2 % (c) The University of Glasgow 2006
3 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
10 tcSimplifyInfer, tcSimplifyInferCheck,
11 tcSimplifyCheck, tcSimplifyRestricted,
12 tcSimplifyRuleLhs, tcSimplifyIPs,
13 tcSimplifySuperClasses,
14 tcSimplifyTop, tcSimplifyInteractive,
15 tcSimplifyBracket, tcSimplifyCheckPat,
17 tcSimplifyDeriv, tcSimplifyDefault,
18 bindInstsOfLocalFuns, bindIrreds,
21 #include "HsVersions.h"
23 import {-# SOURCE #-} TcUnify( unifyType )
59 %************************************************************************
63 %************************************************************************
65 --------------------------------------
66 Notes on functional dependencies (a bug)
67 --------------------------------------
74 instance D a b => C a b -- Undecidable
75 -- (Not sure if it's crucial to this eg)
76 f :: C a b => a -> Bool
79 g :: C a b => a -> Bool
82 Here f typechecks, but g does not!! Reason: before doing improvement,
83 we reduce the (C a b1) constraint from the call of f to (D a b1).
85 Here is a more complicated example:
87 | > class Foo a b | a->b
89 | > class Bar a b | a->b
93 | > instance Bar Obj Obj
95 | > instance (Bar a b) => Foo a b
97 | > foo:: (Foo a b) => a -> String
100 | > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
106 | Could not deduce (Bar a b) from the context (Foo a b)
107 | arising from use of `foo' at <interactive>:1
109 | Add (Bar a b) to the expected type of an expression
110 | In the first argument of `runFoo', namely `foo'
111 | In the definition of `it': it = runFoo foo
113 | Why all of the sudden does GHC need the constraint Bar a b? The
114 | function foo didn't ask for that...
116 The trouble is that to type (runFoo foo), GHC has to solve the problem:
118 Given constraint Foo a b
119 Solve constraint Foo a b'
121 Notice that b and b' aren't the same. To solve this, just do
122 improvement and then they are the same. But GHC currently does
127 That is usually fine, but it isn't here, because it sees that Foo a b is
128 not the same as Foo a b', and so instead applies the instance decl for
129 instance Bar a b => Foo a b. And that's where the Bar constraint comes
132 The Right Thing is to improve whenever the constraint set changes at
133 all. Not hard in principle, but it'll take a bit of fiddling to do.
137 --------------------------------------
138 Notes on quantification
139 --------------------------------------
141 Suppose we are about to do a generalisation step.
145 T the type of the RHS
146 C the constraints from that RHS
148 The game is to figure out
150 Q the set of type variables over which to quantify
151 Ct the constraints we will *not* quantify over
152 Cq the constraints we will quantify over
154 So we're going to infer the type
158 and float the constraints Ct further outwards.
160 Here are the things that *must* be true:
162 (A) Q intersect fv(G) = EMPTY limits how big Q can be
163 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
165 (A) says we can't quantify over a variable that's free in the
166 environment. (B) says we must quantify over all the truly free
167 variables in T, else we won't get a sufficiently general type. We do
168 not *need* to quantify over any variable that is fixed by the free
169 vars of the environment G.
171 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
173 Example: class H x y | x->y where ...
175 fv(G) = {a} C = {H a b, H c d}
178 (A) Q intersect {a} is empty
179 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
181 So Q can be {c,d}, {b,c,d}
183 Other things being equal, however, we'd like to quantify over as few
184 variables as possible: smaller types, fewer type applications, more
185 constraints can get into Ct instead of Cq.
188 -----------------------------------------
191 fv(T) the free type vars of T
193 oclose(vs,C) The result of extending the set of tyvars vs
194 using the functional dependencies from C
196 grow(vs,C) The result of extend the set of tyvars vs
197 using all conceivable links from C.
199 E.g. vs = {a}, C = {H [a] b, K (b,Int) c, Eq e}
200 Then grow(vs,C) = {a,b,c}
202 Note that grow(vs,C) `superset` grow(vs,simplify(C))
203 That is, simplfication can only shrink the result of grow.
206 oclose is conservative one way: v `elem` oclose(vs,C) => v is definitely fixed by vs
207 grow is conservative the other way: if v might be fixed by vs => v `elem` grow(vs,C)
210 -----------------------------------------
212 Note [Choosing which variables to quantify]
213 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
214 Here's a good way to choose Q:
216 Q = grow( fv(T), C ) \ oclose( fv(G), C )
218 That is, quantify over all variable that that MIGHT be fixed by the
219 call site (which influences T), but which aren't DEFINITELY fixed by
220 G. This choice definitely quantifies over enough type variables,
221 albeit perhaps too many.
223 Why grow( fv(T), C ) rather than fv(T)? Consider
225 class H x y | x->y where ...
230 If we used fv(T) = {c} we'd get the type
232 forall c. H c d => c -> b
234 And then if the fn was called at several different c's, each of
235 which fixed d differently, we'd get a unification error, because
236 d isn't quantified. Solution: quantify d. So we must quantify
237 everything that might be influenced by c.
239 Why not oclose( fv(T), C )? Because we might not be able to see
240 all the functional dependencies yet:
242 class H x y | x->y where ...
243 instance H x y => Eq (T x y) where ...
248 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
249 apparent yet, and that's wrong. We must really quantify over d too.
252 There really isn't any point in quantifying over any more than
253 grow( fv(T), C ), because the call sites can't possibly influence
254 any other type variables.
258 -------------------------------------
260 -------------------------------------
262 It's very hard to be certain when a type is ambiguous. Consider
266 instance H x y => K (x,y)
268 Is this type ambiguous?
269 forall a b. (K (a,b), Eq b) => a -> a
271 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
272 now we see that a fixes b. So we can't tell about ambiguity for sure
273 without doing a full simplification. And even that isn't possible if
274 the context has some free vars that may get unified. Urgle!
276 Here's another example: is this ambiguous?
277 forall a b. Eq (T b) => a -> a
278 Not if there's an insance decl (with no context)
279 instance Eq (T b) where ...
281 You may say of this example that we should use the instance decl right
282 away, but you can't always do that:
284 class J a b where ...
285 instance J Int b where ...
287 f :: forall a b. J a b => a -> a
289 (Notice: no functional dependency in J's class decl.)
290 Here f's type is perfectly fine, provided f is only called at Int.
291 It's premature to complain when meeting f's signature, or even
292 when inferring a type for f.
296 However, we don't *need* to report ambiguity right away. It'll always
297 show up at the call site.... and eventually at main, which needs special
298 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
300 So here's the plan. We WARN about probable ambiguity if
302 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
304 (all tested before quantification).
305 That is, all the type variables in Cq must be fixed by the the variables
306 in the environment, or by the variables in the type.
308 Notice that we union before calling oclose. Here's an example:
310 class J a b c | a b -> c
314 forall b c. (J a b c) => b -> b
316 Only if we union {a} from G with {b} from T before using oclose,
317 do we see that c is fixed.
319 It's a bit vague exactly which C we should use for this oclose call. If we
320 don't fix enough variables we might complain when we shouldn't (see
321 the above nasty example). Nothing will be perfect. That's why we can
322 only issue a warning.
325 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
327 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
329 then c is a "bubble"; there's no way it can ever improve, and it's
330 certainly ambiguous. UNLESS it is a constant (sigh). And what about
335 instance H x y => K (x,y)
337 Is this type ambiguous?
338 forall a b. (K (a,b), Eq b) => a -> a
340 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
341 is a "bubble" that's a set of constraints
343 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
345 Hence another idea. To decide Q start with fv(T) and grow it
346 by transitive closure in Cq (no functional dependencies involved).
347 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
348 The definitely-ambiguous can then float out, and get smashed at top level
349 (which squashes out the constants, like Eq (T a) above)
352 --------------------------------------
353 Notes on principal types
354 --------------------------------------
359 f x = let g y = op (y::Int) in True
361 Here the principal type of f is (forall a. a->a)
362 but we'll produce the non-principal type
363 f :: forall a. C Int => a -> a
366 --------------------------------------
367 The need for forall's in constraints
368 --------------------------------------
370 [Exchange on Haskell Cafe 5/6 Dec 2000]
372 class C t where op :: t -> Bool
373 instance C [t] where op x = True
375 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
376 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
378 The definitions of p and q differ only in the order of the components in
379 the pair on their right-hand sides. And yet:
381 ghc and "Typing Haskell in Haskell" reject p, but accept q;
382 Hugs rejects q, but accepts p;
383 hbc rejects both p and q;
384 nhc98 ... (Malcolm, can you fill in the blank for us!).
386 The type signature for f forces context reduction to take place, and
387 the results of this depend on whether or not the type of y is known,
388 which in turn depends on which component of the pair the type checker
391 Solution: if y::m a, float out the constraints
392 Monad m, forall c. C (m c)
393 When m is later unified with [], we can solve both constraints.
396 --------------------------------------
397 Notes on implicit parameters
398 --------------------------------------
400 Note [Inheriting implicit parameters]
401 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
406 where f is *not* a top-level binding.
407 From the RHS of f we'll get the constraint (?y::Int).
408 There are two types we might infer for f:
412 (so we get ?y from the context of f's definition), or
414 f :: (?y::Int) => Int -> Int
416 At first you might think the first was better, becuase then
417 ?y behaves like a free variable of the definition, rather than
418 having to be passed at each call site. But of course, the WHOLE
419 IDEA is that ?y should be passed at each call site (that's what
420 dynamic binding means) so we'd better infer the second.
422 BOTTOM LINE: when *inferring types* you *must* quantify
423 over implicit parameters. See the predicate isFreeWhenInferring.
