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 )
60 %************************************************************************
64 %************************************************************************
66 --------------------------------------
67 Notes on functional dependencies (a bug)
68 --------------------------------------
75 instance D a b => C a b -- Undecidable
76 -- (Not sure if it's crucial to this eg)
77 f :: C a b => a -> Bool
80 g :: C a b => a -> Bool
83 Here f typechecks, but g does not!! Reason: before doing improvement,
84 we reduce the (C a b1) constraint from the call of f to (D a b1).
86 Here is a more complicated example:
88 | > class Foo a b | a->b
90 | > class Bar a b | a->b
94 | > instance Bar Obj Obj
96 | > instance (Bar a b) => Foo a b
98 | > foo:: (Foo a b) => a -> String
101 | > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
107 | Could not deduce (Bar a b) from the context (Foo a b)
108 | arising from use of `foo' at <interactive>:1
110 | Add (Bar a b) to the expected type of an expression
111 | In the first argument of `runFoo', namely `foo'
112 | In the definition of `it': it = runFoo foo
114 | Why all of the sudden does GHC need the constraint Bar a b? The
115 | function foo didn't ask for that...
117 The trouble is that to type (runFoo foo), GHC has to solve the problem:
119 Given constraint Foo a b
120 Solve constraint Foo a b'
122 Notice that b and b' aren't the same. To solve this, just do
123 improvement and then they are the same. But GHC currently does
128 That is usually fine, but it isn't here, because it sees that Foo a b is
129 not the same as Foo a b', and so instead applies the instance decl for
130 instance Bar a b => Foo a b. And that's where the Bar constraint comes
133 The Right Thing is to improve whenever the constraint set changes at
134 all. Not hard in principle, but it'll take a bit of fiddling to do.
136 Note [Choosing which variables to quantify]
137 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
138 Suppose we are about to do a generalisation step. We have in our hand
141 T the type of the RHS
142 C the constraints from that RHS
144 The game is to figure out
146 Q the set of type variables over which to quantify
147 Ct the constraints we will *not* quantify over
148 Cq the constraints we will quantify over
150 So we're going to infer the type
154 and float the constraints Ct further outwards.
156 Here are the things that *must* be true:
158 (A) Q intersect fv(G) = EMPTY limits how big Q can be
159 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
161 (A) says we can't quantify over a variable that's free in the environment.
162 (B) says we must quantify over all the truly free variables in T, else
163 we won't get a sufficiently general type.
165 We do not *need* to quantify over any variable that is fixed by the
166 free vars of the environment G.
168 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
170 Example: class H x y | x->y where ...
172 fv(G) = {a} C = {H a b, H c d}
175 (A) Q intersect {a} is empty
176 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
178 So Q can be {c,d}, {b,c,d}
180 In particular, it's perfectly OK to quantify over more type variables
181 than strictly necessary; there is no need to quantify over 'b', since
182 it is determined by 'a' which is free in the envt, but it's perfectly
183 OK to do so. However we must not quantify over 'a' itself.
185 Other things being equal, however, we'd like to quantify over as few
186 variables as possible: smaller types, fewer type applications, more
187 constraints can get into Ct instead of Cq. Here's a good way to
190 Q = grow( fv(T), C ) \ oclose( fv(G), C )
192 That is, quantify over all variable that that MIGHT be fixed by the
193 call site (which influences T), but which aren't DEFINITELY fixed by
194 G. This choice definitely quantifies over enough type variables,
195 albeit perhaps too many.
197 Why grow( fv(T), C ) rather than fv(T)? Consider
199 class H x y | x->y where ...
204 If we used fv(T) = {c} we'd get the type
206 forall c. H c d => c -> b
208 And then if the fn was called at several different c's, each of
209 which fixed d differently, we'd get a unification error, because
210 d isn't quantified. Solution: quantify d. So we must quantify
211 everything that might be influenced by c.
213 Why not oclose( fv(T), C )? Because we might not be able to see
214 all the functional dependencies yet:
216 class H x y | x->y where ...
217 instance H x y => Eq (T x y) where ...
222 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
223 apparent yet, and that's wrong. We must really quantify over d too.
225 There really isn't any point in quantifying over any more than
226 grow( fv(T), C ), because the call sites can't possibly influence
227 any other type variables.
231 -------------------------------------
233 -------------------------------------
235 It's very hard to be certain when a type is ambiguous. Consider
239 instance H x y => K (x,y)
241 Is this type ambiguous?
242 forall a b. (K (a,b), Eq b) => a -> a
244 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
245 now we see that a fixes b. So we can't tell about ambiguity for sure
246 without doing a full simplification. And even that isn't possible if
247 the context has some free vars that may get unified. Urgle!
249 Here's another example: is this ambiguous?
250 forall a b. Eq (T b) => a -> a
251 Not if there's an insance decl (with no context)
252 instance Eq (T b) where ...
254 You may say of this example that we should use the instance decl right
255 away, but you can't always do that:
257 class J a b where ...
258 instance J Int b where ...
260 f :: forall a b. J a b => a -> a
262 (Notice: no functional dependency in J's class decl.)
263 Here f's type is perfectly fine, provided f is only called at Int.
264 It's premature to complain when meeting f's signature, or even
265 when inferring a type for f.
269 However, we don't *need* to report ambiguity right away. It'll always
270 show up at the call site.... and eventually at main, which needs special
271 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
273 So here's the plan. We WARN about probable ambiguity if
275 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
277 (all tested before quantification).
278 That is, all the type variables in Cq must be fixed by the the variables
279 in the environment, or by the variables in the type.
281 Notice that we union before calling oclose. Here's an example:
283 class J a b c | a b -> c
287 forall b c. (J a b c) => b -> b
289 Only if we union {a} from G with {b} from T before using oclose,
290 do we see that c is fixed.
292 It's a bit vague exactly which C we should use for this oclose call. If we
293 don't fix enough variables we might complain when we shouldn't (see
294 the above nasty example). Nothing will be perfect. That's why we can
295 only issue a warning.
298 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
300 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
302 then c is a "bubble"; there's no way it can ever improve, and it's
303 certainly ambiguous. UNLESS it is a constant (sigh). And what about
308 instance H x y => K (x,y)
310 Is this type ambiguous?
311 forall a b. (K (a,b), Eq b) => a -> a
313 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
314 is a "bubble" that's a set of constraints
316 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
318 Hence another idea. To decide Q start with fv(T) and grow it
319 by transitive closure in Cq (no functional dependencies involved).
320 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
321 The definitely-ambiguous can then float out, and get smashed at top level
322 (which squashes out the constants, like Eq (T a) above)
325 --------------------------------------
326 Notes on principal types
327 --------------------------------------
332 f x = let g y = op (y::Int) in True
334 Here the principal type of f is (forall a. a->a)
335 but we'll produce the non-principal type
336 f :: forall a. C Int => a -> a
339 --------------------------------------
340 The need for forall's in constraints
341 --------------------------------------
343 [Exchange on Haskell Cafe 5/6 Dec 2000]
345 class C t where op :: t -> Bool
346 instance C [t] where op x = True
348 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
349 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
351 The definitions of p and q differ only in the order of the components in
352 the pair on their right-hand sides. And yet:
354 ghc and "Typing Haskell in Haskell" reject p, but accept q;
355 Hugs rejects q, but accepts p;
356 hbc rejects both p and q;
357 nhc98 ... (Malcolm, can you fill in the blank for us!).
