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 = growInstsTyVars wanted' 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 $$ ppr (growInstsTyVars wanted' tau_tvs1) $$ 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 = growInstsTyVars irreds2 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 ---------------------------------------------------
719 ; let (free, irreds3) = partition (isFreeWhenInferring qtvs) irreds2
722 -- Turn the quantified meta-type variables into real type variables
723 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems qtvs)
725 -- We can't abstract over any remaining unsolved
726 -- implications so instead just float them outwards. Ugh.
727 ; let (q_dicts0, implics) = partition isAbstractableInst irreds3
728 ; loc <- getInstLoc (ImplicOrigin doc)
729 ; implic_bind <- bindIrreds loc qtvs2 q_dicts0 implics
731 -- Prepare equality instances for quantification
732 ; let (q_eqs0,q_dicts) = partition isEqInst q_dicts0
733 ; q_eqs <- mapM finalizeEqInst q_eqs0
735 ; return (qtvs2, q_eqs ++ q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
736 -- NB: when we are done, we might have some bindings, but
737 -- the final qtvs might be empty. See Note [NO TYVARS] below.
739 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
740 -- Note [Inference and implication constraints]
741 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
742 -- - fetching any dicts inside them that are free
743 -- - using those dicts as cruder constraints, to solve the implications
744 -- - returning the extra ones too
746 approximateImplications doc want_dict irreds
748 = return (irreds, emptyBag)
750 = do { extra_dicts' <- mapM cloneDict extra_dicts
751 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
752 -- By adding extra_dicts', we make them
753 -- available to solve the implication constraints
755 extra_dicts = get_dicts (filter isImplicInst irreds)
757 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
758 -- Find the wanted constraints in implication constraints that satisfy
759 -- want_dict, and are not bound by forall's in the constraint itself
760 get_dicts ds = concatMap get_dict ds
762 get_dict d@(Dict {}) | want_dict d = [d]
764 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
765 = [ d | let tv_set = mkVarSet tvs
766 , d <- get_dicts wanteds
767 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
768 get_dict i@(EqInst {}) | want_dict i = [i]
770 get_dict other = pprPanic "approximateImplications" (ppr other)
773 Note [Inference and implication constraints]
774 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
775 Suppose we have a wanted implication constraint (perhaps arising from
776 a nested pattern match) like
778 and we are now trying to quantify over 'a' when inferring the type for
779 a function. In principle it's possible that there might be an instance
780 instance (C a, E a) => D [a]
781 so the context (E a) would suffice. The Right Thing is to abstract over
782 the implication constraint, but we don't do that (a) because it'll be
783 surprising to programmers and (b) because we don't have the machinery to deal
784 with 'given' implications.
786 So our best approximation is to make (D [a]) part of the inferred
787 context, so we can use that to discharge the implication. Hence
788 the strange function get_dicts in approximateImplications.
790 The common cases are more clear-cut, when we have things like
792 Here, abstracting over (C b) is not an approximation at all -- but see
793 Note [Freeness and implications].
795 See Trac #1430 and test tc228.
799 -----------------------------------------------------------
800 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
801 -- against, but we don't know the type variables over which we are going to quantify.
802 -- This happens when we have a type signature for a mutually recursive group
805 -> TcTyVarSet -- fv(T)
808 -> TcM ([TyVar], -- Fully zonked, and quantified
809 TcDictBinds) -- Bindings
811 tcSimplifyInferCheck loc tau_tvs givens wanteds
812 = do { traceTc (text "tcSimplifyInferCheck <-" <+> ppr wanteds)
813 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
815 -- Figure out which type variables to quantify over
816 -- You might think it should just be the signature tyvars,
817 -- but in bizarre cases you can get extra ones
818 -- f :: forall a. Num a => a -> a
819 -- f x = fst (g (x, head [])) + 1
821 -- Here we infer g :: forall a b. a -> b -> (b,a)
822 -- We don't want g to be monomorphic in b just because
823 -- f isn't quantified over b.
824 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
825 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
826 ; gbl_tvs <- tcGetGlobalTyVars
827 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
828 -- We could close gbl_tvs, but its not necessary for
829 -- soundness, and it'll only affect which tyvars, not which
830 -- dictionaries, we quantify over
832 ; qtvs' <- zonkQuantifiedTyVars qtvs
834 -- Now we are back to normal (c.f. tcSimplCheck)
835 ; implic_bind <- bindIrreds loc qtvs' givens irreds
837 ; traceTc (text "tcSimplifyInferCheck ->" <+> ppr (implic_bind))
838 ; return (qtvs', binds `unionBags` implic_bind) }
841 Note [Squashing methods]
842 ~~~~~~~~~~~~~~~~~~~~~~~~~
843 Be careful if you want to float methods more:
844 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
845 From an application (truncate f i) we get
848 If we have also have a second occurrence of truncate, we get
851 When simplifying with i,f free, we might still notice that
852 t1=t3; but alas, the binding for t2 (which mentions t1)
853 may continue to float out!
858 class Y a b | a -> b where
861 instance Y [[a]] a where
864 k :: X a -> X a -> X a
866 g :: Num a => [X a] -> [X a]
869 h ys = ys ++ map (k (y [[0]])) xs
871 The excitement comes when simplifying the bindings for h. Initially
872 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
873 From this we get t1~t2, but also various bindings. We can't forget
874 the bindings (because of [LOOP]), but in fact t1 is what g is
877 The net effect of [NO TYVARS]
880 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
881 isFreeWhenInferring qtvs inst
882 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
883 && isInheritableInst inst -- and no implicit parameter involved
884 -- see Note [Inheriting implicit parameters]
886 {- No longer used (with implication constraints)
887 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
888 -> NameSet -- Quantified implicit parameters
890 isFreeWhenChecking qtvs ips inst
891 = isFreeWrtTyVars qtvs inst
892 && isFreeWrtIPs ips inst
895 isFreeWrtTyVars :: VarSet -> Inst -> Bool
896 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
897 isFreeWrtIPs :: NameSet -> Inst -> Bool
898 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
902 %************************************************************************
904 \subsection{tcSimplifyCheck}
906 %************************************************************************
908 @tcSimplifyCheck@ is used when we know exactly the set of variables
909 we are going to quantify over. For example, a class or instance declaration.
912 -----------------------------------------------------------
913 -- tcSimplifyCheck is used when checking expression type signatures,
914 -- class decls, instance decls etc.
915 tcSimplifyCheck :: InstLoc
916 -> [TcTyVar] -- Quantify over these
919 -> TcM TcDictBinds -- Bindings
920 tcSimplifyCheck loc qtvs givens wanteds
921 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
922 do { traceTc (text "tcSimplifyCheck")
923 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
924 ; implic_bind <- bindIrreds loc qtvs givens irreds
925 ; return (binds `unionBags` implic_bind) }
927 -----------------------------------------------------------
928 -- tcSimplifyCheckPat is used for existential pattern match
929 tcSimplifyCheckPat :: InstLoc
930 -> [TcTyVar] -- Quantify over these
933 -> TcM TcDictBinds -- Bindings
934 tcSimplifyCheckPat loc qtvs givens wanteds
935 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
936 do { traceTc (text "tcSimplifyCheckPat")
937 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
938 ; implic_bind <- bindIrredsR loc qtvs givens irreds
939 ; return (binds `unionBags` implic_bind) }
941 -----------------------------------------------------------
942 bindIrreds :: InstLoc -> [TcTyVar]
945 bindIrreds loc qtvs givens irreds
946 = bindIrredsR loc qtvs givens irreds
948 bindIrredsR :: InstLoc -> [TcTyVar] -> [Inst] -> [Inst] -> TcM TcDictBinds
949 -- Make a binding that binds 'irreds', by generating an implication
950 -- constraint for them, *and* throwing the constraint into the LIE
951 bindIrredsR loc qtvs givens irreds
955 = do { let givens' = filter isAbstractableInst givens
956 -- The givens can (redundantly) include methods
957 -- We want to retain both EqInsts and Dicts
958 -- There should be no implicadtion constraints
959 -- See Note [Pruning the givens in an implication constraint]
961 -- If there are no 'givens', then it's safe to
962 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
963 -- See Note [Freeness and implications]
964 ; irreds' <- if null givens'
966 { let qtv_set = mkVarSet qtvs
967 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
969 ; return real_irreds }
972 ; (implics, bind) <- makeImplicationBind loc qtvs givens' irreds'
973 -- This call does the real work
974 -- If irreds' is empty, it does something sensible
979 makeImplicationBind :: InstLoc -> [TcTyVar]
981 -> TcM ([Inst], TcDictBinds)
982 -- Make a binding that binds 'irreds', by generating an implication
983 -- constraint for them.