426 Note [Implicit parameters and ambiguity]
427 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
428 What type should we infer for this?
429 f x = (show ?y, x::Int)
430 Since we must quantify over the ?y, the most plausible type is
431 f :: (Show a, ?y::a) => Int -> (String, Int)
432 But notice that the type of the RHS is (String,Int), with no type
433 varibables mentioned at all! The type of f looks ambiguous. But
434 it isn't, because at a call site we might have
435 let ?y = 5::Int in f 7
436 and all is well. In effect, implicit parameters are, well, parameters,
437 so we can take their type variables into account as part of the
438 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
441 Question 2: type signatures
442 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
443 BUT WATCH OUT: When you supply a type signature, we can't force you
444 to quantify over implicit parameters. For example:
448 This is perfectly reasonable. We do not want to insist on
450 (?x + 1) :: (?x::Int => Int)
452 That would be silly. Here, the definition site *is* the occurrence site,
453 so the above strictures don't apply. Hence the difference between
454 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
455 and tcSimplifyCheckBind (which does not).
457 What about when you supply a type signature for a binding?
458 Is it legal to give the following explicit, user type
459 signature to f, thus:
464 At first sight this seems reasonable, but it has the nasty property
465 that adding a type signature changes the dynamic semantics.
468 (let f x = (x::Int) + ?y
469 in (f 3, f 3 with ?y=5)) with ?y = 6
475 in (f 3, f 3 with ?y=5)) with ?y = 6
479 Indeed, simply inlining f (at the Haskell source level) would change the
482 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
483 semantics for a Haskell program without knowing its typing, so if you
484 change the typing you may change the semantics.
486 To make things consistent in all cases where we are *checking* against
487 a supplied signature (as opposed to inferring a type), we adopt the
490 a signature does not need to quantify over implicit params.
492 [This represents a (rather marginal) change of policy since GHC 5.02,
493 which *required* an explicit signature to quantify over all implicit
494 params for the reasons mentioned above.]
496 But that raises a new question. Consider
498 Given (signature) ?x::Int
499 Wanted (inferred) ?x::Int, ?y::Bool
501 Clearly we want to discharge the ?x and float the ?y out. But
502 what is the criterion that distinguishes them? Clearly it isn't
503 what free type variables they have. The Right Thing seems to be
504 to float a constraint that
505 neither mentions any of the quantified type variables
506 nor any of the quantified implicit parameters
508 See the predicate isFreeWhenChecking.
511 Question 3: monomorphism
512 ~~~~~~~~~~~~~~~~~~~~~~~~
513 There's a nasty corner case when the monomorphism restriction bites:
517 The argument above suggests that we *must* generalise
518 over the ?y parameter, to get
519 z :: (?y::Int) => Int,
520 but the monomorphism restriction says that we *must not*, giving
522 Why does the momomorphism restriction say this? Because if you have
524 let z = x + ?y in z+z
526 you might not expect the addition to be done twice --- but it will if
527 we follow the argument of Question 2 and generalise over ?y.
530 Question 4: top level
531 ~~~~~~~~~~~~~~~~~~~~~
532 At the top level, monomorhism makes no sense at all.
535 main = let ?x = 5 in print foo
539 woggle :: (?x :: Int) => Int -> Int
542 We definitely don't want (foo :: Int) with a top-level implicit parameter
543 (?x::Int) becuase there is no way to bind it.
548 (A) Always generalise over implicit parameters
549 Bindings that fall under the monomorphism restriction can't
553 * Inlining remains valid
554 * No unexpected loss of sharing
555 * But simple bindings like
557 will be rejected, unless you add an explicit type signature
558 (to avoid the monomorphism restriction)
559 z :: (?y::Int) => Int
561 This seems unacceptable
563 (B) Monomorphism restriction "wins"
564 Bindings that fall under the monomorphism restriction can't
566 Always generalise over implicit parameters *except* for bindings
567 that fall under the monomorphism restriction
570 * Inlining isn't valid in general
571 * No unexpected loss of sharing
572 * Simple bindings like
574 accepted (get value of ?y from binding site)
576 (C) Always generalise over implicit parameters
577 Bindings that fall under the monomorphism restriction can't
578 be generalised, EXCEPT for implicit parameters
580 * Inlining remains valid
581 * Unexpected loss of sharing (from the extra generalisation)
582 * Simple bindings like
584 accepted (get value of ?y from occurrence sites)
589 None of these choices seems very satisfactory. But at least we should
590 decide which we want to do.
592 It's really not clear what is the Right Thing To Do. If you see
596 would you expect the value of ?y to be got from the *occurrence sites*
597 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
598 case of function definitions, the answer is clearly the former, but
599 less so in the case of non-fucntion definitions. On the other hand,
600 if we say that we get the value of ?y from the definition site of 'z',
601 then inlining 'z' might change the semantics of the program.
603 Choice (C) really says "the monomorphism restriction doesn't apply
604 to implicit parameters". Which is fine, but remember that every
605 innocent binding 'x = ...' that mentions an implicit parameter in
606 the RHS becomes a *function* of that parameter, called at each
607 use of 'x'. Now, the chances are that there are no intervening 'with'
608 clauses that bind ?y, so a decent compiler should common up all
609 those function calls. So I think I strongly favour (C). Indeed,
610 one could make a similar argument for abolishing the monomorphism
611 restriction altogether.
613 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
617 %************************************************************************
619 \subsection{tcSimplifyInfer}
621 %************************************************************************
623 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
625 1. Compute Q = grow( fvs(T), C )
627 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
628 predicates will end up in Ct; we deal with them at the top level
630 3. Try improvement, using functional dependencies
632 4. If Step 3 did any unification, repeat from step 1
633 (Unification can change the result of 'grow'.)
635 Note: we don't reduce dictionaries in step 2. For example, if we have
636 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
637 after step 2. However note that we may therefore quantify over more
638 type variables than we absolutely have to.
640 For the guts, we need a loop, that alternates context reduction and
641 improvement with unification. E.g. Suppose we have
643 class C x y | x->y where ...
645 and tcSimplify is called with:
647 Then improvement unifies a with b, giving
650 If we need to unify anything, we rattle round the whole thing all over
657 -> TcTyVarSet -- fv(T); type vars
659 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
660 [Inst], -- Dict Ids that must be bound here (zonked)
661 TcDictBinds) -- Bindings
662 -- Any free (escaping) Insts are tossed into the environment
667 tcSimplifyInfer doc tau_tvs wanted
668 = do { tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
669 ; wanted' <- mappM zonkInst wanted -- Zonk before deciding quantified tyvars
670 ; gbl_tvs <- tcGetGlobalTyVars
671 ; let preds = fdPredsOfInsts wanted'
672 qtvs = grow preds tau_tvs' `minusVarSet` oclose preds gbl_tvs
673 -- See Note [Choosing which variables to quantify]
675 -- To maximise sharing, remove from consideration any
676 -- constraints that don't mention qtvs at all
677 ; let (free1, bound) = partition (isFreeWhenInferring qtvs) wanted'
680 -- To make types simple, reduce as much as possible
681 ; traceTc (text "infer" <+> (ppr preds $$ ppr (grow preds tau_tvs') $$ ppr gbl_tvs $$
682 ppr (oclose preds gbl_tvs) $$ ppr free1 $$ ppr bound))
683 ; (irreds1, binds1) <- tryHardCheckLoop doc bound
685 -- Note [Inference and implication constraints]
686 ; let want_dict d = tyVarsOfInst d `intersectsVarSet` qtvs
687 ; (irreds2, binds2) <- approximateImplications doc want_dict irreds1
689 -- By now improvment may have taken place, and we must *not*
690 -- quantify over any variable free in the environment
691 -- tc137 (function h inside g) is an example
692 ; gbl_tvs <- tcGetGlobalTyVars
693 ; qtvs1 <- zonkTcTyVarsAndFV (varSetElems qtvs)
694 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems (qtvs1 `minusVarSet` gbl_tvs))
696 -- Do not quantify over constraints that *now* do not
697 -- mention quantified type variables, because they are
698 -- simply ambiguous (or might be bound further out). Example:
699 -- f :: Eq b => a -> (a, b)
701 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
702 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
703 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
704 -- constraint (Eq beta), which we dump back into the free set
705 -- See test tcfail181
706 ; let (free3, irreds3) = partition (isFreeWhenInferring (mkVarSet qtvs2)) irreds2
709 -- We can't abstract over any remaining unsolved
710 -- implications so instead just float them outwards. Ugh.
711 ; let (q_dicts, implics) = partition isDict irreds3
712 ; loc <- getInstLoc (ImplicOrigin doc)
713 ; implic_bind <- bindIrreds loc qtvs2 q_dicts implics
715 ; return (qtvs2, q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
716 -- NB: when we are done, we might have some bindings, but
717 -- the final qtvs might be empty. See Note [NO TYVARS] below.
719 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
720 -- Note [Inference and implication constraints]
721 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
722 -- - fetching any dicts inside them that are free
723 -- - using those dicts as cruder constraints, to solve the implications
724 -- - returning the extra ones too
726 approximateImplications doc want_dict irreds
728 = return (irreds, emptyBag)
730 = do { extra_dicts' <- mapM cloneDict extra_dicts
731 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
732 -- By adding extra_dicts', we make them
733 -- available to solve the implication constraints
735 extra_dicts = get_dicts (filter isImplicInst irreds)
737 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
738 -- Find the wanted constraints in implication constraints that satisfy
739 -- want_dict, and are not bound by forall's in the constraint itself
740 get_dicts ds = concatMap get_dict ds
742 get_dict d@(Dict {}) | want_dict d = [d]
744 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
745 = [ d | let tv_set = mkVarSet tvs
746 , d <- get_dicts wanteds
747 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
748 get_dict other = pprPanic "approximateImplications" (ppr other)
751 Note [Inference and implication constraints]
752 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
753 Suppose we have a wanted implication constraint (perhaps arising from
754 a nested pattern match) like
756 and we are now trying to quantify over 'a' when inferring the type for
757 a function. In principle it's possible that there might be an instance
758 instance (C a, E a) => D [a]
759 so the context (E a) would suffice. The Right Thing is to abstract over
760 the implication constraint, but we don't do that (a) because it'll be
761 surprising to programmers and (b) because we don't have the machinery to deal
762 with 'given' implications.
764 So our best approximation is to make (D [a]) part of the inferred
765 context, so we can use that to discharge the implication. Hence
766 the strange function getImplicWanteds.