359 The type signature for f forces context reduction to take place, and
360 the results of this depend on whether or not the type of y is known,
361 which in turn depends on which component of the pair the type checker
364 Solution: if y::m a, float out the constraints
365 Monad m, forall c. C (m c)
366 When m is later unified with [], we can solve both constraints.
369 --------------------------------------
370 Notes on implicit parameters
371 --------------------------------------
373 Note [Inheriting implicit parameters]
374 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
379 where f is *not* a top-level binding.
380 From the RHS of f we'll get the constraint (?y::Int).
381 There are two types we might infer for f:
385 (so we get ?y from the context of f's definition), or
387 f :: (?y::Int) => Int -> Int
389 At first you might think the first was better, becuase then
390 ?y behaves like a free variable of the definition, rather than
391 having to be passed at each call site. But of course, the WHOLE
392 IDEA is that ?y should be passed at each call site (that's what
393 dynamic binding means) so we'd better infer the second.
395 BOTTOM LINE: when *inferring types* you *must* quantify
396 over implicit parameters. See the predicate isFreeWhenInferring.
399 Note [Implicit parameters and ambiguity]
400 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
401 What type should we infer for this?
402 f x = (show ?y, x::Int)
403 Since we must quantify over the ?y, the most plausible type is
404 f :: (Show a, ?y::a) => Int -> (String, Int)
405 But notice that the type of the RHS is (String,Int), with no type
406 varibables mentioned at all! The type of f looks ambiguous. But
407 it isn't, because at a call site we might have
408 let ?y = 5::Int in f 7
409 and all is well. In effect, implicit parameters are, well, parameters,
410 so we can take their type variables into account as part of the
411 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
414 Question 2: type signatures
415 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
416 BUT WATCH OUT: When you supply a type signature, we can't force you
417 to quantify over implicit parameters. For example:
421 This is perfectly reasonable. We do not want to insist on
423 (?x + 1) :: (?x::Int => Int)
425 That would be silly. Here, the definition site *is* the occurrence site,
426 so the above strictures don't apply. Hence the difference between
427 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
428 and tcSimplifyCheckBind (which does not).
430 What about when you supply a type signature for a binding?
431 Is it legal to give the following explicit, user type
432 signature to f, thus:
437 At first sight this seems reasonable, but it has the nasty property
438 that adding a type signature changes the dynamic semantics.
441 (let f x = (x::Int) + ?y
442 in (f 3, f 3 with ?y=5)) with ?y = 6
448 in (f 3, f 3 with ?y=5)) with ?y = 6
452 Indeed, simply inlining f (at the Haskell source level) would change the
455 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
456 semantics for a Haskell program without knowing its typing, so if you
457 change the typing you may change the semantics.
459 To make things consistent in all cases where we are *checking* against
460 a supplied signature (as opposed to inferring a type), we adopt the
463 a signature does not need to quantify over implicit params.
465 [This represents a (rather marginal) change of policy since GHC 5.02,
466 which *required* an explicit signature to quantify over all implicit
467 params for the reasons mentioned above.]
469 But that raises a new question. Consider
471 Given (signature) ?x::Int
472 Wanted (inferred) ?x::Int, ?y::Bool
474 Clearly we want to discharge the ?x and float the ?y out. But
475 what is the criterion that distinguishes them? Clearly it isn't
476 what free type variables they have. The Right Thing seems to be
477 to float a constraint that
478 neither mentions any of the quantified type variables
479 nor any of the quantified implicit parameters
481 See the predicate isFreeWhenChecking.
484 Question 3: monomorphism
485 ~~~~~~~~~~~~~~~~~~~~~~~~
486 There's a nasty corner case when the monomorphism restriction bites:
490 The argument above suggests that we *must* generalise
491 over the ?y parameter, to get
492 z :: (?y::Int) => Int,
493 but the monomorphism restriction says that we *must not*, giving
495 Why does the momomorphism restriction say this? Because if you have
497 let z = x + ?y in z+z
499 you might not expect the addition to be done twice --- but it will if
500 we follow the argument of Question 2 and generalise over ?y.
503 Question 4: top level
504 ~~~~~~~~~~~~~~~~~~~~~
505 At the top level, monomorhism makes no sense at all.
508 main = let ?x = 5 in print foo
512 woggle :: (?x :: Int) => Int -> Int
515 We definitely don't want (foo :: Int) with a top-level implicit parameter
516 (?x::Int) becuase there is no way to bind it.
521 (A) Always generalise over implicit parameters
522 Bindings that fall under the monomorphism restriction can't
526 * Inlining remains valid
527 * No unexpected loss of sharing
528 * But simple bindings like
530 will be rejected, unless you add an explicit type signature
531 (to avoid the monomorphism restriction)
532 z :: (?y::Int) => Int
534 This seems unacceptable
536 (B) Monomorphism restriction "wins"
537 Bindings that fall under the monomorphism restriction can't
539 Always generalise over implicit parameters *except* for bindings
540 that fall under the monomorphism restriction
543 * Inlining isn't valid in general
544 * No unexpected loss of sharing
545 * Simple bindings like
547 accepted (get value of ?y from binding site)
549 (C) Always generalise over implicit parameters
550 Bindings that fall under the monomorphism restriction can't
551 be generalised, EXCEPT for implicit parameters
553 * Inlining remains valid
554 * Unexpected loss of sharing (from the extra generalisation)
555 * Simple bindings like
557 accepted (get value of ?y from occurrence sites)
562 None of these choices seems very satisfactory. But at least we should
563 decide which we want to do.
565 It's really not clear what is the Right Thing To Do. If you see
569 would you expect the value of ?y to be got from the *occurrence sites*
570 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
571 case of function definitions, the answer is clearly the former, but
572 less so in the case of non-fucntion definitions. On the other hand,
573 if we say that we get the value of ?y from the definition site of 'z',
574 then inlining 'z' might change the semantics of the program.
576 Choice (C) really says "the monomorphism restriction doesn't apply
577 to implicit parameters". Which is fine, but remember that every
578 innocent binding 'x = ...' that mentions an implicit parameter in
579 the RHS becomes a *function* of that parameter, called at each
580 use of 'x'. Now, the chances are that there are no intervening 'with'
581 clauses that bind ?y, so a decent compiler should common up all
582 those function calls. So I think I strongly favour (C). Indeed,
583 one could make a similar argument for abolishing the monomorphism
584 restriction altogether.
586 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
590 %************************************************************************
592 \subsection{tcSimplifyInfer}
594 %************************************************************************
596 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
598 1. Compute Q = grow( fvs(T), C )
600 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
601 predicates will end up in Ct; we deal with them at the top level
603 3. Try improvement, using functional dependencies
605 4. If Step 3 did any unification, repeat from step 1
606 (Unification can change the result of 'grow'.)
608 Note: we don't reduce dictionaries in step 2. For example, if we have
609 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
610 after step 2. However note that we may therefore quantify over more
611 type variables than we absolutely have to.
613 For the guts, we need a loop, that alternates context reduction and
614 improvement with unification. E.g. Suppose we have
616 class C x y | x->y where ...
