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,
23 #include "HsVersions.h"
25 import {-# SOURCE #-} TcUnify( unifyType )
29 import TcHsSyn ( hsLPatType )
37 import DsUtils -- Big-tuple functions
66 %************************************************************************
70 %************************************************************************
72 --------------------------------------
73 Notes on functional dependencies (a bug)
74 --------------------------------------
81 instance D a b => C a b -- Undecidable
82 -- (Not sure if it's crucial to this eg)
83 f :: C a b => a -> Bool
86 g :: C a b => a -> Bool
89 Here f typechecks, but g does not!! Reason: before doing improvement,
90 we reduce the (C a b1) constraint from the call of f to (D a b1).
92 Here is a more complicated example:
95 > class Foo a b | a->b
97 > class Bar a b | a->b
101 > instance Bar Obj Obj
103 > instance (Bar a b) => Foo a b
105 > foo:: (Foo a b) => a -> String
108 > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
114 Could not deduce (Bar a b) from the context (Foo a b)
115 arising from use of `foo' at <interactive>:1
117 Add (Bar a b) to the expected type of an expression
118 In the first argument of `runFoo', namely `foo'
119 In the definition of `it': it = runFoo foo
121 Why all of the sudden does GHC need the constraint Bar a b? The
122 function foo didn't ask for that...
125 The trouble is that to type (runFoo foo), GHC has to solve the problem:
127 Given constraint Foo a b
128 Solve constraint Foo a b'
130 Notice that b and b' aren't the same. To solve this, just do
131 improvement and then they are the same. But GHC currently does
136 That is usually fine, but it isn't here, because it sees that Foo a b is
137 not the same as Foo a b', and so instead applies the instance decl for
138 instance Bar a b => Foo a b. And that's where the Bar constraint comes
141 The Right Thing is to improve whenever the constraint set changes at
142 all. Not hard in principle, but it'll take a bit of fiddling to do.
144 Note [Choosing which variables to quantify]
145 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
146 Suppose we are about to do a generalisation step. We have in our hand
149 T the type of the RHS
150 C the constraints from that RHS
152 The game is to figure out
154 Q the set of type variables over which to quantify
155 Ct the constraints we will *not* quantify over
156 Cq the constraints we will quantify over
158 So we're going to infer the type
162 and float the constraints Ct further outwards.
164 Here are the things that *must* be true:
166 (A) Q intersect fv(G) = EMPTY limits how big Q can be
167 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
169 (A) says we can't quantify over a variable that's free in the environment.
170 (B) says we must quantify over all the truly free variables in T, else
171 we won't get a sufficiently general type.
173 We do not *need* to quantify over any variable that is fixed by the
174 free vars of the environment G.
176 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
178 Example: class H x y | x->y where ...
180 fv(G) = {a} C = {H a b, H c d}
183 (A) Q intersect {a} is empty
184 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
186 So Q can be {c,d}, {b,c,d}
188 In particular, it's perfectly OK to quantify over more type variables
189 than strictly necessary; there is no need to quantify over 'b', since
190 it is determined by 'a' which is free in the envt, but it's perfectly
191 OK to do so. However we must not quantify over 'a' itself.
193 Other things being equal, however, we'd like to quantify over as few
194 variables as possible: smaller types, fewer type applications, more
195 constraints can get into Ct instead of Cq. Here's a good way to
198 Q = grow( fv(T), C ) \ oclose( fv(G), C )
200 That is, quantify over all variable that that MIGHT be fixed by the
201 call site (which influences T), but which aren't DEFINITELY fixed by
202 G. This choice definitely quantifies over enough type variables,
203 albeit perhaps too many.
205 Why grow( fv(T), C ) rather than fv(T)? Consider
207 class H x y | x->y where ...
212 If we used fv(T) = {c} we'd get the type
214 forall c. H c d => c -> b
216 And then if the fn was called at several different c's, each of
217 which fixed d differently, we'd get a unification error, because
218 d isn't quantified. Solution: quantify d. So we must quantify
219 everything that might be influenced by c.
221 Why not oclose( fv(T), C )? Because we might not be able to see
222 all the functional dependencies yet:
224 class H x y | x->y where ...
225 instance H x y => Eq (T x y) where ...
230 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
231 apparent yet, and that's wrong. We must really quantify over d too.
233 There really isn't any point in quantifying over any more than
234 grow( fv(T), C ), because the call sites can't possibly influence
235 any other type variables.
239 -------------------------------------
241 -------------------------------------
243 It's very hard to be certain when a type is ambiguous. Consider
247 instance H x y => K (x,y)
249 Is this type ambiguous?
250 forall a b. (K (a,b), Eq b) => a -> a
252 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
253 now we see that a fixes b. So we can't tell about ambiguity for sure
254 without doing a full simplification. And even that isn't possible if
255 the context has some free vars that may get unified. Urgle!
257 Here's another example: is this ambiguous?
258 forall a b. Eq (T b) => a -> a
259 Not if there's an insance decl (with no context)
260 instance Eq (T b) where ...
262 You may say of this example that we should use the instance decl right
263 away, but you can't always do that:
265 class J a b where ...
266 instance J Int b where ...
268 f :: forall a b. J a b => a -> a
270 (Notice: no functional dependency in J's class decl.)
271 Here f's type is perfectly fine, provided f is only called at Int.
272 It's premature to complain when meeting f's signature, or even
273 when inferring a type for f.
277 However, we don't *need* to report ambiguity right away. It'll always
278 show up at the call site.... and eventually at main, which needs special
279 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
281 So here's the plan. We WARN about probable ambiguity if
283 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
285 (all tested before quantification).
286 That is, all the type variables in Cq must be fixed by the the variables
287 in the environment, or by the variables in the type.
289 Notice that we union before calling oclose. Here's an example:
291 class J a b c | a b -> c
295 forall b c. (J a b c) => b -> b
297 Only if we union {a} from G with {b} from T before using oclose,
298 do we see that c is fixed.
300 It's a bit vague exactly which C we should use for this oclose call. If we
301 don't fix enough variables we might complain when we shouldn't (see
302 the above nasty example). Nothing will be perfect. That's why we can
303 only issue a warning.
306 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
308 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
310 then c is a "bubble"; there's no way it can ever improve, and it's
311 certainly ambiguous. UNLESS it is a constant (sigh). And what about
316 instance H x y => K (x,y)
318 Is this type ambiguous?
319 forall a b. (K (a,b), Eq b) => a -> a
321 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
322 is a "bubble" that's a set of constraints
324 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
326 Hence another idea. To decide Q start with fv(T) and grow it
327 by transitive closure in Cq (no functional dependencies involved).
328 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
329 The definitely-ambiguous can then float out, and get smashed at top level
330 (which squashes out the constants, like Eq (T a) above)
333 --------------------------------------
334 Notes on principal types
335 --------------------------------------
340 f x = let g y = op (y::Int) in True
342 Here the principal type of f is (forall a. a->a)
343 but we'll produce the non-principal type
344 f :: forall a. C Int => a -> a
347 --------------------------------------
348 The need for forall's in constraints
349 --------------------------------------
351 [Exchange on Haskell Cafe 5/6 Dec 2000]
353 class C t where op :: t -> Bool
354 instance C [t] where op x = True
356 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
357 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
359 The definitions of p and q differ only in the order of the components in
360 the pair on their right-hand sides. And yet:
362 ghc and "Typing Haskell in Haskell" reject p, but accept q;
363 Hugs rejects q, but accepts p;
364 hbc rejects both p and q;
365 nhc98 ... (Malcolm, can you fill in the blank for us!).
367 The type signature for f forces context reduction to take place, and
368 the results of this depend on whether or not the type of y is known,
369 which in turn depends on which component of the pair the type checker
372 Solution: if y::m a, float out the constraints
373 Monad m, forall c. C (m c)
374 When m is later unified with [], we can solve both constraints.
377 --------------------------------------
378 Notes on implicit parameters
379 --------------------------------------
381 Note [Inheriting implicit parameters]
382 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
387 where f is *not* a top-level binding.
388 From the RHS of f we'll get the constraint (?y::Int).
389 There are two types we might infer for f:
393 (so we get ?y from the context of f's definition), or
395 f :: (?y::Int) => Int -> Int
397 At first you might think the first was better, becuase then
398 ?y behaves like a free variable of the definition, rather than
399 having to be passed at each call site. But of course, the WHOLE
400 IDEA is that ?y should be passed at each call site (that's what
401 dynamic binding means) so we'd better infer the second.
403 BOTTOM LINE: when *inferring types* you *must* quantify
404 over implicit parameters. See the predicate isFreeWhenInferring.
407 Note [Implicit parameters and ambiguity]
408 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
409 Only a *class* predicate can give rise to ambiguity
410 An *implicit parameter* cannot. For example:
411 foo :: (?x :: [a]) => Int
413 is fine. The call site will suppply a particular 'x'
415 Furthermore, the type variables fixed by an implicit parameter
416 propagate to the others. E.g.
417 foo :: (Show a, ?x::[a]) => Int
419 The type of foo looks ambiguous. But it isn't, because at a call site
421 let ?x = 5::Int in foo
422 and all is well. In effect, implicit parameters are, well, parameters,
423 so we can take their type variables into account as part of the
424 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
427 Question 2: type signatures
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
429 BUT WATCH OUT: When you supply a type signature, we can't force you
430 to quantify over implicit parameters. For example:
434 This is perfectly reasonable. We do not want to insist on
436 (?x + 1) :: (?x::Int => Int)
438 That would be silly. Here, the definition site *is* the occurrence site,
439 so the above strictures don't apply. Hence the difference between
440 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
441 and tcSimplifyCheckBind (which does not).
443 What about when you supply a type signature for a binding?
444 Is it legal to give the following explicit, user type
445 signature to f, thus:
450 At first sight this seems reasonable, but it has the nasty property
451 that adding a type signature changes the dynamic semantics.
454 (let f x = (x::Int) + ?y
455 in (f 3, f 3 with ?y=5)) with ?y = 6
461 in (f 3, f 3 with ?y=5)) with ?y = 6
465 Indeed, simply inlining f (at the Haskell source level) would change the
468 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
469 semantics for a Haskell program without knowing its typing, so if you
470 change the typing you may change the semantics.
472 To make things consistent in all cases where we are *checking* against
473 a supplied signature (as opposed to inferring a type), we adopt the
476 a signature does not need to quantify over implicit params.
478 [This represents a (rather marginal) change of policy since GHC 5.02,
479 which *required* an explicit signature to quantify over all implicit
480 params for the reasons mentioned above.]
482 But that raises a new question. Consider
484 Given (signature) ?x::Int
485 Wanted (inferred) ?x::Int, ?y::Bool
487 Clearly we want to discharge the ?x and float the ?y out. But
488 what is the criterion that distinguishes them? Clearly it isn't
489 what free type variables they have. The Right Thing seems to be
490 to float a constraint that
491 neither mentions any of the quantified type variables
492 nor any of the quantified implicit parameters
494 See the predicate isFreeWhenChecking.
497 Question 3: monomorphism
498 ~~~~~~~~~~~~~~~~~~~~~~~~
499 There's a nasty corner case when the monomorphism restriction bites:
503 The argument above suggests that we *must* generalise
504 over the ?y parameter, to get
505 z :: (?y::Int) => Int,
506 but the monomorphism restriction says that we *must not*, giving
508 Why does the momomorphism restriction say this? Because if you have
510 let z = x + ?y in z+z
512 you might not expect the addition to be done twice --- but it will if
513 we follow the argument of Question 2 and generalise over ?y.
516 Question 4: top level
517 ~~~~~~~~~~~~~~~~~~~~~
518 At the top level, monomorhism makes no sense at all.
521 main = let ?x = 5 in print foo
525 woggle :: (?x :: Int) => Int -> Int
528 We definitely don't want (foo :: Int) with a top-level implicit parameter
529 (?x::Int) becuase there is no way to bind it.
534 (A) Always generalise over implicit parameters
535 Bindings that fall under the monomorphism restriction can't
539 * Inlining remains valid
540 * No unexpected loss of sharing
541 * But simple bindings like
543 will be rejected, unless you add an explicit type signature
544 (to avoid the monomorphism restriction)
545 z :: (?y::Int) => Int
547 This seems unacceptable
549 (B) Monomorphism restriction "wins"
550 Bindings that fall under the monomorphism restriction can't
552 Always generalise over implicit parameters *except* for bindings
553 that fall under the monomorphism restriction
556 * Inlining isn't valid in general
557 * No unexpected loss of sharing
558 * Simple bindings like
560 accepted (get value of ?y from binding site)
562 (C) Always generalise over implicit parameters
563 Bindings that fall under the monomorphism restriction can't
564 be generalised, EXCEPT for implicit parameters
566 * Inlining remains valid
567 * Unexpected loss of sharing (from the extra generalisation)
568 * Simple bindings like
570 accepted (get value of ?y from occurrence sites)
575 None of these choices seems very satisfactory. But at least we should
576 decide which we want to do.
