders of
2300 -- implication constraints, it is crucial that these given equalities
2301 -- always have the form
2303 -- where cotv is a simple coercion type variable (and not a more
2304 -- complex coercion term). We require that the extra_givens always
2305 -- have this form and exploit the special form when generating binders.
2306 reduceImplication env
2307 orig_implic@(ImplicInst { tci_name = name, tci_loc = inst_loc,
2309 tci_given = extra_givens, tci_wanted = wanteds
2311 = do { -- Solve the sub-problem
2312 ; let try_me _ = ReduceMe -- Note [Freeness and implications]
2313 env' = env { red_givens = extra_givens ++ red_givens env
2314 , red_doc = sep [ptext (sLit "reduceImplication for")
2316 nest 2 (parens $ ptext (sLit "within")
2318 , red_try_me = try_me }
2320 ; traceTc (text "reduceImplication" <+> vcat
2321 [ ppr (red_givens env), ppr extra_givens,
2323 ; (irreds, binds) <- checkLoop env' wanteds
2325 ; traceTc (text "reduceImplication result" <+> vcat
2326 [ppr irreds, ppr binds])
2328 ; -- extract superclass binds
2329 -- (sc_binds,_) <- extractResults avails []
2330 -- ; traceTc (text "reduceImplication sc_binds" <+> vcat
2331 -- [ppr sc_binds, ppr avails])
2334 -- SLPJ Sept 07: what if improvement happened inside the checkLoop?
2335 -- Then we must iterate the outer loop too!
2337 ; didntSolveWantedEqs <- allM wantedEqInstIsUnsolved wanteds
2338 -- we solve wanted eqs by side effect!
2340 -- Progress is no longer measered by the number of bindings
2341 -- If there are any irreds, but no bindings and no solved
2342 -- equalities, we back off and do nothing
2343 ; let backOff = isEmptyLHsBinds binds && -- no new bindings
2344 (not $ null irreds) && -- but still some irreds
2345 didntSolveWantedEqs -- no instantiated cotv
2347 ; if backOff then -- No progress
2348 return (emptyBag, [orig_implic])
2350 { (simpler_implic_insts, bind)
2351 <- makeImplicationBind inst_loc tvs extra_givens irreds
2352 -- This binding is useless if the recursive simplification
2353 -- made no progress; but currently we don't try to optimise that
2354 -- case. After all, we only try hard to reduce at top level, or
2355 -- when inferring types.
2357 ; let -- extract Id binders for dicts and CoTyVar binders for eqs;
2358 -- see Note [Binders for equalities]
2359 (extra_eq_givens, extra_dict_givens) = partition isEqInst
2361 eq_cotvs = map instToVar extra_eq_givens
2362 dict_ids = map instToId extra_dict_givens
2365 <.> mkWpTyLams eq_cotvs
2366 <.> mkWpLams dict_ids
2367 <.> WpLet (binds `unionBags` bind)
2368 rhs = mkLHsWrap co payload
2369 loc = instLocSpan inst_loc
2370 -- wanted equalities are solved by updating their
2371 -- cotv; we don't generate bindings for them
2372 dict_bndrs = map (L loc . HsVar . instToId)
2373 . filter (not . isEqInst)
2375 payload = mkBigLHsTup dict_bndrs
2377 ; traceTc (vcat [text "reduceImplication" <+> ppr name,
2378 ppr simpler_implic_insts,
2379 text "->" <+> ppr rhs])
2380 ; return (unitBag (L loc (VarBind { var_id= instToId orig_implic
2382 , var_inline = not (null dict_ids) }
2383 -- See Note [Always inline implication constraints]
2385 simpler_implic_insts)
2388 reduceImplication _ i = pprPanic "reduceImplication" (ppr i)
2391 Note [Always inline implication constraints]
2392 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2393 Suppose an implication constraint floats out of an INLINE function.
2394 Then although the implication has a single call site, it won't be
2395 inlined. And that is bad because it means that even if there is really
2396 *no* overloading (type signatures specify the exact types) there will
2397 still be dictionary passing in the resulting code. To avert this,
2398 we mark the implication constraints themselves as INLINE, at least when
2399 there is no loss of sharing as a result.
2401 Note [Freeness and implications]
2402 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2403 It's hard to say when an implication constraint can be floated out. Consider
2404 forall {} Eq a => Foo [a]
2405 The (Foo [a]) doesn't mention any of the quantified variables, but it
2406 still might be partially satisfied by the (Eq a).