768 The common cases are more clear-cut, when we have things like
770 Here, abstracting over (C b) is not an approximation at all -- but see
771 Note [Freeness and implications].
773 See Trac #1430 and test tc228.
777 -----------------------------------------------------------
778 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
779 -- against, but we don't know the type variables over which we are going to quantify.
780 -- This happens when we have a type signature for a mutually recursive group
783 -> TcTyVarSet -- fv(T)
786 -> TcM ([TyVar], -- Fully zonked, and quantified
787 TcDictBinds) -- Bindings
789 tcSimplifyInferCheck loc tau_tvs givens wanteds
790 = do { (irreds, binds) <- gentleCheckLoop loc givens wanteds
792 -- Figure out which type variables to quantify over
793 -- You might think it should just be the signature tyvars,
794 -- but in bizarre cases you can get extra ones
795 -- f :: forall a. Num a => a -> a
796 -- f x = fst (g (x, head [])) + 1
798 -- Here we infer g :: forall a b. a -> b -> (b,a)
799 -- We don't want g to be monomorphic in b just because
800 -- f isn't quantified over b.
801 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
802 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
803 ; gbl_tvs <- tcGetGlobalTyVars
804 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
805 -- We could close gbl_tvs, but its not necessary for
806 -- soundness, and it'll only affect which tyvars, not which
807 -- dictionaries, we quantify over
809 ; qtvs' <- zonkQuantifiedTyVars qtvs
811 -- Now we are back to normal (c.f. tcSimplCheck)
812 ; implic_bind <- bindIrreds loc qtvs' givens irreds
814 ; return (qtvs', binds `unionBags` implic_bind) }
817 Note [Squashing methods]
818 ~~~~~~~~~~~~~~~~~~~~~~~~~
819 Be careful if you want to float methods more:
820 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
821 From an application (truncate f i) we get
824 If we have also have a second occurrence of truncate, we get
827 When simplifying with i,f free, we might still notice that
828 t1=t3; but alas, the binding for t2 (which mentions t1)
829 may continue to float out!
834 class Y a b | a -> b where
837 instance Y [[a]] a where
840 k :: X a -> X a -> X a
842 g :: Num a => [X a] -> [X a]
845 h ys = ys ++ map (k (y [[0]])) xs
847 The excitement comes when simplifying the bindings for h. Initially
848 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
849 From this we get t1:=:t2, but also various bindings. We can't forget
850 the bindings (because of [LOOP]), but in fact t1 is what g is
853 The net effect of [NO TYVARS]
856 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
857 isFreeWhenInferring qtvs inst
858 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
859 && isInheritableInst inst -- and no implicit parameter involved
860 -- see Note [Inheriting implicit parameters]
862 {- No longer used (with implication constraints)
863 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
864 -> NameSet -- Quantified implicit parameters
866 isFreeWhenChecking qtvs ips inst
867 = isFreeWrtTyVars qtvs inst
868 && isFreeWrtIPs ips inst
871 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
872 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
876 %************************************************************************
878 \subsection{tcSimplifyCheck}
880 %************************************************************************
882 @tcSimplifyCheck@ is used when we know exactly the set of variables
883 we are going to quantify over. For example, a class or instance declaration.
886 -----------------------------------------------------------
887 -- tcSimplifyCheck is used when checking expression type signatures,
888 -- class decls, instance decls etc.
889 tcSimplifyCheck :: InstLoc
890 -> [TcTyVar] -- Quantify over these
893 -> TcM TcDictBinds -- Bindings
894 tcSimplifyCheck loc qtvs givens wanteds
895 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
896 do { (irreds, binds) <- gentleCheckLoop loc givens wanteds
897 ; implic_bind <- bindIrreds loc qtvs givens irreds
898 ; return (binds `unionBags` implic_bind) }
900 -----------------------------------------------------------
901 -- tcSimplifyCheckPat is used for existential pattern match
902 tcSimplifyCheckPat :: InstLoc
903 -> [CoVar] -> Refinement
904 -> [TcTyVar] -- Quantify over these
907 -> TcM TcDictBinds -- Bindings
908 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
909 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
910 do { (irreds, binds) <- gentleCheckLoop loc givens wanteds
911 ; implic_bind <- bindIrredsR loc qtvs co_vars reft
913 ; return (binds `unionBags` implic_bind) }
915 -----------------------------------------------------------
916 bindIrreds :: InstLoc -> [TcTyVar]
919 bindIrreds loc qtvs givens irreds
920 = bindIrredsR loc qtvs [] emptyRefinement givens irreds
922 bindIrredsR :: InstLoc -> [TcTyVar] -> [CoVar]
923 -> Refinement -> [Inst] -> [Inst]
925 -- Make a binding that binds 'irreds', by generating an implication
926 -- constraint for them, *and* throwing the constraint into the LIE
927 bindIrredsR loc qtvs co_vars reft givens irreds
931 = do { let givens' = filter isDict givens
932 -- The givens can include methods
933 -- See Note [Pruning the givens in an implication constraint]
935 -- If there are no 'givens' *and* the refinement is empty
936 -- (the refinement is like more givens), then it's safe to
937 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
938 -- See Note [Freeness and implications]
939 ; irreds' <- if null givens' && isEmptyRefinement reft
941 { let qtv_set = mkVarSet qtvs
942 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
944 ; return real_irreds }
947 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
948 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
949 -- This call does the real work
950 -- If irreds' is empty, it does something sensible
955 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
957 -> TcM ([Inst], TcDictBinds)
958 -- Make a binding that binds 'irreds', by generating an implication
959 -- constraint for them, *and* throwing the constraint into the LIE
960 -- The binding looks like
961 -- (ir1, .., irn) = f qtvs givens
962 -- where f is (evidence for) the new implication constraint
963 -- f :: forall qtvs. {reft} givens => (ir1, .., irn)
964 -- qtvs includes coercion variables
966 -- This binding must line up the 'rhs' in reduceImplication
967 makeImplicationBind loc all_tvs reft
968 givens -- Guaranteed all Dicts
970 | null irreds -- If there are no irreds, we are done
971 = return ([], emptyBag)
972 | otherwise -- Otherwise we must generate a binding
973 = do { uniq <- newUnique
974 ; span <- getSrcSpanM
975 ; let name = mkInternalName uniq (mkVarOcc "ic") span
976 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
977 tci_tyvars = all_tvs,
979 tci_wanted = irreds, tci_loc = loc }
981 ; let n_irreds = length irreds
982 irred_ids = map instToId irreds
983 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
984 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
985 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
986 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
987 bind | n_irreds==1 = VarBind (head irred_ids) rhs
988 | otherwise = PatBind { pat_lhs = L span pat,
989 pat_rhs = unguardedGRHSs rhs,
991 bind_fvs = placeHolderNames }
992 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
993 return ([implic_inst], unitBag (L span bind)) }
995 -----------------------------------------------------------
996 tryHardCheckLoop :: SDoc
998 -> TcM ([Inst], TcDictBinds)
1000 tryHardCheckLoop doc wanteds
1001 = checkLoop (mkRedEnv doc try_me []) wanteds
1003 try_me inst = ReduceMe AddSCs
1004 -- Here's the try-hard bit
1006 -----------------------------------------------------------
1007 gentleCheckLoop :: InstLoc
1010 -> TcM ([Inst], TcDictBinds)
1012 gentleCheckLoop inst_loc givens wanteds
1013 = checkLoop env wanteds
1015 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1017 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
1019 -- When checking against a given signature
1020 -- we MUST be very gentle: Note [Check gently]
1024 ~~~~~~~~~~~~~~~~~~~~
1025 We have to very careful about not simplifying too vigorously
1030 f :: Show b => T b -> b
1031 f (MkT x) = show [x]
1033 Inside the pattern match, which binds (a:*, x:a), we know that
1035 Hence we have a dictionary for Show [a] available; and indeed we
1036 need it. We are going to build an implication contraint
1037 forall a. (b~[a]) => Show [a]
1038 Later, we will solve this constraint using the knowledg e(Show b)
1040 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1041 thing becomes insoluble. So we simplify gently (get rid of literals
1042 and methods only, plus common up equal things), deferring the real
1043 work until top level, when we solve the implication constraint
1044 with tryHardCheckLooop.
1048 -----------------------------------------------------------
1051 -> TcM ([Inst], TcDictBinds)
1052 -- Precondition: givens are completely rigid
1053 -- Postcondition: returned Insts are zonked
1055 checkLoop env wanteds
1056 = do { -- Givens are skolems, so no need to zonk them
1057 wanteds' <- mappM zonkInst wanteds
1059 ; (improved, binds, irreds) <- reduceContext env wanteds'
1061 ; if not improved then
1062 return (irreds, binds)
1065 -- If improvement did some unification, we go round again.
1066 -- We start again with irreds, not wanteds
1067 -- Using an instance decl might have introduced a fresh type variable
1068 -- which might have been unified, so we'd get an infinite loop
1069 -- if we started again with wanteds! See Note [LOOP]
1070 { (irreds1, binds1) <- checkLoop env irreds
1071 ; return (irreds1, binds `unionBags` binds1) } }
1076 class If b t e r | b t e -> r
1079 class Lte a b c | a b -> c where lte :: a -> b -> c
1081 instance (Lte a b l,If l b a c) => Max a b c
1083 Wanted: Max Z (S x) y
1085 Then we'll reduce using the Max instance to:
1086 (Lte Z (S x) l, If l (S x) Z y)
1087 and improve by binding l->T, after which we can do some reduction
1088 on both the Lte and If constraints. What we *can't* do is start again
1089 with (Max Z (S x) y)!