578 It's really not clear what is the Right Thing To Do. If you see
582 would you expect the value of ?y to be got from the *occurrence sites*
583 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
584 case of function definitions, the answer is clearly the former, but
585 less so in the case of non-fucntion definitions. On the other hand,
586 if we say that we get the value of ?y from the definition site of 'z',
587 then inlining 'z' might change the semantics of the program.
589 Choice (C) really says "the monomorphism restriction doesn't apply
590 to implicit parameters". Which is fine, but remember that every
591 innocent binding 'x = ...' that mentions an implicit parameter in
592 the RHS becomes a *function* of that parameter, called at each
593 use of 'x'. Now, the chances are that there are no intervening 'with'
594 clauses that bind ?y, so a decent compiler should common up all
595 those function calls. So I think I strongly favour (C). Indeed,
596 one could make a similar argument for abolishing the monomorphism
597 restriction altogether.
599 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
603 %************************************************************************
605 \subsection{tcSimplifyInfer}
607 %************************************************************************
609 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
611 1. Compute Q = grow( fvs(T), C )
613 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
614 predicates will end up in Ct; we deal with them at the top level
616 3. Try improvement, using functional dependencies
618 4. If Step 3 did any unification, repeat from step 1
619 (Unification can change the result of 'grow'.)
621 Note: we don't reduce dictionaries in step 2. For example, if we have
622 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
623 after step 2. However note that we may therefore quantify over more
624 type variables than we absolutely have to.
626 For the guts, we need a loop, that alternates context reduction and
627 improvement with unification. E.g. Suppose we have
629 class C x y | x->y where ...
631 and tcSimplify is called with:
633 Then improvement unifies a with b, giving
636 If we need to unify anything, we rattle round the whole thing all over
643 -> TcTyVarSet -- fv(T); type vars
645 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
646 [Inst], -- Dict Ids that must be bound here (zonked)
647 TcDictBinds) -- Bindings
648 -- Any free (escaping) Insts are tossed into the environment
653 tcSimplifyInfer doc tau_tvs wanted
654 = do { tau_tvs1 <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
655 ; wanted' <- mapM zonkInst wanted -- Zonk before deciding quantified tyvars
656 ; gbl_tvs <- tcGetGlobalTyVars
657 ; let preds1 = fdPredsOfInsts wanted'
658 gbl_tvs1 = oclose preds1 gbl_tvs
659 qtvs = grow preds1 tau_tvs1 `minusVarSet` gbl_tvs1
660 -- See Note [Choosing which variables to quantify]
662 -- To maximise sharing, remove from consideration any
663 -- constraints that don't mention qtvs at all
664 ; let (free, bound) = partition (isFreeWhenInferring qtvs) wanted'
667 -- To make types simple, reduce as much as possible
668 ; traceTc (text "infer" <+> (ppr preds1 $$ ppr (grow preds1 tau_tvs1) $$ ppr gbl_tvs $$
669 ppr gbl_tvs1 $$ ppr free $$ ppr bound))
670 ; (irreds1, binds1) <- tryHardCheckLoop doc bound
672 -- Note [Inference and implication constraints]
673 ; let want_dict d = tyVarsOfInst d `intersectsVarSet` qtvs
674 ; (irreds2, binds2) <- approximateImplications doc want_dict irreds1
676 -- Now work out all over again which type variables to quantify,
677 -- exactly in the same way as before, but starting from irreds2. Why?
678 -- a) By now improvment may have taken place, and we must *not*
679 -- quantify over any variable free in the environment
680 -- tc137 (function h inside g) is an example
682 -- b) Do not quantify over constraints that *now* do not
683 -- mention quantified type variables, because they are
684 -- simply ambiguous (or might be bound further out). Example:
685 -- f :: Eq b => a -> (a, b)
687 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
688 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
689 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
690 -- constraint (Eq beta), which we dump back into the free set
691 -- See test tcfail181
693 -- c) irreds may contain type variables not previously mentioned,
694 -- e.g. instance D a x => Foo [a]
696 -- Then after simplifying we'll get (D a x), and x is fresh
697 -- We must quantify over x else it'll be totally unbound
698 ; tau_tvs2 <- zonkTcTyVarsAndFV (varSetElems tau_tvs1)
699 ; gbl_tvs2 <- zonkTcTyVarsAndFV (varSetElems gbl_tvs1)
700 -- Note that we start from gbl_tvs1
701 -- We use tcGetGlobalTyVars, then oclose wrt preds2, because
702 -- we've already put some of the original preds1 into frees
703 -- E.g. wanteds = C a b (where a->b)
706 -- Then b is fixed by gbl_tvs, so (C a b) will be in free, and
707 -- irreds2 will be empty. But we don't want to generalise over b!
708 ; let preds2 = fdPredsOfInsts irreds2 -- irreds2 is zonked
709 qtvs = grow preds2 tau_tvs2 `minusVarSet` oclose preds2 gbl_tvs2
710 ; let (free, irreds3) = partition (isFreeWhenInferring qtvs) irreds2
713 -- Turn the quantified meta-type variables into real type variables
714 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems qtvs)
716 -- We can't abstract over any remaining unsolved
717 -- implications so instead just float them outwards. Ugh.
718 ; let (q_dicts0, implics) = partition isAbstractableInst irreds3
719 ; loc <- getInstLoc (ImplicOrigin doc)
720 ; implic_bind <- bindIrreds loc qtvs2 q_dicts0 implics
722 -- Prepare equality instances for quantification
723 ; let (q_eqs0,q_dicts) = partition isEqInst q_dicts0
724 ; q_eqs <- mapM finalizeEqInst q_eqs0
726 ; return (qtvs2, q_eqs ++ q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
727 -- NB: when we are done, we might have some bindings, but
728 -- the final qtvs might be empty. See Note [NO TYVARS] below.
730 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
731 -- Note [Inference and implication constraints]
732 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
733 -- - fetching any dicts inside them that are free
734 -- - using those dicts as cruder constraints, to solve the implications
735 -- - returning the extra ones too
737 approximateImplications doc want_dict irreds
739 = return (irreds, emptyBag)
741 = do { extra_dicts' <- mapM cloneDict extra_dicts
742 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
743 -- By adding extra_dicts', we make them
744 -- available to solve the implication constraints
746 extra_dicts = get_dicts (filter isImplicInst irreds)
748 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
749 -- Find the wanted constraints in implication constraints that satisfy
750 -- want_dict, and are not bound by forall's in the constraint itself
751 get_dicts ds = concatMap get_dict ds
753 get_dict d@(Dict {}) | want_dict d = [d]
755 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
756 = [ d | let tv_set = mkVarSet tvs
757 , d <- get_dicts wanteds
758 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
759 get_dict i@(EqInst {}) | want_dict i = [i]
761 get_dict other = pprPanic "approximateImplications" (ppr other)
764 Note [Inference and implication constraints]
765 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
766 Suppose we have a wanted implication constraint (perhaps arising from
767 a nested pattern match) like
769 and we are now trying to quantify over 'a' when inferring the type for
770 a function. In principle it's possible that there might be an instance
771 instance (C a, E a) => D [a]
772 so the context (E a) would suffice. The Right Thing is to abstract over
773 the implication constraint, but we don't do that (a) because it'll be
774 surprising to programmers and (b) because we don't have the machinery to deal
775 with 'given' implications.
777 So our best approximation is to make (D [a]) part of the inferred
778 context, so we can use that to discharge the implication. Hence
779 the strange function get_dicts in approximateImplications.
781 The common cases are more clear-cut, when we have things like
783 Here, abstracting over (C b) is not an approximation at all -- but see
784 Note [Freeness and implications].
786 See Trac #1430 and test tc228.
790 -----------------------------------------------------------
791 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
792 -- against, but we don't know the type variables over which we are going to quantify.
793 -- This happens when we have a type signature for a mutually recursive group
796 -> TcTyVarSet -- fv(T)
799 -> TcM ([TyVar], -- Fully zonked, and quantified
800 TcDictBinds) -- Bindings
802 tcSimplifyInferCheck loc tau_tvs givens wanteds
803 = do { traceTc (text "tcSimplifyInferCheck <-" <+> ppr wanteds)
804 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
806 -- Figure out which type variables to quantify over
807 -- You might think it should just be the signature tyvars,
808 -- but in bizarre cases you can get extra ones
809 -- f :: forall a. Num a => a -> a
810 -- f x = fst (g (x, head [])) + 1
812 -- Here we infer g :: forall a b. a -> b -> (b,a)
813 -- We don't want g to be monomorphic in b just because
814 -- f isn't quantified over b.
815 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
816 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
817 ; gbl_tvs <- tcGetGlobalTyVars
818 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
819 -- We could close gbl_tvs, but its not necessary for
820 -- soundness, and it'll only affect which tyvars, not which
821 -- dictionaries, we quantify over
823 ; qtvs' <- zonkQuantifiedTyVars qtvs
825 -- Now we are back to normal (c.f. tcSimplCheck)
826 ; implic_bind <- bindIrreds loc qtvs' givens irreds
828 ; traceTc (text "tcSimplifyInferCheck ->" <+> ppr (implic_bind))
829 ; return (qtvs', binds `unionBags` implic_bind) }
832 Note [Squashing methods]
833 ~~~~~~~~~~~~~~~~~~~~~~~~~
834 Be careful if you want to float methods more:
835 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
836 From an application (truncate f i) we get
839 If we have also have a second occurrence of truncate, we get
842 When simplifying with i,f free, we might still notice that
843 t1=t3; but alas, the binding for t2 (which mentions t1)
844 may continue to float out!
849 class Y a b | a -> b where
852 instance Y [[a]] a where
855 k :: X a -> X a -> X a
857 g :: Num a => [X a] -> [X a]
860 h ys = ys ++ map (k (y [[0]])) xs
862 The excitement comes when simplifying the bindings for h. Initially
863 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
864 From this we get t1:=:t2, but also various bindings. We can't forget
865 the bindings (because of [LOOP]), but in fact t1 is what g is
868 The net effect of [NO TYVARS]
871 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
872 isFreeWhenInferring qtvs inst
873 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
874 && isInheritableInst inst -- and no implicit parameter involved
875 -- see Note [Inheriting implicit parameters]
877 {- No longer used (with implication constraints)
878 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
879 -> NameSet -- Quantified implicit parameters
881 isFreeWhenChecking qtvs ips inst
882 = isFreeWrtTyVars qtvs inst
883 && isFreeWrtIPs ips inst
886 isFreeWrtTyVars :: VarSet -> Inst -> Bool
887 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
888 isFreeWrtIPs :: NameSet -> Inst -> Bool
889 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
893 %************************************************************************
895 \subsection{tcSimplifyCheck}
897 %************************************************************************
899 @tcSimplifyCheck@ is used when we know exactly the set of variables
900 we are going to quantify over. For example, a class or instance declaration.
903 -----------------------------------------------------------
904 -- tcSimplifyCheck is used when checking expression type signatures,
905 -- class decls, instance decls etc.