2408 There is a useful special case when it *is* easy to partition the
2409 constraints, namely when there are no 'givens'. Consider
2410 forall {a}. () => Bar b
2411 There are no 'givens', and so there is no reason to capture (Bar b).
2412 We can let it float out. But if there is even one constraint we
2413 must be much more careful:
2414 forall {a}. C a b => Bar (m b)
2415 because (C a b) might have a superclass (D b), from which we might
2416 deduce (Bar [b]) when m later gets instantiated to []. Ha!
2418 Here is an even more exotic example
2420 Now consider the constraint
2421 forall b. D Int b => C Int
2422 We can satisfy the (C Int) from the superclass of D, so we don't want
2423 to float the (C Int) out, even though it mentions no type variable in
2426 One more example: the constraint
2428 instance (C a, E c) => E (a,c)
2430 constraint: forall b. D Int b => E (Int,c)
2432 You might think that the (D Int b) can't possibly contribute
2433 to solving (E (Int,c)), since the latter mentions 'c'. But
2434 in fact it can, because solving the (E (Int,c)) constraint needs
2437 and the (C Int) can be satisfied from the superclass of (D Int b).
2438 So we must still not float (E (Int,c)) out.
2440 To think about: special cases for unary type classes?
2442 Note [Pruning the givens in an implication constraint]
2443 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2444 Suppose we are about to form the implication constraint
2445 forall tvs. Eq a => Ord b
2446 The (Eq a) cannot contribute to the (Ord b), because it has no access to
2447 the type variable 'b'. So we could filter out the (Eq a) from the givens.
2448 But BE CAREFUL of the examples above in [Freeness and implications].
2450 Doing so would be a bit tidier, but all the implication constraints get
2451 simplified away by the optimiser, so it's no great win. So I don't take
2452 advantage of that at the moment.
2454 If you do, BE CAREFUL of wobbly type variables.
2457 %************************************************************************
2459 Avails and AvailHow: the pool of evidence
2461 %************************************************************************
2465 data Avails = Avails !ImprovementDone !AvailEnv
2467 type ImprovementDone = Bool -- True <=> some unification has happened
2468 -- so some Irreds might now be reducible
2469 -- keys that are now
2471 type AvailEnv = FiniteMap Inst AvailHow
2473 = IsIrred -- Used for irreducible dictionaries,
2474 -- which are going to be lambda bound
2476 | Given Inst -- Used for dictionaries for which we have a binding
2477 -- e.g. those "given" in a signature
2479 | Rhs -- Used when there is a RHS
2480 (LHsExpr TcId) -- The RHS
2481 [Inst] -- Insts free in the RHS; we need these too
2483 instance Outputable Avails where
2486 pprAvails :: Avails -> SDoc
2487 pprAvails (Avails imp avails)
2488 = vcat [ ptext (sLit "Avails") <> (if imp then ptext (sLit "[improved]") else empty)
2490 vcat [ sep [ppr inst, nest 2 (equals <+> ppr avail)]
2491 | (inst,avail) <- fmToList avails ]]
2493 instance Outputable AvailHow where
2496 -------------------------
2497 pprAvail :: AvailHow -> SDoc
2498 pprAvail IsIrred = text "Irred"
2499 pprAvail (Given x) = text "Given" <+> ppr x
2500 pprAvail (Rhs rhs bs) = sep [text "Rhs" <+> ppr bs,
2503 -------------------------
2504 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2505 extendAvailEnv env inst avail = addToFM env inst avail
2507 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2508 findAvailEnv env wanted = lookupFM env wanted
2509 -- NB 1: the Ord instance of Inst compares by the class/type info
2510 -- *not* by unique. So
2511 -- d1::C Int == d2::C Int
2513 emptyAvails :: Avails
2514 emptyAvails = Avails False emptyFM
2516 findAvail :: Avails -> Inst -> Maybe AvailHow
2517 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2519 elemAvails :: Inst -> Avails -> Bool
2520 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2522 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2524 extendAvails avails@(Avails imp env) inst avail
2525 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2526 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2528 availsInsts :: Avails -> [Inst]