1093 %************************************************************************
1095 tcSimplifySuperClasses
1097 %************************************************************************
1099 Note [SUPERCLASS-LOOP 1]
1100 ~~~~~~~~~~~~~~~~~~~~~~~~
1101 We have to be very, very careful when generating superclasses, lest we
1102 accidentally build a loop. Here's an example:
1106 class S a => C a where { opc :: a -> a }
1107 class S b => D b where { opd :: b -> b }
1109 instance C Int where
1112 instance D Int where
1115 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1116 Simplifying, we may well get:
1117 $dfCInt = :C ds1 (opd dd)
1120 Notice that we spot that we can extract ds1 from dd.
1122 Alas! Alack! We can do the same for (instance D Int):
1124 $dfDInt = :D ds2 (opc dc)
1128 And now we've defined the superclass in terms of itself.
1130 Solution: never generate a superclass selectors at all when
1131 satisfying the superclass context of an instance declaration.
1133 Two more nasty cases are in
1138 tcSimplifySuperClasses
1143 tcSimplifySuperClasses loc givens sc_wanteds
1144 = do { (irreds, binds1) <- checkLoop env sc_wanteds
1145 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1146 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1149 env = mkRedEnv (pprInstLoc loc) try_me givens
1150 try_me inst = ReduceMe NoSCs
1151 -- Like tryHardCheckLoop, but with NoSCs
1155 %************************************************************************
1157 \subsection{tcSimplifyRestricted}
1159 %************************************************************************
1161 tcSimplifyRestricted infers which type variables to quantify for a
1162 group of restricted bindings. This isn't trivial.
1165 We want to quantify over a to get id :: forall a. a->a
1168 We do not want to quantify over a, because there's an Eq a
1169 constraint, so we get eq :: a->a->Bool (notice no forall)
1172 RHS has type 'tau', whose free tyvars are tau_tvs
1173 RHS has constraints 'wanteds'
1176 Quantify over (tau_tvs \ ftvs(wanteds))
1177 This is bad. The constraints may contain (Monad (ST s))
1178 where we have instance Monad (ST s) where...
1179 so there's no need to be monomorphic in s!
1181 Also the constraint might be a method constraint,
1182 whose type mentions a perfectly innocent tyvar:
1183 op :: Num a => a -> b -> a
1184 Here, b is unconstrained. A good example would be
1186 We want to infer the polymorphic type
1187 foo :: forall b. b -> b
1190 Plan B (cunning, used for a long time up to and including GHC 6.2)
1191 Step 1: Simplify the constraints as much as possible (to deal
1192 with Plan A's problem). Then set
1193 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1195 Step 2: Now simplify again, treating the constraint as 'free' if
1196 it does not mention qtvs, and trying to reduce it otherwise.
1197 The reasons for this is to maximise sharing.
1199 This fails for a very subtle reason. Suppose that in the Step 2
1200 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1201 In the Step 1 this constraint might have been simplified, perhaps to
1202 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1203 This won't happen in Step 2... but that in turn might prevent some other
1204 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1205 and that in turn breaks the invariant that no constraints are quantified over.
1207 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1212 Step 1: Simplify the constraints as much as possible (to deal
1213 with Plan A's problem). Then set
1214 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1215 Return the bindings from Step 1.
1218 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1221 instance (HasBinary ty IO) => HasCodedValue ty
1223 foo :: HasCodedValue a => String -> IO a
1225 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1226 doDecodeIO codedValue view
1227 = let { act = foo "foo" } in act
1229 You might think this should work becuase the call to foo gives rise to a constraint
1230 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1231 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1232 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1234 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1238 Plan D (a variant of plan B)
1239 Step 1: Simplify the constraints as much as possible (to deal
1240 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1241 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1243 Step 2: Now simplify again, treating the constraint as 'free' if
1244 it does not mention qtvs, and trying to reduce it otherwise.
1246 The point here is that it's generally OK to have too few qtvs; that is,
1247 to make the thing more monomorphic than it could be. We don't want to
1248 do that in the common cases, but in wierd cases it's ok: the programmer
1249 can always add a signature.
1251 Too few qtvs => too many wanteds, which is what happens if you do less
1256 tcSimplifyRestricted -- Used for restricted binding groups
1257 -- i.e. ones subject to the monomorphism restriction
1260 -> [Name] -- Things bound in this group
1261 -> TcTyVarSet -- Free in the type of the RHSs
1262 -> [Inst] -- Free in the RHSs
1263 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1264 TcDictBinds) -- Bindings
1265 -- tcSimpifyRestricted returns no constraints to
1266 -- quantify over; by definition there are none.
1267 -- They are all thrown back in the LIE
1269 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1270 -- Zonk everything in sight
1271 = do { wanteds' <- mappM zonkInst wanteds
1273 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1274 -- dicts; the idea is to get rid of as many type
1275 -- variables as possible, and we don't want to stop
1276 -- at (say) Monad (ST s), because that reduces
1277 -- immediately, with no constraint on s.
1279 -- BUT do no improvement! See Plan D above
1280 -- HOWEVER, some unification may take place, if we instantiate
1281 -- a method Inst with an equality constraint
1282 ; let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1283 ; (_imp, _binds, constrained_dicts) <- reduceContext env wanteds'
1285 -- Next, figure out the tyvars we will quantify over
1286 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1287 ; gbl_tvs' <- tcGetGlobalTyVars
1288 ; constrained_dicts' <- mappM zonkInst constrained_dicts
1290 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1291 -- As in tcSimplifyInfer
1293 -- Do not quantify over constrained type variables:
1294 -- this is the monomorphism restriction
1295 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1296 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1297 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1300 ; warn_mono <- doptM Opt_WarnMonomorphism
1301 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1302 (vcat[ ptext SLIT("the Monomorphism Restriction applies to the binding")
1303 <> plural bndrs <+> ptext SLIT("for") <+> pp_bndrs,
1304 ptext SLIT("Consider giving a type signature for") <+> pp_bndrs])
1306 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1307 pprInsts wanteds, pprInsts constrained_dicts',
1309 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1311 -- The first step may have squashed more methods than
1312 -- necessary, so try again, this time more gently, knowing the exact
1313 -- set of type variables to quantify over.
1315 -- We quantify only over constraints that are captured by qtvs;
1316 -- these will just be a subset of non-dicts. This in contrast
1317 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1318 -- all *non-inheritable* constraints too. This implements choice
1319 -- (B) under "implicit parameter and monomorphism" above.
1321 -- Remember that we may need to do *some* simplification, to
1322 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1323 -- just to float all constraints
1325 -- At top level, we *do* squash methods becuase we want to
1326 -- expose implicit parameters to the test that follows
1327 ; let is_nested_group = isNotTopLevel top_lvl
1328 try_me inst | isFreeWrtTyVars qtvs inst,
1329 (is_nested_group || isDict inst) = Stop
1330 | otherwise = ReduceMe AddSCs
1331 env = mkNoImproveRedEnv doc try_me
1332 ; (_imp, binds, irreds) <- reduceContext env wanteds'
1334 -- See "Notes on implicit parameters, Question 4: top level"
1335 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1336 if is_nested_group then
1338 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1339 ; addTopIPErrs bndrs bad_ips
1340 ; extendLIEs non_ips }
1342 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1343 ; return (qtvs', binds) }
1347 %************************************************************************
1351 %************************************************************************
1353 On the LHS of transformation rules we only simplify methods and constants,
1354 getting dictionaries. We want to keep all of them unsimplified, to serve
1355 as the available stuff for the RHS of the rule.
1357 Example. Consider the following left-hand side of a rule
1359 f (x == y) (y > z) = ...
1361 If we typecheck this expression we get constraints
1363 d1 :: Ord a, d2 :: Eq a
1365 We do NOT want to "simplify" to the LHS
1367 forall x::a, y::a, z::a, d1::Ord a.
1368 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1372 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1373 f ((==) d2 x y) ((>) d1 y z) = ...
1375 Here is another example:
1377 fromIntegral :: (Integral a, Num b) => a -> b
1378 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1380 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1381 we *dont* want to get
1383 forall dIntegralInt.
1384 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1386 because the scsel will mess up RULE matching. Instead we want
1388 forall dIntegralInt, dNumInt.
1389 fromIntegral Int Int dIntegralInt dNumInt = id Int
1393 g (x == y) (y == z) = ..
1395 where the two dictionaries are *identical*, we do NOT WANT
1397 forall x::a, y::a, z::a, d1::Eq a
1398 f ((==) d1 x y) ((>) d1 y z) = ...
1400 because that will only match if the dict args are (visibly) equal.
1401 Instead we want to quantify over the dictionaries separately.
1403 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1404 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1405 from scratch, rather than further parameterise simpleReduceLoop etc
1408 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1409 tcSimplifyRuleLhs wanteds
1410 = go [] emptyBag wanteds
1413 = return (dicts, binds)
1414 go dicts binds (w:ws)
1416 = go (w:dicts) binds ws
1418 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1419 -- to fromInteger; this looks fragile to me
1420 ; lookup_result <- lookupSimpleInst w'
1421 ; case lookup_result of
1422 GenInst ws' rhs -> go dicts (addBind binds (instToId w) rhs) (ws' ++ ws)
1423 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1427 tcSimplifyBracket is used when simplifying the constraints arising from
1428 a Template Haskell bracket [| ... |]. We want to check that there aren't
1429 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1430 Show instance), but we aren't otherwise interested in the results.
1431 Nor do we care about ambiguous dictionaries etc. We will type check
1432 this bracket again at its usage site.
1435 tcSimplifyBracket :: [Inst] -> TcM ()
1436 tcSimplifyBracket wanteds
1437 = do { tryHardCheckLoop doc wanteds
1440 doc = text "tcSimplifyBracket"
1444 %************************************************************************
1446 \subsection{Filtering at a dynamic binding}
1448 %************************************************************************
1453 we must discharge all the ?x constraints from B. We also do an improvement
1454 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1456 Actually, the constraints from B might improve the types in ?x. For example
1458 f :: (?x::Int) => Char -> Char
1461 then the constraint (?x::Int) arising from the call to f will
1462 force the binding for ?x to be of type Int.
1465 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1468 -- We need a loop so that we do improvement, and then
1469 -- (next time round) generate a binding to connect the two
1471 -- Here the two ?x's have different types, and improvement
1472 -- makes them the same.