985 -- The binding looks like
986 -- (ir1, .., irn) = f qtvs givens
987 -- where f is (evidence for) the new implication constraint
988 -- f :: forall qtvs. givens => (ir1, .., irn)
989 -- qtvs includes coercion variables
991 -- This binding must line up the 'rhs' in reduceImplication
992 makeImplicationBind loc all_tvs
993 givens -- Guaranteed all Dicts or EqInsts
995 | null irreds -- If there are no irreds, we are done
996 = return ([], emptyBag)
997 | otherwise -- Otherwise we must generate a binding
998 = do { uniq <- newUnique
999 ; span <- getSrcSpanM
1000 ; let (eq_givens, dict_givens) = partition isEqInst givens
1002 -- extract equality binders
1003 eq_cotvs = map eqInstType eq_givens
1005 -- make the implication constraint instance
1006 name = mkInternalName uniq (mkVarOcc "ic") span
1007 implic_inst = ImplicInst { tci_name = name,
1008 tci_tyvars = all_tvs,
1009 tci_given = eq_givens ++ dict_givens,
1010 -- same order as binders
1011 tci_wanted = irreds,
1014 -- create binders for the irreducible dictionaries
1015 dict_irreds = filter (not . isEqInst) irreds
1016 dict_irred_ids = map instToId dict_irreds
1017 lpat = mkBigLHsPatTup (map (L span . VarPat) dict_irred_ids)
1019 -- create the binding
1020 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
1021 co = mkWpApps (map instToId dict_givens)
1022 <.> mkWpTyApps eq_cotvs
1023 <.> mkWpTyApps (mkTyVarTys all_tvs)
1024 bind | [dict_irred_id] <- dict_irred_ids
1025 = VarBind dict_irred_id rhs
1027 = PatBind { pat_lhs = lpat
1028 , pat_rhs = unguardedGRHSs rhs
1029 , pat_rhs_ty = hsLPatType lpat
1030 , bind_fvs = placeHolderNames
1033 ; traceTc $ text "makeImplicationBind" <+> ppr implic_inst
1034 ; return ([implic_inst], unitBag (L span bind))
1037 -----------------------------------------------------------
1038 tryHardCheckLoop :: SDoc
1040 -> TcM ([Inst], TcDictBinds)
1042 tryHardCheckLoop doc wanteds
1043 = do { (irreds,binds) <- checkLoop (mkInferRedEnv doc try_me) wanteds
1044 ; return (irreds,binds)
1048 -- Here's the try-hard bit
1050 -----------------------------------------------------------
1051 gentleCheckLoop :: InstLoc
1054 -> TcM ([Inst], TcDictBinds)
1056 gentleCheckLoop inst_loc givens wanteds
1057 = do { (irreds,binds) <- checkLoop env wanteds
1058 ; return (irreds,binds)
1061 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1063 try_me inst | isMethodOrLit inst = ReduceMe
1065 -- When checking against a given signature
1066 -- we MUST be very gentle: Note [Check gently]
1068 gentleInferLoop :: SDoc -> [Inst]
1069 -> TcM ([Inst], TcDictBinds)
1070 gentleInferLoop doc wanteds
1071 = do { (irreds, binds) <- checkLoop env wanteds
1072 ; return (irreds, binds) }
1074 env = mkInferRedEnv doc try_me
1075 try_me inst | isMethodOrLit inst = ReduceMe
1080 ~~~~~~~~~~~~~~~~~~~~
1081 We have to very careful about not simplifying too vigorously
1086 f :: Show b => T b -> b
1087 f (MkT x) = show [x]
1089 Inside the pattern match, which binds (a:*, x:a), we know that
1091 Hence we have a dictionary for Show [a] available; and indeed we
1092 need it. We are going to build an implication contraint
1093 forall a. (b~[a]) => Show [a]
1094 Later, we will solve this constraint using the knowledge (Show b)
1096 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1097 thing becomes insoluble. So we simplify gently (get rid of literals
1098 and methods only, plus common up equal things), deferring the real
1099 work until top level, when we solve the implication constraint
1100 with tryHardCheckLooop.
1104 -----------------------------------------------------------
1107 -> TcM ([Inst], TcDictBinds)
1108 -- Precondition: givens are completely rigid
1109 -- Postcondition: returned Insts are zonked
1111 checkLoop env wanteds
1113 where go env wanteds
1114 = do { -- We do need to zonk the givens; cf Note [Zonking RedEnv]
1115 ; env' <- zonkRedEnv env
1116 ; wanteds' <- zonkInsts wanteds
1118 ; (improved, tybinds, binds, irreds)
1119 <- reduceContext env' wanteds'
1120 ; execTcTyVarBinds tybinds
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.
1278 There is a complication though. Suppose there are equalities
1279 instance (Eq a, a~b) => Num (a,b)
1280 Then we normalise the 'givens' wrt the equalities, so the original
1281 given "this" dictionary is cast to one of a different type. So it's a
1282 bit trickier than before to identify the "special" dictionary whose
1283 superclasses must not be added. See test
1284 indexed-types/should_run/EqInInstance
1286 We need a persistent property of the dictionary to record this
1287 special-ness. Current I'm using the InstLocOrigin (a bit of a hack,
1288 but cool), which is maintained by dictionary normalisation.
1289 Specifically, the InstLocOrigin is
1291 then the no-superclass thing kicks in. WATCH OUT if you fiddle
1295 tcSimplifySuperClasses
1297 -> Inst -- The dict whose superclasses
1298 -- are being figured out
1302 tcSimplifySuperClasses loc this givens sc_wanteds
1303 = do { traceTc (text "tcSimplifySuperClasses")
1305 -- Note [Recursive instances and superclases]
1306 ; no_sc_loc <- getInstLoc NoScOrigin
1307 ; let no_sc_this = setInstLoc this no_sc_loc
1309 ; let env = RedEnv { red_doc = pprInstLoc loc,
1310 red_try_me = try_me,
1311 red_givens = no_sc_this : givens,
1313 red_improve = False } -- No unification vars
1316 ; (irreds,binds1) <- checkLoop env sc_wanteds
1317 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1318 ; reportNoInstances tidy_env (Just (loc, givens)) [] tidy_irreds
1321 try_me _ = ReduceMe -- Try hard, so we completely solve the superclass
1322 -- constraints right here. See Note [SUPERCLASS-LOOP 1]
1326 %************************************************************************
1328 \subsection{tcSimplifyRestricted}
1330 %************************************************************************
1332 tcSimplifyRestricted infers which type variables to quantify for a
1333 group of restricted bindings. This isn't trivial.
1336 We want to quantify over a to get id :: forall a. a->a
1339 We do not want to quantify over a, because there's an Eq a
1340 constraint, so we get eq :: a->a->Bool (notice no forall)
1343 RHS has type 'tau', whose free tyvars are tau_tvs
1344 RHS has constraints 'wanteds'
1347 Quantify over (tau_tvs \ ftvs(wanteds))
1348 This is bad. The constraints may contain (Monad (ST s))
1349 where we have instance Monad (ST s) where...
1350 so there's no need to be monomorphic in s!
1352 Also the constraint might be a method constraint,
1353 whose type mentions a perfectly innocent tyvar:
1354 op :: Num a => a -> b -> a
1355 Here, b is unconstrained. A good example would be
1357 We want to infer the polymorphic type
1358 foo :: forall b. b -> b
1361 Plan B (cunning, used for a long time up to and including GHC 6.2)
1362 Step 1: Simplify the constraints as much as possible (to deal
1363 with Plan A's problem). Then set
1364 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1366 Step 2: Now simplify again, treating the constraint as 'free' if
1367 it does not mention qtvs, and trying to reduce it otherwise.
1368 The reasons for this is to maximise sharing.
1370 This fails for a very subtle reason. Suppose that in the Step 2
1371 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1372 In the Step 1 this constraint might have been simplified, perhaps to
1373 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1374 This won't happen in Step 2... but that in turn might prevent some other
1375 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1376 and that in turn breaks the invariant that no constraints are quantified over.
1378 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1383 Step 1: Simplify the constraints as much as possible (to deal
1384 with Plan A's problem). Then set
1385 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1386 Return the bindings from Step 1.
1389 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1392 instance (HasBinary ty IO) => HasCodedValue ty
1394 foo :: HasCodedValue a => String -> IO a
1396 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1397 doDecodeIO codedValue view
1398 = let { act = foo "foo" } in act
1400 You might think this should work becuase the call to foo gives rise to a constraint
1401 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1402 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1403 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1405 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1409 Plan D (a variant of plan B)
1410 Step 1: Simplify the constraints as much as possible (to deal
1411 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1412 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1414 Step 2: Now simplify again, treating the constraint as 'free' if
1415 it does not mention qtvs, and trying to reduce it otherwise.
1417 The point here is that it's generally OK to have too few qtvs; that is,
1418 to make the thing more monomorphic than it could be. We don't want to
1419 do that in the common cases, but in wierd cases it's ok: the programmer
1420 can always add a signature.
1422 Too few qtvs => too many wanteds, which is what happens if you do less
1427 tcSimplifyRestricted -- Used for restricted binding groups
1428 -- i.e. ones subject to the monomorphism restriction
1431 -> [Name] -- Things bound in this group
1432 -> TcTyVarSet -- Free in the type of the RHSs
1433 -> [Inst] -- Free in the RHSs
1434 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1435 TcDictBinds) -- Bindings
1436 -- tcSimpifyRestricted returns no constraints to
1437 -- quantify over; by definition there are none.
1438 -- They are all thrown back in the LIE
1440 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1441 -- Zonk everything in sight
1442 = do { traceTc (text "tcSimplifyRestricted")
1443 ; wanteds_z <- zonkInsts wanteds
1445 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1446 -- dicts; the idea is to get rid of as many type
1447 -- variables as possible, and we don't want to stop
1448 -- at (say) Monad (ST s), because that reduces
1449 -- immediately, with no constraint on s.
1451 -- BUT do no improvement! See Plan D above
1452 -- HOWEVER, some unification may take place, if we instantiate
1453 -- a method Inst with an equality constraint
1454 ; let env = mkNoImproveRedEnv doc (\_ -> ReduceMe)
1455 ; (_imp, _tybinds, _binds, constrained_dicts)
1456 <- reduceContext env wanteds_z
1458 -- Next, figure out the tyvars we will quantify over
1459 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1460 ; gbl_tvs' <- tcGetGlobalTyVars
1461 ; constrained_dicts' <- zonkInsts constrained_dicts
1463 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1464 -- As in tcSimplifyInfer
1466 -- Do not quantify over constrained type variables:
1467 -- this is the monomorphism restriction
1468 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1469 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1470 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1473 ; warn_mono <- doptM Opt_WarnMonomorphism
1474 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1475 (vcat[ ptext (sLit "the Monomorphism Restriction applies to the binding")
1476 <> plural bndrs <+> ptext (sLit "for") <+> pp_bndrs,
1477 ptext (sLit "Consider giving a type signature for") <+> pp_bndrs])
1479 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1480 pprInsts wanteds, pprInsts constrained_dicts',
1482 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1484 -- The first step may have squashed more methods than
1485 -- necessary, so try again, this time more gently, knowing the exact
1486 -- set of type variables to quantify over.
1488 -- We quantify only over constraints that are captured by qtvs;
1489 -- these will just be a subset of non-dicts. This in contrast
1490 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1491 -- all *non-inheritable* constraints too. This implements choice
1492 -- (B) under "implicit parameter and monomorphism" above.
1494 -- Remember that we may need to do *some* simplification, to
1495 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1496 -- just to float all constraints
1498 -- At top level, we *do* squash methods becuase we want to
1499 -- expose implicit parameters to the test that follows
1500 ; let is_nested_group = isNotTopLevel top_lvl
1501 try_me inst | isFreeWrtTyVars qtvs inst,
1502 (is_nested_group || isDict inst) = Stop
1503 | otherwise = ReduceMe
1504 env = mkNoImproveRedEnv doc try_me
1505 ; (_imp, tybinds, binds, irreds) <- reduceContext env wanteds_z
1506 ; execTcTyVarBinds tybinds
1508 -- See "Notes on implicit parameters, Question 4: top level"
1509 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1510 if is_nested_group then
1512 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1513 ; addTopIPErrs bndrs bad_ips
1514 ; extendLIEs non_ips }
1516 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1517 ; return (qtvs', binds) }
1521 %************************************************************************
1525 %************************************************************************
1527 On the LHS of transformation rules we only simplify methods and constants,
1528 getting dictionaries. We want to keep all of them unsimplified, to serve
1529 as the available stuff for the RHS of the rule.