906 tcSimplifyCheck :: InstLoc
907 -> [TcTyVar] -- Quantify over these
910 -> TcM TcDictBinds -- Bindings
911 tcSimplifyCheck loc qtvs givens wanteds
912 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
913 do { traceTc (text "tcSimplifyCheck")
914 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
915 ; implic_bind <- bindIrreds loc qtvs givens irreds
916 ; return (binds `unionBags` implic_bind) }
918 -----------------------------------------------------------
919 -- tcSimplifyCheckPat is used for existential pattern match
920 tcSimplifyCheckPat :: InstLoc
921 -> [TcTyVar] -- Quantify over these
924 -> TcM TcDictBinds -- Bindings
925 tcSimplifyCheckPat loc qtvs givens wanteds
926 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
927 do { traceTc (text "tcSimplifyCheckPat")
928 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
929 ; implic_bind <- bindIrredsR loc qtvs givens irreds
930 ; return (binds `unionBags` implic_bind) }
932 -----------------------------------------------------------
933 bindIrreds :: InstLoc -> [TcTyVar]
936 bindIrreds loc qtvs givens irreds
937 = bindIrredsR loc qtvs givens irreds
939 bindIrredsR :: InstLoc -> [TcTyVar] -> [Inst] -> [Inst] -> TcM TcDictBinds
940 -- Make a binding that binds 'irreds', by generating an implication
941 -- constraint for them, *and* throwing the constraint into the LIE
942 bindIrredsR loc qtvs givens irreds
946 = do { let givens' = filter isAbstractableInst givens
947 -- The givens can (redundantly) include methods
948 -- We want to retain both EqInsts and Dicts
949 -- There should be no implicadtion constraints
950 -- See Note [Pruning the givens in an implication constraint]
952 -- If there are no 'givens', then it's safe to
953 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
954 -- See Note [Freeness and implications]
955 ; irreds' <- if null givens'
957 { let qtv_set = mkVarSet qtvs
958 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
960 ; return real_irreds }
963 ; (implics, bind) <- makeImplicationBind loc qtvs givens' irreds'
964 -- This call does the real work
965 -- If irreds' is empty, it does something sensible
970 makeImplicationBind :: InstLoc -> [TcTyVar]
972 -> TcM ([Inst], TcDictBinds)
973 -- Make a binding that binds 'irreds', by generating an implication
974 -- constraint for them, *and* throwing the constraint into the LIE
975 -- The binding looks like
976 -- (ir1, .., irn) = f qtvs givens
977 -- where f is (evidence for) the new implication constraint
978 -- f :: forall qtvs. {reft} givens => (ir1, .., irn)
979 -- qtvs includes coercion variables
981 -- This binding must line up the 'rhs' in reduceImplication
982 makeImplicationBind loc all_tvs
983 givens -- Guaranteed all Dicts
986 | null irreds -- If there are no irreds, we are done
987 = return ([], emptyBag)
988 | otherwise -- Otherwise we must generate a binding
989 = do { uniq <- newUnique
990 ; span <- getSrcSpanM
991 ; let (eq_givens, dict_givens) = partition isEqInst givens
992 eq_tyvar_cos = mkTyVarTys (varSetElems $ tyVarsOfTypes $ map eqInstType eq_givens)
993 -- Urgh! See line 2187 or thereabouts. I believe that all these
994 -- 'givens' must be a simple CoVar. This MUST be cleaned up.
996 ; let name = mkInternalName uniq (mkVarOcc "ic") span
997 implic_inst = ImplicInst { tci_name = name,
998 tci_tyvars = all_tvs,
999 tci_given = (eq_givens ++ dict_givens),
1000 tci_wanted = irreds, tci_loc = loc }
1001 ; let -- only create binder for dict_irreds
1002 (_, dict_irreds) = partition isEqInst irreds
1003 dict_irred_ids = map instToId dict_irreds
1004 lpat = mkBigLHsPatTup (map (L span . VarPat) dict_irred_ids)
1005 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
1006 co = mkWpApps (map instToId dict_givens)
1007 <.> mkWpTyApps eq_tyvar_cos
1008 <.> mkWpTyApps (mkTyVarTys all_tvs)
1009 bind | [dict_irred_id] <- dict_irred_ids = VarBind dict_irred_id rhs
1010 | otherwise = PatBind { pat_lhs = lpat,
1011 pat_rhs = unguardedGRHSs rhs,
1012 pat_rhs_ty = hsLPatType lpat,
1013 bind_fvs = placeHolderNames }
1014 ; traceTc $ text "makeImplicationBind" <+> ppr implic_inst
1015 ; return ([implic_inst], unitBag (L span bind))
1018 -----------------------------------------------------------
1019 tryHardCheckLoop :: SDoc
1021 -> TcM ([Inst], TcDictBinds)
1023 tryHardCheckLoop doc wanteds
1024 = do { (irreds,binds) <- checkLoop (mkInferRedEnv doc try_me) wanteds
1025 ; return (irreds,binds)
1029 -- Here's the try-hard bit
1031 -----------------------------------------------------------
1032 gentleCheckLoop :: InstLoc
1035 -> TcM ([Inst], TcDictBinds)
1037 gentleCheckLoop inst_loc givens wanteds
1038 = do { (irreds,binds) <- checkLoop env wanteds
1039 ; return (irreds,binds)
1042 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1044 try_me inst | isMethodOrLit inst = ReduceMe
1046 -- When checking against a given signature
1047 -- we MUST be very gentle: Note [Check gently]
1049 gentleInferLoop :: SDoc -> [Inst]
1050 -> TcM ([Inst], TcDictBinds)
1051 gentleInferLoop doc wanteds
1052 = do { (irreds, binds) <- checkLoop env wanteds
1053 ; return (irreds, binds) }
1055 env = mkInferRedEnv doc try_me
1056 try_me inst | isMethodOrLit inst = ReduceMe
1061 ~~~~~~~~~~~~~~~~~~~~
1062 We have to very careful about not simplifying too vigorously
1067 f :: Show b => T b -> b
1068 f (MkT x) = show [x]
1070 Inside the pattern match, which binds (a:*, x:a), we know that
1072 Hence we have a dictionary for Show [a] available; and indeed we
1073 need it. We are going to build an implication contraint
1074 forall a. (b~[a]) => Show [a]
1075 Later, we will solve this constraint using the knowledge (Show b)
1077 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1078 thing becomes insoluble. So we simplify gently (get rid of literals
1079 and methods only, plus common up equal things), deferring the real
1080 work until top level, when we solve the implication constraint
1081 with tryHardCheckLooop.
1085 -----------------------------------------------------------
1088 -> TcM ([Inst], TcDictBinds)
1089 -- Precondition: givens are completely rigid
1090 -- Postcondition: returned Insts are zonked
1092 checkLoop env wanteds
1094 where go env wanteds
1095 = do { -- We do need to zonk the givens; cf Note [Zonking RedEnv]
1096 ; env' <- zonkRedEnv env
1097 ; wanteds' <- zonkInsts wanteds
1099 ; (improved, binds, irreds) <- reduceContext env' wanteds'
1101 ; if null irreds || not improved then
1102 return (irreds, binds)
1105 -- If improvement did some unification, we go round again.
1106 -- We start again with irreds, not wanteds
1107 -- Using an instance decl might have introduced a fresh type
1108 -- variable which might have been unified, so we'd get an
1109 -- infinite loop if we started again with wanteds!
1111 { (irreds1, binds1) <- go env' irreds
1112 ; return (irreds1, binds `unionBags` binds1) } }
1115 Note [Zonking RedEnv]
1116 ~~~~~~~~~~~~~~~~~~~~~
1117 It might appear as if the givens in RedEnv are always rigid, but that is not
1118 necessarily the case for programs involving higher-rank types that have class
1119 contexts constraining the higher-rank variables. An example from tc237 in the
1122 class Modular s a | s -> a
1124 wim :: forall a w. Integral a
1125 => a -> (forall s. Modular s a => M s w) -> w
1126 wim i k = error "urk"
1128 test5 :: (Modular s a, Integral a) => M s a
1131 test4 = wim 4 test4'
1133 Notice how the variable 'a' of (Modular s a) in the rank-2 type of wim is
1134 quantified further outside. When type checking test4, we have to check
1135 whether the signature of test5 is an instance of
1137 (forall s. Modular s a => M s w)
1139 Consequently, we will get (Modular s t_a), where t_a is a TauTv into the
1142 Given the FD of Modular in this example, class improvement will instantiate
1143 t_a to 'a', where 'a' is the skolem from test5's signatures (due to the
1144 Modular s a predicate in that signature). If we don't zonk (Modular s t_a) in
1145 the givens, we will get into a loop as improveOne uses the unification engine
1146 Unify.tcUnifyTys, which doesn't know about mutable type variables.
1151 class If b t e r | b t e -> r
1154 class Lte a b c | a b -> c where lte :: a -> b -> c
1156 instance (Lte a b l,If l b a c) => Max a b c
1158 Wanted: Max Z (S x) y
1160 Then we'll reduce using the Max instance to:
1161 (Lte Z (S x) l, If l (S x) Z y)
1162 and improve by binding l->T, after which we can do some reduction
1163 on both the Lte and If constraints. What we *can't* do is start again
1164 with (Max Z (S x) y)!
1168 %************************************************************************
1170 tcSimplifySuperClasses
1172 %************************************************************************
1174 Note [SUPERCLASS-LOOP 1]
1175 ~~~~~~~~~~~~~~~~~~~~~~~~
1176 We have to be very, very careful when generating superclasses, lest we
1177 accidentally build a loop. Here's an example:
1181 class S a => C a where { opc :: a -> a }
1182 class S b => D b where { opd :: b -> b }
1184 instance C Int where
1187 instance D Int where
1190 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1191 Simplifying, we may well get:
1192 $dfCInt = :C ds1 (opd dd)
1195 Notice that we spot that we can extract ds1 from dd.
1197 Alas! Alack! We can do the same for (instance D Int):
1199 $dfDInt = :D ds2 (opc dc)
1203 And now we've defined the superclass in terms of itself.
1204 Two more nasty cases are in
1209 - Satisfy the superclass context *all by itself*
1210 (tcSimplifySuperClasses)
1211 - And do so completely; i.e. no left-over constraints
1212 to mix with the constraints arising from method declarations
1215 Note [Recursive instances and superclases]
1216 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1217 Consider this code, which arises in the context of "Scrap Your
1218 Boilerplate with Class".
1222 instance Sat (ctx Char) => Data ctx Char
1223 instance (Sat (ctx [a]), Data ctx a) => Data ctx [a]
1225 class Data Maybe a => Foo a
1227 instance Foo t => Sat (Maybe t)
1229 instance Data Maybe a => Foo a
1230 instance Foo a => Foo [a]
1233 In the instance for Foo [a], when generating evidence for the superclasses
1234 (ie in tcSimplifySuperClasses) we need a superclass (Data Maybe [a]).
1235 Using the instance for Data, we therefore need
1236 (Sat (Maybe [a], Data Maybe a)
1237 But we are given (Foo a), and hence its superclass (Data Maybe a).
1238 So that leaves (Sat (Maybe [a])). Using the instance for Sat means
1239 we need (Foo [a]). And that is the very dictionary we are bulding
1240 an instance for! So we must put that in the "givens". So in this
1242 Given: Foo a, Foo [a]
1243 Watend: Data Maybe [a]
1245 BUT we must *not not not* put the *superclasses* of (Foo [a]) in
1246 the givens, which is what 'addGiven' would normally do. Why? Because
1247 (Data Maybe [a]) is the superclass, so we'd "satisfy" the wanted
1248 by selecting a superclass from Foo [a], which simply makes a loop.
1250 On the other hand we *must* put the superclasses of (Foo a) in
1251 the givens, as you can see from the derivation described above.
1253 Conclusion: in the very special case of tcSimplifySuperClasses
1254 we have one 'given' (namely the "this" dictionary) whose superclasses
1255 must not be added to 'givens' by addGiven. That is the *whole* reason
1256 for the red_given_scs field in RedEnv, and the function argument to
1260 tcSimplifySuperClasses
1262 -> Inst -- The dict whose superclasses
1263 -- are being figured out
1267 tcSimplifySuperClasses loc this givens sc_wanteds
1268 = do { traceTc (text "tcSimplifySuperClasses")
1269 ; (irreds,binds1) <- checkLoop env sc_wanteds
1270 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1271 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1274 env = RedEnv { red_doc = pprInstLoc loc,
1275 red_try_me = try_me,
1276 red_givens = this:givens,
1277 red_given_scs = add_scs,
1279 red_improve = False } -- No unification vars
1280 add_scs g | g==this = NoSCs
1281 | otherwise = AddSCs
1283 try_me _ = ReduceMe -- Try hard, so we completely solve the superclass
1284 -- constraints right here. See Note [SUPERCLASS-LOOP 1]
1288 %************************************************************************
1290 \subsection{tcSimplifyRestricted}
1292 %************************************************************************
1294 tcSimplifyRestricted infers which type variables to quantify for a
1295 group of restricted bindings. This isn't trivial.
1298 We want to quantify over a to get id :: forall a. a->a
1301 We do not want to quantify over a, because there's an Eq a
1302 constraint, so we get eq :: a->a->Bool (notice no forall)
1305 RHS has type 'tau', whose free tyvars are tau_tvs
1306 RHS has constraints 'wanteds'
1309 Quantify over (tau_tvs \ ftvs(wanteds))
1310 This is bad. The constraints may contain (Monad (ST s))
1311 where we have instance Monad (ST s) where...
1312 so there's no need to be monomorphic in s!
1314 Also the constraint might be a method constraint,
1315 whose type mentions a perfectly innocent tyvar:
1316 op :: Num a => a -> b -> a
1317 Here, b is unconstrained. A good example would be
1319 We want to infer the polymorphic type
1320 foo :: forall b. b -> b
1323 Plan B (cunning, used for a long time up to and including GHC 6.2)
1324 Step 1: Simplify the constraints as much as possible (to deal
1325 with Plan A's problem). Then set
1326 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1328 Step 2: Now simplify again, treating the constraint as 'free' if
1329 it does not mention qtvs, and trying to reduce it otherwise.