1474 tcSimplifyIPs given_ips wanteds
1475 = do { wanteds' <- mappM zonkInst wanteds
1476 ; given_ips' <- mappM zonkInst given_ips
1477 -- Unusually for checking, we *must* zonk the given_ips
1479 ; let env = mkRedEnv doc try_me given_ips'
1480 ; (improved, binds, irreds) <- reduceContext env wanteds'
1482 ; if not improved then
1483 ASSERT( all is_free irreds )
1484 do { extendLIEs irreds
1487 tcSimplifyIPs given_ips wanteds }
1489 doc = text "tcSimplifyIPs" <+> ppr given_ips
1490 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1491 is_free inst = isFreeWrtIPs ip_set inst
1493 -- Simplify any methods that mention the implicit parameter
1494 try_me inst | is_free inst = Stop
1495 | otherwise = ReduceMe NoSCs
1499 %************************************************************************
1501 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1503 %************************************************************************
1505 When doing a binding group, we may have @Insts@ of local functions.
1506 For example, we might have...
1508 let f x = x + 1 -- orig local function (overloaded)
1509 f.1 = f Int -- two instances of f
1514 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1515 where @f@ is in scope; those @Insts@ must certainly not be passed
1516 upwards towards the top-level. If the @Insts@ were binding-ified up
1517 there, they would have unresolvable references to @f@.
1519 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1520 For each method @Inst@ in the @init_lie@ that mentions one of the
1521 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1522 @LIE@), as well as the @HsBinds@ generated.
1525 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1526 -- Simlifies only MethodInsts, and generate only bindings of form
1528 -- We're careful not to even generate bindings of the form
1530 -- You'd think that'd be fine, but it interacts with what is
1531 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1533 bindInstsOfLocalFuns wanteds local_ids
1534 | null overloaded_ids
1536 = extendLIEs wanteds `thenM_`
1537 returnM emptyLHsBinds
1540 = do { (irreds, binds) <- checkLoop env for_me
1541 ; extendLIEs not_for_me
1545 env = mkRedEnv doc try_me []
1546 doc = text "bindInsts" <+> ppr local_ids
1547 overloaded_ids = filter is_overloaded local_ids
1548 is_overloaded id = isOverloadedTy (idType id)
1549 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1551 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1552 -- so it's worth building a set, so that
1553 -- lookup (in isMethodFor) is faster
1554 try_me inst | isMethod inst = ReduceMe NoSCs
1559 %************************************************************************
1561 \subsection{Data types for the reduction mechanism}
1563 %************************************************************************
1565 The main control over context reduction is here
1569 = RedEnv { red_doc :: SDoc -- The context
1570 , red_try_me :: Inst -> WhatToDo
1571 , red_improve :: Bool -- True <=> do improvement
1572 , red_givens :: [Inst] -- All guaranteed rigid
1574 -- but see Note [Rigidity]
1575 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1576 -- See Note [RedStack]
1580 -- The red_givens are rigid so far as cmpInst is concerned.
1581 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1582 -- let ?x = e in ...
1583 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1584 -- But that doesn't affect the comparison, which is based only on mame.
1587 -- The red_stack pair (n,insts) pair is just used for error reporting.
1588 -- 'n' is always the depth of the stack.
1589 -- The 'insts' is the stack of Insts being reduced: to produce X
1590 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1593 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1594 mkRedEnv doc try_me givens
1595 = RedEnv { red_doc = doc, red_try_me = try_me,
1596 red_givens = givens, red_stack = (0,[]),
1597 red_improve = True }
1599 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1600 -- Do not do improvement; no givens
1601 mkNoImproveRedEnv doc try_me
1602 = RedEnv { red_doc = doc, red_try_me = try_me,
1603 red_givens = [], red_stack = (0,[]),
1604 red_improve = True }
1607 = ReduceMe WantSCs -- Try to reduce this
1608 -- If there's no instance, add the inst to the
1609 -- irreductible ones, but don't produce an error
1610 -- message of any kind.
1611 -- It might be quite legitimate such as (Eq a)!
1613 | Stop -- Return as irreducible unless it can
1614 -- be reduced to a constant in one step
1615 -- Do not add superclasses; see
1617 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1618 -- of a predicate when adding it to the avails
1619 -- The reason for this flag is entirely the super-class loop problem
1620 -- Note [SUPER-CLASS LOOP 1]
1623 %************************************************************************
1625 \subsection[reduce]{@reduce@}
1627 %************************************************************************
1631 reduceContext :: RedEnv
1633 -> TcM (ImprovementDone,
1634 TcDictBinds, -- Dictionary bindings
1635 [Inst]) -- Irreducible
1637 reduceContext env wanteds
1638 = do { traceTc (text "reduceContext" <+> (vcat [
1639 text "----------------------",
1641 text "given" <+> ppr (red_givens env),
1642 text "wanted" <+> ppr wanteds,
1643 text "----------------------"
1646 -- Build the Avail mapping from "givens"
1647 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1650 -- Process non-implication constraints first, so that they are
1651 -- available to help solving the implication constraints
1652 -- ToDo: seems a bit inefficient and ad-hoc
1653 ; let (implics, rest) = partition isImplicInst wanteds
1654 ; avails <- reduceList env (rest ++ implics) init_state
1656 ; let improved = availsImproved avails
1657 ; (binds, irreds) <- extractResults avails wanteds
1659 ; traceTc (text "reduceContext end" <+> (vcat [
1660 text "----------------------",
1662 text "given" <+> ppr (red_givens env),
1663 text "wanted" <+> ppr wanteds,
1665 text "avails" <+> pprAvails avails,
1666 text "improved =" <+> ppr improved,
1667 text "irreds = " <+> ppr irreds,
1668 text "----------------------"
1671 ; return (improved, binds, irreds) }
1673 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1674 tcImproveOne avails inst
1675 | not (isDict inst) = return False
1677 = do { inst_envs <- tcGetInstEnvs
1678 ; let eqns = improveOne (classInstances inst_envs)
1679 (dictPred inst, pprInstArising inst)
1680 [ (dictPred p, pprInstArising p)
1681 | p <- availsInsts avails, isDict p ]
1682 -- Avails has all the superclasses etc (good)
1683 -- It also has all the intermediates of the deduction (good)
1684 -- It does not have duplicates (good)
1685 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1686 -- so that improve will see them separate
1687 ; traceTc (text "improveOne" <+> ppr inst)
1690 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1691 -> TcM ImprovementDone
1692 unifyEqns [] = return False
1694 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1698 unify ((qtvs, pairs), what1, what2)
1699 = addErrCtxtM (mkEqnMsg what1 what2) $
1700 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1701 mapM_ (unif_pr tenv) pairs
1702 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1704 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1706 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1707 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1708 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1709 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1710 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1711 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1712 ; return (tidy_env, msg) }
1715 The main context-reduction function is @reduce@. Here's its game plan.
1718 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1719 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1720 = do { dopts <- getDOpts
1723 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1724 2 (ifPprDebug (nest 2 (pprStack stk))))
1727 ; if n >= ctxtStkDepth dopts then
1728 failWithTc (reduceDepthErr n stk)
1732 go [] state = return state
1733 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1736 -- Base case: we're done!
1737 reduce env wanted avails
1738 -- It's the same as an existing inst, or a superclass thereof
1739 | Just avail <- findAvail avails wanted
1743 = case red_try_me env wanted of {
1744 ; Stop -> try_simple (addIrred NoSCs) -- See Note [No superclasses for Stop]
1746 ; ReduceMe want_scs -> -- It should be reduced
1747 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1748 case lookup_result of
1749 NoInstance -> -- No such instance!
1750 -- Add it and its superclasses
1751 addIrred want_scs avails wanted
1753 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1755 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1756 ; avails2 <- reduceList env wanteds' avails1
1757 ; addWanted want_scs avails2 wanted rhs wanteds' }
1758 -- Temporarily do addIrred *before* the reduceList,
1759 -- which has the effect of adding the thing we are trying
1760 -- to prove to the database before trying to prove the things it
1761 -- needs. See note [RECURSIVE DICTIONARIES]
1762 -- NB: we must not do an addWanted before, because that adds the
1763 -- superclasses too, and thaat can lead to a spurious loop; see
1764 -- the examples in [SUPERCLASS-LOOP]
1765 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1769 -- First, see if the inst can be reduced to a constant in one step
1770 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1771 -- Don't bother for implication constraints, which take real work
1772 try_simple do_this_otherwise
1773 = do { res <- lookupSimpleInst wanted
1775 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1776 other -> do_this_otherwise avails wanted }
1780 Note [SUPERCLASS-LOOP 2]
1781 ~~~~~~~~~~~~~~~~~~~~~~~~
1782 But the above isn't enough. Suppose we are *given* d1:Ord a,
1783 and want to deduce (d2:C [a]) where
1785 class Ord a => C a where
1786 instance Ord [a] => C [a] where ...
1788 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1789 superclasses of C [a] to avails. But we must not overwrite the binding
1790 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1793 Here's another variant, immortalised in tcrun020
1794 class Monad m => C1 m
1795 class C1 m => C2 m x
1796 instance C2 Maybe Bool
1797 For the instance decl we need to build (C1 Maybe), and it's no good if
1798 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1799 before we search for C1 Maybe.
1801 Here's another example
1802 class Eq b => Foo a b
1803 instance Eq a => Foo [a] a
1807 we'll first deduce that it holds (via the instance decl). We must not
1808 then overwrite the Eq t constraint with a superclass selection!
1810 At first I had a gross hack, whereby I simply did not add superclass constraints
1811 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1812 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1813 I found a very obscure program (now tcrun021) in which improvement meant the
1814 simplifier got two bites a the cherry... so something seemed to be an Stop
1815 first time, but reducible next time.
1817 Now we implement the Right Solution, which is to check for loops directly
1818 when adding superclasses. It's a bit like the occurs check in unification.