1531 Example. Consider the following left-hand side of a rule
1533 f (x == y) (y > z) = ...
1535 If we typecheck this expression we get constraints
1537 d1 :: Ord a, d2 :: Eq a
1539 We do NOT want to "simplify" to the LHS
1541 forall x::a, y::a, z::a, d1::Ord a.
1542 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1546 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1547 f ((==) d2 x y) ((>) d1 y z) = ...
1549 Here is another example:
1551 fromIntegral :: (Integral a, Num b) => a -> b
1552 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1554 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1555 we *dont* want to get
1557 forall dIntegralInt.
1558 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1560 because the scsel will mess up RULE matching. Instead we want
1562 forall dIntegralInt, dNumInt.
1563 fromIntegral Int Int dIntegralInt dNumInt = id Int
1567 g (x == y) (y == z) = ..
1569 where the two dictionaries are *identical*, we do NOT WANT
1571 forall x::a, y::a, z::a, d1::Eq a
1572 f ((==) d1 x y) ((>) d1 y z) = ...
1574 because that will only match if the dict args are (visibly) equal.
1575 Instead we want to quantify over the dictionaries separately.
1577 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1578 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1579 from scratch, rather than further parameterise simpleReduceLoop etc.
1580 Simpler, maybe, but alas not simple (see Trac #2494)
1582 * Type errors may give rise to an (unsatisfiable) equality constraint
1584 * Applications of a higher-rank function on the LHS may give
1585 rise to an implication constraint, esp if there are unsatisfiable
1586 equality constraints inside.
1589 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1590 tcSimplifyRuleLhs wanteds
1591 = do { wanteds' <- zonkInsts wanteds
1592 ; (irreds, binds) <- go [] emptyBag wanteds'
1593 ; let (dicts, bad_irreds) = partition isDict irreds
1594 ; traceTc (text "tcSimplifyrulelhs" <+> pprInsts bad_irreds)
1595 ; addNoInstanceErrs (nub bad_irreds)
1596 -- The nub removes duplicates, which has
1597 -- not happened otherwise (see notes above)
1598 ; return (dicts, binds) }
1600 go :: [Inst] -> TcDictBinds -> [Inst] -> TcM ([Inst], TcDictBinds)
1602 = return (irreds, binds)
1603 go irreds binds (w:ws)
1605 = go (w:irreds) binds ws
1606 | isImplicInst w -- Have a go at reducing the implication
1607 = do { (binds1, irreds1) <- reduceImplication red_env w
1608 ; let (bad_irreds, ok_irreds) = partition isImplicInst irreds1
1609 ; go (bad_irreds ++ irreds)
1610 (binds `unionBags` binds1)
1613 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1614 -- to fromInteger; this looks fragile to me
1615 ; lookup_result <- lookupSimpleInst w'
1616 ; case lookup_result of
1617 NoInstance -> go (w:irreds) binds ws
1618 GenInst ws' rhs -> go irreds binds' (ws' ++ ws)
1620 binds' = addInstToDictBind binds w rhs
1623 -- Sigh: we need to reduce inside implications
1624 red_env = mkInferRedEnv doc try_me
1625 doc = ptext (sLit "Implication constraint in RULE lhs")
1626 try_me inst | isMethodOrLit inst = ReduceMe
1627 | otherwise = Stop -- Be gentle
1630 tcSimplifyBracket is used when simplifying the constraints arising from
1631 a Template Haskell bracket [| ... |]. We want to check that there aren't
1632 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1633 Show instance), but we aren't otherwise interested in the results.
1634 Nor do we care about ambiguous dictionaries etc. We will type check
1635 this bracket again at its usage site.
1638 tcSimplifyBracket :: [Inst] -> TcM ()
1639 tcSimplifyBracket wanteds
1640 = do { tryHardCheckLoop doc wanteds
1643 doc = text "tcSimplifyBracket"
1647 %************************************************************************
1649 \subsection{Filtering at a dynamic binding}
1651 %************************************************************************
1656 we must discharge all the ?x constraints from B. We also do an improvement
1657 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1659 Actually, the constraints from B might improve the types in ?x. For example
1661 f :: (?x::Int) => Char -> Char
1664 then the constraint (?x::Int) arising from the call to f will
1665 force the binding for ?x to be of type Int.
1668 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1671 -- We need a loop so that we do improvement, and then
1672 -- (next time round) generate a binding to connect the two
1674 -- Here the two ?x's have different types, and improvement
1675 -- makes them the same.
1677 tcSimplifyIPs given_ips wanteds
1678 = do { wanteds' <- zonkInsts wanteds
1679 ; given_ips' <- zonkInsts given_ips
1680 -- Unusually for checking, we *must* zonk the given_ips
1682 ; let env = mkRedEnv doc try_me given_ips'
1683 ; (improved, tybinds, binds, irreds) <- reduceContext env wanteds'
1684 ; execTcTyVarBinds tybinds
1686 ; if null irreds || not improved then
1687 ASSERT( all is_free irreds )
1688 do { extendLIEs irreds
1691 -- If improvement did some unification, we go round again.
1692 -- We start again with irreds, not wanteds
1693 -- Using an instance decl might have introduced a fresh type
1694 -- variable which might have been unified, so we'd get an
1695 -- infinite loop if we started again with wanteds!
1697 { binds1 <- tcSimplifyIPs given_ips' irreds
1698 ; return $ binds `unionBags` binds1
1701 doc = text "tcSimplifyIPs" <+> ppr given_ips
1702 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1703 is_free inst = isFreeWrtIPs ip_set inst
1705 -- Simplify any methods that mention the implicit parameter
1706 try_me inst | is_free inst = Stop
1707 | otherwise = ReduceMe
1711 %************************************************************************
1713 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1715 %************************************************************************
1717 When doing a binding group, we may have @Insts@ of local functions.
1718 For example, we might have...
1720 let f x = x + 1 -- orig local function (overloaded)
1721 f.1 = f Int -- two instances of f
1726 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1727 where @f@ is in scope; those @Insts@ must certainly not be passed
1728 upwards towards the top-level. If the @Insts@ were binding-ified up
1729 there, they would have unresolvable references to @f@.
1731 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1732 For each method @Inst@ in the @init_lie@ that mentions one of the
1733 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1734 @LIE@), as well as the @HsBinds@ generated.
1737 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1738 -- Simlifies only MethodInsts, and generate only bindings of form
1740 -- We're careful not to even generate bindings of the form
1742 -- You'd think that'd be fine, but it interacts with what is
1743 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1745 bindInstsOfLocalFuns wanteds local_ids
1746 | null overloaded_ids = do
1749 return emptyLHsBinds
1752 = do { (irreds, binds) <- gentleInferLoop doc for_me
1753 ; extendLIEs not_for_me
1757 doc = text "bindInsts" <+> ppr local_ids
1758 overloaded_ids = filter is_overloaded local_ids
1759 is_overloaded id = isOverloadedTy (idType id)
1760 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1762 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1763 -- so it's worth building a set, so that
1764 -- lookup (in isMethodFor) is faster
1768 %************************************************************************
1770 \subsection{Data types for the reduction mechanism}
1772 %************************************************************************
1774 The main control over context reduction is here
1778 = RedEnv { red_doc :: SDoc -- The context
1779 , red_try_me :: Inst -> WhatToDo
1780 , red_improve :: Bool -- True <=> do improvement
1781 , red_givens :: [Inst] -- All guaranteed rigid
1782 -- Always dicts & equalities
1783 -- but see Note [Rigidity]
1785 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1786 -- See Note [RedStack]
1790 -- The red_givens are rigid so far as cmpInst is concerned.
1791 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1792 -- let ?x = e in ...
1793 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1794 -- But that doesn't affect the comparison, which is based only on mame.
1797 -- The red_stack pair (n,insts) pair is just used for error reporting.
1798 -- 'n' is always the depth of the stack.
1799 -- The 'insts' is the stack of Insts being reduced: to produce X
1800 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1803 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1804 mkRedEnv doc try_me givens
1805 = RedEnv { red_doc = doc, red_try_me = try_me,
1806 red_givens = givens,
1808 red_improve = True }
1810 mkInferRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1812 mkInferRedEnv doc try_me
1813 = RedEnv { red_doc = doc, red_try_me = try_me,
1816 red_improve = True }
1818 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1819 -- Do not do improvement; no givens
1820 mkNoImproveRedEnv doc try_me
1821 = RedEnv { red_doc = doc, red_try_me = try_me,
1824 red_improve = True }
1827 = ReduceMe -- Try to reduce this
1828 -- If there's no instance, add the inst to the
1829 -- irreductible ones, but don't produce an error
1830 -- message of any kind.
1831 -- It might be quite legitimate such as (Eq a)!
1833 | Stop -- Return as irreducible unless it can
1834 -- be reduced to a constant in one step
1835 -- Do not add superclasses; see
1837 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1838 -- of a predicate when adding it to the avails
1839 -- The reason for this flag is entirely the super-class loop problem
1840 -- Note [SUPER-CLASS LOOP 1]
1842 zonkRedEnv :: RedEnv -> TcM RedEnv
1844 = do { givens' <- mapM zonkInst (red_givens env)
1845 ; return $ env {red_givens = givens'}
1850 %************************************************************************
1852 \subsection[reduce]{@reduce@}
1854 %************************************************************************
1856 Note [Ancestor Equalities]
1857 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1858 During context reduction, we add to the wanted equalities also those
1859 equalities that (transitively) occur in superclass contexts of wanted
1860 class constraints. Consider the following code
1862 class a ~ Int => C a
1865 If (C a) is wanted, we want to add (a ~ Int), which will be discharged by
1866 substituting Int for a. Hence, we ultimately want (C Int), which we
1867 discharge with the explicit instance.