1330 The reasons for this is to maximise sharing.
1332 This fails for a very subtle reason. Suppose that in the Step 2
1333 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1334 In the Step 1 this constraint might have been simplified, perhaps to
1335 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1336 This won't happen in Step 2... but that in turn might prevent some other
1337 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1338 and that in turn breaks the invariant that no constraints are quantified over.
1340 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1345 Step 1: Simplify the constraints as much as possible (to deal
1346 with Plan A's problem). Then set
1347 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1348 Return the bindings from Step 1.
1351 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1354 instance (HasBinary ty IO) => HasCodedValue ty
1356 foo :: HasCodedValue a => String -> IO a
1358 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1359 doDecodeIO codedValue view
1360 = let { act = foo "foo" } in act
1362 You might think this should work becuase the call to foo gives rise to a constraint
1363 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1364 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1365 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1367 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1371 Plan D (a variant of plan B)
1372 Step 1: Simplify the constraints as much as possible (to deal
1373 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1374 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1376 Step 2: Now simplify again, treating the constraint as 'free' if
1377 it does not mention qtvs, and trying to reduce it otherwise.
1379 The point here is that it's generally OK to have too few qtvs; that is,
1380 to make the thing more monomorphic than it could be. We don't want to
1381 do that in the common cases, but in wierd cases it's ok: the programmer
1382 can always add a signature.
1384 Too few qtvs => too many wanteds, which is what happens if you do less
1389 tcSimplifyRestricted -- Used for restricted binding groups
1390 -- i.e. ones subject to the monomorphism restriction
1393 -> [Name] -- Things bound in this group
1394 -> TcTyVarSet -- Free in the type of the RHSs
1395 -> [Inst] -- Free in the RHSs
1396 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1397 TcDictBinds) -- Bindings
1398 -- tcSimpifyRestricted returns no constraints to
1399 -- quantify over; by definition there are none.
1400 -- They are all thrown back in the LIE
1402 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1403 -- Zonk everything in sight
1404 = do { traceTc (text "tcSimplifyRestricted")
1405 ; wanteds' <- zonkInsts wanteds
1407 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1408 -- dicts; the idea is to get rid of as many type
1409 -- variables as possible, and we don't want to stop
1410 -- at (say) Monad (ST s), because that reduces
1411 -- immediately, with no constraint on s.
1413 -- BUT do no improvement! See Plan D above
1414 -- HOWEVER, some unification may take place, if we instantiate
1415 -- a method Inst with an equality constraint
1416 ; let env = mkNoImproveRedEnv doc (\_ -> ReduceMe)
1417 ; (_imp, _binds, constrained_dicts) <- reduceContext env wanteds'
1419 -- Next, figure out the tyvars we will quantify over
1420 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1421 ; gbl_tvs' <- tcGetGlobalTyVars
1422 ; constrained_dicts' <- zonkInsts constrained_dicts
1424 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1425 -- As in tcSimplifyInfer
1427 -- Do not quantify over constrained type variables:
1428 -- this is the monomorphism restriction
1429 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1430 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1431 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1434 ; warn_mono <- doptM Opt_WarnMonomorphism
1435 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1436 (vcat[ ptext (sLit "the Monomorphism Restriction applies to the binding")
1437 <> plural bndrs <+> ptext (sLit "for") <+> pp_bndrs,
1438 ptext (sLit "Consider giving a type signature for") <+> pp_bndrs])
1440 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1441 pprInsts wanteds, pprInsts constrained_dicts',
1443 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1445 -- The first step may have squashed more methods than
1446 -- necessary, so try again, this time more gently, knowing the exact
1447 -- set of type variables to quantify over.
1449 -- We quantify only over constraints that are captured by qtvs;
1450 -- these will just be a subset of non-dicts. This in contrast
1451 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1452 -- all *non-inheritable* constraints too. This implements choice
1453 -- (B) under "implicit parameter and monomorphism" above.
1455 -- Remember that we may need to do *some* simplification, to
1456 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1457 -- just to float all constraints
1459 -- At top level, we *do* squash methods becuase we want to
1460 -- expose implicit parameters to the test that follows
1461 ; let is_nested_group = isNotTopLevel top_lvl
1462 try_me inst | isFreeWrtTyVars qtvs inst,
1463 (is_nested_group || isDict inst) = Stop
1464 | otherwise = ReduceMe
1465 env = mkNoImproveRedEnv doc try_me
1466 ; (_imp, binds, irreds) <- reduceContext env wanteds'
1468 -- See "Notes on implicit parameters, Question 4: top level"
1469 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1470 if is_nested_group then
1472 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1473 ; addTopIPErrs bndrs bad_ips
1474 ; extendLIEs non_ips }
1476 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1477 ; return (qtvs', binds) }
1481 %************************************************************************
1485 %************************************************************************
1487 On the LHS of transformation rules we only simplify methods and constants,
1488 getting dictionaries. We want to keep all of them unsimplified, to serve
1489 as the available stuff for the RHS of the rule.
1491 Example. Consider the following left-hand side of a rule
1493 f (x == y) (y > z) = ...
1495 If we typecheck this expression we get constraints
1497 d1 :: Ord a, d2 :: Eq a
1499 We do NOT want to "simplify" to the LHS
1501 forall x::a, y::a, z::a, d1::Ord a.
1502 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1506 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1507 f ((==) d2 x y) ((>) d1 y z) = ...
1509 Here is another example:
1511 fromIntegral :: (Integral a, Num b) => a -> b
1512 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1514 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1515 we *dont* want to get
1517 forall dIntegralInt.
1518 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1520 because the scsel will mess up RULE matching. Instead we want
1522 forall dIntegralInt, dNumInt.
1523 fromIntegral Int Int dIntegralInt dNumInt = id Int
1527 g (x == y) (y == z) = ..
1529 where the two dictionaries are *identical*, we do NOT WANT
1531 forall x::a, y::a, z::a, d1::Eq a
1532 f ((==) d1 x y) ((>) d1 y z) = ...
1534 because that will only match if the dict args are (visibly) equal.
1535 Instead we want to quantify over the dictionaries separately.
1537 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1538 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1539 from scratch, rather than further parameterise simpleReduceLoop etc.
1540 Simpler, maybe, but alas not simple (see Trac #2494)
1542 * Type errors may give rise to an (unsatisfiable) equality constraint
1544 * Applications of a higher-rank function on the LHS may give
1545 rise to an implication constraint, esp if there are unsatisfiable
1546 equality constraints inside.
1549 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1550 tcSimplifyRuleLhs wanteds
1551 = do { wanteds' <- zonkInsts wanteds
1552 ; (irreds, binds) <- go [] emptyBag wanteds'
1553 ; let (dicts, bad_irreds) = partition isDict irreds
1554 ; traceTc (text "tcSimplifyrulelhs" <+> pprInsts bad_irreds)
1555 ; addNoInstanceErrs (nub bad_irreds)
1556 -- The nub removes duplicates, which has
1557 -- not happened otherwise (see notes above)
1558 ; return (dicts, binds) }
1560 go :: [Inst] -> TcDictBinds -> [Inst] -> TcM ([Inst], TcDictBinds)
1562 = return (irreds, binds)
1563 go irreds binds (w:ws)
1565 = go (w:irreds) binds ws
1566 | isImplicInst w -- Have a go at reducing the implication
1567 = do { (binds1, irreds1) <- reduceImplication red_env w
1568 ; let (bad_irreds, ok_irreds) = partition isImplicInst irreds1
1569 ; go (bad_irreds ++ irreds)
1570 (binds `unionBags` binds1)
1573 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1574 -- to fromInteger; this looks fragile to me
1575 ; lookup_result <- lookupSimpleInst w'
1576 ; case lookup_result of
1577 NoInstance -> go (w:irreds) binds ws
1578 GenInst ws' rhs -> go irreds binds' (ws' ++ ws)
1580 binds' = addInstToDictBind binds w rhs
1583 -- Sigh: we need to reduce inside implications
1584 red_env = mkInferRedEnv doc try_me
1585 doc = ptext (sLit "Implication constraint in RULE lhs")
1586 try_me inst | isMethodOrLit inst = ReduceMe
1587 | otherwise = Stop -- Be gentle
1590 tcSimplifyBracket is used when simplifying the constraints arising from
1591 a Template Haskell bracket [| ... |]. We want to check that there aren't
1592 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1593 Show instance), but we aren't otherwise interested in the results.
1594 Nor do we care about ambiguous dictionaries etc. We will type check
1595 this bracket again at its usage site.
1598 tcSimplifyBracket :: [Inst] -> TcM ()
1599 tcSimplifyBracket wanteds
1600 = do { tryHardCheckLoop doc wanteds
1603 doc = text "tcSimplifyBracket"
1607 %************************************************************************
1609 \subsection{Filtering at a dynamic binding}
1611 %************************************************************************
1616 we must discharge all the ?x constraints from B. We also do an improvement
1617 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1619 Actually, the constraints from B might improve the types in ?x. For example
1621 f :: (?x::Int) => Char -> Char
1624 then the constraint (?x::Int) arising from the call to f will
1625 force the binding for ?x to be of type Int.
1628 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1631 -- We need a loop so that we do improvement, and then
1632 -- (next time round) generate a binding to connect the two
1634 -- Here the two ?x's have different types, and improvement
1635 -- makes them the same.
1637 tcSimplifyIPs given_ips wanteds
1638 = do { wanteds' <- zonkInsts wanteds
1639 ; given_ips' <- zonkInsts given_ips
1640 -- Unusually for checking, we *must* zonk the given_ips
1642 ; let env = mkRedEnv doc try_me given_ips'
1643 ; (improved, binds, irreds) <- reduceContext env wanteds'
1645 ; if not improved then
1646 ASSERT( all is_free irreds )
1647 do { extendLIEs irreds
1650 tcSimplifyIPs given_ips wanteds }
1652 doc = text "tcSimplifyIPs" <+> ppr given_ips
1653 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1654 is_free inst = isFreeWrtIPs ip_set inst
1656 -- Simplify any methods that mention the implicit parameter
1657 try_me inst | is_free inst = Stop
1658 | otherwise = ReduceMe
1662 %************************************************************************
1664 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1666 %************************************************************************
1668 When doing a binding group, we may have @Insts@ of local functions.
1669 For example, we might have...
1671 let f x = x + 1 -- orig local function (overloaded)
1672 f.1 = f Int -- two instances of f
1677 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1678 where @f@ is in scope; those @Insts@ must certainly not be passed
1679 upwards towards the top-level. If the @Insts@ were binding-ified up
1680 there, they would have unresolvable references to @f@.
1682 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1683 For each method @Inst@ in the @init_lie@ that mentions one of the
1684 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1685 @LIE@), as well as the @HsBinds@ generated.
1688 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1689 -- Simlifies only MethodInsts, and generate only bindings of form
1691 -- We're careful not to even generate bindings of the form
1693 -- You'd think that'd be fine, but it interacts with what is
1694 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1696 bindInstsOfLocalFuns wanteds local_ids
1697 | null overloaded_ids = do
1700 return emptyLHsBinds
1703 = do { (irreds, binds) <- gentleInferLoop doc for_me
1704 ; extendLIEs not_for_me
1708 doc = text "bindInsts" <+> ppr local_ids
1709 overloaded_ids = filter is_overloaded local_ids
1710 is_overloaded id = isOverloadedTy (idType id)
1711 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1713 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1714 -- so it's worth building a set, so that
1715 -- lookup (in isMethodFor) is faster
1719 %************************************************************************
1721 \subsection{Data types for the reduction mechanism}
1723 %************************************************************************
1725 The main control over context reduction is here
1729 = RedEnv { red_doc :: SDoc -- The context
1730 , red_try_me :: Inst -> WhatToDo
1731 , red_improve :: Bool -- True <=> do improvement
1732 , red_givens :: [Inst] -- All guaranteed rigid
1733 -- Always dicts & equalities
1734 -- but see Note [Rigidity]
1736 , red_given_scs :: Inst -> WantSCs -- See Note [Recursive instances and superclases]
1738 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1739 -- See Note [RedStack]
1743 -- The red_givens are rigid so far as cmpInst is concerned.
1744 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1745 -- let ?x = e in ...
1746 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1747 -- But that doesn't affect the comparison, which is based only on mame.
1750 -- The red_stack pair (n,insts) pair is just used for error reporting.