2529 availsInsts (Avails _ avails) = keysFM avails
2531 availsImproved :: Avails -> ImprovementDone
2532 availsImproved (Avails imp _) = imp
2535 Extracting the bindings from a bunch of Avails.
2536 The bindings do *not* come back sorted in dependency order.
2537 We assume that they'll be wrapped in a big Rec, so that the
2538 dependency analyser can sort them out later
2541 type DoneEnv = FiniteMap Inst [Id]
2542 -- Tracks which things we have evidence for
2544 extractResults :: Avails
2546 -> TcM (TcDictBinds, -- Bindings
2547 [Inst], -- The insts bound by the bindings
2548 [Inst]) -- Irreducible ones
2549 -- Note [Reducing implication constraints]
2551 extractResults (Avails _ avails) wanteds
2552 = go emptyBag [] [] emptyFM wanteds
2554 go :: TcDictBinds -- Bindings for dicts
2555 -> [Inst] -- Bound by the bindings
2557 -> DoneEnv -- Has an entry for each inst in the above three sets
2559 -> TcM (TcDictBinds, [Inst], [Inst])
2560 go binds bound_dicts irreds _ []
2561 = return (binds, bound_dicts, irreds)
2563 go binds bound_dicts irreds done (w:ws)
2565 = go binds bound_dicts (w:irreds) done' ws
2567 | Just done_ids@(done_id : rest_done_ids) <- lookupFM done w
2568 = if w_id `elem` done_ids then
2569 go binds bound_dicts irreds done ws
2571 go (add_bind (nlHsVar done_id)) bound_dicts irreds
2572 (addToFM done w (done_id : w_id : rest_done_ids)) ws
2574 | otherwise -- Not yet done
2575 = case findAvailEnv avails w of
2576 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2577 go binds bound_dicts irreds done ws
2579 Just IsIrred -> go binds bound_dicts (w:irreds) done' ws
2581 Just (Rhs rhs ws') -> go (add_bind rhs) (w:bound_dicts) irreds done' (ws' ++ ws)
2583 Just (Given g) -> go binds' bound_dicts irreds (addToFM done w [g_id]) ws
2586 binds' | w_id == g_id = binds
2587 | otherwise = add_bind (nlHsVar g_id)
2590 done' = addToFM done w [w_id]
2591 add_bind rhs = addInstToDictBind binds w rhs
2595 Note [No superclasses for Stop]
2596 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2597 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2598 add it to avails, so that any other equal Insts will be commoned up
2599 right here. However, we do *not* add superclasses. If we have
2602 but a is not bound here, then we *don't* want to derive dn from df
2603 here lest we lose sharing.
2606 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2607 addWanted want_scs avails wanted rhs_expr wanteds
2608 = addAvailAndSCs want_scs avails wanted avail
2610 avail = Rhs rhs_expr wanteds
2612 addGiven :: (Inst -> WantSCs) -> Avails -> Inst -> TcM Avails
2613 addGiven want_scs avails given = addAvailAndSCs (want_scs given) avails given (Given given)
2614 -- Conditionally add superclasses for 'givens'
2615 -- See Note [Recursive instances and superclases]
2617 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2618 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2619 -- so the assert isn't true
2623 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2624 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2625 addAvailAndSCs want_scs avails irred IsIrred
2627 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2628 addAvailAndSCs want_scs avails inst avail
2629 | not (isClassDict inst) = extendAvails avails inst avail
2630 | NoSCs <- want_scs = extendAvails avails inst avail
2631 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2632 ; avails' <- extendAvails avails inst avail
2633 ; addSCs is_loop avails' inst }
2635 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2636 -- Note: this compares by *type*, not by Unique
2637 deps = findAllDeps (unitVarSet (instToVar inst)) avail
2638 dep_tys = map idType (varSetElems deps)
2640 findAllDeps :: IdSet -> AvailHow -> IdSet
2641 -- Find all the Insts that this one depends on
2642 -- See Note [SUPERCLASS-LOOP 2]
2643 -- Watch out, though. Since the avails may contain loops
2644 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2645 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2646 findAllDeps so_far _ = so_far
2648 find_all :: IdSet -> Inst -> IdSet
2650 | isEqInst kid = so_far
2651 | kid_id `elemVarSet` so_far = so_far
2652 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2653 | otherwise = so_far'
2655 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2656 kid_id = instToId kid
2658 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2659 -- Add all the superclasses of the Inst to Avails
2660 -- The first param says "don't do this because the original thing
2661 -- depends on this one, so you'd build a loop"
2662 -- Invariant: the Inst is already in Avails.
2664 addSCs is_loop avails dict
2665 = ASSERT( isDict dict )
2666 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2667 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2669 (clas, tys) = getDictClassTys dict
2670 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2671 sc_theta' = filter (not . isEqPred) $
2672 substTheta (zipTopTvSubst tyvars tys) sc_theta
2674 add_sc avails (sc_dict, sc_sel)
2675 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2676 | is_given sc_dict = return avails
2677 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2678 ; addSCs is_loop avails' sc_dict }
2680 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2681 co_fn = WpApp (instToVar dict) <.> mkWpTyApps tys
2683 is_given :: Inst -> Bool
2684 is_given sc_dict = case findAvail avails sc_dict of
2685 Just (Given _) -> True -- Given is cheaper than superclass selection
2688 -- From the a set of insts obtain all equalities that (transitively) occur in
2689 -- superclass contexts of class constraints (aka the ancestor equalities).