1821 Note [RECURSIVE DICTIONARIES]
1822 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1824 data D r = ZeroD | SuccD (r (D r));
1826 instance (Eq (r (D r))) => Eq (D r) where
1827 ZeroD == ZeroD = True
1828 (SuccD a) == (SuccD b) = a == b
1831 equalDC :: D [] -> D [] -> Bool;
1834 We need to prove (Eq (D [])). Here's how we go:
1838 by instance decl, holds if
1842 by instance decl of Eq, holds if
1844 where d2 = dfEqList d3
1847 But now we can "tie the knot" to give
1853 and it'll even run! The trick is to put the thing we are trying to prove
1854 (in this case Eq (D []) into the database before trying to prove its
1855 contributing clauses.
1858 %************************************************************************
1860 Reducing a single constraint
1862 %************************************************************************
1865 ---------------------------------------------
1866 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1867 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1868 tci_given = extra_givens, tci_wanted = wanteds })
1869 = reduceImplication env avails reft tvs extra_givens wanteds loc
1871 reduceInst env avails other_inst
1872 = do { result <- lookupSimpleInst other_inst
1873 ; return (avails, result) }
1877 ---------------------------------------------
1878 reduceImplication :: RedEnv
1880 -> Refinement -- May refine the givens; often empty
1881 -> [TcTyVar] -- Quantified type variables; all skolems
1882 -> [Inst] -- Extra givens; all rigid
1885 -> TcM (Avails, LookupInstResult)
1888 Suppose we are simplifying the constraint
1889 forall bs. extras => wanted
1890 in the context of an overall simplification problem with givens 'givens',
1891 and refinment 'reft'.
1894 * The refinement is often empty
1896 * The 'extra givens' need not mention any of the quantified type variables
1897 e.g. forall {}. Eq a => Eq [a]
1898 forall {}. C Int => D (Tree Int)
1900 This happens when you have something like
1902 T1 :: Eq a => a -> T a
1905 f x = ...(case x of { T1 v -> v==v })...
1908 -- ToDo: should we instantiate tvs? I think it's not necessary
1910 -- ToDo: what about improvement? There may be some improvement
1911 -- exposed as a result of the simplifications done by reduceList
1912 -- which are discarded if we back off.
1913 -- This is almost certainly Wrong, but we'll fix it when dealing
1914 -- better with equality constraints
1915 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1916 = do { -- Add refined givens, and the extra givens
1917 (refined_red_givens, avails)
1918 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1919 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1920 ; avails <- foldlM addGiven avails extra_givens
1922 -- Solve the sub-problem
1923 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1924 env' = env { red_givens = refined_red_givens ++ extra_givens
1925 , red_try_me = try_me }
1927 ; traceTc (text "reduceImplication" <+> vcat
1929 ppr (red_givens env), ppr extra_givens,
1930 ppr reft, ppr wanteds, ppr avails ])
1931 ; avails <- reduceList env' wanteds avails
1933 -- Extract the binding
1934 ; (binds, irreds) <- extractResults avails wanteds
1936 ; traceTc (text "reduceImplication result" <+> vcat
1937 [ ppr irreds, ppr binds])
1939 -- We always discard the extra avails we've generated;
1940 -- but we remember if we have done any (global) improvement
1941 ; let ret_avails = updateImprovement orig_avails avails
1943 ; if isEmptyLHsBinds binds then -- No progress
1944 return (ret_avails, NoInstance)
1946 { (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
1948 ; let dict_ids = map instToId extra_givens
1949 co = mkWpTyLams tvs <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
1950 rhs = mkHsWrap co payload
1951 loc = instLocSpan inst_loc
1952 payload | [wanted] <- wanteds = HsVar (instToId wanted)
1953 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) wanteds) Boxed
1955 -- If there are any irreds, we back off and return NoInstance
1956 ; return (ret_avails, GenInst implic_insts (L loc rhs))
1960 Note [Freeness and implications]
1961 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1962 It's hard to say when an implication constraint can be floated out. Consider
1963 forall {} Eq a => Foo [a]
1964 The (Foo [a]) doesn't mention any of the quantified variables, but it
1965 still might be partially satisfied by the (Eq a).
1967 There is a useful special case when it *is* easy to partition the
1968 constraints, namely when there are no 'givens'. Consider
1969 forall {a}. () => Bar b
1970 There are no 'givens', and so there is no reason to capture (Bar b).
1971 We can let it float out. But if there is even one constraint we
1972 must be much more careful:
1973 forall {a}. C a b => Bar (m b)
1974 because (C a b) might have a superclass (D b), from which we might
1975 deduce (Bar [b]) when m later gets instantiated to []. Ha!
1977 Here is an even more exotic example
1979 Now consider the constraint
1980 forall b. D Int b => C Int
1981 We can satisfy the (C Int) from the superclass of D, so we don't want
1982 to float the (C Int) out, even though it mentions no type variable in
1985 Note [Pruning the givens in an implication constraint]
1986 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1987 Suppose we are about to form the implication constraint
1988 forall tvs. Eq a => Ord b
1989 The (Eq a) cannot contribute to the (Ord b), because it has no access to
1990 the type variable 'b'. So we could filter out the (Eq a) from the givens.
1992 Doing so would be a bit tidier, but all the implication constraints get
1993 simplified away by the optimiser, so it's no great win. So I don't take
1994 advantage of that at the moment.
1996 If you do, BE CAREFUL of wobbly type variables.
1999 %************************************************************************
2001 Avails and AvailHow: the pool of evidence
2003 %************************************************************************
2007 data Avails = Avails !ImprovementDone !AvailEnv
2009 type ImprovementDone = Bool -- True <=> some unification has happened
2010 -- so some Irreds might now be reducible
2011 -- keys that are now
2013 type AvailEnv = FiniteMap Inst AvailHow
2015 = IsIrred TcId -- Used for irreducible dictionaries,
2016 -- which are going to be lambda bound
2018 | Given TcId -- Used for dictionaries for which we have a binding
2019 -- e.g. those "given" in a signature
2021 | Rhs -- Used when there is a RHS
2022 (LHsExpr TcId) -- The RHS
2023 [Inst] -- Insts free in the RHS; we need these too
2025 instance Outputable Avails where
2028 pprAvails (Avails imp avails)
2029 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
2030 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
2031 | (inst,avail) <- fmToList avails ])]
2033 instance Outputable AvailHow where
2036 -------------------------
2037 pprAvail :: AvailHow -> SDoc
2038 pprAvail (IsIrred x) = text "Irred" <+> ppr x
2039 pprAvail (Given x) = text "Given" <+> ppr x
2040 pprAvail (Rhs rhs bs) = text "Rhs" <+> ppr rhs <+> braces (ppr bs)
2042 -------------------------
2043 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2044 extendAvailEnv env inst avail = addToFM env inst avail
2046 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2047 findAvailEnv env wanted = lookupFM env wanted
2048 -- NB 1: the Ord instance of Inst compares by the class/type info
2049 -- *not* by unique. So
2050 -- d1::C Int == d2::C Int
2052 emptyAvails :: Avails
2053 emptyAvails = Avails False emptyFM
2055 findAvail :: Avails -> Inst -> Maybe AvailHow
2056 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2058 elemAvails :: Inst -> Avails -> Bool
2059 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2061 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2063 extendAvails avails@(Avails imp env) inst avail
2064 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2065 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2067 availsInsts :: Avails -> [Inst]
2068 availsInsts (Avails _ avails) = keysFM avails
2070 availsImproved (Avails imp _) = imp
2072 updateImprovement :: Avails -> Avails -> Avails
2073 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
2074 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
2077 Extracting the bindings from a bunch of Avails.
2078 The bindings do *not* come back sorted in dependency order.
2079 We assume that they'll be wrapped in a big Rec, so that the
2080 dependency analyser can sort them out later
2083 extractResults :: Avails
2085 -> TcM ( TcDictBinds, -- Bindings
2086 [Inst]) -- Irreducible ones
2088 extractResults (Avails _ avails) wanteds
2089 = go avails emptyBag [] wanteds
2091 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
2092 -> TcM (TcDictBinds, [Inst])
2093 go avails binds irreds []
2094 = returnM (binds, irreds)
2096 go avails binds irreds (w:ws)
2097 = case findAvailEnv avails w of
2098 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2099 go avails binds irreds ws
2102 | id == w_id -> go avails binds irreds ws
2103 | otherwise -> go avails (addBind binds w_id (nlHsVar id)) irreds ws
2104 -- The sought Id can be one of the givens, via a superclass chain
2105 -- and then we definitely don't want to generate an x=x binding!
2108 | id == w_id -> go (add_given avails w) binds (w:irreds) ws
2109 | otherwise -> go avails (addBind binds w_id (nlHsVar id)) irreds ws
2110 -- The add_given handles the case where we want (Ord a, Eq a), and we
2111 -- don't want to emit *two* Irreds for Ord a, one via the superclass chain
2112 -- This showed up in a dupliated Ord constraint in the error message for
2115 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds (ws' ++ ws)
2117 new_binds = addBind binds w_id rhs
2121 add_given avails w = extendAvailEnv avails w (Given (instToId w))
2122 -- Don't add the same binding twice
2124 addBind binds id rhs = binds `unionBags` unitBag (L (getSrcSpan id) (VarBind id rhs))
2128 Note [No superclasses for Stop]
2129 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2130 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2131 add it to avails, so that any other equal Insts will be commoned up
2132 right here. However, we do *not* add superclasses. If we have
2135 but a is not bound here, then we *don't* want to derive dn from df
2136 here lest we lose sharing.
2139 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2140 addWanted want_scs avails wanted rhs_expr wanteds
2141 = addAvailAndSCs want_scs avails wanted avail
2143 avail = Rhs rhs_expr wanteds
2145 addGiven :: Avails -> Inst -> TcM Avails
2146 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2147 -- Always add superclasses for 'givens'
2149 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2150 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2151 -- so the assert isn't true
2153 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2154 addRefinedGiven reft (refined_givens, avails) given
2155 | isDict given -- We sometimes have 'given' methods, but they
2156 -- are always optional, so we can drop them
2157 , let pred = dictPred given
2158 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2159 , Just (co, pred) <- refinePred reft pred
2160 = do { new_given <- newDictBndr (instLoc given) pred
2161 ; let rhs = L (instSpan given) $
2162 HsWrap (WpCo co) (HsVar (instToId given))
2163 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2164 ; return (new_given:refined_givens, avails) }
2165 -- ToDo: the superclasses of the original given all exist in Avails
2166 -- so we could really just cast them, but it's more awkward to do,
2167 -- and hopefully the optimiser will spot the duplicated work
2169 = return (refined_givens, avails)
2172 Note [ImplicInst rigidity]
2173 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2175 C :: forall ab. (Eq a, Ord b) => b -> T a
2177 ...(case x of C v -> <body>)...