1870 reduceContext :: RedEnv
1872 -> TcM (ImprovementDone,
1873 TcTyVarBinds, -- Type variable bindings
1874 TcDictBinds, -- Dictionary bindings
1875 [Inst]) -- Irreducible
1877 reduceContext env wanteds0
1878 = do { traceTc (text "reduceContext" <+> (vcat [
1879 text "----------------------",
1881 text "given" <+> ppr (red_givens env),
1882 text "wanted" <+> ppr wanteds0,
1883 text "----------------------"
1886 -- We want to add as wanted equalities those that (transitively)
1887 -- occur in superclass contexts of wanted class constraints.
1888 -- See Note [Ancestor Equalities]
1889 ; ancestor_eqs <- ancestorEqualities wanteds0
1890 ; traceTc $ text "reduceContext: ancestor eqs" <+> ppr ancestor_eqs
1892 -- Normalise and solve all equality constraints as far as possible
1893 -- and normalise all dictionary constraints wrt to the reduced
1894 -- equalities. The returned wanted constraints include the
1895 -- irreducible wanted equalities.
1896 ; let wanteds = wanteds0 ++ ancestor_eqs
1897 givens = red_givens env
1901 normalise_binds) <- tcReduceEqs givens wanteds
1902 ; traceTc $ text "reduceContext: tcReduceEqs result" <+> vcat
1903 [ppr givens', ppr wanteds', ppr tybinds,
1904 ppr normalise_binds]
1906 -- Build the Avail mapping from "given_dicts"
1907 ; (init_state, _) <- getLIE $ do
1908 { init_state <- foldlM addGiven emptyAvails givens'
1912 -- Solve the *wanted* *dictionary* constraints (not implications)
1913 -- This may expose some further equational constraints in the course
1914 -- of improvement due to functional dependencies if any of the
1915 -- involved unifications gets deferred.
1916 ; let (wanted_implics, wanted_dicts) = partition isImplicInst wanteds'
1917 ; (avails, extra_eqs) <- getLIE (reduceList env wanted_dicts init_state)
1918 -- The getLIE is reqd because reduceList does improvement
1919 -- (via extendAvails) which may in turn do unification
1922 dict_irreds) <- extractResults avails wanted_dicts
1923 ; traceTc $ text "reduceContext: extractResults" <+> vcat
1924 [ppr avails, ppr wanted_dicts, ppr dict_binds]
1926 -- Solve the wanted *implications*. In doing so, we can provide
1927 -- as "given" all the dicts that were originally given,
1928 -- *or* for which we now have bindings,
1929 -- *or* which are now irreds
1930 -- NB: Equality irreds need to be converted, as the recursive
1931 -- invocation of the solver will still treat them as wanteds
1933 ; let implic_env = env { red_givens
1934 = givens ++ bound_dicts ++
1935 map wantedToLocalEqInst dict_irreds }
1936 ; (implic_binds_s, implic_irreds_s)
1937 <- mapAndUnzipM (reduceImplication implic_env) wanted_implics
1938 ; let implic_binds = unionManyBags implic_binds_s
1939 implic_irreds = concat implic_irreds_s
1941 -- Collect all irreducible instances, and determine whether we should
1942 -- go round again. We do so in either of two cases:
1943 -- (1) If dictionary reduction or equality solving led to
1944 -- improvement (i.e., bindings for type variables).
1945 -- (2) If we reduced dictionaries (i.e., got dictionary bindings),
1946 -- they may have exposed further opportunities to normalise
1947 -- family applications. See Note [Dictionary Improvement]
1949 -- NB: We do *not* go around for new extra_eqs. Morally, we should,
1950 -- but we can't without risking non-termination (see #2688). By
1951 -- not going around, we miss some legal programs mixing FDs and
1952 -- TFs, but we never claimed to support such programs in the
1953 -- current implementation anyway.
1955 ; let all_irreds = dict_irreds ++ implic_irreds ++ extra_eqs
1956 avails_improved = availsImproved avails
1957 eq_improved = anyBag (not . isCoVarBind) tybinds
1958 improvedFlexible = avails_improved || eq_improved
1959 reduced_dicts = not (isEmptyBag dict_binds)
1960 improved = improvedFlexible || reduced_dicts
1962 improvedHint = (if avails_improved then " [AVAILS]" else "") ++
1963 (if eq_improved then " [EQ]" else "")
1965 ; traceTc (text "reduceContext end" <+> (vcat [
1966 text "----------------------",
1968 text "given" <+> ppr givens,
1969 text "wanted" <+> ppr wanteds0,
1971 text "tybinds" <+> ppr tybinds,
1972 text "avails" <+> pprAvails avails,
1973 text "improved =" <+> ppr improved <+> text improvedHint,
1974 text "(all) irreds = " <+> ppr all_irreds,
1975 text "dict-binds = " <+> ppr dict_binds,
1976 text "implic-binds = " <+> ppr implic_binds,
1977 text "----------------------"
1982 normalise_binds `unionBags` dict_binds
1983 `unionBags` implic_binds,
1987 isCoVarBind (TcTyVarBind tv _) = isCoVar tv
1989 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1990 tcImproveOne avails inst
1991 | not (isDict inst) = return False
1993 = do { inst_envs <- tcGetInstEnvs
1994 ; let eqns = improveOne (classInstances inst_envs)
1995 (dictPred inst, pprInstArising inst)
1996 [ (dictPred p, pprInstArising p)
1997 | p <- availsInsts avails, isDict p ]
1998 -- Avails has all the superclasses etc (good)
1999 -- It also has all the intermediates of the deduction (good)
2000 -- It does not have duplicates (good)
2001 -- NB that (?x::t1) and (?x::t2) will be held separately in
2002 -- avails so that improve will see them separate
2003 ; traceTc (text "improveOne" <+> ppr inst)
2006 unifyEqns :: [(Equation, (PredType, SDoc), (PredType, SDoc))]
2007 -> TcM ImprovementDone
2008 unifyEqns [] = return False
2010 = do { traceTc (ptext (sLit "Improve:") <+> vcat (map pprEquationDoc eqns))
2011 ; improved <- mapM unify eqns
2012 ; return $ or improved
2015 unify ((qtvs, pairs), what1, what2)
2016 = addErrCtxtM (mkEqnMsg what1 what2) $
2017 do { let freeTyVars = unionVarSets (map tvs_pr pairs)
2019 ; (_, _, tenv) <- tcInstTyVars (varSetElems qtvs)
2020 ; mapM_ (unif_pr tenv) pairs
2021 ; anyM isFilledMetaTyVar $ varSetElems freeTyVars
2024 unif_pr tenv (ty1, ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
2026 tvs_pr (ty1, ty2) = tyVarsOfType ty1 `unionVarSet` tyVarsOfType ty2
2028 pprEquationDoc :: (Equation, (PredType, SDoc), (PredType, SDoc)) -> SDoc
2029 pprEquationDoc (eqn, (p1, _), (p2, _))
2030 = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
2032 mkEqnMsg :: (TcPredType, SDoc) -> (TcPredType, SDoc) -> TidyEnv
2033 -> TcM (TidyEnv, SDoc)
2034 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
2035 = do { pred1' <- zonkTcPredType pred1
2036 ; pred2' <- zonkTcPredType pred2
2037 ; let { pred1'' = tidyPred tidy_env pred1'
2038 ; pred2'' = tidyPred tidy_env pred2' }
2039 ; let msg = vcat [ptext (sLit "When using functional dependencies to combine"),
2040 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
2041 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
2042 ; return (tidy_env, msg) }
2045 Note [Dictionary Improvement]
2046 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2047 In reduceContext, we first reduce equalities and then class constraints.
2048 However, the letter may expose further opportunities for the former. Hence,
2049 we need to go around again if dictionary reduction produced any dictionary
2050 bindings. The following example demonstrated the point:
2052 data EX _x _y (p :: * -> *)
2057 class Base (Def p) => Prop p where
2061 instance Prop () where
2064 instance (Base (Def (p ANY))) => Base (EX _x _y p)
2065 instance (Prop (p ANY)) => Prop (EX _x _y p) where
2066 type Def (EX _x _y p) = EX _x _y p
2069 instance Prop (FOO x) where
2070 type Def (FOO x) = ()
2073 instance Prop BAR where
2074 type Def BAR = EX () () FOO
2076 During checking the last instance declaration, we need to check the superclass
2077 cosntraint Base (Def BAR), which family normalisation reduced to
2078 Base (EX () () FOO). Chasing the instance for Base (EX _x _y p), gives us
2079 Base (Def (FOO ANY)), which again requires family normalisation of Def to
2080 Base () before we can finish.
2083 The main context-reduction function is @reduce@. Here's its game plan.
2086 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
2087 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
2088 = do { traceTc (text "reduceList " <+> (ppr wanteds $$ ppr state))
2090 ; when (debugIsOn && (n > 8)) $ do
2091 debugDumpTcRn (hang (ptext (sLit "Interesting! Context reduction stack depth") <+> int n)
2092 2 (ifPprDebug (nest 2 (pprStack stk))))
2093 ; if n >= ctxtStkDepth dopts then
2094 failWithTc (reduceDepthErr n stk)
2098 go [] state = return state
2099 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
2102 -- Base case: we're done!
2103 reduce :: RedEnv -> Inst -> Avails -> TcM Avails
2104 reduce env wanted avails
2106 -- We don't reduce equalities here (and they must not end up as irreds
2111 -- It's the same as an existing inst, or a superclass thereof
2112 | Just _ <- findAvail avails wanted
2113 = do { traceTc (text "reduce: found " <+> ppr wanted)
2118 = do { traceTc (text "reduce" <+> ppr wanted $$ ppr avails)
2119 ; case red_try_me env wanted of {
2120 Stop -> try_simple (addIrred NoSCs);
2121 -- See Note [No superclasses for Stop]
2123 ReduceMe -> do -- It should be reduced
2124 { (avails, lookup_result) <- reduceInst env avails wanted
2125 ; case lookup_result of
2126 NoInstance -> addIrred AddSCs avails wanted
2127 -- Add it and its superclasses
2129 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
2131 GenInst wanteds' rhs
2132 -> do { avails1 <- addIrred NoSCs avails wanted
2133 ; avails2 <- reduceList env wanteds' avails1
2134 ; addWanted AddSCs avails2 wanted rhs wanteds' } }
2135 -- Temporarily do addIrred *before* the reduceList,
2136 -- which has the effect of adding the thing we are trying
2137 -- to prove to the database before trying to prove the things it
2138 -- needs. See note [RECURSIVE DICTIONARIES]
2139 -- NB: we must not do an addWanted before, because that adds the
2140 -- superclasses too, and that can lead to a spurious loop; see
2141 -- the examples in [SUPERCLASS-LOOP]
2142 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
2145 -- First, see if the inst can be reduced to a constant in one step
2146 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
2147 -- Don't bother for implication constraints, which take real work
2148 try_simple do_this_otherwise
2149 = do { res <- lookupSimpleInst wanted
2151 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
2152 _ -> do_this_otherwise avails wanted }
2156 Note [RECURSIVE DICTIONARIES]
2157 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2159 data D r = ZeroD | SuccD (r (D r));
2161 instance (Eq (r (D r))) => Eq (D r) where
2162 ZeroD == ZeroD = True
2163 (SuccD a) == (SuccD b) = a == b
2166 equalDC :: D [] -> D [] -> Bool;
2169 We need to prove (Eq (D [])). Here's how we go:
2173 by instance decl, holds if
2177 by instance decl of Eq, holds if
2179 where d2 = dfEqList d3
2182 But now we can "tie the knot" to give
2188 and it'll even run! The trick is to put the thing we are trying to prove
2189 (in this case Eq (D []) into the database before trying to prove its
2190 contributing clauses.