618 and tcSimplify is called with:
620 Then improvement unifies a with b, giving
623 If we need to unify anything, we rattle round the whole thing all over
630 -> TcTyVarSet -- fv(T); type vars
632 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
633 [Inst], -- Dict Ids that must be bound here (zonked)
634 TcDictBinds) -- Bindings
635 -- Any free (escaping) Insts are tossed into the environment
640 tcSimplifyInfer doc tau_tvs wanted
641 = do { tau_tvs1 <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
642 ; wanted' <- mappM zonkInst wanted -- Zonk before deciding quantified tyvars
643 ; gbl_tvs <- tcGetGlobalTyVars
644 ; let preds = fdPredsOfInsts wanted'
645 qtvs = grow preds tau_tvs1 `minusVarSet` oclose preds gbl_tvs
646 -- See Note [Choosing which variables to quantify]
648 -- To maximise sharing, remove from consideration any
649 -- constraints that don't mention qtvs at all
650 ; let (free, bound) = partition (isFreeWhenInferring qtvs) wanted'
653 -- To make types simple, reduce as much as possible
654 ; traceTc (text "infer" <+> (ppr preds $$ ppr (grow preds tau_tvs1) $$ ppr gbl_tvs $$
655 ppr (oclose preds gbl_tvs) $$ ppr free $$ ppr bound))
656 ; (irreds1, binds1) <- tryHardCheckLoop doc bound
658 -- Note [Inference and implication constraints]
659 ; let want_dict d = tyVarsOfInst d `intersectsVarSet` qtvs
660 ; (irreds2, binds2) <- approximateImplications doc want_dict irreds1
662 -- Now work out all over again which type variables to quantify,
663 -- exactly in the same way as before, but starting from irreds2. Why?
664 -- a) By now improvment may have taken place, and we must *not*
665 -- quantify over any variable free in the environment
666 -- tc137 (function h inside g) is an example
668 -- b) Do not quantify over constraints that *now* do not
669 -- mention quantified type variables, because they are
670 -- simply ambiguous (or might be bound further out). Example:
671 -- f :: Eq b => a -> (a, b)
673 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
674 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
675 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
676 -- constraint (Eq beta), which we dump back into the free set
677 -- See test tcfail181
679 -- c) irreds may contain type variables not previously mentioned,
680 -- e.g. instance D a x => Foo [a]
682 -- Then after simplifying we'll get (D a x), and x is fresh
683 -- We must quantify over x else it'll be totally unbound
684 ; tau_tvs2 <- zonkTcTyVarsAndFV (varSetElems tau_tvs1)
685 ; gbl_tvs <- tcGetGlobalTyVars
686 ; let preds = fdPredsOfInsts irreds2 -- irreds2 is zonked
687 qtvs = grow preds tau_tvs2 `minusVarSet` oclose preds gbl_tvs
688 ; let (free, irreds3) = partition (isFreeWhenInferring qtvs) irreds2
691 -- Turn the quantified meta-type variables into real type variables
692 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems qtvs)
694 -- We can't abstract over any remaining unsolved
695 -- implications so instead just float them outwards. Ugh.
696 ; let (q_dicts, implics) = partition isDict irreds3
697 ; loc <- getInstLoc (ImplicOrigin doc)
698 ; implic_bind <- bindIrreds loc qtvs2 q_dicts implics
700 ; return (qtvs2, q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
701 -- NB: when we are done, we might have some bindings, but
702 -- the final qtvs might be empty. See Note [NO TYVARS] below.
704 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
705 -- Note [Inference and implication constraints]
706 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
707 -- - fetching any dicts inside them that are free
708 -- - using those dicts as cruder constraints, to solve the implications
709 -- - returning the extra ones too
711 approximateImplications doc want_dict irreds
713 = return (irreds, emptyBag)
715 = do { extra_dicts' <- mapM cloneDict extra_dicts
716 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
717 -- By adding extra_dicts', we make them
718 -- available to solve the implication constraints
720 extra_dicts = get_dicts (filter isImplicInst irreds)
722 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
723 -- Find the wanted constraints in implication constraints that satisfy
724 -- want_dict, and are not bound by forall's in the constraint itself
725 get_dicts ds = concatMap get_dict ds
727 get_dict d@(Dict {}) | want_dict d = [d]
729 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
730 = [ d | let tv_set = mkVarSet tvs
731 , d <- get_dicts wanteds
732 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
733 get_dict other = pprPanic "approximateImplications" (ppr other)
736 Note [Inference and implication constraints]
737 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
738 Suppose we have a wanted implication constraint (perhaps arising from
739 a nested pattern match) like
741 and we are now trying to quantify over 'a' when inferring the type for
742 a function. In principle it's possible that there might be an instance
743 instance (C a, E a) => D [a]
744 so the context (E a) would suffice. The Right Thing is to abstract over
745 the implication constraint, but we don't do that (a) because it'll be
746 surprising to programmers and (b) because we don't have the machinery to deal
747 with 'given' implications.
749 So our best approximation is to make (D [a]) part of the inferred
750 context, so we can use that to discharge the implication. Hence
751 the strange function get_dictsin approximateImplications.
753 The common cases are more clear-cut, when we have things like
755 Here, abstracting over (C b) is not an approximation at all -- but see
756 Note [Freeness and implications].
758 See Trac #1430 and test tc228.
762 -----------------------------------------------------------
763 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
764 -- against, but we don't know the type variables over which we are going to quantify.
765 -- This happens when we have a type signature for a mutually recursive group
768 -> TcTyVarSet -- fv(T)
771 -> TcM ([TyVar], -- Fully zonked, and quantified
772 TcDictBinds) -- Bindings
774 tcSimplifyInferCheck loc tau_tvs givens wanteds
775 = do { (irreds, binds) <- gentleCheckLoop loc givens wanteds
777 -- Figure out which type variables to quantify over
778 -- You might think it should just be the signature tyvars,
779 -- but in bizarre cases you can get extra ones
780 -- f :: forall a. Num a => a -> a
781 -- f x = fst (g (x, head [])) + 1
783 -- Here we infer g :: forall a b. a -> b -> (b,a)
784 -- We don't want g to be monomorphic in b just because
785 -- f isn't quantified over b.
786 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
787 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
788 ; gbl_tvs <- tcGetGlobalTyVars
789 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
790 -- We could close gbl_tvs, but its not necessary for
791 -- soundness, and it'll only affect which tyvars, not which
792 -- dictionaries, we quantify over
794 ; qtvs' <- zonkQuantifiedTyVars qtvs
796 -- Now we are back to normal (c.f. tcSimplCheck)
797 ; implic_bind <- bindIrreds loc qtvs' givens irreds
799 ; return (qtvs', binds `unionBags` implic_bind) }
802 Note [Squashing methods]
803 ~~~~~~~~~~~~~~~~~~~~~~~~~
804 Be careful if you want to float methods more:
805 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
806 From an application (truncate f i) we get
809 If we have also have a second occurrence of truncate, we get
812 When simplifying with i,f free, we might still notice that
813 t1=t3; but alas, the binding for t2 (which mentions t1)
814 may continue to float out!
819 class Y a b | a -> b where
822 instance Y [[a]] a where
825 k :: X a -> X a -> X a
827 g :: Num a => [X a] -> [X a]
830 h ys = ys ++ map (k (y [[0]])) xs
832 The excitement comes when simplifying the bindings for h. Initially
833 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
834 From this we get t1:=:t2, but also various bindings. We can't forget
835 the bindings (because of [LOOP]), but in fact t1 is what g is
838 The net effect of [NO TYVARS]
841 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
842 isFreeWhenInferring qtvs inst
843 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
844 && isInheritableInst inst -- and no implicit parameter involved
845 -- see Note [Inheriting implicit parameters]
847 {- No longer used (with implication constraints)
848 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
849 -> NameSet -- Quantified implicit parameters
851 isFreeWhenChecking qtvs ips inst
852 = isFreeWrtTyVars qtvs inst
853 && isFreeWrtIPs ips inst
856 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
857 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
861 %************************************************************************
863 \subsection{tcSimplifyCheck}
865 %************************************************************************
867 @tcSimplifyCheck@ is used when we know exactly the set of variables
868 we are going to quantify over. For example, a class or instance declaration.
871 -----------------------------------------------------------
872 -- tcSimplifyCheck is used when checking expression type signatures,
873 -- class decls, instance decls etc.