2179 From the case (where x::T ty) we'll get an implication constraint
2180 forall b. (Eq ty, Ord b) => <body-constraints>
2181 Now suppose <body-constraints> itself has an implication constraint
2183 forall c. <reft> => <payload>
2184 Then, we can certainly apply the refinement <reft> to the Ord b, becuase it is
2185 existential, but we probably should not apply it to the (Eq ty) because it may
2186 be wobbly. Hence the isRigidInst
2188 @Insts@ are ordered by their class/type info, rather than by their
2189 unique. This allows the context-reduction mechanism to use standard finite
2190 maps to do their stuff. It's horrible that this code is here, rather
2191 than with the Avails handling stuff in TcSimplify
2194 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2195 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2196 addAvailAndSCs want_scs avails irred (IsIrred (instToId irred))
2198 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2199 addAvailAndSCs want_scs avails inst avail
2200 | not (isClassDict inst) = extendAvails avails inst avail
2201 | NoSCs <- want_scs = extendAvails avails inst avail
2202 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2203 ; avails' <- extendAvails avails inst avail
2204 ; addSCs is_loop avails' inst }
2206 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2207 -- Note: this compares by *type*, not by Unique
2208 deps = findAllDeps (unitVarSet (instToId inst)) avail
2209 dep_tys = map idType (varSetElems deps)
2211 findAllDeps :: IdSet -> AvailHow -> IdSet
2212 -- Find all the Insts that this one depends on
2213 -- See Note [SUPERCLASS-LOOP 2]
2214 -- Watch out, though. Since the avails may contain loops
2215 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2216 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2217 findAllDeps so_far other = so_far
2219 find_all :: IdSet -> Inst -> IdSet
2221 | kid_id `elemVarSet` so_far = so_far
2222 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2223 | otherwise = so_far'
2225 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2226 kid_id = instToId kid
2228 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2229 -- Add all the superclasses of the Inst to Avails
2230 -- The first param says "dont do this because the original thing
2231 -- depends on this one, so you'd build a loop"
2232 -- Invariant: the Inst is already in Avails.
2234 addSCs is_loop avails dict
2235 = ASSERT( isDict dict )
2236 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2237 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2239 (clas, tys) = getDictClassTys dict
2240 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2241 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2243 add_sc avails (sc_dict, sc_sel)
2244 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2245 | is_given sc_dict = return avails
2246 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2247 ; addSCs is_loop avails' sc_dict }
2249 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2250 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2252 is_given :: Inst -> Bool
2253 is_given sc_dict = case findAvail avails sc_dict of
2254 Just (Given _) -> True -- Given is cheaper than superclass selection
2258 %************************************************************************
2260 \section{tcSimplifyTop: defaulting}
2262 %************************************************************************
2265 @tcSimplifyTop@ is called once per module to simplify all the constant
2266 and ambiguous Insts.
2268 We need to be careful of one case. Suppose we have
2270 instance Num a => Num (Foo a b) where ...
2272 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2273 to (Num x), and default x to Int. But what about y??
2275 It's OK: the final zonking stage should zap y to (), which is fine.
2279 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2280 tcSimplifyTop wanteds
2281 = tc_simplify_top doc False wanteds
2283 doc = text "tcSimplifyTop"
2285 tcSimplifyInteractive wanteds
2286 = tc_simplify_top doc True wanteds
2288 doc = text "tcSimplifyInteractive"
2290 -- The TcLclEnv should be valid here, solely to improve
2291 -- error message generation for the monomorphism restriction
2292 tc_simplify_top doc interactive wanteds
2293 = do { dflags <- getDOpts
2294 ; wanteds <- mapM zonkInst wanteds
2295 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2297 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2298 ; (irreds2, binds2) <- approximateImplications doc2 (\d -> True) irreds1
2300 -- Use the defaulting rules to do extra unification
2301 -- NB: irreds2 are already zonked
2302 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2304 -- Deal with implicit parameters
2305 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2306 (ambigs, others) = partition isTyVarDict non_ips
2308 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2310 ; addNoInstanceErrs others
2311 ; addTopAmbigErrs ambigs
2313 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2315 doc1 = doc <+> ptext SLIT("(first round)")
2316 doc2 = doc <+> ptext SLIT("(approximate)")
2317 doc3 = doc <+> ptext SLIT("(disambiguate)")
2320 If a dictionary constrains a type variable which is
2321 * not mentioned in the environment
2322 * and not mentioned in the type of the expression
2323 then it is ambiguous. No further information will arise to instantiate
2324 the type variable; nor will it be generalised and turned into an extra
2325 parameter to a function.
2327 It is an error for this to occur, except that Haskell provided for
2328 certain rules to be applied in the special case of numeric types.
2330 * at least one of its classes is a numeric class, and
2331 * all of its classes are numeric or standard
2332 then the type variable can be defaulted to the first type in the
2333 default-type list which is an instance of all the offending classes.
2335 So here is the function which does the work. It takes the ambiguous
2336 dictionaries and either resolves them (producing bindings) or
2337 complains. It works by splitting the dictionary list by type
2338 variable, and using @disambigOne@ to do the real business.
2340 @disambigOne@ assumes that its arguments dictionaries constrain all
2341 the same type variable.
2343 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2344 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2345 the most common use of defaulting is code like:
2347 _ccall_ foo `seqPrimIO` bar
2349 Since we're not using the result of @foo@, the result if (presumably)
2353 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2354 -- Just does unification to fix the default types
2355 -- The Insts are assumed to be pre-zonked
2356 disambiguate doc interactive dflags insts
2358 = return (insts, emptyBag)
2360 | null defaultable_groups
2361 = do { traceTc (text "disambigutate" <+> vcat [ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2362 ; return (insts, emptyBag) }
2365 = do { -- Figure out what default types to use
2366 default_tys <- getDefaultTys extended_defaulting ovl_strings
2368 ; traceTc (text "disambigutate" <+> vcat [ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2369 ; mapM_ (disambigGroup default_tys) defaultable_groups
2371 -- disambigGroup does unification, hence try again
2372 ; tryHardCheckLoop doc insts }
2375 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2376 ovl_strings = dopt Opt_OverloadedStrings dflags
2378 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2379 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2380 (unaries, bad_tvs_s) = partitionWith find_unary insts
2381 bad_tvs = unionVarSets bad_tvs_s
2383 -- Finds unary type-class constraints
2384 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2385 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2386 find_unary inst = Right (tyVarsOfInst inst)
2388 -- Group by type variable
2389 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2390 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2391 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2393 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2394 defaultable_group ds@((_,_,tv):_)
2395 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2396 && not (tv `elemVarSet` bad_tvs)
2397 && defaultable_classes [c | (_,c,_) <- ds]
2398 defaultable_group [] = panic "defaultable_group"
2400 defaultable_classes clss
2401 | extended_defaulting = any isInteractiveClass clss
2402 | otherwise = all is_std_class clss && (any is_num_class clss)
2404 -- In interactive mode, or with -fextended-default-rules,
2405 -- we default Show a to Show () to avoid graututious errors on "show []"
2406 isInteractiveClass cls
2407 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2409 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2410 -- is_num_class adds IsString to the standard numeric classes,
2411 -- when -foverloaded-strings is enabled
2413 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2414 -- Similarly is_std_class
2416 -----------------------
2417 disambigGroup :: [Type] -- The default types
2418 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2419 -> TcM () -- Just does unification, to fix the default types
2421 disambigGroup default_tys dicts
2422 = try_default default_tys
2424 (_,_,tyvar) = head dicts -- Should be non-empty
2425 classes = [c | (_,c,_) <- dicts]
2427 try_default [] = return ()
2428 try_default (default_ty : default_tys)
2429 = tryTcLIE_ (try_default default_tys) $
2430 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2431 -- This may fail; then the tryTcLIE_ kicks in
2432 -- Failure here is caused by there being no type in the
2433 -- default list which can satisfy all the ambiguous classes.
2434 -- For example, if Real a is reqd, but the only type in the
2435 -- default list is Int.
2437 -- After this we can't fail
2438 ; warnDefault dicts default_ty
2439 ; unifyType default_ty (mkTyVarTy tyvar) }
2442 -----------------------
2443 getDefaultTys :: Bool -> Bool -> TcM [Type]
2444 getDefaultTys extended_deflts ovl_strings
2445 = do { mb_defaults <- getDeclaredDefaultTys
2446 ; case mb_defaults of {
2447 Just tys -> return tys ; -- User-supplied defaults
2450 -- No use-supplied default
2451 -- Use [Integer, Double], plus modifications
2452 { integer_ty <- tcMetaTy integerTyConName
2453 ; checkWiredInTyCon doubleTyCon
2454 ; string_ty <- tcMetaTy stringTyConName
2455 ; return (opt_deflt extended_deflts unitTy
2456 -- Note [Default unitTy]
2458 [integer_ty,doubleTy]
2460 opt_deflt ovl_strings string_ty) } } }
2462 opt_deflt True ty = [ty]
2463 opt_deflt False ty = []
2466 Note [Default unitTy]
2467 ~~~~~~~~~~~~~~~~~~~~~
2468 In interative mode (or with -fextended-default-rules) we add () as the first type we
2469 try when defaulting. This has very little real impact, except in the following case.
2471 Text.Printf.printf "hello"
2472 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2473 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2474 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2475 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2476 () to the list of defaulting types. See Trac #1200.