2192 Note [SUPERCLASS-LOOP 2]
2193 ~~~~~~~~~~~~~~~~~~~~~~~~
2194 We need to be careful when adding "the constaint we are trying to prove".
2195 Suppose we are *given* d1:Ord a, and want to deduce (d2:C [a]) where
2197 class Ord a => C a where
2198 instance Ord [a] => C [a] where ...
2200 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
2201 superclasses of C [a] to avails. But we must not overwrite the binding
2202 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
2205 Here's another variant, immortalised in tcrun020
2206 class Monad m => C1 m
2207 class C1 m => C2 m x
2208 instance C2 Maybe Bool
2209 For the instance decl we need to build (C1 Maybe), and it's no good if
2210 we run around and add (C2 Maybe Bool) and its superclasses to the avails
2211 before we search for C1 Maybe.
2213 Here's another example
2214 class Eq b => Foo a b
2215 instance Eq a => Foo [a] a
2219 we'll first deduce that it holds (via the instance decl). We must not
2220 then overwrite the Eq t constraint with a superclass selection!
2222 At first I had a gross hack, whereby I simply did not add superclass constraints
2223 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
2224 becuase it lost legitimate superclass sharing, and it still didn't do the job:
2225 I found a very obscure program (now tcrun021) in which improvement meant the
2226 simplifier got two bites a the cherry... so something seemed to be an Stop
2227 first time, but reducible next time.
2229 Now we implement the Right Solution, which is to check for loops directly
2230 when adding superclasses. It's a bit like the occurs check in unification.
2234 %************************************************************************
2236 Reducing a single constraint
2238 %************************************************************************
2241 ---------------------------------------------
2242 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
2243 reduceInst _ avails other_inst
2244 = do { result <- lookupSimpleInst other_inst
2245 ; return (avails, result) }
2248 Note [Equational Constraints in Implication Constraints]
2249 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2251 An implication constraint is of the form
2253 where Given and Wanted may contain both equational and dictionary
2254 constraints. The delay and reduction of these two kinds of constraints
2257 -) In the generated code, wanted Dictionary constraints are wrapped up in an
2258 implication constraint that is created at the code site where the wanted
2259 dictionaries can be reduced via a let-binding. This let-bound implication
2260 constraint is deconstructed at the use-site of the wanted dictionaries.
2262 -) While the reduction of equational constraints is also delayed, the delay
2263 is not manifest in the generated code. The required evidence is generated
2264 in the code directly at the use-site. There is no let-binding and deconstruction
2265 necessary. The main disadvantage is that we cannot exploit sharing as the
2266 same evidence may be generated at multiple use-sites. However, this disadvantage
2267 is limited because it only concerns coercions which are erased.
2269 The different treatment is motivated by the different in representation. Dictionary
2270 constraints require manifest runtime dictionaries, while equations require coercions
2274 ---------------------------------------------
2275 reduceImplication :: RedEnv
2277 -> TcM (TcDictBinds, [Inst])
2280 Suppose we are simplifying the constraint
2281 forall bs. extras => wanted
2282 in the context of an overall simplification problem with givens 'givens'.
2285 * The 'givens' need not mention any of the quantified type variables
2286 e.g. forall {}. Eq a => Eq [a]
2287 forall {}. C Int => D (Tree Int)
2289 This happens when you have something like
2291 T1 :: Eq a => a -> T a
2294 f x = ...(case x of { T1 v -> v==v })...
2297 -- ToDo: should we instantiate tvs? I think it's not necessary
2299 -- Note on coercion variables:
2301 -- The extra given coercion variables are bound at two different
2304 -- -) in the creation context of the implication constraint
2305 -- the solved equational constraints use these binders
2307 -- -) at the solving site of the implication constraint
2308 -- the solved dictionaries use these binders;
2309 -- these binders are generated by reduceImplication
2311 -- Note [Binders for equalities]
2312 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2313 -- To reuse the binders of local/given equalities in the binders of
2314 -- implication constraints, it is crucial that these given equalities
2315 -- always have the form
2317 -- where cotv is a simple coercion type variable (and not a more
2318 -- complex coercion term). We require that the extra_givens always
2319 -- have this form and exploit the special form when generating binders.
2320 reduceImplication env
2321 orig_implic@(ImplicInst { tci_name = name, tci_loc = inst_loc,
2323 tci_given = extra_givens, tci_wanted = wanteds
2325 = do { -- Solve the sub-problem
2326 ; let try_me _ = ReduceMe -- Note [Freeness and implications]
2327 env' = env { red_givens = extra_givens ++ red_givens env
2328 , red_doc = sep [ptext (sLit "reduceImplication for")
2330 nest 2 (parens $ ptext (sLit "within")
2332 , red_try_me = try_me }
2334 ; traceTc (text "reduceImplication" <+> vcat
2335 [ ppr (red_givens env), ppr extra_givens,
2337 ; (irreds, binds) <- checkLoop env' wanteds
2339 ; traceTc (text "reduceImplication result" <+> vcat
2340 [ppr irreds, ppr binds])
2342 ; -- extract superclass binds
2343 -- (sc_binds,_) <- extractResults avails []
2344 -- ; traceTc (text "reduceImplication sc_binds" <+> vcat
2345 -- [ppr sc_binds, ppr avails])
2348 -- SLPJ Sept 07: what if improvement happened inside the checkLoop?
2349 -- Then we must iterate the outer loop too!
2351 ; didntSolveWantedEqs <- allM wantedEqInstIsUnsolved wanteds
2352 -- we solve wanted eqs by side effect!
2354 -- Progress is no longer measered by the number of bindings
2355 -- If there are any irreds, but no bindings and no solved
2356 -- equalities, we back off and do nothing
2357 ; let backOff = isEmptyLHsBinds binds && -- no new bindings
2358 (not $ null irreds) && -- but still some irreds
2359 didntSolveWantedEqs -- no instantiated cotv
2361 ; if backOff then -- No progress
2362 return (emptyBag, [orig_implic])
2364 { (simpler_implic_insts, bind)
2365 <- makeImplicationBind inst_loc tvs extra_givens irreds
2366 -- This binding is useless if the recursive simplification
2367 -- made no progress; but currently we don't try to optimise that
2368 -- case. After all, we only try hard to reduce at top level, or
2369 -- when inferring types.
2371 ; let -- extract Id binders for dicts and CoTyVar binders for eqs;
2372 -- see Note [Binders for equalities]
2373 (extra_eq_givens, extra_dict_givens) = partition isEqInst
2375 eq_cotvs = map instToVar extra_eq_givens
2376 dict_ids = map instToId extra_dict_givens
2378 -- Note [Always inline implication constraints]
2379 wrap_inline | null dict_ids = idHsWrapper
2380 | otherwise = WpInline
2383 <.> mkWpTyLams eq_cotvs
2384 <.> mkWpLams dict_ids
2385 <.> WpLet (binds `unionBags` bind)
2386 rhs = mkLHsWrap co payload
2387 loc = instLocSpan inst_loc
2388 -- wanted equalities are solved by updating their
2389 -- cotv; we don't generate bindings for them
2390 dict_bndrs = map (L loc . HsVar . instToId)
2391 . filter (not . isEqInst)
2393 payload = mkBigLHsTup dict_bndrs
2396 ; traceTc (vcat [text "reduceImplication" <+> ppr name,
2397 ppr simpler_implic_insts,
2398 text "->" <+> ppr rhs])
2399 ; return (unitBag (L loc (VarBind (instToId orig_implic) rhs)),
2400 simpler_implic_insts)
2403 reduceImplication _ i = pprPanic "reduceImplication" (ppr i)
2406 Note [Always inline implication constraints]
2407 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2408 Suppose an implication constraint floats out of an INLINE function.
2409 Then although the implication has a single call site, it won't be
2410 inlined. And that is bad because it means that even if there is really
2411 *no* overloading (type signatures specify the exact types) there will
2412 still be dictionary passing in the resulting code. To avert this,
2413 we mark the implication constraints themselves as INLINE, at least when
2414 there is no loss of sharing as a result.
2416 Note [Freeness and implications]
2417 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2418 It's hard to say when an implication constraint can be floated out. Consider
2419 forall {} Eq a => Foo [a]
2420 The (Foo [a]) doesn't mention any of the quantified variables, but it
2421 still might be partially satisfied by the (Eq a).
2423 There is a useful special case when it *is* easy to partition the
2424 constraints, namely when there are no 'givens'. Consider
2425 forall {a}. () => Bar b
2426 There are no 'givens', and so there is no reason to capture (Bar b).
2427 We can let it float out. But if there is even one constraint we
2428 must be much more careful:
2429 forall {a}. C a b => Bar (m b)
2430 because (C a b) might have a superclass (D b), from which we might
2431 deduce (Bar [b]) when m later gets instantiated to []. Ha!
2433 Here is an even more exotic example
2435 Now consider the constraint
2436 forall b. D Int b => C Int
2437 We can satisfy the (C Int) from the superclass of D, so we don't want
2438 to float the (C Int) out, even though it mentions no type variable in
2441 One more example: the constraint
2443 instance (C a, E c) => E (a,c)
2445 constraint: forall b. D Int b => E (Int,c)
2447 You might think that the (D Int b) can't possibly contribute
2448 to solving (E (Int,c)), since the latter mentions 'c'. But
2449 in fact it can, because solving the (E (Int,c)) constraint needs
2452 and the (C Int) can be satisfied from the superclass of (D Int b).