1751 -- 'n' is always the depth of the stack.
1752 -- The 'insts' is the stack of Insts being reduced: to produce X
1753 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1756 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1757 mkRedEnv doc try_me givens
1758 = RedEnv { red_doc = doc, red_try_me = try_me,
1759 red_givens = givens,
1760 red_given_scs = const AddSCs,
1762 red_improve = True }
1764 mkInferRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1766 mkInferRedEnv doc try_me
1767 = RedEnv { red_doc = doc, red_try_me = try_me,
1769 red_given_scs = const AddSCs,
1771 red_improve = True }
1773 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1774 -- Do not do improvement; no givens
1775 mkNoImproveRedEnv doc try_me
1776 = RedEnv { red_doc = doc, red_try_me = try_me,
1778 red_given_scs = const AddSCs,
1780 red_improve = True }
1783 = ReduceMe -- Try to reduce this
1784 -- If there's no instance, add the inst to the
1785 -- irreductible ones, but don't produce an error
1786 -- message of any kind.
1787 -- It might be quite legitimate such as (Eq a)!
1789 | Stop -- Return as irreducible unless it can
1790 -- be reduced to a constant in one step
1791 -- Do not add superclasses; see
1793 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1794 -- of a predicate when adding it to the avails
1795 -- The reason for this flag is entirely the super-class loop problem
1796 -- Note [SUPER-CLASS LOOP 1]
1798 zonkRedEnv :: RedEnv -> TcM RedEnv
1800 = do { givens' <- mapM zonkInst (red_givens env)
1801 ; return $ env {red_givens = givens'}
1806 %************************************************************************
1808 \subsection[reduce]{@reduce@}
1810 %************************************************************************
1812 Note [Ancestor Equalities]
1813 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1814 During context reduction, we add to the wanted equalities also those
1815 equalities that (transitively) occur in superclass contexts of wanted
1816 class constraints. Consider the following code
1818 class a ~ Int => C a
1821 If (C a) is wanted, we want to add (a ~ Int), which will be discharged by
1822 substituting Int for a. Hence, we ultimately want (C Int), which we
1823 discharge with the explicit instance.
1826 reduceContext :: RedEnv
1828 -> TcM (ImprovementDone,
1829 TcDictBinds, -- Dictionary bindings
1830 [Inst]) -- Irreducible
1832 reduceContext env wanteds0
1833 = do { traceTc (text "reduceContext" <+> (vcat [
1834 text "----------------------",
1836 text "given" <+> ppr (red_givens env),
1837 text "wanted" <+> ppr wanteds0,
1838 text "----------------------"
1841 -- We want to add as wanted equalities those that (transitively)
1842 -- occur in superclass contexts of wanted class constraints.
1843 -- See Note [Ancestor Equalities]
1844 ; ancestor_eqs <- ancestorEqualities wanteds0
1845 ; traceTc $ text "reduceContext: ancestor eqs" <+> ppr ancestor_eqs
1847 -- Normalise and solve all equality constraints as far as possible
1848 -- and normalise all dictionary constraints wrt to the reduced
1849 -- equalities. The returned wanted constraints include the
1850 -- irreducible wanted equalities.
1851 ; let wanteds = wanteds0 ++ ancestor_eqs
1852 givens = red_givens env
1856 eq_improved) <- tcReduceEqs givens wanteds
1857 ; traceTc $ text "reduceContext: tcReduceEqs result" <+> vcat
1858 [ppr givens', ppr wanteds', ppr normalise_binds]
1860 -- Build the Avail mapping from "given_dicts"
1861 ; (init_state, _) <- getLIE $ do
1862 { init_state <- foldlM (addGiven (red_given_scs env))
1867 -- Solve the *wanted* *dictionary* constraints (not implications)
1868 -- This may expose some further equational constraints...
1869 ; let (wanted_implics, wanted_dicts) = partition isImplicInst wanteds'
1870 ; (avails, extra_eqs) <- getLIE (reduceList env wanted_dicts init_state)
1873 dict_irreds) <- extractResults avails wanted_dicts
1874 ; traceTc $ text "reduceContext: extractResults" <+> vcat
1875 [ppr avails, ppr wanted_dicts, ppr dict_binds]
1877 -- Solve the wanted *implications*. In doing so, we can provide
1878 -- as "given" all the dicts that were originally given,
1879 -- *or* for which we now have bindings,
1880 -- *or* which are now irreds
1881 -- NB: Equality irreds need to be converted, as the recursive
1882 -- invocation of the solver will still treat them as wanteds
1884 ; let implic_env = env { red_givens
1885 = givens ++ bound_dicts ++
1886 map wantedToLocalEqInst dict_irreds }
1887 ; (implic_binds_s, implic_irreds_s)
1888 <- mapAndUnzipM (reduceImplication implic_env) wanted_implics
1889 ; let implic_binds = unionManyBags implic_binds_s
1890 implic_irreds = concat implic_irreds_s
1892 -- Collect all irreducible instances, and determine whether we should
1893 -- go round again. We do so in either of two cases:
1894 -- (1) If dictionary reduction or equality solving led to
1895 -- improvement (i.e., instantiated type variables).
1896 -- (2) If we uncovered extra equalities. We will try to solve them
1897 -- in the next iteration.
1899 ; let all_irreds = dict_irreds ++ implic_irreds ++ extra_eqs
1900 avails_improved = availsImproved avails
1901 improvedFlexible = avails_improved || eq_improved
1902 extraEqs = (not . null) extra_eqs
1903 improved = improvedFlexible || extraEqs
1905 improvedHint = (if avails_improved then " [AVAILS]" else "") ++
1906 (if eq_improved then " [EQ]" else "") ++
1907 (if extraEqs then " [EXTRA EQS]" else "")
1909 ; traceTc (text "reduceContext end" <+> (vcat [
1910 text "----------------------",
1912 text "given" <+> ppr givens,
1913 text "wanted" <+> ppr wanteds0,
1915 text "avails" <+> pprAvails avails,
1916 text "improved =" <+> ppr improved <+> text improvedHint,
1917 text "(all) irreds = " <+> ppr all_irreds,
1918 text "dict-binds = " <+> ppr dict_binds,
1919 text "implic-binds = " <+> ppr implic_binds,
1920 text "----------------------"
1924 normalise_binds `unionBags` dict_binds
1925 `unionBags` implic_binds,
1929 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1930 tcImproveOne avails inst
1931 | not (isDict inst) = return False
1933 = do { inst_envs <- tcGetInstEnvs
1934 ; let eqns = improveOne (classInstances inst_envs)
1935 (dictPred inst, pprInstArising inst)
1936 [ (dictPred p, pprInstArising p)
1937 | p <- availsInsts avails, isDict p ]
1938 -- Avails has all the superclasses etc (good)
1939 -- It also has all the intermediates of the deduction (good)
1940 -- It does not have duplicates (good)
1941 -- NB that (?x::t1) and (?x::t2) will be held separately in
1942 -- avails so that improve will see them separate
1943 ; traceTc (text "improveOne" <+> ppr inst)
1946 unifyEqns :: [(Equation, (PredType, SDoc), (PredType, SDoc))]
1947 -> TcM ImprovementDone
1948 unifyEqns [] = return False
1950 = do { traceTc (ptext (sLit "Improve:") <+> vcat (map pprEquationDoc eqns))
1951 ; improved <- mapM unify eqns
1952 ; return $ or improved
1955 unify ((qtvs, pairs), what1, what2)
1956 = addErrCtxtM (mkEqnMsg what1 what2) $
1957 do { let freeTyVars = unionVarSets (map tvs_pr pairs)
1959 ; (_, _, tenv) <- tcInstTyVars (varSetElems qtvs)
1960 ; mapM_ (unif_pr tenv) pairs
1961 ; anyM isFilledMetaTyVar $ varSetElems freeTyVars
1964 unif_pr tenv (ty1, ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1966 tvs_pr (ty1, ty2) = tyVarsOfType ty1 `unionVarSet` tyVarsOfType ty2
1968 pprEquationDoc :: (Equation, (PredType, SDoc), (PredType, SDoc)) -> SDoc
1969 pprEquationDoc (eqn, (p1, _), (p2, _))
1970 = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1972 mkEqnMsg :: (TcPredType, SDoc) -> (TcPredType, SDoc) -> TidyEnv
1973 -> TcM (TidyEnv, SDoc)
1974 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1975 = do { pred1' <- zonkTcPredType pred1
1976 ; pred2' <- zonkTcPredType pred2
1977 ; let { pred1'' = tidyPred tidy_env pred1'
1978 ; pred2'' = tidyPred tidy_env pred2' }
1979 ; let msg = vcat [ptext (sLit "When using functional dependencies to combine"),
1980 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1981 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1982 ; return (tidy_env, msg) }
1985 The main context-reduction function is @reduce@. Here's its game plan.
1988 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1989 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1990 = do { traceTc (text "reduceList " <+> (ppr wanteds $$ ppr state))
1992 ; when (debugIsOn && (n > 8)) $ do
1993 debugDumpTcRn (hang (ptext (sLit "Interesting! Context reduction stack depth") <+> int n)
1994 2 (ifPprDebug (nest 2 (pprStack stk))))
1995 ; if n >= ctxtStkDepth dopts then
1996 failWithTc (reduceDepthErr n stk)
2000 go [] state = return state
2001 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
2004 -- Base case: we're done!
2005 reduce :: RedEnv -> Inst -> Avails -> TcM Avails
2006 reduce env wanted avails
2008 -- We don't reduce equalities here (and they must not end up as irreds
2013 -- It's the same as an existing inst, or a superclass thereof
2014 | Just _ <- findAvail avails wanted
2015 = do { traceTc (text "reduce: found " <+> ppr wanted)
2020 = do { traceTc (text "reduce" <+> ppr wanted $$ ppr avails)
2021 ; case red_try_me env wanted of {
2022 Stop -> try_simple (addIrred NoSCs);
2023 -- See Note [No superclasses for Stop]
2025 ReduceMe -> do -- It should be reduced
2026 { (avails, lookup_result) <- reduceInst env avails wanted
2027 ; case lookup_result of
2028 NoInstance -> addIrred AddSCs avails wanted
2029 -- Add it and its superclasses
2031 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
2033 GenInst wanteds' rhs
2034 -> do { avails1 <- addIrred NoSCs avails wanted
2035 ; avails2 <- reduceList env wanteds' avails1
2036 ; addWanted AddSCs avails2 wanted rhs wanteds' } }
2037 -- Temporarily do addIrred *before* the reduceList,
2038 -- which has the effect of adding the thing we are trying
2039 -- to prove to the database before trying to prove the things it
2040 -- needs. See note [RECURSIVE DICTIONARIES]
2041 -- NB: we must not do an addWanted before, because that adds the
2042 -- superclasses too, and that can lead to a spurious loop; see
2043 -- the examples in [SUPERCLASS-LOOP]
2044 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
2047 -- First, see if the inst can be reduced to a constant in one step
2048 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
2049 -- Don't bother for implication constraints, which take real work
2050 try_simple do_this_otherwise
2051 = do { res <- lookupSimpleInst wanted
2053 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
2054 _ -> do_this_otherwise avails wanted }
2058 Note [RECURSIVE DICTIONARIES]
2059 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2061 data D r = ZeroD | SuccD (r (D r));
2063 instance (Eq (r (D r))) => Eq (D r) where
2064 ZeroD == ZeroD = True
2065 (SuccD a) == (SuccD b) = a == b
2068 equalDC :: D [] -> D [] -> Bool;
2071 We need to prove (Eq (D [])). Here's how we go:
2075 by instance decl, holds if
2079 by instance decl of Eq, holds if
2081 where d2 = dfEqList d3
2084 But now we can "tie the knot" to give
2090 and it'll even run! The trick is to put the thing we are trying to prove
2091 (in this case Eq (D []) into the database before trying to prove its
2092 contributing clauses.
2094 Note [SUPERCLASS-LOOP 2]
2095 ~~~~~~~~~~~~~~~~~~~~~~~~
2096 We need to be careful when adding "the constaint we are trying to prove".
2097 Suppose we are *given* d1:Ord a, and want to deduce (d2:C [a]) where
2099 class Ord a => C a where
2100 instance Ord [a] => C [a] where ...
2102 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
2103 superclasses of C [a] to avails. But we must not overwrite the binding
2104 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
2107 Here's another variant, immortalised in tcrun020
2108 class Monad m => C1 m
2109 class C1 m => C2 m x
2110 instance C2 Maybe Bool
2111 For the instance decl we need to build (C1 Maybe), and it's no good if
2112 we run around and add (C2 Maybe Bool) and its superclasses to the avails
2113 before we search for C1 Maybe.