2691 ancestorEqualities :: [Inst] -> TcM [Inst]
2693 = mapM mkWantedEqInst -- turn only equality predicates..
2694 . filter isEqPred -- ..into wanted equality insts
2696 . addAEsToBag emptyBag -- collect the superclass constraints..
2697 . map dictPred -- ..of all predicates in a bag
2698 . filter isClassDict
2700 addAEsToBag :: Bag PredType -> [PredType] -> Bag PredType
2701 addAEsToBag bag [] = bag
2702 addAEsToBag bag (pred:preds)
2703 | pred `elemBag` bag = addAEsToBag bag preds
2704 | isEqPred pred = addAEsToBag bagWithPred preds
2705 | isClassPred pred = addAEsToBag bagWithPred predsWithSCs
2706 | otherwise = addAEsToBag bag preds
2708 bagWithPred = bag `snocBag` pred
2709 predsWithSCs = preds ++ substTheta (zipTopTvSubst tyvars tys) sc_theta
2711 (tyvars, sc_theta, _, _) = classBigSig clas
2712 (clas, tys) = getClassPredTys pred
2716 %************************************************************************
2718 \section{tcSimplifyTop: defaulting}
2720 %************************************************************************
2723 @tcSimplifyTop@ is called once per module to simplify all the constant
2724 and ambiguous Insts.
2726 We need to be careful of one case. Suppose we have
2728 instance Num a => Num (Foo a b) where ...
2730 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2731 to (Num x), and default x to Int. But what about y??
2733 It's OK: the final zonking stage should zap y to (), which is fine.
2737 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2738 tcSimplifyTop wanteds
2739 = tc_simplify_top doc False wanteds
2741 doc = text "tcSimplifyTop"
2743 tcSimplifyInteractive wanteds
2744 = tc_simplify_top doc True wanteds
2746 doc = text "tcSimplifyInteractive"
2748 -- The TcLclEnv should be valid here, solely to improve
2749 -- error message generation for the monomorphism restriction
2750 tc_simplify_top :: SDoc -> Bool -> [Inst] -> TcM (Bag (LHsBind TcId))
2751 tc_simplify_top doc interactive wanteds
2752 = do { dflags <- getDOpts
2753 ; wanteds <- zonkInsts wanteds
2754 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2756 ; traceTc (text "tc_simplify_top 0: " <+> ppr wanteds)
2757 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2758 -- ; (irreds1, binds1) <- gentleInferLoop doc1 wanteds
2759 ; traceTc (text "tc_simplify_top 1: " <+> ppr irreds1)
2760 ; (irreds2, binds2) <- approximateImplications doc2 (\_ -> True) irreds1
2761 ; traceTc (text "tc_simplify_top 2: " <+> ppr irreds2)
2763 -- Use the defaulting rules to do extra unification
2764 -- NB: irreds2 are already zonked
2765 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2767 -- Deal with implicit parameters
2768 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2769 (ambigs, others) = partition isTyVarDict non_ips
2771 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2773 ; addNoInstanceErrs others
2774 ; addTopAmbigErrs ambigs
2776 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2778 doc1 = doc <+> ptext (sLit "(first round)")
2779 doc2 = doc <+> ptext (sLit "(approximate)")
2780 doc3 = doc <+> ptext (sLit "(disambiguate)")
2783 If a dictionary constrains a type variable which is
2784 * not mentioned in the environment
2785 * and not mentioned in the type of the expression
2786 then it is ambiguous. No further information will arise to instantiate
2787 the type variable; nor will it be generalised and turned into an extra
2788 parameter to a function.
2790 It is an error for this to occur, except that Haskell provided for
2791 certain rules to be applied in the special case of numeric types.
2793 * at least one of its classes is a numeric class, and
2794 * all of its classes are numeric or standard
2795 then the type variable can be defaulted to the first type in the
2796 default-type list which is an instance of all the offending classes.
2798 So here is the function which does the work. It takes the ambiguous
2799 dictionaries and either resolves them (producing bindings) or
2800 complains. It works by splitting the dictionary list by type
2801 variable, and using @disambigOne@ to do the real business.