2453 So we must still not float (E (Int,c)) out.
2455 To think about: special cases for unary type classes?
2457 Note [Pruning the givens in an implication constraint]
2458 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2459 Suppose we are about to form the implication constraint
2460 forall tvs. Eq a => Ord b
2461 The (Eq a) cannot contribute to the (Ord b), because it has no access to
2462 the type variable 'b'. So we could filter out the (Eq a) from the givens.
2463 But BE CAREFUL of the examples above in [Freeness and implications].
2465 Doing so would be a bit tidier, but all the implication constraints get
2466 simplified away by the optimiser, so it's no great win. So I don't take
2467 advantage of that at the moment.
2469 If you do, BE CAREFUL of wobbly type variables.
2472 %************************************************************************
2474 Avails and AvailHow: the pool of evidence
2476 %************************************************************************
2480 data Avails = Avails !ImprovementDone !AvailEnv
2482 type ImprovementDone = Bool -- True <=> some unification has happened
2483 -- so some Irreds might now be reducible
2484 -- keys that are now
2486 type AvailEnv = FiniteMap Inst AvailHow
2488 = IsIrred -- Used for irreducible dictionaries,
2489 -- which are going to be lambda bound
2491 | Given Inst -- Used for dictionaries for which we have a binding
2492 -- e.g. those "given" in a signature
2494 | Rhs -- Used when there is a RHS
2495 (LHsExpr TcId) -- The RHS
2496 [Inst] -- Insts free in the RHS; we need these too
2498 instance Outputable Avails where
2501 pprAvails :: Avails -> SDoc
2502 pprAvails (Avails imp avails)
2503 = vcat [ ptext (sLit "Avails") <> (if imp then ptext (sLit "[improved]") else empty)
2505 vcat [ sep [ppr inst, nest 2 (equals <+> ppr avail)]
2506 | (inst,avail) <- fmToList avails ]]
2508 instance Outputable AvailHow where
2511 -------------------------
2512 pprAvail :: AvailHow -> SDoc
2513 pprAvail IsIrred = text "Irred"
2514 pprAvail (Given x) = text "Given" <+> ppr x
2515 pprAvail (Rhs rhs bs) = sep [text "Rhs" <+> ppr bs,
2518 -------------------------
2519 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2520 extendAvailEnv env inst avail = addToFM env inst avail
2522 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2523 findAvailEnv env wanted = lookupFM env wanted
2524 -- NB 1: the Ord instance of Inst compares by the class/type info
2525 -- *not* by unique. So
2526 -- d1::C Int == d2::C Int
2528 emptyAvails :: Avails
2529 emptyAvails = Avails False emptyFM
2531 findAvail :: Avails -> Inst -> Maybe AvailHow
2532 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2534 elemAvails :: Inst -> Avails -> Bool
2535 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2537 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2539 extendAvails avails@(Avails imp env) inst avail
2540 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2541 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2543 availsInsts :: Avails -> [Inst]
2544 availsInsts (Avails _ avails) = keysFM avails
2546 availsImproved :: Avails -> ImprovementDone
2547 availsImproved (Avails imp _) = imp
2550 Extracting the bindings from a bunch of Avails.
2551 The bindings do *not* come back sorted in dependency order.
2552 We assume that they'll be wrapped in a big Rec, so that the
2553 dependency analyser can sort them out later
2556 type DoneEnv = FiniteMap Inst [Id]
2557 -- Tracks which things we have evidence for
2559 extractResults :: Avails
2561 -> TcM (TcDictBinds, -- Bindings
2562 [Inst], -- The insts bound by the bindings
2563 [Inst]) -- Irreducible ones
2564 -- Note [Reducing implication constraints]
2566 extractResults (Avails _ avails) wanteds
2567 = go emptyBag [] [] emptyFM wanteds
2569 go :: TcDictBinds -- Bindings for dicts
2570 -> [Inst] -- Bound by the bindings
2572 -> DoneEnv -- Has an entry for each inst in the above three sets
2574 -> TcM (TcDictBinds, [Inst], [Inst])
2575 go binds bound_dicts irreds _ []
2576 = return (binds, bound_dicts, irreds)
2578 go binds bound_dicts irreds done (w:ws)
2580 = go binds bound_dicts (w:irreds) done' ws
2582 | Just done_ids@(done_id : rest_done_ids) <- lookupFM done w
2583 = if w_id `elem` done_ids then
2584 go binds bound_dicts irreds done ws
2586 go (add_bind (nlHsVar done_id)) bound_dicts irreds
2587 (addToFM done w (done_id : w_id : rest_done_ids)) ws
2589 | otherwise -- Not yet done
2590 = case findAvailEnv avails w of
2591 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2592 go binds bound_dicts irreds done ws
2594 Just IsIrred -> go binds bound_dicts (w:irreds) done' ws
2596 Just (Rhs rhs ws') -> go (add_bind rhs) (w:bound_dicts) irreds done' (ws' ++ ws)
2598 Just (Given g) -> go binds' bound_dicts irreds (addToFM done w [g_id]) ws
2601 binds' | w_id == g_id = binds
2602 | otherwise = add_bind (nlHsVar g_id)
2605 done' = addToFM done w [w_id]
2606 add_bind rhs = addInstToDictBind binds w rhs
2610 Note [No superclasses for Stop]
2611 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2612 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2613 add it to avails, so that any other equal Insts will be commoned up
2614 right here. However, we do *not* add superclasses. If we have
2617 but a is not bound here, then we *don't* want to derive dn from df
2618 here lest we lose sharing.
2621 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2622 addWanted want_scs avails wanted rhs_expr wanteds
2623 = addAvailAndSCs want_scs avails wanted avail
2625 avail = Rhs rhs_expr wanteds
2627 addGiven :: Avails -> Inst -> TcM Avails
2628 addGiven avails given
2629 = addAvailAndSCs want_scs avails given (Given given)
2631 want_scs = case instLocOrigin (instLoc given) of
2634 -- Conditionally add superclasses for 'given'
2635 -- See Note [Recursive instances and superclases]
2637 -- No ASSERT( not (given `elemAvails` avails) ) because in an
2638 -- instance decl for Ord t we can add both Ord t and Eq t as
2639 -- 'givens', so the assert isn't true
2643 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2644 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2645 addAvailAndSCs want_scs avails irred IsIrred
2647 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2648 addAvailAndSCs want_scs avails inst avail
2649 | not (isClassDict inst) = extendAvails avails inst avail
2650 | NoSCs <- want_scs = extendAvails avails inst avail
2651 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2652 ; avails' <- extendAvails avails inst avail
2653 ; addSCs is_loop avails' inst }
2655 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2656 -- Note: this compares by *type*, not by Unique
2657 deps = findAllDeps (unitVarSet (instToVar inst)) avail
2658 dep_tys = map idType (varSetElems deps)
2660 findAllDeps :: IdSet -> AvailHow -> IdSet
2661 -- Find all the Insts that this one depends on
2662 -- See Note [SUPERCLASS-LOOP 2]
2663 -- Watch out, though. Since the avails may contain loops
2664 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2665 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2666 findAllDeps so_far _ = so_far
2668 find_all :: IdSet -> Inst -> IdSet
2670 | isEqInst kid = so_far
2671 | kid_id `elemVarSet` so_far = so_far
2672 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2673 | otherwise = so_far'
2675 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2676 kid_id = instToId kid
2678 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2679 -- Add all the superclasses of the Inst to Avails
2680 -- The first param says "don't do this because the original thing
2681 -- depends on this one, so you'd build a loop"
2682 -- Invariant: the Inst is already in Avails.
2684 addSCs is_loop avails dict
2685 = ASSERT( isDict dict )
2686 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2687 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2689 (clas, tys) = getDictClassTys dict
2690 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2691 sc_theta' = filter (not . isEqPred) $
2692 substTheta (zipTopTvSubst tyvars tys) sc_theta
2694 add_sc avails (sc_dict, sc_sel)
2695 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2696 | is_given sc_dict = return avails
2697 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2698 ; addSCs is_loop avails' sc_dict }
2700 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2701 co_fn = WpApp (instToVar dict) <.> mkWpTyApps tys
2703 is_given :: Inst -> Bool
2704 is_given sc_dict = case findAvail avails sc_dict of
2705 Just (Given _) -> True -- Given is cheaper than superclass selection
2708 -- From the a set of insts obtain all equalities that (transitively) occur in
2709 -- superclass contexts of class constraints (aka the ancestor equalities).
2711 ancestorEqualities :: [Inst] -> TcM [Inst]
2713 = mapM mkWantedEqInst -- turn only equality predicates..
2714 . filter isEqPred -- ..into wanted equality insts
2716 . addAEsToBag emptyBag -- collect the superclass constraints..
2717 . map dictPred -- ..of all predicates in a bag
2718 . filter isClassDict
2720 addAEsToBag :: Bag PredType -> [PredType] -> Bag PredType
2721 addAEsToBag bag [] = bag
2722 addAEsToBag bag (pred:preds)
2723 | pred `elemBag` bag = addAEsToBag bag preds
2724 | isEqPred pred = addAEsToBag bagWithPred preds
2725 | isClassPred pred = addAEsToBag bagWithPred predsWithSCs
2726 | otherwise = addAEsToBag bag preds
2728 bagWithPred = bag `snocBag` pred
2729 predsWithSCs = preds ++ substTheta (zipTopTvSubst tyvars tys) sc_theta
2731 (tyvars, sc_theta, _, _) = classBigSig clas
2732 (clas, tys) = getClassPredTys pred
2736 %************************************************************************
2738 \section{tcSimplifyTop: defaulting}
2740 %************************************************************************
2743 @tcSimplifyTop@ is called once per module to simplify all the constant
2744 and ambiguous Insts.
2746 We need to be careful of one case. Suppose we have
2748 instance Num a => Num (Foo a b) where ...
2750 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2751 to (Num x), and default x to Int. But what about y??