874 tcSimplifyCheck :: InstLoc
875 -> [TcTyVar] -- Quantify over these
878 -> TcM TcDictBinds -- Bindings
879 tcSimplifyCheck loc qtvs givens wanteds
880 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
881 do { (irreds, binds) <- gentleCheckLoop loc givens wanteds
882 ; implic_bind <- bindIrreds loc qtvs givens irreds
883 ; return (binds `unionBags` implic_bind) }
885 -----------------------------------------------------------
886 -- tcSimplifyCheckPat is used for existential pattern match
887 tcSimplifyCheckPat :: InstLoc
888 -> [CoVar] -> Refinement
889 -> [TcTyVar] -- Quantify over these
892 -> TcM TcDictBinds -- Bindings
893 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
894 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
895 do { (irreds, binds) <- gentleCheckLoop loc givens wanteds
896 ; implic_bind <- bindIrredsR loc qtvs co_vars reft
898 ; return (binds `unionBags` implic_bind) }
900 -----------------------------------------------------------
901 bindIrreds :: InstLoc -> [TcTyVar]
904 bindIrreds loc qtvs givens irreds
905 = bindIrredsR loc qtvs [] emptyRefinement givens irreds
907 bindIrredsR :: InstLoc -> [TcTyVar] -> [CoVar]
908 -> Refinement -> [Inst] -> [Inst]
910 -- Make a binding that binds 'irreds', by generating an implication
911 -- constraint for them, *and* throwing the constraint into the LIE
912 bindIrredsR loc qtvs co_vars reft givens irreds
916 = do { let givens' = filter isDict givens
917 -- The givens can include methods
918 -- See Note [Pruning the givens in an implication constraint]
920 -- If there are no 'givens' *and* the refinement is empty
921 -- (the refinement is like more givens), then it's safe to
922 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
923 -- See Note [Freeness and implications]
924 ; irreds' <- if null givens' && isEmptyRefinement reft
926 { let qtv_set = mkVarSet qtvs
927 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
929 ; return real_irreds }
932 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
933 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
934 -- This call does the real work
935 -- If irreds' is empty, it does something sensible
940 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
942 -> TcM ([Inst], TcDictBinds)
943 -- Make a binding that binds 'irreds', by generating an implication
944 -- constraint for them, *and* throwing the constraint into the LIE
945 -- The binding looks like
946 -- (ir1, .., irn) = f qtvs givens
947 -- where f is (evidence for) the new implication constraint
948 -- f :: forall qtvs. {reft} givens => (ir1, .., irn)
949 -- qtvs includes coercion variables
951 -- This binding must line up the 'rhs' in reduceImplication
952 makeImplicationBind loc all_tvs reft
953 givens -- Guaranteed all Dicts
955 | null irreds -- If there are no irreds, we are done
956 = return ([], emptyBag)
957 | otherwise -- Otherwise we must generate a binding
958 = do { uniq <- newUnique
959 ; span <- getSrcSpanM
960 ; let name = mkInternalName uniq (mkVarOcc "ic") span
961 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
962 tci_tyvars = all_tvs,
964 tci_wanted = irreds, tci_loc = loc }
966 ; let n_irreds = length irreds
967 irred_ids = map instToId irreds
968 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
969 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
970 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
971 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
972 bind | n_irreds==1 = VarBind (head irred_ids) rhs
973 | otherwise = PatBind { pat_lhs = L span pat,
974 pat_rhs = unguardedGRHSs rhs,
976 bind_fvs = placeHolderNames }
977 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
978 return ([implic_inst], unitBag (L span bind)) }
980 -----------------------------------------------------------
981 tryHardCheckLoop :: SDoc
983 -> TcM ([Inst], TcDictBinds)
985 tryHardCheckLoop doc wanteds
986 = checkLoop (mkRedEnv doc try_me []) wanteds
988 try_me inst = ReduceMe AddSCs
989 -- Here's the try-hard bit
991 -----------------------------------------------------------
992 gentleCheckLoop :: InstLoc
995 -> TcM ([Inst], TcDictBinds)
997 gentleCheckLoop inst_loc givens wanteds
998 = checkLoop env wanteds
1000 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1002 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
1004 -- When checking against a given signature
1005 -- we MUST be very gentle: Note [Check gently]
1009 ~~~~~~~~~~~~~~~~~~~~
1010 We have to very careful about not simplifying too vigorously
1015 f :: Show b => T b -> b
1016 f (MkT x) = show [x]
1018 Inside the pattern match, which binds (a:*, x:a), we know that
1020 Hence we have a dictionary for Show [a] available; and indeed we
1021 need it. We are going to build an implication contraint
1022 forall a. (b~[a]) => Show [a]
1023 Later, we will solve this constraint using the knowledg e(Show b)
1025 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1026 thing becomes insoluble. So we simplify gently (get rid of literals
1027 and methods only, plus common up equal things), deferring the real
1028 work until top level, when we solve the implication constraint
1029 with tryHardCheckLooop.
1033 -----------------------------------------------------------
1036 -> TcM ([Inst], TcDictBinds)
1037 -- Precondition: givens are completely rigid
1038 -- Postcondition: returned Insts are zonked
1040 checkLoop env wanteds
1041 = do { -- Givens are skolems, so no need to zonk them
1042 wanteds' <- mappM zonkInst wanteds
1044 ; (improved, binds, irreds) <- reduceContext env wanteds'
1046 ; if not improved then
1047 return (irreds, binds)
1050 -- If improvement did some unification, we go round again.
1051 -- We start again with irreds, not wanteds
1052 -- Using an instance decl might have introduced a fresh type variable
1053 -- which might have been unified, so we'd get an infinite loop
1054 -- if we started again with wanteds! See Note [LOOP]
1055 { (irreds1, binds1) <- checkLoop env irreds
1056 ; return (irreds1, binds `unionBags` binds1) } }
1061 class If b t e r | b t e -> r
1064 class Lte a b c | a b -> c where lte :: a -> b -> c
1066 instance (Lte a b l,If l b a c) => Max a b c
1068 Wanted: Max Z (S x) y
1070 Then we'll reduce using the Max instance to:
1071 (Lte Z (S x) l, If l (S x) Z y)
1072 and improve by binding l->T, after which we can do some reduction
1073 on both the Lte and If constraints. What we *can't* do is start again
1074 with (Max Z (S x) y)!
1078 %************************************************************************
1080 tcSimplifySuperClasses
1082 %************************************************************************
1084 Note [SUPERCLASS-LOOP 1]
1085 ~~~~~~~~~~~~~~~~~~~~~~~~
1086 We have to be very, very careful when generating superclasses, lest we
1087 accidentally build a loop. Here's an example:
1091 class S a => C a where { opc :: a -> a }
1092 class S b => D b where { opd :: b -> b }
1094 instance C Int where
1097 instance D Int where
1100 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1101 Simplifying, we may well get:
1102 $dfCInt = :C ds1 (opd dd)
1105 Notice that we spot that we can extract ds1 from dd.
1107 Alas! Alack! We can do the same for (instance D Int):
1109 $dfDInt = :D ds2 (opc dc)
1113 And now we've defined the superclass in terms of itself.
1115 Solution: never generate a superclass selectors at all when
1116 satisfying the superclass context of an instance declaration.
1118 Two more nasty cases are in
1123 tcSimplifySuperClasses
1128 tcSimplifySuperClasses loc givens sc_wanteds
1129 = do { (irreds, binds1) <- checkLoop env sc_wanteds
1130 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1131 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1134 env = mkRedEnv (pprInstLoc loc) try_me givens
1135 try_me inst = ReduceMe NoSCs
1136 -- Like tryHardCheckLoop, but with NoSCs
1140 %************************************************************************
1142 \subsection{tcSimplifyRestricted}
1144 %************************************************************************
1146 tcSimplifyRestricted infers which type variables to quantify for a
1147 group of restricted bindings. This isn't trivial.