2478 Note [Avoiding spurious errors]
2479 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2480 When doing the unification for defaulting, we check for skolem
2481 type variables, and simply don't default them. For example:
2482 f = (*) -- Monomorphic
2483 g :: Num a => a -> a
2485 Here, we get a complaint when checking the type signature for g,
2486 that g isn't polymorphic enough; but then we get another one when
2487 dealing with the (Num a) context arising from f's definition;
2488 we try to unify a with Int (to default it), but find that it's
2489 already been unified with the rigid variable from g's type sig
2492 %************************************************************************
2494 \subsection[simple]{@Simple@ versions}
2496 %************************************************************************
2498 Much simpler versions when there are no bindings to make!
2500 @tcSimplifyThetas@ simplifies class-type constraints formed by
2501 @deriving@ declarations and when specialising instances. We are
2502 only interested in the simplified bunch of class/type constraints.
2504 It simplifies to constraints of the form (C a b c) where
2505 a,b,c are type variables. This is required for the context of
2506 instance declarations.
2509 tcSimplifyDeriv :: InstOrigin
2511 -> ThetaType -- Wanted
2512 -> TcM ThetaType -- Needed
2513 -- Given instance (wanted) => C inst_ty
2514 -- Simplify 'wanted' as much as possible
2515 -- The inst_ty is needed only for the termination check
2517 tcSimplifyDeriv orig tyvars theta
2518 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2519 -- The main loop may do unification, and that may crash if
2520 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2521 -- ToDo: what if two of them do get unified?
2522 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2523 ; (irreds, _) <- tryHardCheckLoop doc wanteds
2525 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2526 simpl_theta = substTheta rev_env (map dictPred irreds)
2527 -- This reverse-mapping is a pain, but the result
2528 -- should mention the original TyVars not TcTyVars
2530 -- NB: the caller will further check the tv_dicts for
2531 -- legal instance-declaration form
2533 ; return simpl_theta }
2535 doc = ptext SLIT("deriving classes for a data type")
2540 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2541 used with \tr{default} declarations. We are only interested in
2542 whether it worked or not.
2545 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2548 tcSimplifyDefault theta
2549 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2550 tryHardCheckLoop doc wanteds `thenM` \ (irreds, _) ->
2551 addNoInstanceErrs irreds `thenM_`
2557 doc = ptext SLIT("default declaration")
2561 %************************************************************************
2563 \section{Errors and contexts}
2565 %************************************************************************
2567 ToDo: for these error messages, should we note the location as coming
2568 from the insts, or just whatever seems to be around in the monad just
2572 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2573 -> [Inst] -- The offending Insts
2575 -- Group together insts with the same origin
2576 -- We want to report them together in error messages
2578 groupErrs report_err []
2580 groupErrs report_err (inst:insts)
2581 = do_one (inst:friends) `thenM_`
2582 groupErrs report_err others
2585 -- (It may seem a bit crude to compare the error messages,
2586 -- but it makes sure that we combine just what the user sees,
2587 -- and it avoids need equality on InstLocs.)
2588 (friends, others) = partition is_friend insts
2589 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2590 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2591 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2592 -- Add location and context information derived from the Insts
2594 -- Add the "arising from..." part to a message about bunch of dicts
2595 addInstLoc :: [Inst] -> Message -> Message
2596 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2598 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2599 addTopIPErrs bndrs []
2601 addTopIPErrs bndrs ips
2602 = do { dflags <- getDOpts
2603 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
2605 (tidy_env, tidy_ips) = tidyInsts ips
2607 = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2608 nest 2 (ptext SLIT("the monomorphic top-level binding")
2609 <> plural bndrs <+> ptext SLIT("of")
2610 <+> pprBinders bndrs <> colon)],
2611 nest 2 (vcat (map ppr_ip ips)),
2612 monomorphism_fix dflags]
2613 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2615 topIPErrs :: [Inst] -> TcM ()
2617 = groupErrs report tidy_dicts
2619 (tidy_env, tidy_dicts) = tidyInsts dicts
2620 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2621 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2622 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2624 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2626 addNoInstanceErrs insts
2627 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2628 ; reportNoInstances tidy_env Nothing tidy_insts }
2632 -> Maybe (InstLoc, [Inst]) -- Context
2633 -- Nothing => top level
2634 -- Just (d,g) => d describes the construct
2636 -> [Inst] -- What is wanted (can include implications)
2639 reportNoInstances tidy_env mb_what insts
2640 = groupErrs (report_no_instances tidy_env mb_what) insts
2642 report_no_instances tidy_env mb_what insts
2643 = do { inst_envs <- tcGetInstEnvs
2644 ; let (implics, insts1) = partition isImplicInst insts
2645 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2646 ; traceTc (text "reportNoInstnces" <+> vcat
2647 [ppr implics, ppr insts1, ppr insts2])
2648 ; mapM_ complain_implic implics
2649 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2650 ; groupErrs complain_no_inst insts2 }
2652 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2654 complain_implic inst -- Recurse!
2655 = reportNoInstances tidy_env
2656 (Just (tci_loc inst, tci_given inst))
2659 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2660 -- Right msg => overlap message
2661 -- Left inst => no instance
2662 check_overlap inst_envs wanted
2663 | not (isClassDict wanted) = Left wanted
2665 = case lookupInstEnv inst_envs clas tys of
2666 -- The case of exactly one match and no unifiers means
2667 -- a successful lookup. That can't happen here, becuase
2668 -- dicts only end up here if they didn't match in Inst.lookupInst
2670 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2672 ([], _) -> Left wanted -- No match
2673 res -> Right (mk_overlap_msg wanted res)
2675 (clas,tys) = getDictClassTys wanted
2677 mk_overlap_msg dict (matches, unifiers)
2678 = vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2679 <+> pprPred (dictPred dict))),
2680 sep [ptext SLIT("Matching instances") <> colon,
2681 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2682 ASSERT( not (null matches) )
2683 if not (isSingleton matches)
2684 then -- Two or more matches
2686 else -- One match, plus some unifiers
2687 ASSERT( not (null unifiers) )
2688 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2689 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2690 ptext SLIT("Use -fallow-incoherent-instances to use the first choice above")])]
2692 ispecs = [ispec | (ispec, _) <- matches]
2694 mk_no_inst_err insts
2695 | null insts = empty
2697 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2698 not (isEmptyVarSet (tyVarsOfInsts insts))
2699 = vcat [ addInstLoc insts $
2700 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2701 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2702 , show_fixes (fix1 loc : fixes2) ]
2704 | otherwise -- Top level
2705 = vcat [ addInstLoc insts $
2706 ptext SLIT("No instance") <> plural insts
2707 <+> ptext SLIT("for") <+> pprDictsTheta insts
2708 , show_fixes fixes2 ]
2711 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2712 <+> ptext SLIT("to the context of"),
2713 nest 2 (ppr (instLocOrigin loc)) ]
2714 -- I'm not sure it helps to add the location
2715 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2717 fixes2 | null instance_dicts = []
2718 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2719 pprDictsTheta instance_dicts]]
2720 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2721 -- Insts for which it is worth suggesting an adding an instance declaration
2722 -- Exclude implicit parameters, and tyvar dicts
2724 show_fixes :: [SDoc] -> SDoc
2725 show_fixes [] = empty
2726 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2727 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2729 addTopAmbigErrs dicts
2730 -- Divide into groups that share a common set of ambiguous tyvars
2731 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2732 -- See Note [Avoiding spurious errors]
2733 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2735 (tidy_env, tidy_dicts) = tidyInsts dicts
2737 tvs_of :: Inst -> [TcTyVar]
2738 tvs_of d = varSetElems (tyVarsOfInst d)
2739 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2741 report :: [(Inst,[TcTyVar])] -> TcM ()
2742 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2743 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2744 setSrcSpan (instSpan inst) $
2745 -- the location of the first one will do for the err message
2746 addErrTcM (tidy_env, msg $$ mono_msg)
2748 dicts = map fst pairs
2749 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2750 pprQuotedList tvs <+> in_msg,
2751 nest 2 (pprDictsInFull dicts)]
2752 in_msg = text "in the constraint" <> plural dicts <> colon
2753 report [] = panic "addTopAmbigErrs"
2756 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2757 -- There's an error with these Insts; if they have free type variables
2758 -- it's probably caused by the monomorphism restriction.
2759 -- Try to identify the offending variable
2760 -- ASSUMPTION: the Insts are fully zonked
2761 mkMonomorphismMsg tidy_env inst_tvs
2762 = do { dflags <- getDOpts
2763 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
2764 ; return (tidy_env, mk_msg dflags docs) }
2766 mk_msg _ _ | any isRuntimeUnk inst_tvs
2767 = vcat [ptext SLIT("Cannot resolve unknown runtime types:") <+>
2768 (pprWithCommas ppr inst_tvs),
2769 ptext SLIT("Use :print or :force to determine these types")]
2770 mk_msg _ [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2771 -- This happens in things like
2772 -- f x = show (read "foo")
2773 -- where monomorphism doesn't play any role
2775 = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2777 monomorphism_fix dflags]
2779 isRuntimeUnk :: TcTyVar -> Bool
2780 isRuntimeUnk x | SkolemTv RuntimeUnkSkol <- tcTyVarDetails x = True
2783 monomorphism_fix :: DynFlags -> SDoc
2784 monomorphism_fix dflags
2785 = ptext SLIT("Probable fix:") <+> vcat
2786 [ptext SLIT("give these definition(s) an explicit type signature"),
2787 if dopt Opt_MonomorphismRestriction dflags
2788 then ptext SLIT("or use -fno-monomorphism-restriction")
2789 else empty] -- Only suggest adding "-fno-monomorphism-restriction"
2790 -- if it is not already set!
2792 warnDefault ups default_ty
2793 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2794 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2796 dicts = [d | (d,_,_) <- ups]
2799 (_, tidy_dicts) = tidyInsts dicts
2800 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2801 quotes (ppr default_ty),
2802 pprDictsInFull tidy_dicts]
2804 reduceDepthErr n stack
2805 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2806 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2807 nest 4 (pprStack stack)]
2809 pprStack stack = vcat (map pprInstInFull stack)