2753 It's OK: the final zonking stage should zap y to (), which is fine.
2757 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2758 tcSimplifyTop wanteds
2759 = tc_simplify_top doc False wanteds
2761 doc = text "tcSimplifyTop"
2763 tcSimplifyInteractive wanteds
2764 = tc_simplify_top doc True wanteds
2766 doc = text "tcSimplifyInteractive"
2768 -- The TcLclEnv should be valid here, solely to improve
2769 -- error message generation for the monomorphism restriction
2770 tc_simplify_top :: SDoc -> Bool -> [Inst] -> TcM (Bag (LHsBind TcId))
2771 tc_simplify_top doc interactive wanteds
2772 = do { dflags <- getDOpts
2773 ; wanteds <- zonkInsts wanteds
2774 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2776 ; traceTc (text "tc_simplify_top 0: " <+> ppr wanteds)
2777 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2778 -- ; (irreds1, binds1) <- gentleInferLoop doc1 wanteds
2779 ; traceTc (text "tc_simplify_top 1: " <+> ppr irreds1)
2780 ; (irreds2, binds2) <- approximateImplications doc2 (\_ -> True) irreds1
2781 ; traceTc (text "tc_simplify_top 2: " <+> ppr irreds2)
2783 -- Use the defaulting rules to do extra unification
2784 -- NB: irreds2 are already zonked
2785 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2787 -- Deal with implicit parameters
2788 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2789 (ambigs, others) = partition isTyVarDict non_ips
2791 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2793 ; addNoInstanceErrs others
2794 ; addTopAmbigErrs ambigs
2796 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2798 doc1 = doc <+> ptext (sLit "(first round)")
2799 doc2 = doc <+> ptext (sLit "(approximate)")
2800 doc3 = doc <+> ptext (sLit "(disambiguate)")
2803 If a dictionary constrains a type variable which is
2804 * not mentioned in the environment
2805 * and not mentioned in the type of the expression
2806 then it is ambiguous. No further information will arise to instantiate
2807 the type variable; nor will it be generalised and turned into an extra
2808 parameter to a function.
2810 It is an error for this to occur, except that Haskell provided for
2811 certain rules to be applied in the special case of numeric types.
2813 * at least one of its classes is a numeric class, and
2814 * all of its classes are numeric or standard
2815 then the type variable can be defaulted to the first type in the
2816 default-type list which is an instance of all the offending classes.
2818 So here is the function which does the work. It takes the ambiguous
2819 dictionaries and either resolves them (producing bindings) or
2820 complains. It works by splitting the dictionary list by type
2821 variable, and using @disambigOne@ to do the real business.
2823 @disambigOne@ assumes that its arguments dictionaries constrain all
2824 the same type variable.
2826 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2827 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2828 the most common use of defaulting is code like:
2830 _ccall_ foo `seqPrimIO` bar
2832 Since we're not using the result of @foo@, the result if (presumably)
2836 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2837 -- Just does unification to fix the default types
2838 -- The Insts are assumed to be pre-zonked
2839 disambiguate doc interactive dflags insts
2841 = return (insts, emptyBag)
2843 | null defaultable_groups
2844 = do { traceTc (text "disambigutate, no defaultable groups" <+> vcat [ppr unaries, ppr insts, ppr bad_tvs, ppr defaultable_groups])
2845 ; return (insts, emptyBag) }
2848 = do { -- Figure out what default types to use
2849 default_tys <- getDefaultTys extended_defaulting ovl_strings
2851 ; traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2852 ; mapM_ (disambigGroup default_tys) defaultable_groups
2854 -- disambigGroup does unification, hence try again
2855 ; tryHardCheckLoop doc insts }
2858 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2859 ovl_strings = dopt Opt_OverloadedStrings dflags
2861 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2862 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2863 (unaries, bad_tvs_s) = partitionWith find_unary insts
2864 bad_tvs = unionVarSets bad_tvs_s
2866 -- Finds unary type-class constraints
2867 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2868 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2869 find_unary inst = Right (tyVarsOfInst inst)
2871 -- Group by type variable
2872 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2873 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2874 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2876 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2877 defaultable_group ds@((_,_,tv):_)
2878 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2879 && not (tv `elemVarSet` bad_tvs)
2880 && defaultable_classes [c | (_,c,_) <- ds]
2881 defaultable_group [] = panic "defaultable_group"
2883 defaultable_classes clss
2884 | extended_defaulting = any isInteractiveClass clss
2885 | otherwise = all is_std_class clss && (any is_num_class clss)
2887 -- In interactive mode, or with -XExtendedDefaultRules,
2888 -- we default Show a to Show () to avoid graututious errors on "show []"
2889 isInteractiveClass cls
2890 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2892 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2893 -- is_num_class adds IsString to the standard numeric classes,
2894 -- when -foverloaded-strings is enabled
2896 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2897 -- Similarly is_std_class
2899 -----------------------
2900 disambigGroup :: [Type] -- The default types
2901 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2902 -> TcM () -- Just does unification, to fix the default types
2904 disambigGroup default_tys dicts
2905 = try_default default_tys
2907 (_,_,tyvar) = ASSERT(not (null dicts)) head dicts -- Should be non-empty
2908 classes = [c | (_,c,_) <- dicts]
2910 try_default [] = return ()
2911 try_default (default_ty : default_tys)
2912 = tryTcLIE_ (try_default default_tys) $
2913 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2914 -- This may fail; then the tryTcLIE_ kicks in
2915 -- Failure here is caused by there being no type in the
2916 -- default list which can satisfy all the ambiguous classes.
2917 -- For example, if Real a is reqd, but the only type in the
2918 -- default list is Int.
2920 -- After this we can't fail
2921 ; warnDefault dicts default_ty
2922 ; unifyType default_ty (mkTyVarTy tyvar)
2923 ; return () -- TOMDO: do something with the coercion
2927 -----------------------
2928 getDefaultTys :: Bool -> Bool -> TcM [Type]
2929 getDefaultTys extended_deflts ovl_strings
2930 = do { mb_defaults <- getDeclaredDefaultTys
2931 ; case mb_defaults of {
2932 Just tys -> return tys ; -- User-supplied defaults
2935 -- No use-supplied default
2936 -- Use [Integer, Double], plus modifications
2937 { integer_ty <- tcMetaTy integerTyConName
2938 ; checkWiredInTyCon doubleTyCon
2939 ; string_ty <- tcMetaTy stringTyConName
2940 ; return (opt_deflt extended_deflts unitTy
2941 -- Note [Default unitTy]
2943 [integer_ty,doubleTy]
2945 opt_deflt ovl_strings string_ty) } } }
2947 opt_deflt True ty = [ty]
2948 opt_deflt False _ = []
2951 Note [Default unitTy]
2952 ~~~~~~~~~~~~~~~~~~~~~
2953 In interative mode (or with -XExtendedDefaultRules) we add () as the first type we
2954 try when defaulting. This has very little real impact, except in the following case.
2956 Text.Printf.printf "hello"
2957 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2958 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2959 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2960 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2961 () to the list of defaulting types. See Trac #1200.
2963 Note [Avoiding spurious errors]
2964 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2965 When doing the unification for defaulting, we check for skolem
2966 type variables, and simply don't default them. For example:
2967 f = (*) -- Monomorphic
2968 g :: Num a => a -> a
2970 Here, we get a complaint when checking the type signature for g,
2971 that g isn't polymorphic enough; but then we get another one when
2972 dealing with the (Num a) context arising from f's definition;
2973 we try to unify a with Int (to default it), but find that it's
2974 already been unified with the rigid variable from g's type sig
2977 %************************************************************************
2979 \subsection[simple]{@Simple@ versions}
2981 %************************************************************************
2983 Much simpler versions when there are no bindings to make!
2985 @tcSimplifyThetas@ simplifies class-type constraints formed by
2986 @deriving@ declarations and when specialising instances. We are
2987 only interested in the simplified bunch of class/type constraints.
2989 It simplifies to constraints of the form (C a b c) where
2990 a,b,c are type variables. This is required for the context of
2991 instance declarations.
2994 tcSimplifyDeriv :: InstOrigin
2996 -> ThetaType -- Wanted
2997 -> TcM ThetaType -- Needed
2998 -- Given instance (wanted) => C inst_ty
2999 -- Simplify 'wanted' as much as possible
3001 tcSimplifyDeriv orig tyvars theta
3002 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
3003 -- The main loop may do unification, and that may crash if
3004 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
3005 -- ToDo: what if two of them do get unified?
3006 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
3007 ; (irreds, _) <- tryHardCheckLoop doc wanteds
3009 ; let (tv_dicts, others) = partition ok irreds
3010 (tidy_env, tidy_insts) = tidyInsts others
3011 ; reportNoInstances tidy_env Nothing [alt_fix] tidy_insts
3012 -- See Note [Exotic derived instance contexts] in TcMType
3014 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
3015 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
3016 -- This reverse-mapping is a pain, but the result
3017 -- should mention the original TyVars not TcTyVars
3019 ; return simpl_theta }
3021 doc = ptext (sLit "deriving classes for a data type")
3023 ok dict | isDict dict = validDerivPred (dictPred dict)
3025 alt_fix = vcat [ptext (sLit "use a standalone 'deriving instance' declaration instead,"),
3026 ptext (sLit "so you can specify the instance context yourself")]
3030 @tcSimplifyDefault@ just checks class-type constraints, essentially;
3031 used with \tr{default} declarations. We are only interested in
3032 whether it worked or not.
3035 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
3038 tcSimplifyDefault theta = do
3039 wanteds <- newDictBndrsO DefaultOrigin theta
3040 (irreds, _) <- tryHardCheckLoop doc wanteds
3041 addNoInstanceErrs irreds
3045 traceTc (ptext (sLit "tcSimplifyDefault failing")) >> failM
3047 doc = ptext (sLit "default declaration")
3050 @tcSimplifyStagedExpr@ performs a simplification but does so at a new
3051 stage. This is used when typechecking annotations and splices.
3055 tcSimplifyStagedExpr :: ThStage -> TcM a -> TcM (a, TcDictBinds)
3056 -- Type check an expression that runs at a top level stage as if
3057 -- it were going to be spliced and then simplify it
3058 tcSimplifyStagedExpr stage tc_action
3059 = setStage stage $ do {
3060 -- Typecheck the expression
3061 (thing', lie) <- getLIE tc_action
3063 -- Solve the constraints
3064 ; const_binds <- tcSimplifyTop lie
3066 ; return (thing', const_binds) }
3071 %************************************************************************
3073 \section{Errors and contexts}
3075 %************************************************************************
3077 ToDo: for these error messages, should we note the location as coming
3078 from the insts, or just whatever seems to be around in the monad just
3082 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
3083 -> [Inst] -- The offending Insts
3085 -- Group together insts with the same origin
3086 -- We want to report them together in error messages
3090 groupErrs report_err (inst:insts)
3091 = do { do_one (inst:friends)
3092 ; groupErrs report_err others }
3094 -- (It may seem a bit crude to compare the error messages,
3095 -- but it makes sure that we combine just what the user sees,
3096 -- and it avoids need equality on InstLocs.)