2115 Here's another example
2116 class Eq b => Foo a b
2117 instance Eq a => Foo [a] a
2121 we'll first deduce that it holds (via the instance decl). We must not
2122 then overwrite the Eq t constraint with a superclass selection!
2124 At first I had a gross hack, whereby I simply did not add superclass constraints
2125 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
2126 becuase it lost legitimate superclass sharing, and it still didn't do the job:
2127 I found a very obscure program (now tcrun021) in which improvement meant the
2128 simplifier got two bites a the cherry... so something seemed to be an Stop
2129 first time, but reducible next time.
2131 Now we implement the Right Solution, which is to check for loops directly
2132 when adding superclasses. It's a bit like the occurs check in unification.
2136 %************************************************************************
2138 Reducing a single constraint
2140 %************************************************************************
2143 ---------------------------------------------
2144 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
2145 reduceInst _ avails other_inst
2146 = do { result <- lookupSimpleInst other_inst
2147 ; return (avails, result) }
2150 Note [Equational Constraints in Implication Constraints]
2151 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2153 An implication constraint is of the form
2155 where Given and Wanted may contain both equational and dictionary
2156 constraints. The delay and reduction of these two kinds of constraints
2159 -) In the generated code, wanted Dictionary constraints are wrapped up in an
2160 implication constraint that is created at the code site where the wanted
2161 dictionaries can be reduced via a let-binding. This let-bound implication
2162 constraint is deconstructed at the use-site of the wanted dictionaries.
2164 -) While the reduction of equational constraints is also delayed, the delay
2165 is not manifest in the generated code. The required evidence is generated
2166 in the code directly at the use-site. There is no let-binding and deconstruction
2167 necessary. The main disadvantage is that we cannot exploit sharing as the
2168 same evidence may be generated at multiple use-sites. However, this disadvantage
2169 is limited because it only concerns coercions which are erased.
2171 The different treatment is motivated by the different in representation. Dictionary
2172 constraints require manifest runtime dictionaries, while equations require coercions
2176 ---------------------------------------------
2177 reduceImplication :: RedEnv
2179 -> TcM (TcDictBinds, [Inst])
2182 Suppose we are simplifying the constraint
2183 forall bs. extras => wanted
2184 in the context of an overall simplification problem with givens 'givens'.
2187 * The 'givens' need not mention any of the quantified type variables
2188 e.g. forall {}. Eq a => Eq [a]
2189 forall {}. C Int => D (Tree Int)
2191 This happens when you have something like
2193 T1 :: Eq a => a -> T a
2196 f x = ...(case x of { T1 v -> v==v })...
2199 -- ToDo: should we instantiate tvs? I think it's not necessary
2201 -- Note on coercion variables:
2203 -- The extra given coercion variables are bound at two different sites:
2204 -- -) in the creation context of the implication constraint
2205 -- the solved equational constraints use these binders
2207 -- -) at the solving site of the implication constraint
2208 -- the solved dictionaries use these binders
2209 -- these binders are generated by reduceImplication
2211 reduceImplication env
2212 orig_implic@(ImplicInst { tci_name = name, tci_loc = inst_loc,
2214 tci_given = extra_givens, tci_wanted = wanteds })
2215 = do { -- Solve the sub-problem
2216 ; let try_me _ = ReduceMe -- Note [Freeness and implications]
2217 env' = env { red_givens = extra_givens ++ red_givens env
2218 , red_doc = sep [ptext (sLit "reduceImplication for")
2220 nest 2 (parens $ ptext (sLit "within")
2222 , red_try_me = try_me }
2224 ; traceTc (text "reduceImplication" <+> vcat
2225 [ ppr (red_givens env), ppr extra_givens,
2227 ; (irreds, binds) <- checkLoop env' wanteds
2228 ; let (extra_eq_givens, extra_dict_givens) = partition isEqInst extra_givens
2229 -- SLPJ Sept 07: I think this is bogus; currently
2230 -- there are no Eqinsts in extra_givens
2231 dict_ids = map instToId extra_dict_givens
2233 -- Note [Reducing implication constraints]
2234 -- Tom -- update note, put somewhere!
2236 ; traceTc (text "reduceImplication result" <+> vcat
2237 [ppr irreds, ppr binds])
2239 ; -- extract superclass binds
2240 -- (sc_binds,_) <- extractResults avails []
2241 -- ; traceTc (text "reduceImplication sc_binds" <+> vcat
2242 -- [ppr sc_binds, ppr avails])
2245 -- SLPJ Sept 07: what if improvement happened inside the checkLoop?
2246 -- Then we must iterate the outer loop too!
2248 ; let backOff = isEmptyLHsBinds binds && -- no new bindings
2249 (not $ null irreds) && -- but still some irreds
2250 all (not . isEqInst) wanteds
2251 -- we may have instantiated a cotv
2252 -- => must make a new implication constraint!
2254 ; traceTc $ text "reduceImplication condition" <+> ppr backOff
2256 -- Progress is no longer measered by the number of bindings
2257 ; if backOff then -- No progress
2258 -- If there are any irreds, we back off and do nothing
2259 return (emptyBag, [orig_implic])
2261 { (simpler_implic_insts, bind)
2262 <- makeImplicationBind inst_loc tvs extra_givens irreds
2263 -- This binding is useless if the recursive simplification
2264 -- made no progress; but currently we don't try to optimise that
2265 -- case. After all, we only try hard to reduce at top level, or
2266 -- when inferring types.
2268 ; let dict_wanteds = filter (not . isEqInst) wanteds
2269 -- TOMDO: given equational constraints bug!
2270 -- we need a different evidence for given
2271 -- equations depending on whether we solve
2272 -- dictionary constraints or equational constraints
2274 eq_tyvars = varSetElems $ tyVarsOfTypes $ map eqInstType extra_eq_givens
2275 -- SLPJ Sept07: this looks Utterly Wrong to me, but I think
2276 -- that current extra_givens has no EqInsts, so
2277 -- it makes no difference
2278 co = wrap_inline -- Note [Always inline implication constraints]
2280 <.> mkWpLams eq_tyvars
2281 <.> mkWpLams dict_ids
2282 <.> WpLet (binds `unionBags` bind)
2283 wrap_inline | null dict_ids = idHsWrapper
2284 | otherwise = WpInline
2285 rhs = mkLHsWrap co payload
2286 loc = instLocSpan inst_loc
2287 payload = mkBigLHsTup (map (L loc . HsVar . instToId) dict_wanteds)
2290 ; traceTc (vcat [text "reduceImplication" <+> ppr name,
2291 ppr simpler_implic_insts,
2292 text "->" <+> ppr rhs])
2293 ; return (unitBag (L loc (VarBind (instToId orig_implic) rhs)),
2294 simpler_implic_insts)
2297 reduceImplication _ i = pprPanic "reduceImplication" (ppr i)
2300 Note [Always inline implication constraints]
2301 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2302 Suppose an implication constraint floats out of an INLINE function.
2303 Then although the implication has a single call site, it won't be
2304 inlined. And that is bad because it means that even if there is really
2305 *no* overloading (type signatures specify the exact types) there will
2306 still be dictionary passing in the resulting code. To avert this,
2307 we mark the implication constraints themselves as INLINE, at least when
2308 there is no loss of sharing as a result.
2310 Note [Freeness and implications]
2311 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2312 It's hard to say when an implication constraint can be floated out. Consider
2313 forall {} Eq a => Foo [a]
2314 The (Foo [a]) doesn't mention any of the quantified variables, but it
2315 still might be partially satisfied by the (Eq a).
2317 There is a useful special case when it *is* easy to partition the
2318 constraints, namely when there are no 'givens'. Consider
2319 forall {a}. () => Bar b
2320 There are no 'givens', and so there is no reason to capture (Bar b).
2321 We can let it float out. But if there is even one constraint we
2322 must be much more careful:
2323 forall {a}. C a b => Bar (m b)
2324 because (C a b) might have a superclass (D b), from which we might
2325 deduce (Bar [b]) when m later gets instantiated to []. Ha!
2327 Here is an even more exotic example
2329 Now consider the constraint
2330 forall b. D Int b => C Int
2331 We can satisfy the (C Int) from the superclass of D, so we don't want
2332 to float the (C Int) out, even though it mentions no type variable in
2335 One more example: the constraint
2337 instance (C a, E c) => E (a,c)
2339 constraint: forall b. D Int b => E (Int,c)
2341 You might think that the (D Int b) can't possibly contribute
2342 to solving (E (Int,c)), since the latter mentions 'c'. But
2343 in fact it can, because solving the (E (Int,c)) constraint needs
2346 and the (C Int) can be satisfied from the superclass of (D Int b).
2347 So we must still not float (E (Int,c)) out.
2349 To think about: special cases for unary type classes?
2351 Note [Pruning the givens in an implication constraint]
2352 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2353 Suppose we are about to form the implication constraint
2354 forall tvs. Eq a => Ord b
2355 The (Eq a) cannot contribute to the (Ord b), because it has no access to
2356 the type variable 'b'. So we could filter out the (Eq a) from the givens.
2357 But BE CAREFUL of the examples above in [Freeness and implications].
2359 Doing so would be a bit tidier, but all the implication constraints get
2360 simplified away by the optimiser, so it's no great win. So I don't take
2361 advantage of that at the moment.
2363 If you do, BE CAREFUL of wobbly type variables.
2366 %************************************************************************
2368 Avails and AvailHow: the pool of evidence
2370 %************************************************************************
2374 data Avails = Avails !ImprovementDone !AvailEnv
2376 type ImprovementDone = Bool -- True <=> some unification has happened
2377 -- so some Irreds might now be reducible
2378 -- keys that are now
2380 type AvailEnv = FiniteMap Inst AvailHow
2382 = IsIrred -- Used for irreducible dictionaries,
2383 -- which are going to be lambda bound
2385 | Given Inst -- Used for dictionaries for which we have a binding
2386 -- e.g. those "given" in a signature
2388 | Rhs -- Used when there is a RHS
2389 (LHsExpr TcId) -- The RHS
2390 [Inst] -- Insts free in the RHS; we need these too
2392 instance Outputable Avails where
2395 pprAvails :: Avails -> SDoc
2396 pprAvails (Avails imp avails)
2397 = vcat [ ptext (sLit "Avails") <> (if imp then ptext (sLit "[improved]") else empty)
2399 vcat [ sep [ppr inst, nest 2 (equals <+> ppr avail)]
2400 | (inst,avail) <- fmToList avails ]]
2402 instance Outputable AvailHow where
2405 -------------------------
2406 pprAvail :: AvailHow -> SDoc
2407 pprAvail IsIrred = text "Irred"
2408 pprAvail (Given x) = text "Given" <+> ppr x
2409 pprAvail (Rhs rhs bs) = sep [text "Rhs" <+> ppr bs,
2412 -------------------------
2413 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2414 extendAvailEnv env inst avail = addToFM env inst avail
2416 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2417 findAvailEnv env wanted = lookupFM env wanted
2418 -- NB 1: the Ord instance of Inst compares by the class/type info
2419 -- *not* by unique. So
2420 -- d1::C Int == d2::C Int
2422 emptyAvails :: Avails
2423 emptyAvails = Avails False emptyFM
2425 findAvail :: Avails -> Inst -> Maybe AvailHow
2426 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2428 elemAvails :: Inst -> Avails -> Bool
2429 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2431 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2433 extendAvails avails@(Avails imp env) inst avail
2434 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2435 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2437 availsInsts :: Avails -> [Inst]
2438 availsInsts (Avails _ avails) = keysFM avails
2440 availsImproved :: Avails -> ImprovementDone
2441 availsImproved (Avails imp _) = imp
2444 Extracting the bindings from a bunch of Avails.
2445 The bindings do *not* come back sorted in dependency order.
2446 We assume that they'll be wrapped in a big Rec, so that the
2447 dependency analyser can sort them out later
2450 type DoneEnv = FiniteMap Inst [Id]
2451 -- Tracks which things we have evidence for
2453 extractResults :: Avails
2455 -> TcM (TcDictBinds, -- Bindings
2456 [Inst], -- The insts bound by the bindings
2457 [Inst]) -- Irreducible ones
2458 -- Note [Reducing implication constraints]
2460 extractResults (Avails _ avails) wanteds
2461 = go emptyBag [] [] emptyFM wanteds
2463 go :: TcDictBinds -- Bindings for dicts
2464 -> [Inst] -- Bound by the bindings
2466 -> DoneEnv -- Has an entry for each inst in the above three sets
2468 -> TcM (TcDictBinds, [Inst], [Inst])
2469 go binds bound_dicts irreds _ []
2470 = return (binds, bound_dicts, irreds)
2472 go binds bound_dicts irreds done (w:ws)
2474 = go binds bound_dicts (w:irreds) done' ws
2476 | Just done_ids@(done_id : rest_done_ids) <- lookupFM done w
2477 = if w_id `elem` done_ids then
2478 go binds bound_dicts irreds done ws
2480 go (add_bind (nlHsVar done_id)) bound_dicts irreds
2481 (addToFM done w (done_id : w_id : rest_done_ids)) ws
2483 | otherwise -- Not yet done
2484 = case findAvailEnv avails w of
2485 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2486 go binds bound_dicts irreds done ws
2488 Just IsIrred -> go binds bound_dicts (w:irreds) done' ws
2490 Just (Rhs rhs ws') -> go (add_bind rhs) (w:bound_dicts) irreds done' (ws' ++ ws)
2492 Just (Given g) -> go binds' bound_dicts irreds (addToFM done w [g_id]) ws
2495 binds' | w_id == g_id = binds
2496 | otherwise = add_bind (nlHsVar g_id)
2499 done' = addToFM done w [w_id]
2500 add_bind rhs = addInstToDictBind binds w rhs
2504 Note [No superclasses for Stop]
2505 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2506 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2507 add it to avails, so that any other equal Insts will be commoned up
2508 right here. However, we do *not* add superclasses. If we have
2511 but a is not bound here, then we *don't* want to derive dn from df
2512 here lest we lose sharing.