2803 @disambigOne@ assumes that its arguments dictionaries constrain all
2804 the same type variable.
2806 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2807 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2808 the most common use of defaulting is code like:
2810 _ccall_ foo `seqPrimIO` bar
2812 Since we're not using the result of @foo@, the result if (presumably)
2816 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2817 -- Just does unification to fix the default types
2818 -- The Insts are assumed to be pre-zonked
2819 disambiguate doc interactive dflags insts
2821 = return (insts, emptyBag)
2823 | null defaultable_groups
2824 = do { traceTc (text "disambigutate, no defaultable groups" <+> vcat [ppr unaries, ppr insts, ppr bad_tvs, ppr defaultable_groups])
2825 ; return (insts, emptyBag) }
2828 = do { -- Figure out what default types to use
2829 default_tys <- getDefaultTys extended_defaulting ovl_strings
2831 ; traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2832 ; mapM_ (disambigGroup default_tys) defaultable_groups
2834 -- disambigGroup does unification, hence try again
2835 ; tryHardCheckLoop doc insts }
2838 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2839 ovl_strings = dopt Opt_OverloadedStrings dflags
2841 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2842 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2843 (unaries, bad_tvs_s) = partitionWith find_unary insts
2844 bad_tvs = unionVarSets bad_tvs_s
2846 -- Finds unary type-class constraints
2847 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2848 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2849 find_unary inst = Right (tyVarsOfInst inst)
2851 -- Group by type variable
2852 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2853 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2854 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2856 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2857 defaultable_group ds@((_,_,tv):_)
2858 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2859 && not (tv `elemVarSet` bad_tvs)
2860 && defaultable_classes [c | (_,c,_) <- ds]
2861 defaultable_group [] = panic "defaultable_group"
2863 defaultable_classes clss
2864 | extended_defaulting = any isInteractiveClass clss
2865 | otherwise = all is_std_class clss && (any is_num_class clss)
2867 -- In interactive mode, or with -XExtendedDefaultRules,
2868 -- we default Show a to Show () to avoid graututious errors on "show []"
2869 isInteractiveClass cls
2870 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2872 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2873 -- is_num_class adds IsString to the standard numeric classes,
2874 -- when -foverloaded-strings is enabled
2876 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2877 -- Similarly is_std_class
2879 -----------------------
2880 disambigGroup :: [Type] -- The default types
2881 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2882 -> TcM () -- Just does unification, to fix the default types
2884 disambigGroup default_tys dicts
2885 = try_default default_tys
2887 (_,_,tyvar) = ASSERT(not (null dicts)) head dicts -- Should be non-empty
2888 classes = [c | (_,c,_) <- dicts]
2890 try_default [] = return ()
2891 try_default (default_ty : default_tys)
2892 = tryTcLIE_ (try_default default_tys) $
2893 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2894 -- This may fail; then the tryTcLIE_ kicks in
2895 -- Failure here is caused by there being no type in the
2896 -- default list which can satisfy all the ambiguous classes.
2897 -- For example, if Real a is reqd, but the only type in the
2898 -- default list is Int.
2900 -- After this we can't fail
2901 ; warnDefault dicts default_ty
2902 ; unifyType default_ty (mkTyVarTy tyvar)
2903 ; return () -- TOMDO: do something with the coercion
2907 -----------------------
2908 getDefaultTys :: Bool -> Bool -> TcM [Type]
2909 getDefaultTys extended_deflts ovl_strings
2910 = do { mb_defaults <- getDeclaredDefaultTys
2911 ; case mb_defaults of {
2912 Just tys -> return tys ; -- User-supplied defaults
2915 -- No use-supplied default
2916 -- Use [Integer, Double], plus modifications
2917 { integer_ty <- tcMetaTy integerTyConName
2918 ; checkWiredInTyCon doubleTyCon
2919 ; string_ty <- tcMetaTy stringTyConName
2920 ; return (opt_deflt extended_deflts unitTy
2921 -- Note [Default unitTy]
2923 [integer_ty,doubleTy]
2925 opt_deflt ovl_strings string_ty) } } }
2927 opt_deflt True ty = [ty]
2928 opt_deflt False _ = []
2931 Note [Default unitTy]
2932 ~~~~~~~~~~~~~~~~~~~~~
2933 In interative mode (or with -XExtendedDefaultRules) we add () as the first type we
2934 try when defaulting. This has very little real impact, except in the following case.
2936 Text.Printf.printf "hello"
2937 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2938 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2939 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2940 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2941 () to the list of defaulting types. See Trac #1200.