3097 (friends, others) = partition is_friend insts
3098 loc_msg = showSDoc (pprInstLoc (instLoc inst))
3099 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
3100 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
3101 -- Add location and context information derived from the Insts
3103 -- Add the "arising from..." part to a message about bunch of dicts
3104 addInstLoc :: [Inst] -> Message -> Message
3105 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
3107 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
3110 addTopIPErrs bndrs ips
3111 = do { dflags <- getDOpts
3112 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
3114 (tidy_env, tidy_ips) = tidyInsts ips
3116 = vcat [sep [ptext (sLit "Implicit parameters escape from"),
3117 nest 2 (ptext (sLit "the monomorphic top-level binding")
3118 <> plural bndrs <+> ptext (sLit "of")
3119 <+> pprBinders bndrs <> colon)],
3120 nest 2 (vcat (map ppr_ip ips)),
3121 monomorphism_fix dflags]
3122 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
3124 topIPErrs :: [Inst] -> TcM ()
3126 = groupErrs report tidy_dicts
3128 (tidy_env, tidy_dicts) = tidyInsts dicts
3129 report dicts = addErrTcM (tidy_env, mk_msg dicts)
3130 mk_msg dicts = addInstLoc dicts (ptext (sLit "Unbound implicit parameter") <>
3131 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
3133 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
3135 addNoInstanceErrs insts
3136 = do { let (tidy_env, tidy_insts) = tidyInsts insts
3137 ; reportNoInstances tidy_env Nothing [] tidy_insts }
3141 -> Maybe (InstLoc, [Inst]) -- Context
3142 -- Nothing => top level
3143 -- Just (d,g) => d describes the construct
3145 -> [SDoc] -- Alternative fix for no-such-instance
3146 -> [Inst] -- What is wanted (can include implications)
3149 reportNoInstances tidy_env mb_what alt_fix insts
3150 = groupErrs (report_no_instances tidy_env mb_what alt_fix) insts
3152 report_no_instances :: TidyEnv -> Maybe (InstLoc, [Inst]) -> [SDoc] -> [Inst] -> TcM ()
3153 report_no_instances tidy_env mb_what alt_fixes insts
3154 = do { inst_envs <- tcGetInstEnvs
3155 ; let (implics, insts1) = partition isImplicInst insts
3156 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
3157 (eqInsts, insts3) = partition isEqInst insts2
3158 ; traceTc (text "reportNoInstances" <+> vcat
3159 [ppr insts, ppr implics, ppr insts1, ppr insts2])
3160 ; mapM_ complain_implic implics
3161 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
3162 ; groupErrs complain_no_inst insts3
3163 ; mapM_ (addErrTcM . mk_eq_err) eqInsts
3166 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
3168 complain_implic inst -- Recurse!
3169 = reportNoInstances tidy_env
3170 (Just (tci_loc inst, tci_given inst))
3171 alt_fixes (tci_wanted inst)
3173 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
3174 -- Right msg => overlap message
3175 -- Left inst => no instance
3176 check_overlap inst_envs wanted
3177 | not (isClassDict wanted) = Left wanted
3179 = case lookupInstEnv inst_envs clas tys of
3180 ([], _) -> Left wanted -- No match
3181 -- The case of exactly one match and no unifiers means a
3182 -- successful lookup. That can't happen here, because dicts
3183 -- only end up here if they didn't match in Inst.lookupInst
3185 | debugIsOn -> pprPanic "reportNoInstance" (ppr wanted)
3186 res -> Right (mk_overlap_msg wanted res)
3188 (clas,tys) = getDictClassTys wanted
3190 mk_overlap_msg dict (matches, unifiers)
3191 = ASSERT( not (null matches) )
3192 vcat [ addInstLoc [dict] ((ptext (sLit "Overlapping instances for")
3193 <+> pprPred (dictPred dict))),
3194 sep [ptext (sLit "Matching instances") <> colon,
3195 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
3196 if not (isSingleton matches)
3197 then -- Two or more matches
3199 else -- One match, plus some unifiers
3200 ASSERT( not (null unifiers) )
3201 parens (vcat [ptext (sLit "The choice depends on the instantiation of") <+>
3202 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
3203 ptext (sLit "To pick the first instance above, use -XIncoherentInstances"),
3204 ptext (sLit "when compiling the other instance declarations")])]
3206 ispecs = [ispec | (ispec, _) <- matches]
3208 mk_eq_err :: Inst -> (TidyEnv, SDoc)
3209 mk_eq_err inst = misMatchMsg tidy_env (eqInstTys inst)
3211 mk_no_inst_err insts
3212 | null insts = empty
3214 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
3215 not (isEmptyVarSet (tyVarsOfInsts insts))
3216 = vcat [ addInstLoc insts $
3217 sep [ ptext (sLit "Could not deduce") <+> pprDictsTheta insts
3218 , nest 2 $ ptext (sLit "from the context") <+> pprDictsTheta givens]
3219 , show_fixes (fix1 loc : fixes2 ++ alt_fixes) ]
3221 | otherwise -- Top level
3222 = vcat [ addInstLoc insts $
3223 ptext (sLit "No instance") <> plural insts
3224 <+> ptext (sLit "for") <+> pprDictsTheta insts
3225 , show_fixes (fixes2 ++ alt_fixes) ]
3228 fix1 loc = sep [ ptext (sLit "add") <+> pprDictsTheta insts
3229 <+> ptext (sLit "to the context of"),
3230 nest 2 (ppr (instLocOrigin loc)) ]
3231 -- I'm not sure it helps to add the location
3232 -- nest 2 (ptext (sLit "at") <+> ppr (instLocSpan loc)) ]
3234 fixes2 | null instance_dicts = []
3235 | otherwise = [sep [ptext (sLit "add an instance declaration for"),
3236 pprDictsTheta instance_dicts]]
3237 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
3238 -- Insts for which it is worth suggesting an adding an instance declaration
3239 -- Exclude implicit parameters, and tyvar dicts
3241 show_fixes :: [SDoc] -> SDoc
3242 show_fixes [] = empty
3243 show_fixes (f:fs) = sep [ptext (sLit "Possible fix:"),
3244 nest 2 (vcat (f : map (ptext (sLit "or") <+>) fs))]
3246 addTopAmbigErrs :: [Inst] -> TcRn ()
3247 addTopAmbigErrs dicts
3248 -- Divide into groups that share a common set of ambiguous tyvars
3249 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
3250 -- See Note [Avoiding spurious errors]
3251 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
3253 (tidy_env, tidy_dicts) = tidyInsts dicts
3255 tvs_of :: Inst -> [TcTyVar]
3256 tvs_of d = varSetElems (tyVarsOfInst d)
3257 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
3259 report :: [(Inst,[TcTyVar])] -> TcM ()
3260 report pairs@((inst,tvs) : _) = do -- The pairs share a common set of ambiguous tyvars
3261 (tidy_env, mono_msg) <- mkMonomorphismMsg tidy_env tvs
3262 setSrcSpan (instSpan inst) $
3263 -- the location of the first one will do for the err message
3264 addErrTcM (tidy_env, msg $$ mono_msg)
3266 dicts = map fst pairs
3267 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
3268 pprQuotedList tvs <+> in_msg,
3269 nest 2 (pprDictsInFull dicts)]
3270 in_msg = text "in the constraint" <> plural dicts <> colon
3271 report [] = panic "addTopAmbigErrs"
3274 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
3275 -- There's an error with these Insts; if they have free type variables
3276 -- it's probably caused by the monomorphism restriction.
3277 -- Try to identify the offending variable
3278 -- ASSUMPTION: the Insts are fully zonked
3279 mkMonomorphismMsg tidy_env inst_tvs
3280 = do { dflags <- getDOpts
3281 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
3282 ; return (tidy_env, mk_msg dflags docs) }
3284 mk_msg _ _ | any isRuntimeUnk inst_tvs
3285 = vcat [ptext (sLit "Cannot resolve unknown runtime types:") <+>
3286 (pprWithCommas ppr inst_tvs),
3287 ptext (sLit "Use :print or :force to determine these types")]
3288 mk_msg _ [] = ptext (sLit "Probable fix: add a type signature that fixes these type variable(s)")
3289 -- This happens in things like
3290 -- f x = show (read "foo")
3291 -- where monomorphism doesn't play any role
3293 = vcat [ptext (sLit "Possible cause: the monomorphism restriction applied to the following:"),
3295 monomorphism_fix dflags]
3297 monomorphism_fix :: DynFlags -> SDoc
3298 monomorphism_fix dflags
3299 = ptext (sLit "Probable fix:") <+> vcat
3300 [ptext (sLit "give these definition(s) an explicit type signature"),
3301 if dopt Opt_MonomorphismRestriction dflags
3302 then ptext (sLit "or use -XNoMonomorphismRestriction")
3303 else empty] -- Only suggest adding "-XNoMonomorphismRestriction"
3304 -- if it is not already set!
3306 warnDefault :: [(Inst, Class, Var)] -> Type -> TcM ()
3307 warnDefault ups default_ty = do
3308 warn_flag <- doptM Opt_WarnTypeDefaults
3309 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
3311 dicts = [d | (d,_,_) <- ups]
3314 (_, tidy_dicts) = tidyInsts dicts
3315 warn_msg = vcat [ptext (sLit "Defaulting the following constraint(s) to type") <+>
3316 quotes (ppr default_ty),
3317 pprDictsInFull tidy_dicts]
3319 reduceDepthErr :: Int -> [Inst] -> SDoc
3320 reduceDepthErr n stack
3321 = vcat [ptext (sLit "Context reduction stack overflow; size =") <+> int n,
3322 ptext (sLit "Use -fcontext-stack=N to increase stack size to N"),
3323 nest 4 (pprStack stack)]
3325 pprStack :: [Inst] -> SDoc
3326 pprStack stack = vcat (map pprInstInFull stack)