2515 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2516 addWanted want_scs avails wanted rhs_expr wanteds
2517 = addAvailAndSCs want_scs avails wanted avail
2519 avail = Rhs rhs_expr wanteds
2521 addGiven :: (Inst -> WantSCs) -> Avails -> Inst -> TcM Avails
2522 addGiven want_scs avails given = addAvailAndSCs (want_scs given) avails given (Given given)
2523 -- Conditionally add superclasses for 'givens'
2524 -- See Note [Recursive instances and superclases]
2526 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2527 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2528 -- so the assert isn't true
2532 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2533 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2534 addAvailAndSCs want_scs avails irred IsIrred
2536 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2537 addAvailAndSCs want_scs avails inst avail
2538 | not (isClassDict inst) = extendAvails avails inst avail
2539 | NoSCs <- want_scs = extendAvails avails inst avail
2540 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2541 ; avails' <- extendAvails avails inst avail
2542 ; addSCs is_loop avails' inst }
2544 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2545 -- Note: this compares by *type*, not by Unique
2546 deps = findAllDeps (unitVarSet (instToVar inst)) avail
2547 dep_tys = map idType (varSetElems deps)
2549 findAllDeps :: IdSet -> AvailHow -> IdSet
2550 -- Find all the Insts that this one depends on
2551 -- See Note [SUPERCLASS-LOOP 2]
2552 -- Watch out, though. Since the avails may contain loops
2553 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2554 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2555 findAllDeps so_far _ = so_far
2557 find_all :: IdSet -> Inst -> IdSet
2559 | isEqInst kid = so_far
2560 | kid_id `elemVarSet` so_far = so_far
2561 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2562 | otherwise = so_far'
2564 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2565 kid_id = instToId kid
2567 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2568 -- Add all the superclasses of the Inst to Avails
2569 -- The first param says "don't do this because the original thing
2570 -- depends on this one, so you'd build a loop"
2571 -- Invariant: the Inst is already in Avails.
2573 addSCs is_loop avails dict
2574 = ASSERT( isDict dict )
2575 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2576 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2578 (clas, tys) = getDictClassTys dict
2579 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2580 sc_theta' = filter (not . isEqPred) $
2581 substTheta (zipTopTvSubst tyvars tys) sc_theta
2583 add_sc avails (sc_dict, sc_sel)
2584 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2585 | is_given sc_dict = return avails
2586 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2587 ; addSCs is_loop avails' sc_dict }
2589 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2590 co_fn = WpApp (instToVar dict) <.> mkWpTyApps tys
2592 is_given :: Inst -> Bool
2593 is_given sc_dict = case findAvail avails sc_dict of
2594 Just (Given _) -> True -- Given is cheaper than superclass selection
2597 -- From the a set of insts obtain all equalities that (transitively) occur in
2598 -- superclass contexts of class constraints (aka the ancestor equalities).
2600 ancestorEqualities :: [Inst] -> TcM [Inst]
2602 = mapM mkWantedEqInst -- turn only equality predicates..
2603 . filter isEqPred -- ..into wanted equality insts
2605 . addAEsToBag emptyBag -- collect the superclass constraints..
2606 . map dictPred -- ..of all predicates in a bag
2607 . filter isClassDict
2609 addAEsToBag :: Bag PredType -> [PredType] -> Bag PredType
2610 addAEsToBag bag [] = bag
2611 addAEsToBag bag (pred:preds)
2612 | pred `elemBag` bag = addAEsToBag bag preds
2613 | isEqPred pred = addAEsToBag bagWithPred preds
2614 | isClassPred pred = addAEsToBag bagWithPred predsWithSCs
2615 | otherwise = addAEsToBag bag preds
2617 bagWithPred = bag `snocBag` pred
2618 predsWithSCs = preds ++ substTheta (zipTopTvSubst tyvars tys) sc_theta
2620 (tyvars, sc_theta, _, _) = classBigSig clas
2621 (clas, tys) = getClassPredTys pred
2625 %************************************************************************
2627 \section{tcSimplifyTop: defaulting}
2629 %************************************************************************
2632 @tcSimplifyTop@ is called once per module to simplify all the constant
2633 and ambiguous Insts.
2635 We need to be careful of one case. Suppose we have
2637 instance Num a => Num (Foo a b) where ...
2639 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2640 to (Num x), and default x to Int. But what about y??
2642 It's OK: the final zonking stage should zap y to (), which is fine.
2646 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2647 tcSimplifyTop wanteds
2648 = tc_simplify_top doc False wanteds
2650 doc = text "tcSimplifyTop"
2652 tcSimplifyInteractive wanteds
2653 = tc_simplify_top doc True wanteds
2655 doc = text "tcSimplifyInteractive"
2657 -- The TcLclEnv should be valid here, solely to improve
2658 -- error message generation for the monomorphism restriction
2659 tc_simplify_top :: SDoc -> Bool -> [Inst] -> TcM (Bag (LHsBind TcId))
2660 tc_simplify_top doc interactive wanteds
2661 = do { dflags <- getDOpts
2662 ; wanteds <- zonkInsts wanteds
2663 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2665 ; traceTc (text "tc_simplify_top 0: " <+> ppr wanteds)
2666 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2667 -- ; (irreds1, binds1) <- gentleInferLoop doc1 wanteds
2668 ; traceTc (text "tc_simplify_top 1: " <+> ppr irreds1)
2669 ; (irreds2, binds2) <- approximateImplications doc2 (\_ -> True) irreds1
2670 ; traceTc (text "tc_simplify_top 2: " <+> ppr irreds2)
2672 -- Use the defaulting rules to do extra unification
2673 -- NB: irreds2 are already zonked
2674 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2676 -- Deal with implicit parameters
2677 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2678 (ambigs, others) = partition isTyVarDict non_ips
2680 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2682 ; addNoInstanceErrs others
2683 ; addTopAmbigErrs ambigs
2685 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2687 doc1 = doc <+> ptext (sLit "(first round)")
2688 doc2 = doc <+> ptext (sLit "(approximate)")
2689 doc3 = doc <+> ptext (sLit "(disambiguate)")
2692 If a dictionary constrains a type variable which is
2693 * not mentioned in the environment
2694 * and not mentioned in the type of the expression
2695 then it is ambiguous. No further information will arise to instantiate
2696 the type variable; nor will it be generalised and turned into an extra
2697 parameter to a function.
2699 It is an error for this to occur, except that Haskell provided for
2700 certain rules to be applied in the special case of numeric types.
2702 * at least one of its classes is a numeric class, and
2703 * all of its classes are numeric or standard
2704 then the type variable can be defaulted to the first type in the
2705 default-type list which is an instance of all the offending classes.
2707 So here is the function which does the work. It takes the ambiguous
2708 dictionaries and either resolves them (producing bindings) or
2709 complains. It works by splitting the dictionary list by type
2710 variable, and using @disambigOne@ to do the real business.
2712 @disambigOne@ assumes that its arguments dictionaries constrain all
2713 the same type variable.
2715 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2716 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2717 the most common use of defaulting is code like:
2719 _ccall_ foo `seqPrimIO` bar
2721 Since we're not using the result of @foo@, the result if (presumably)
2725 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2726 -- Just does unification to fix the default types
2727 -- The Insts are assumed to be pre-zonked
2728 disambiguate doc interactive dflags insts
2730 = return (insts, emptyBag)
2732 | null defaultable_groups
2733 = do { traceTc (text "disambigutate, no defaultable groups" <+> vcat [ppr unaries, ppr insts, ppr bad_tvs, ppr defaultable_groups])
2734 ; return (insts, emptyBag) }
2737 = do { -- Figure out what default types to use
2738 default_tys <- getDefaultTys extended_defaulting ovl_strings
2740 ; traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2741 ; mapM_ (disambigGroup default_tys) defaultable_groups
2743 -- disambigGroup does unification, hence try again
2744 ; tryHardCheckLoop doc insts }
2747 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2748 ovl_strings = dopt Opt_OverloadedStrings dflags
2750 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2751 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2752 (unaries, bad_tvs_s) = partitionWith find_unary insts
2753 bad_tvs = unionVarSets bad_tvs_s
2755 -- Finds unary type-class constraints
2756 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2757 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2758 find_unary inst = Right (tyVarsOfInst inst)
2760 -- Group by type variable
2761 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2762 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2763 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2765 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2766 defaultable_group ds@((_,_,tv):_)
2767 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2768 && not (tv `elemVarSet` bad_tvs)
2769 && defaultable_classes [c | (_,c,_) <- ds]
2770 defaultable_group [] = panic "defaultable_group"
2772 defaultable_classes clss
2773 | extended_defaulting = any isInteractiveClass clss
2774 | otherwise = all is_std_class clss && (any is_num_class clss)
2776 -- In interactive mode, or with -XExtendedDefaultRules,
2777 -- we default Show a to Show () to avoid graututious errors on "show []"
2778 isInteractiveClass cls
2779 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2781 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2782 -- is_num_class adds IsString to the standard numeric classes,
2783 -- when -foverloaded-strings is enabled
2785 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2786 -- Similarly is_std_class
2788 -----------------------
2789 disambigGroup :: [Type] -- The default types
2790 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2791 -> TcM () -- Just does unification, to fix the default types
2793 disambigGroup default_tys dicts
2794 = try_default default_tys
2796 (_,_,tyvar) = ASSERT(not (null dicts)) head dicts -- Should be non-empty
2797 classes = [c | (_,c,_) <- dicts]
2799 try_default [] = return ()
2800 try_default (default_ty : default_tys)
2801 = tryTcLIE_ (try_default default_tys) $
2802 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2803 -- This may fail; then the tryTcLIE_ kicks in
2804 -- Failure here is caused by there being no type in the
2805 -- default list which can satisfy all the ambiguous classes.
2806 -- For example, if Real a is reqd, but the only type in the
2807 -- default list is Int.
2809 -- After this we can't fail
2810 ; warnDefault dicts default_ty
2811 ; unifyType default_ty (mkTyVarTy tyvar)
2812 ; return () -- TOMDO: do something with the coercion
2816 -----------------------
2817 getDefaultTys :: Bool -> Bool -> TcM [Type]
2818 getDefaultTys extended_deflts ovl_strings
2819 = do { mb_defaults <- getDeclaredDefaultTys
2820 ; case mb_defaults of {
2821 Just tys -> return tys ; -- User-supplied defaults
2824 -- No use-supplied default
2825 -- Use [Integer, Double], plus modifications
2826 { integer_ty <- tcMetaTy integerTyConName
2827 ; checkWiredInTyCon doubleTyCon
2828 ; string_ty <- tcMetaTy stringTyConName
2829 ; return (opt_deflt extended_deflts unitTy
2830 -- Note [Default unitTy]
2832 [integer_ty,doubleTy]
2834 opt_deflt ovl_strings string_ty) } } }
2836 opt_deflt True ty = [ty]
2837 opt_deflt False _ = []
2840 Note [Default unitTy]
2841 ~~~~~~~~~~~~~~~~~~~~~
2842 In interative mode (or with -XExtendedDefaultRules) we add () as the first type we
2843 try when defaulting. This has very little real impact, except in the following case.