2943 Note [Avoiding spurious errors]
2944 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2945 When doing the unification for defaulting, we check for skolem
2946 type variables, and simply don't default them. For example:
2947 f = (*) -- Monomorphic
2948 g :: Num a => a -> a
2950 Here, we get a complaint when checking the type signature for g,
2951 that g isn't polymorphic enough; but then we get another one when
2952 dealing with the (Num a) context arising from f's definition;
2953 we try to unify a with Int (to default it), but find that it's
2954 already been unified with the rigid variable from g's type sig
2957 %************************************************************************
2959 \subsection[simple]{@Simple@ versions}
2961 %************************************************************************
2963 Much simpler versions when there are no bindings to make!
2965 @tcSimplifyThetas@ simplifies class-type constraints formed by
2966 @deriving@ declarations and when specialising instances. We are
2967 only interested in the simplified bunch of class/type constraints.
2969 It simplifies to constraints of the form (C a b c) where
2970 a,b,c are type variables. This is required for the context of
2971 instance declarations.
2974 tcSimplifyDeriv :: InstOrigin
2976 -> ThetaType -- Wanted
2977 -> TcM ThetaType -- Needed
2978 -- Given instance (wanted) => C inst_ty
2979 -- Simplify 'wanted' as much as possible
2981 tcSimplifyDeriv orig tyvars theta
2982 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2983 -- The main loop may do unification, and that may crash if
2984 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2985 -- ToDo: what if two of them do get unified?
2986 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2987 ; (irreds, _) <- tryHardCheckLoop doc wanteds
2989 ; let (tv_dicts, others) = partition ok irreds
2990 ; addNoInstanceErrs others
2991 -- See Note [Exotic derived instance contexts] in TcMType
2993 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2994 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
2995 -- This reverse-mapping is a pain, but the result
2996 -- should mention the original TyVars not TcTyVars
2998 ; return simpl_theta }
3000 doc = ptext (sLit "deriving classes for a data type")
3002 ok dict | isDict dict = validDerivPred (dictPred dict)
3007 @tcSimplifyDefault@ just checks class-type constraints, essentially;
3008 used with \tr{default} declarations. We are only interested in
3009 whether it worked or not.
3012 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
3015 tcSimplifyDefault theta = do
3016 wanteds <- newDictBndrsO DefaultOrigin theta
3017 (irreds, _) <- tryHardCheckLoop doc wanteds
3018 addNoInstanceErrs irreds
3022 traceTc (ptext (sLit "tcSimplifyDefault failing")) >> failM
3024 doc = ptext (sLit "default declaration")
3027 @tcSimplifyStagedExpr@ performs a simplification but does so at a new
3028 stage. This is used when typechecking annotations and splices.
3032 tcSimplifyStagedExpr :: ThStage -> TcM a -> TcM (a, TcDictBinds)
3033 -- Type check an expression that runs at a top level stage as if
3034 -- it were going to be spliced and then simplify it
3035 tcSimplifyStagedExpr stage tc_action
3036 = setStage stage $ do {
3037 -- Typecheck the expression
3038 (thing', lie) <- getLIE tc_action
3040 -- Solve the constraints
3041 ; const_binds <- tcSimplifyTop lie
3043 ; return (thing', const_binds) }
3048 %************************************************************************
3050 \section{Errors and contexts}
3052 %************************************************************************
3054 ToDo: for these error messages, should we note the location as coming
3055 from the insts, or just whatever seems to be around in the monad just
3059 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
3060 -> [Inst] -- The offending Insts
3062 -- Group together insts with the same origin
3063 -- We want to report them together in error messages
3067 groupErrs report_err (inst:insts)
3068 = do { do_one (inst:friends)
3069 ; groupErrs report_err others }
3071 -- (It may seem a bit crude to compare the error messages,
3072 -- but it makes sure that we combine just what the user sees,
3073 -- and it avoids need equality on InstLocs.)