1150 We want to quantify over a to get id :: forall a. a->a
1153 We do not want to quantify over a, because there's an Eq a
1154 constraint, so we get eq :: a->a->Bool (notice no forall)
1157 RHS has type 'tau', whose free tyvars are tau_tvs
1158 RHS has constraints 'wanteds'
1161 Quantify over (tau_tvs \ ftvs(wanteds))
1162 This is bad. The constraints may contain (Monad (ST s))
1163 where we have instance Monad (ST s) where...
1164 so there's no need to be monomorphic in s!
1166 Also the constraint might be a method constraint,
1167 whose type mentions a perfectly innocent tyvar:
1168 op :: Num a => a -> b -> a
1169 Here, b is unconstrained. A good example would be
1171 We want to infer the polymorphic type
1172 foo :: forall b. b -> b
1175 Plan B (cunning, used for a long time up to and including GHC 6.2)
1176 Step 1: Simplify the constraints as much as possible (to deal
1177 with Plan A's problem). Then set
1178 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1180 Step 2: Now simplify again, treating the constraint as 'free' if
1181 it does not mention qtvs, and trying to reduce it otherwise.
1182 The reasons for this is to maximise sharing.
1184 This fails for a very subtle reason. Suppose that in the Step 2
1185 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1186 In the Step 1 this constraint might have been simplified, perhaps to
1187 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1188 This won't happen in Step 2... but that in turn might prevent some other
1189 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1190 and that in turn breaks the invariant that no constraints are quantified over.
1192 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1197 Step 1: Simplify the constraints as much as possible (to deal
1198 with Plan A's problem). Then set
1199 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1200 Return the bindings from Step 1.
1203 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1206 instance (HasBinary ty IO) => HasCodedValue ty
1208 foo :: HasCodedValue a => String -> IO a
1210 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1211 doDecodeIO codedValue view
1212 = let { act = foo "foo" } in act
1214 You might think this should work becuase the call to foo gives rise to a constraint
1215 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1216 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1217 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1219 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1223 Plan D (a variant of plan B)
1224 Step 1: Simplify the constraints as much as possible (to deal
1225 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1226 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1228 Step 2: Now simplify again, treating the constraint as 'free' if
1229 it does not mention qtvs, and trying to reduce it otherwise.
1231 The point here is that it's generally OK to have too few qtvs; that is,
1232 to make the thing more monomorphic than it could be. We don't want to
1233 do that in the common cases, but in wierd cases it's ok: the programmer
1234 can always add a signature.
1236 Too few qtvs => too many wanteds, which is what happens if you do less
1241 tcSimplifyRestricted -- Used for restricted binding groups
1242 -- i.e. ones subject to the monomorphism restriction
1245 -> [Name] -- Things bound in this group
1246 -> TcTyVarSet -- Free in the type of the RHSs
1247 -> [Inst] -- Free in the RHSs
1248 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1249 TcDictBinds) -- Bindings
1250 -- tcSimpifyRestricted returns no constraints to
1251 -- quantify over; by definition there are none.
1252 -- They are all thrown back in the LIE
1254 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1255 -- Zonk everything in sight
1256 = do { wanteds' <- mappM zonkInst wanteds
1258 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1259 -- dicts; the idea is to get rid of as many type
1260 -- variables as possible, and we don't want to stop
1261 -- at (say) Monad (ST s), because that reduces
1262 -- immediately, with no constraint on s.
1264 -- BUT do no improvement! See Plan D above
1265 -- HOWEVER, some unification may take place, if we instantiate
1266 -- a method Inst with an equality constraint
1267 ; let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1268 ; (_imp, _binds, constrained_dicts) <- reduceContext env wanteds'
1270 -- Next, figure out the tyvars we will quantify over
1271 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1272 ; gbl_tvs' <- tcGetGlobalTyVars
1273 ; constrained_dicts' <- mappM zonkInst constrained_dicts
1275 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1276 -- As in tcSimplifyInfer
1278 -- Do not quantify over constrained type variables:
1279 -- this is the monomorphism restriction
1280 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1281 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1282 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1285 ; warn_mono <- doptM Opt_WarnMonomorphism
1286 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1287 (vcat[ ptext SLIT("the Monomorphism Restriction applies to the binding")
1288 <> plural bndrs <+> ptext SLIT("for") <+> pp_bndrs,
1289 ptext SLIT("Consider giving a type signature for") <+> pp_bndrs])
1291 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1292 pprInsts wanteds, pprInsts constrained_dicts',
1294 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1296 -- The first step may have squashed more methods than
1297 -- necessary, so try again, this time more gently, knowing the exact
1298 -- set of type variables to quantify over.
1300 -- We quantify only over constraints that are captured by qtvs;
1301 -- these will just be a subset of non-dicts. This in contrast
1302 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1303 -- all *non-inheritable* constraints too. This implements choice
1304 -- (B) under "implicit parameter and monomorphism" above.
1306 -- Remember that we may need to do *some* simplification, to
1307 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1308 -- just to float all constraints
1310 -- At top level, we *do* squash methods becuase we want to
1311 -- expose implicit parameters to the test that follows
1312 ; let is_nested_group = isNotTopLevel top_lvl
1313 try_me inst | isFreeWrtTyVars qtvs inst,
1314 (is_nested_group || isDict inst) = Stop
1315 | otherwise = ReduceMe AddSCs
1316 env = mkNoImproveRedEnv doc try_me
1317 ; (_imp, binds, irreds) <- reduceContext env wanteds'
1319 -- See "Notes on implicit parameters, Question 4: top level"
1320 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1321 if is_nested_group then
1323 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1324 ; addTopIPErrs bndrs bad_ips
1325 ; extendLIEs non_ips }
1327 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1328 ; return (qtvs', binds) }
1332 %************************************************************************
1336 %************************************************************************
1338 On the LHS of transformation rules we only simplify methods and constants,
1339 getting dictionaries. We want to keep all of them unsimplified, to serve
1340 as the available stuff for the RHS of the rule.
1342 Example. Consider the following left-hand side of a rule
1344 f (x == y) (y > z) = ...
1346 If we typecheck this expression we get constraints
1348 d1 :: Ord a, d2 :: Eq a
1350 We do NOT want to "simplify" to the LHS
1352 forall x::a, y::a, z::a, d1::Ord a.
1353 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1357 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1358 f ((==) d2 x y) ((>) d1 y z) = ...
1360 Here is another example:
1362 fromIntegral :: (Integral a, Num b) => a -> b
1363 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1365 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1366 we *dont* want to get
1368 forall dIntegralInt.
1369 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1371 because the scsel will mess up RULE matching. Instead we want
1373 forall dIntegralInt, dNumInt.
1374 fromIntegral Int Int dIntegralInt dNumInt = id Int
1378 g (x == y) (y == z) = ..
1380 where the two dictionaries are *identical*, we do NOT WANT
1382 forall x::a, y::a, z::a, d1::Eq a
1383 f ((==) d1 x y) ((>) d1 y z) = ...
1385 because that will only match if the dict args are (visibly) equal.