2845 Text.Printf.printf "hello"
2846 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2847 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2848 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2849 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2850 () to the list of defaulting types. See Trac #1200.
2852 Note [Avoiding spurious errors]
2853 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2854 When doing the unification for defaulting, we check for skolem
2855 type variables, and simply don't default them. For example:
2856 f = (*) -- Monomorphic
2857 g :: Num a => a -> a
2859 Here, we get a complaint when checking the type signature for g,
2860 that g isn't polymorphic enough; but then we get another one when
2861 dealing with the (Num a) context arising from f's definition;
2862 we try to unify a with Int (to default it), but find that it's
2863 already been unified with the rigid variable from g's type sig
2866 %************************************************************************
2868 \subsection[simple]{@Simple@ versions}
2870 %************************************************************************
2872 Much simpler versions when there are no bindings to make!
2874 @tcSimplifyThetas@ simplifies class-type constraints formed by
2875 @deriving@ declarations and when specialising instances. We are
2876 only interested in the simplified bunch of class/type constraints.
2878 It simplifies to constraints of the form (C a b c) where
2879 a,b,c are type variables. This is required for the context of
2880 instance declarations.
2883 tcSimplifyDeriv :: InstOrigin
2885 -> ThetaType -- Wanted
2886 -> TcM ThetaType -- Needed
2887 -- Given instance (wanted) => C inst_ty
2888 -- Simplify 'wanted' as much as possible
2890 tcSimplifyDeriv orig tyvars theta
2891 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2892 -- The main loop may do unification, and that may crash if
2893 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2894 -- ToDo: what if two of them do get unified?
2895 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2896 ; (irreds, _) <- tryHardCheckLoop doc wanteds
2898 ; let (tv_dicts, others) = partition ok irreds
2899 ; addNoInstanceErrs others
2900 -- See Note [Exotic derived instance contexts] in TcMType
2902 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2903 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
2904 -- This reverse-mapping is a pain, but the result
2905 -- should mention the original TyVars not TcTyVars
2907 ; return simpl_theta }
2909 doc = ptext (sLit "deriving classes for a data type")
2911 ok dict | isDict dict = validDerivPred (dictPred dict)
2916 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2917 used with \tr{default} declarations. We are only interested in
2918 whether it worked or not.
2921 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2924 tcSimplifyDefault theta = do
2925 wanteds <- newDictBndrsO DefaultOrigin theta
2926 (irreds, _) <- tryHardCheckLoop doc wanteds
2927 addNoInstanceErrs irreds
2931 traceTc (ptext (sLit "tcSimplifyDefault failing")) >> failM
2933 doc = ptext (sLit "default declaration")
2937 %************************************************************************
2939 \section{Errors and contexts}
2941 %************************************************************************
2943 ToDo: for these error messages, should we note the location as coming
2944 from the insts, or just whatever seems to be around in the monad just
2948 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2949 -> [Inst] -- The offending Insts
2951 -- Group together insts with the same origin
2952 -- We want to report them together in error messages
2956 groupErrs report_err (inst:insts)
2957 = do { do_one (inst:friends)
2958 ; groupErrs report_err others }
2960 -- (It may seem a bit crude to compare the error messages,
2961 -- but it makes sure that we combine just what the user sees,
2962 -- and it avoids need equality on InstLocs.)
2963 (friends, others) = partition is_friend insts
2964 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2965 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2966 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2967 -- Add location and context information derived from the Insts
2969 -- Add the "arising from..." part to a message about bunch of dicts
2970 addInstLoc :: [Inst] -> Message -> Message
2971 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2973 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2976 addTopIPErrs bndrs ips
2977 = do { dflags <- getDOpts
2978 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
2980 (tidy_env, tidy_ips) = tidyInsts ips
2982 = vcat [sep [ptext (sLit "Implicit parameters escape from"),
2983 nest 2 (ptext (sLit "the monomorphic top-level binding")
2984 <> plural bndrs <+> ptext (sLit "of")
2985 <+> pprBinders bndrs <> colon)],
2986 nest 2 (vcat (map ppr_ip ips)),
2987 monomorphism_fix dflags]
2988 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2990 topIPErrs :: [Inst] -> TcM ()
2992 = groupErrs report tidy_dicts
2994 (tidy_env, tidy_dicts) = tidyInsts dicts
2995 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2996 mk_msg dicts = addInstLoc dicts (ptext (sLit "Unbound implicit parameter") <>
2997 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2999 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
3001 addNoInstanceErrs insts
3002 = do { let (tidy_env, tidy_insts) = tidyInsts insts
3003 ; reportNoInstances tidy_env Nothing tidy_insts }
3007 -> Maybe (InstLoc, [Inst]) -- Context
3008 -- Nothing => top level
3009 -- Just (d,g) => d describes the construct
3011 -> [Inst] -- What is wanted (can include implications)
3014 reportNoInstances tidy_env mb_what insts
3015 = groupErrs (report_no_instances tidy_env mb_what) insts
3017 report_no_instances :: TidyEnv -> Maybe (InstLoc, [Inst]) -> [Inst] -> TcM ()
3018 report_no_instances tidy_env mb_what insts
3019 = do { inst_envs <- tcGetInstEnvs
3020 ; let (implics, insts1) = partition isImplicInst insts
3021 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
3022 (eqInsts, insts3) = partition isEqInst insts2
3023 ; traceTc (text "reportNoInstances" <+> vcat
3024 [ppr insts, ppr implics, ppr insts1, ppr insts2])
3025 ; mapM_ complain_implic implics
3026 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
3027 ; groupErrs complain_no_inst insts3
3028 ; mapM_ (addErrTcM . mk_eq_err) eqInsts
3031 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
3033 complain_implic inst -- Recurse!
3034 = reportNoInstances tidy_env
3035 (Just (tci_loc inst, tci_given inst))
3038 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
3039 -- Right msg => overlap message
3040 -- Left inst => no instance
3041 check_overlap inst_envs wanted
3042 | not (isClassDict wanted) = Left wanted
3044 = case lookupInstEnv inst_envs clas tys of
3045 ([], _) -> Left wanted -- No match
3046 -- The case of exactly one match and no unifiers means a
3047 -- successful lookup. That can't happen here, because dicts
3048 -- only end up here if they didn't match in Inst.lookupInst
3050 | debugIsOn -> pprPanic "reportNoInstance" (ppr wanted)
3051 res -> Right (mk_overlap_msg wanted res)
3053 (clas,tys) = getDictClassTys wanted
3055 mk_overlap_msg dict (matches, unifiers)
3056 = ASSERT( not (null matches) )
3057 vcat [ addInstLoc [dict] ((ptext (sLit "Overlapping instances for")
3058 <+> pprPred (dictPred dict))),
3059 sep [ptext (sLit "Matching instances") <> colon,
3060 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
3061 if not (isSingleton matches)
3062 then -- Two or more matches
3064 else -- One match, plus some unifiers
3065 ASSERT( not (null unifiers) )
3066 parens (vcat [ptext (sLit "The choice depends on the instantiation of") <+>
3067 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
3068 ptext (sLit "To pick the first instance above, use -XIncoherentInstances"),
3069 ptext (sLit "when compiling the other instance declarations")])]
3071 ispecs = [ispec | (ispec, _) <- matches]
3073 mk_eq_err :: Inst -> (TidyEnv, SDoc)
3074 mk_eq_err inst = misMatchMsg tidy_env (eqInstTys inst)
3076 mk_no_inst_err insts
3077 | null insts = empty
3079 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
3080 not (isEmptyVarSet (tyVarsOfInsts insts))
3081 = vcat [ addInstLoc insts $
3082 sep [ ptext (sLit "Could not deduce") <+> pprDictsTheta insts
3083 , nest 2 $ ptext (sLit "from the context") <+> pprDictsTheta givens]
3084 , show_fixes (fix1 loc : fixes2) ]
3086 | otherwise -- Top level
3087 = vcat [ addInstLoc insts $
3088 ptext (sLit "No instance") <> plural insts
3089 <+> ptext (sLit "for") <+> pprDictsTheta insts
3090 , show_fixes fixes2 ]
3093 fix1 loc = sep [ ptext (sLit "add") <+> pprDictsTheta insts
3094 <+> ptext (sLit "to the context of"),
3095 nest 2 (ppr (instLocOrigin loc)) ]
3096 -- I'm not sure it helps to add the location
3097 -- nest 2 (ptext (sLit "at") <+> ppr (instLocSpan loc)) ]
3099 fixes2 | null instance_dicts = []
3100 | otherwise = [sep [ptext (sLit "add an instance declaration for"),
3101 pprDictsTheta instance_dicts]]
3102 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
3103 -- Insts for which it is worth suggesting an adding an instance declaration
3104 -- Exclude implicit parameters, and tyvar dicts
3106 show_fixes :: [SDoc] -> SDoc
3107 show_fixes [] = empty
3108 show_fixes (f:fs) = sep [ptext (sLit "Possible fix:"),
3109 nest 2 (vcat (f : map (ptext (sLit "or") <+>) fs))]
3111 addTopAmbigErrs :: [Inst] -> TcRn ()
3112 addTopAmbigErrs dicts
3113 -- Divide into groups that share a common set of ambiguous tyvars
3114 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
3115 -- See Note [Avoiding spurious errors]
3116 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
3118 (tidy_env, tidy_dicts) = tidyInsts dicts
3120 tvs_of :: Inst -> [TcTyVar]
3121 tvs_of d = varSetElems (tyVarsOfInst d)
3122 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
3124 report :: [(Inst,[TcTyVar])] -> TcM ()
3125 report pairs@((inst,tvs) : _) = do -- The pairs share a common set of ambiguous tyvars
3126 (tidy_env, mono_msg) <- mkMonomorphismMsg tidy_env tvs
3127 setSrcSpan (instSpan inst) $
3128 -- the location of the first one will do for the err message
3129 addErrTcM (tidy_env, msg $$ mono_msg)
3131 dicts = map fst pairs
3132 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
3133 pprQuotedList tvs <+> in_msg,
3134 nest 2 (pprDictsInFull dicts)]
3135 in_msg = text "in the constraint" <> plural dicts <> colon
3136 report [] = panic "addTopAmbigErrs"
3139 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
3140 -- There's an error with these Insts; if they have free type variables
3141 -- it's probably caused by the monomorphism restriction.
3142 -- Try to identify the offending variable
3143 -- ASSUMPTION: the Insts are fully zonked
3144 mkMonomorphismMsg tidy_env inst_tvs
3145 = do { dflags <- getDOpts
3146 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
3147 ; return (tidy_env, mk_msg dflags docs) }
3149 mk_msg _ _ | any isRuntimeUnk inst_tvs
3150 = vcat [ptext (sLit "Cannot resolve unknown runtime types:") <+>
3151 (pprWithCommas ppr inst_tvs),
3152 ptext (sLit "Use :print or :force to determine these types")]
3153 mk_msg _ [] = ptext (sLit "Probable fix: add a type signature that fixes these type variable(s)")
3154 -- This happens in things like
3155 -- f x = show (read "foo")
3156 -- where monomorphism doesn't play any role
3158 = vcat [ptext (sLit "Possible cause: the monomorphism restriction applied to the following:"),
3160 monomorphism_fix dflags]
3162 monomorphism_fix :: DynFlags -> SDoc
3163 monomorphism_fix dflags
3164 = ptext (sLit "Probable fix:") <+> vcat
3165 [ptext (sLit "give these definition(s) an explicit type signature"),
3166 if dopt Opt_MonomorphismRestriction dflags
3167 then ptext (sLit "or use -XNoMonomorphismRestriction")
3168 else empty] -- Only suggest adding "-XNoMonomorphismRestriction"
3169 -- if it is not already set!
3171 warnDefault :: [(Inst, Class, Var)] -> Type -> TcM ()
3172 warnDefault ups default_ty = do
3173 warn_flag <- doptM Opt_WarnTypeDefaults
3174 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
3176 dicts = [d | (d,_,_) <- ups]
3179 (_, tidy_dicts) = tidyInsts dicts
3180 warn_msg = vcat [ptext (sLit "Defaulting the following constraint(s) to type") <+>
3181 quotes (ppr default_ty),
3182 pprDictsInFull tidy_dicts]
3184 reduceDepthErr :: Int -> [Inst] -> SDoc
3185 reduceDepthErr n stack
3186 = vcat [ptext (sLit "Context reduction stack overflow; size =") <+> int n,
3187 ptext (sLit "Use -fcontext-stack=N to increase stack size to N"),
3188 nest 4 (pprStack stack)]
3190 pprStack :: [Inst] -> SDoc
3191 pprStack stack = vcat (map pprInstInFull stack)