3074 (friends, others) = partition is_friend insts
3075 loc_msg = showSDoc (pprInstLoc (instLoc inst))
3076 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
3077 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
3078 -- Add location and context information derived from the Insts
3080 -- Add the "arising from..." part to a message about bunch of dicts
3081 addInstLoc :: [Inst] -> Message -> Message
3082 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
3084 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
3087 addTopIPErrs bndrs ips
3088 = do { dflags <- getDOpts
3089 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
3091 (tidy_env, tidy_ips) = tidyInsts ips
3093 = vcat [sep [ptext (sLit "Implicit parameters escape from"),
3094 nest 2 (ptext (sLit "the monomorphic top-level binding")
3095 <> plural bndrs <+> ptext (sLit "of")
3096 <+> pprBinders bndrs <> colon)],
3097 nest 2 (vcat (map ppr_ip ips)),
3098 monomorphism_fix dflags]
3099 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
3101 topIPErrs :: [Inst] -> TcM ()
3103 = groupErrs report tidy_dicts
3105 (tidy_env, tidy_dicts) = tidyInsts dicts
3106 report dicts = addErrTcM (tidy_env, mk_msg dicts)
3107 mk_msg dicts = addInstLoc dicts (ptext (sLit "Unbound implicit parameter") <>
3108 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
3110 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
3112 addNoInstanceErrs insts
3113 = do { let (tidy_env, tidy_insts) = tidyInsts insts
3114 ; reportNoInstances tidy_env Nothing tidy_insts }
3118 -> Maybe (InstLoc, [Inst]) -- Context
3119 -- Nothing => top level
3120 -- Just (d,g) => d describes the construct
3122 -> [Inst] -- What is wanted (can include implications)
3125 reportNoInstances tidy_env mb_what insts
3126 = groupErrs (report_no_instances tidy_env mb_what) insts
3128 report_no_instances :: TidyEnv -> Maybe (InstLoc, [Inst]) -> [Inst] -> TcM ()
3129 report_no_instances tidy_env mb_what insts
3130 = do { inst_envs <- tcGetInstEnvs
3131 ; let (implics, insts1) = partition isImplicInst insts
3132 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
3133 (eqInsts, insts3) = partition isEqInst insts2
3134 ; traceTc (text "reportNoInstances" <+> vcat
3135 [ppr insts, ppr implics, ppr insts1, ppr insts2])
3136 ; mapM_ complain_implic implics
3137 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
3138 ; groupErrs complain_no_inst insts3
3139 ; mapM_ (addErrTcM . mk_eq_err) eqInsts
3142 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
3144 complain_implic inst -- Recurse!
3145 = reportNoInstances tidy_env
3146 (Just (tci_loc inst, tci_given inst))
3149 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
3150 -- Right msg => overlap message
3151 -- Left inst => no instance
3152 check_overlap inst_envs wanted
3153 | not (isClassDict wanted) = Left wanted
3155 = case lookupInstEnv inst_envs clas tys of
3156 ([], _) -> Left wanted -- No match
3157 -- The case of exactly one match and no unifiers means a
3158 -- successful lookup. That can't happen here, because dicts
3159 -- only end up here if they didn't match in Inst.lookupInst
3161 | debugIsOn -> pprPanic "reportNoInstance" (ppr wanted)
3162 res -> Right (mk_overlap_msg wanted res)
3164 (clas,tys) = getDictClassTys wanted
3166 mk_overlap_msg dict (matches, unifiers)
3167 = ASSERT( not (null matches) )
3168 vcat [ addInstLoc [dict] ((ptext (sLit "Overlapping instances for")
3169 <+> pprPred (dictPred dict))),
3170 sep [ptext (sLit "Matching instances") <> colon,
3171 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
3172 if not (isSingleton matches)
3173 then -- Two or more matches
3175 else -- One match, plus some unifiers
3176 ASSERT( not (null unifiers) )
3177 parens (vcat [ptext (sLit "The choice depends on the instantiation of") <+>
3178 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
3179 ptext (sLit "To pick the first instance above, use -XIncoherentInstances"),
3180 ptext (sLit "when compiling the other instance declarations")])]
3182 ispecs = [ispec | (ispec, _) <- matches]
3184 mk_eq_err :: Inst -> (TidyEnv, SDoc)
3185 mk_eq_err inst = misMatchMsg tidy_env (eqInstTys inst)
3187 mk_no_inst_err insts
3188 | null insts = empty
3190 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
3191 not (isEmptyVarSet (tyVarsOfInsts insts))
3192 = vcat [ addInstLoc insts $
3193 sep [ ptext (sLit "Could not deduce") <+> pprDictsTheta insts