1386 Instead we want to quantify over the dictionaries separately.
1388 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1389 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1390 from scratch, rather than further parameterise simpleReduceLoop etc
1393 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1394 tcSimplifyRuleLhs wanteds
1395 = go [] emptyBag wanteds
1398 = return (dicts, binds)
1399 go dicts binds (w:ws)
1401 = go (w:dicts) binds ws
1403 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1404 -- to fromInteger; this looks fragile to me
1405 ; lookup_result <- lookupSimpleInst w'
1406 ; case lookup_result of
1407 GenInst ws' rhs -> go dicts (addBind binds (instToId w) rhs) (ws' ++ ws)
1408 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1412 tcSimplifyBracket is used when simplifying the constraints arising from
1413 a Template Haskell bracket [| ... |]. We want to check that there aren't
1414 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1415 Show instance), but we aren't otherwise interested in the results.
1416 Nor do we care about ambiguous dictionaries etc. We will type check
1417 this bracket again at its usage site.
1420 tcSimplifyBracket :: [Inst] -> TcM ()
1421 tcSimplifyBracket wanteds
1422 = do { tryHardCheckLoop doc wanteds
1425 doc = text "tcSimplifyBracket"
1429 %************************************************************************
1431 \subsection{Filtering at a dynamic binding}
1433 %************************************************************************
1438 we must discharge all the ?x constraints from B. We also do an improvement
1439 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1441 Actually, the constraints from B might improve the types in ?x. For example
1443 f :: (?x::Int) => Char -> Char
1446 then the constraint (?x::Int) arising from the call to f will
1447 force the binding for ?x to be of type Int.
1450 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1453 -- We need a loop so that we do improvement, and then
1454 -- (next time round) generate a binding to connect the two
1456 -- Here the two ?x's have different types, and improvement
1457 -- makes them the same.
1459 tcSimplifyIPs given_ips wanteds
1460 = do { wanteds' <- mappM zonkInst wanteds
1461 ; given_ips' <- mappM zonkInst given_ips
1462 -- Unusually for checking, we *must* zonk the given_ips
1464 ; let env = mkRedEnv doc try_me given_ips'
1465 ; (improved, binds, irreds) <- reduceContext env wanteds'
1467 ; if not improved then
1468 ASSERT( all is_free irreds )
1469 do { extendLIEs irreds
1472 tcSimplifyIPs given_ips wanteds }
1474 doc = text "tcSimplifyIPs" <+> ppr given_ips
1475 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1476 is_free inst = isFreeWrtIPs ip_set inst
1478 -- Simplify any methods that mention the implicit parameter
1479 try_me inst | is_free inst = Stop
1480 | otherwise = ReduceMe NoSCs
1484 %************************************************************************
1486 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1488 %************************************************************************
1490 When doing a binding group, we may have @Insts@ of local functions.
1491 For example, we might have...
1493 let f x = x + 1 -- orig local function (overloaded)
1494 f.1 = f Int -- two instances of f
1499 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1500 where @f@ is in scope; those @Insts@ must certainly not be passed
1501 upwards towards the top-level. If the @Insts@ were binding-ified up
1502 there, they would have unresolvable references to @f@.
1504 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1505 For each method @Inst@ in the @init_lie@ that mentions one of the
1506 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1507 @LIE@), as well as the @HsBinds@ generated.
1510 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1511 -- Simlifies only MethodInsts, and generate only bindings of form
1513 -- We're careful not to even generate bindings of the form
1515 -- You'd think that'd be fine, but it interacts with what is
1516 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1518 bindInstsOfLocalFuns wanteds local_ids
1519 | null overloaded_ids
1521 = extendLIEs wanteds `thenM_`
1522 returnM emptyLHsBinds
1525 = do { (irreds, binds) <- checkLoop env for_me
1526 ; extendLIEs not_for_me
1530 env = mkRedEnv doc try_me []
1531 doc = text "bindInsts" <+> ppr local_ids
1532 overloaded_ids = filter is_overloaded local_ids
1533 is_overloaded id = isOverloadedTy (idType id)
1534 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1536 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1537 -- so it's worth building a set, so that
1538 -- lookup (in isMethodFor) is faster
1539 try_me inst | isMethod inst = ReduceMe NoSCs
1544 %************************************************************************
1546 \subsection{Data types for the reduction mechanism}
1548 %************************************************************************
1550 The main control over context reduction is here
1554 = RedEnv { red_doc :: SDoc -- The context
1555 , red_try_me :: Inst -> WhatToDo
1556 , red_improve :: Bool -- True <=> do improvement
1557 , red_givens :: [Inst] -- All guaranteed rigid
1559 -- but see Note [Rigidity]
1560 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1561 -- See Note [RedStack]
1565 -- The red_givens are rigid so far as cmpInst is concerned.
1566 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1567 -- let ?x = e in ...
1568 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1569 -- But that doesn't affect the comparison, which is based only on mame.
1572 -- The red_stack pair (n,insts) pair is just used for error reporting.
1573 -- 'n' is always the depth of the stack.
1574 -- The 'insts' is the stack of Insts being reduced: to produce X
1575 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1578 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1579 mkRedEnv doc try_me givens
1580 = RedEnv { red_doc = doc, red_try_me = try_me,
1581 red_givens = givens, red_stack = (0,[]),
1582 red_improve = True }
1584 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1585 -- Do not do improvement; no givens
1586 mkNoImproveRedEnv doc try_me
1587 = RedEnv { red_doc = doc, red_try_me = try_me,
1588 red_givens = [], red_stack = (0,[]),
1589 red_improve = True }
1592 = ReduceMe WantSCs -- Try to reduce this
1593 -- If there's no instance, add the inst to the
1594 -- irreductible ones, but don't produce an error
1595 -- message of any kind.
1596 -- It might be quite legitimate such as (Eq a)!
1598 | Stop -- Return as irreducible unless it can
1599 -- be reduced to a constant in one step
1600 -- Do not add superclasses; see
1602 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1603 -- of a predicate when adding it to the avails
1604 -- The reason for this flag is entirely the super-class loop problem
1605 -- Note [SUPER-CLASS LOOP 1]
1608 %************************************************************************
1610 \subsection[reduce]{@reduce@}
1612 %************************************************************************
1616 reduceContext :: RedEnv
1618 -> TcM (ImprovementDone,
1619 TcDictBinds, -- Dictionary bindings
1620 [Inst]) -- Irreducible
1622 reduceContext env wanteds
1623 = do { traceTc (text "reduceContext" <+> (vcat [
1624 text "----------------------",
1626 text "given" <+> ppr (red_givens env),
1627 text "wanted" <+> ppr wanteds,
1628 text "----------------------"
1631 -- Build the Avail mapping from "givens"
1632 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1635 -- Process non-implication constraints first, so that they are
1636 -- available to help solving the implication constraints
1637 -- ToDo: seems a bit inefficient and ad-hoc
1638 ; let (implics, rest) = partition isImplicInst wanteds
1639 ; avails <- reduceList env (rest ++ implics) init_state
1641 ; let improved = availsImproved avails
1642 ; (binds, irreds) <- extractResults avails wanteds
1644 ; traceTc (text "reduceContext end" <+> (vcat [
1645 text "----------------------",
1647 text "given" <+> ppr (red_givens env),
1648 text "wanted" <+> ppr wanteds,
1650 text "avails" <+> pprAvails avails,
1651 text "improved =" <+> ppr improved,
1652 text "irreds = " <+> ppr irreds,
1653 text "----------------------"
1656 ; return (improved, binds, irreds) }
1658 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1659 tcImproveOne avails inst
1660 | not (isDict inst) = return False
1662 = do { inst_envs <- tcGetInstEnvs
1663 ; let eqns = improveOne (classInstances inst_envs)
1664 (dictPred inst, pprInstArising inst)
1665 [ (dictPred p, pprInstArising p)
1666 | p <- availsInsts avails, isDict p ]
1667 -- Avails has all the superclasses etc (good)
1668 -- It also has all the intermediates of the deduction (good)
1669 -- It does not have duplicates (good)
1670 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1671 -- so that improve will see them separate
1672 ; traceTc (text "improveOne" <+> ppr inst)
1675 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1676 -> TcM ImprovementDone
1677 unifyEqns [] = return False
1679 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1683 unify ((qtvs, pairs), what1, what2)
1684 = addErrCtxtM (mkEqnMsg what1 what2) $
1685 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1686 mapM_ (unif_pr tenv) pairs
1687 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1689 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1691 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1692 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1693 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1694 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1695 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1696 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1697 ; return (tidy_env, msg) }
1700 The main context-reduction function is @reduce@. Here's its game plan.
1703 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1704 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1705 = do { dopts <- getDOpts
1708 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1709 2 (ifPprDebug (nest 2 (pprStack stk))))
1712 ; if n >= ctxtStkDepth dopts then
1713 failWithTc (reduceDepthErr n stk)
1717 go [] state = return state
1718 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1721 -- Base case: we're done!
1722 reduce env wanted avails
1723 -- It's the same as an existing inst, or a superclass thereof
1724 | Just avail <- findAvail avails wanted
1728 = case red_try_me env wanted of {
1729 ; Stop -> try_simple (addIrred NoSCs) -- See Note [No superclasses for Stop]
1731 ; ReduceMe want_scs -> -- It should be reduced
1732 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1733 case lookup_result of
1734 NoInstance -> -- No such instance!
1735 -- Add it and its superclasses
1736 addIrred want_scs avails wanted
1738 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1740 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1741 ; avails2 <- reduceList env wanteds' avails1
1742 ; addWanted want_scs avails2 wanted rhs wanteds' }
1743 -- Temporarily do addIrred *before* the reduceList,
1744 -- which has the effect of adding the thing we are trying
1745 -- to prove to the database before trying to prove the things it
1746 -- needs. See note [RECURSIVE DICTIONARIES]
1747 -- NB: we must not do an addWanted before, because that adds the
1748 -- superclasses too, and thaat can lead to a spurious loop; see
1749 -- the examples in [SUPERCLASS-LOOP]
1750 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1754 -- First, see if the inst can be reduced to a constant in one step
1755 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1756 -- Don't bother for implication constraints, which take real work
1757 try_simple do_this_otherwise
1758 = do { res <- lookupSimpleInst wanted
1760 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1761 other -> do_this_otherwise avails wanted }
1765 Note [SUPERCLASS-LOOP 2]
1766 ~~~~~~~~~~~~~~~~~~~~~~~~
1767 But the above isn't enough. Suppose we are *given* d1:Ord a,
1768 and want to deduce (d2:C [a]) where
1770 class Ord a => C a where
1771 instance Ord [a] => C [a] where ...
1773 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1774 superclasses of C [a] to avails. But we must not overwrite the binding
1775 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1778 Here's another variant, immortalised in tcrun020
1779 class Monad m => C1 m
1780 class C1 m => C2 m x
1781 instance C2 Maybe Bool
1782 For the instance decl we need to build (C1 Maybe), and it's no good if
1783 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1784 before we search for C1 Maybe.
1786 Here's another example
1787 class Eq b => Foo a b
1788 instance Eq a => Foo [a] a
1792 we'll first deduce that it holds (via the instance decl). We must not
1793 then overwrite the Eq t constraint with a superclass selection!
1795 At first I had a gross hack, whereby I simply did not add superclass constraints
1796 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1797 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1798 I found a very obscure program (now tcrun021) in which improvement meant the
1799 simplifier got two bites a the cherry... so something seemed to be an Stop
1800 first time, but reducible next time.
1802 Now we implement the Right Solution, which is to check for loops directly
1803 when adding superclasses. It's a bit like the occurs check in unification.
1806 Note [RECURSIVE DICTIONARIES]
1807 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1809 data D r = ZeroD | SuccD (r (D r));
1811 instance (Eq (r (D r))) => Eq (D r) where
1812 ZeroD == ZeroD = True
1813 (SuccD a) == (SuccD b) = a == b
1816 equalDC :: D [] -> D [] -> Bool;
1819 We need to prove (Eq (D [])). Here's how we go:
1823 by instance decl, holds if
1827 by instance decl of Eq, holds if
1829 where d2 = dfEqList d3
1832 But now we can "tie the knot" to give
1838 and it'll even run! The trick is to put the thing we are trying to prove
1839 (in this case Eq (D []) into the database before trying to prove its
1840 contributing clauses.
1843 %************************************************************************
1845 Reducing a single constraint
1847 %************************************************************************
1850 ---------------------------------------------
1851 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1852 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1853 tci_given = extra_givens, tci_wanted = wanteds })
1854 = reduceImplication env avails reft tvs extra_givens wanteds loc
1856 reduceInst env avails other_inst
1857 = do { result <- lookupSimpleInst other_inst
1858 ; return (avails, result) }
1862 ---------------------------------------------
1863 reduceImplication :: RedEnv
1865 -> Refinement -- May refine the givens; often empty
1866 -> [TcTyVar] -- Quantified type variables; all skolems
1867 -> [Inst] -- Extra givens; all rigid
1870 -> TcM (Avails, LookupInstResult)
1873 Suppose we are simplifying the constraint
1874 forall bs. extras => wanted
1875 in the context of an overall simplification problem with givens 'givens',
1876 and refinment 'reft'.
1879 * The refinement is often empty
1881 * The 'extra givens' need not mention any of the quantified type variables
1882 e.g. forall {}. Eq a => Eq [a]
1883 forall {}. C Int => D (Tree Int)
1885 This happens when you have something like
1887 T1 :: Eq a => a -> T a
1890 f x = ...(case x of { T1 v -> v==v })...
1893 -- ToDo: should we instantiate tvs? I think it's not necessary
1895 -- ToDo: what about improvement? There may be some improvement
1896 -- exposed as a result of the simplifications done by reduceList
1897 -- which are discarded if we back off.
1898 -- This is almost certainly Wrong, but we'll fix it when dealing
1899 -- better with equality constraints
1900 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1901 = do { -- Add refined givens, and the extra givens
1902 (refined_red_givens, avails)
1903 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1904 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1905 ; avails <- foldlM addGiven avails extra_givens
1907 -- Solve the sub-problem
1908 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1909 env' = env { red_givens = refined_red_givens ++ extra_givens
1910 , red_try_me = try_me }
1912 ; traceTc (text "reduceImplication" <+> vcat
1914 ppr (red_givens env), ppr extra_givens,
1915 ppr reft, ppr wanteds, ppr avails ])
1916 ; avails <- reduceList env' wanteds avails
1918 -- Extract the results
1919 -- Note [Reducing implication constraints]
1920 ; (binds, irreds) <- extractResults avails wanteds
1921 ; let (outer, inner) = partition (isJust . findAvail orig_avails) irreds
1923 ; traceTc (text "reduceImplication result" <+> vcat
1924 [ ppr outer, ppr inner, ppr binds])
1926 -- We always discard the extra avails we've generated;
1927 -- but we remember if we have done any (global) improvement
1928 ; let ret_avails = updateImprovement orig_avails avails
1930 ; if isEmptyLHsBinds binds && null outer then -- No progress
1931 return (ret_avails, NoInstance)
1933 { (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens inner
1935 ; let dict_ids = map instToId extra_givens
1936 co = mkWpTyLams tvs <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
1937 rhs = mkHsWrap co payload
1938 loc = instLocSpan inst_loc
1939 payload | [wanted] <- wanteds = HsVar (instToId wanted)
1940 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) wanteds) Boxed
1942 ; return (ret_avails, GenInst (implic_insts ++ outer) (L loc rhs))
1946 Note [Reducing implication constraints]
1947 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1948 Suppose we are trying to simplify
1949 (Ord a, forall b. C a b => (W [a] b, D c b))
1951 instance (C a b, Ord a) => W [a] b
1952 When solving the implication constraint, we'll start with
1954 in the Avails. Then we add (C a b -> Given) and solve. Extracting
1955 the results gives us a binding for the (W [a] b), with an Irred of
1956 (Ord a, D c b). Now, the (Ord a) comes from "outside" the implication,
1957 but the (D d b) is from "inside". So we want to generate a Rhs binding
1960 ic = /\b \dc:C a b). (df a b dc do, ic' b dc)
1963 ic' :: forall b. C a b => D c b
1965 The 'depending on' part of the Rhs is important, because it drives
1966 the extractResults code.