3194 , nest 2 $ ptext (sLit "from the context") <+> pprDictsTheta givens]
3195 , show_fixes (fix1 loc : fixes2) ]
3197 | otherwise -- Top level
3198 = vcat [ addInstLoc insts $
3199 ptext (sLit "No instance") <> plural insts
3200 <+> ptext (sLit "for") <+> pprDictsTheta insts
3201 , show_fixes fixes2 ]
3204 fix1 loc = sep [ ptext (sLit "add") <+> pprDictsTheta insts
3205 <+> ptext (sLit "to the context of"),
3206 nest 2 (ppr (instLocOrigin loc)) ]
3207 -- I'm not sure it helps to add the location
3208 -- nest 2 (ptext (sLit "at") <+> ppr (instLocSpan loc)) ]
3210 fixes2 | null instance_dicts = []
3211 | otherwise = [sep [ptext (sLit "add an instance declaration for"),
3212 pprDictsTheta instance_dicts]]
3213 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
3214 -- Insts for which it is worth suggesting an adding an instance declaration
3215 -- Exclude implicit parameters, and tyvar dicts
3217 show_fixes :: [SDoc] -> SDoc
3218 show_fixes [] = empty
3219 show_fixes (f:fs) = sep [ptext (sLit "Possible fix:"),
3220 nest 2 (vcat (f : map (ptext (sLit "or") <+>) fs))]
3222 addTopAmbigErrs :: [Inst] -> TcRn ()
3223 addTopAmbigErrs dicts
3224 -- Divide into groups that share a common set of ambiguous tyvars
3225 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
3226 -- See Note [Avoiding spurious errors]
3227 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
3229 (tidy_env, tidy_dicts) = tidyInsts dicts
3231 tvs_of :: Inst -> [TcTyVar]
3232 tvs_of d = varSetElems (tyVarsOfInst d)
3233 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
3235 report :: [(Inst,[TcTyVar])] -> TcM ()
3236 report pairs@((inst,tvs) : _) = do -- The pairs share a common set of ambiguous tyvars
3237 (tidy_env, mono_msg) <- mkMonomorphismMsg tidy_env tvs
3238 setSrcSpan (instSpan inst) $
3239 -- the location of the first one will do for the err message
3240 addErrTcM (tidy_env, msg $$ mono_msg)
3242 dicts = map fst pairs
3243 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
3244 pprQuotedList tvs <+> in_msg,
3245 nest 2 (pprDictsInFull dicts)]
3246 in_msg = text "in the constraint" <> plural dicts <> colon
3247 report [] = panic "addTopAmbigErrs"
3250 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
3251 -- There's an error with these Insts; if they have free type variables
3252 -- it's probably caused by the monomorphism restriction.
3253 -- Try to identify the offending variable
3254 -- ASSUMPTION: the Insts are fully zonked
3255 mkMonomorphismMsg tidy_env inst_tvs
3256 = do { dflags <- getDOpts
3257 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
3258 ; return (tidy_env, mk_msg dflags docs) }
3260 mk_msg _ _ | any isRuntimeUnk inst_tvs
3261 = vcat [ptext (sLit "Cannot resolve unknown runtime types:") <+>
3262 (pprWithCommas ppr inst_tvs),
3263 ptext (sLit "Use :print or :force to determine these types")]
3264 mk_msg _ [] = ptext (sLit "Probable fix: add a type signature that fixes these type variable(s)")
3265 -- This happens in things like
3266 -- f x = show (read "foo")
3267 -- where monomorphism doesn't play any role
3269 = vcat [ptext (sLit "Possible cause: the monomorphism restriction applied to the following:"),
3271 monomorphism_fix dflags]
3273 monomorphism_fix :: DynFlags -> SDoc
3274 monomorphism_fix dflags
3275 = ptext (sLit "Probable fix:") <+> vcat
3276 [ptext (sLit "give these definition(s) an explicit type signature"),
3277 if dopt Opt_MonomorphismRestriction dflags
3278 then ptext (sLit "or use -XNoMonomorphismRestriction")
3279 else empty] -- Only suggest adding "-XNoMonomorphismRestriction"
3280 -- if it is not already set!
3282 warnDefault :: [(Inst, Class, Var)] -> Type -> TcM ()
3283 warnDefault ups default_ty = do
3284 warn_flag <- doptM Opt_WarnTypeDefaults
3285 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
3287 dicts = [d | (d,_,_) <- ups]
3290 (_, tidy_dicts) = tidyInsts dicts
3291 warn_msg = vcat [ptext (sLit "Defaulting the following constraint(s) to type") <+>
3292 quotes (ppr default_ty),
3293 pprDictsInFull tidy_dicts]
3295 reduceDepthErr :: Int -> [Inst] -> SDoc
3296 reduceDepthErr n stack
3297 = vcat [ptext (sLit "Context reduction stack overflow; size =") <+> int n,
3298 ptext (sLit "Use -fcontext-stack=N to increase stack size to N"),
3299 nest 4 (pprStack stack)]
3301 pprStack :: [Inst] -> SDoc
3302 pprStack stack = vcat (map pprInstInFull stack)