1968 The "inside" and "outside" distinction is what's going on with 'inner' and
1969 'outer' in reduceImplication
1972 Note [Freeness and implications]
1973 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1974 It's hard to say when an implication constraint can be floated out. Consider
1975 forall {} Eq a => Foo [a]
1976 The (Foo [a]) doesn't mention any of the quantified variables, but it
1977 still might be partially satisfied by the (Eq a).
1979 There is a useful special case when it *is* easy to partition the
1980 constraints, namely when there are no 'givens'. Consider
1981 forall {a}. () => Bar b
1982 There are no 'givens', and so there is no reason to capture (Bar b).
1983 We can let it float out. But if there is even one constraint we
1984 must be much more careful:
1985 forall {a}. C a b => Bar (m b)
1986 because (C a b) might have a superclass (D b), from which we might
1987 deduce (Bar [b]) when m later gets instantiated to []. Ha!
1989 Here is an even more exotic example
1991 Now consider the constraint
1992 forall b. D Int b => C Int
1993 We can satisfy the (C Int) from the superclass of D, so we don't want
1994 to float the (C Int) out, even though it mentions no type variable in
1997 Note [Pruning the givens in an implication constraint]
1998 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1999 Suppose we are about to form the implication constraint
2000 forall tvs. Eq a => Ord b
2001 The (Eq a) cannot contribute to the (Ord b), because it has no access to
2002 the type variable 'b'. So we could filter out the (Eq a) from the givens.
2004 Doing so would be a bit tidier, but all the implication constraints get
2005 simplified away by the optimiser, so it's no great win. So I don't take
2006 advantage of that at the moment.
2008 If you do, BE CAREFUL of wobbly type variables.
2011 %************************************************************************
2013 Avails and AvailHow: the pool of evidence
2015 %************************************************************************
2019 data Avails = Avails !ImprovementDone !AvailEnv
2021 type ImprovementDone = Bool -- True <=> some unification has happened
2022 -- so some Irreds might now be reducible
2023 -- keys that are now
2025 type AvailEnv = FiniteMap Inst AvailHow
2027 = IsIrred -- Used for irreducible dictionaries,
2028 -- which are going to be lambda bound
2030 | Given TcId -- Used for dictionaries for which we have a binding
2031 -- e.g. those "given" in a signature
2033 | Rhs -- Used when there is a RHS
2034 (LHsExpr TcId) -- The RHS
2035 [Inst] -- Insts free in the RHS; we need these too
2037 instance Outputable Avails where
2040 pprAvails (Avails imp avails)
2041 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
2042 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
2043 | (inst,avail) <- fmToList avails ])]
2045 instance Outputable AvailHow where
2048 -------------------------
2049 pprAvail :: AvailHow -> SDoc
2050 pprAvail IsIrred = text "Irred"
2051 pprAvail (Given x) = text "Given" <+> ppr x
2052 pprAvail (Rhs rhs bs) = text "Rhs" <+> sep [ppr rhs, braces (ppr bs)]
2054 -------------------------
2055 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2056 extendAvailEnv env inst avail = addToFM env inst avail
2058 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2059 findAvailEnv env wanted = lookupFM env wanted
2060 -- NB 1: the Ord instance of Inst compares by the class/type info
2061 -- *not* by unique. So
2062 -- d1::C Int == d2::C Int
2064 emptyAvails :: Avails
2065 emptyAvails = Avails False emptyFM
2067 findAvail :: Avails -> Inst -> Maybe AvailHow
2068 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2070 elemAvails :: Inst -> Avails -> Bool
2071 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2073 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2075 extendAvails avails@(Avails imp env) inst avail
2076 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2077 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2079 availsInsts :: Avails -> [Inst]
2080 availsInsts (Avails _ avails) = keysFM avails
2082 availsImproved (Avails imp _) = imp
2084 updateImprovement :: Avails -> Avails -> Avails
2085 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
2086 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
2089 Extracting the bindings from a bunch of Avails.
2090 The bindings do *not* come back sorted in dependency order.
2091 We assume that they'll be wrapped in a big Rec, so that the
2092 dependency analyser can sort them out later
2095 extractResults :: Avails
2097 -> TcM ( TcDictBinds, -- Bindings
2098 [Inst]) -- Irreducible ones
2100 extractResults (Avails _ avails) wanteds
2101 = go avails emptyBag [] wanteds
2103 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
2104 -> TcM (TcDictBinds, [Inst])
2105 go avails binds irreds []
2106 = returnM (binds, irreds)
2108 go avails binds irreds (w:ws)
2109 = case findAvailEnv avails w of
2110 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2111 go avails binds irreds ws
2114 | id == w_id -> go avails binds irreds ws
2115 | otherwise -> go avails (addBind binds w_id (nlHsVar id)) irreds ws
2116 -- The sought Id can be one of the givens, via a superclass chain
2117 -- and then we definitely don't want to generate an x=x binding!
2119 Just IsIrred -> go (add_given avails w) binds (w:irreds) ws
2120 -- The add_given handles the case where we want (Ord a, Eq a), and we
2121 -- don't want to emit *two* Irreds for Ord a, one via the superclass chain
2122 -- This showed up in a dupliated Ord constraint in the error message for
2125 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds (ws' ++ ws)
2127 new_binds = addBind binds w_id rhs
2131 add_given avails w = extendAvailEnv avails w (Given (instToId w))
2132 -- Don't add the same binding twice
2134 addBind binds id rhs = binds `unionBags` unitBag (L (getSrcSpan id) (VarBind id rhs))
2138 Note [No superclasses for Stop]
2139 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2140 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2141 add it to avails, so that any other equal Insts will be commoned up
2142 right here. However, we do *not* add superclasses. If we have
2145 but a is not bound here, then we *don't* want to derive dn from df
2146 here lest we lose sharing.
2149 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2150 addWanted want_scs avails wanted rhs_expr wanteds
2151 = addAvailAndSCs want_scs avails wanted avail
2153 avail = Rhs rhs_expr wanteds
2155 addGiven :: Avails -> Inst -> TcM Avails
2156 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2157 -- Always add superclasses for 'givens'
2159 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2160 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2161 -- so the assert isn't true
2163 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2164 addRefinedGiven reft (refined_givens, avails) given
2165 | isDict given -- We sometimes have 'given' methods, but they
2166 -- are always optional, so we can drop them
2167 , let pred = dictPred given
2168 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2169 , Just (co, pred) <- refinePred reft pred
2170 = do { new_given <- newDictBndr (instLoc given) pred
2171 ; let rhs = L (instSpan given) $
2172 HsWrap (WpCo co) (HsVar (instToId given))
2173 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2174 ; return (new_given:refined_givens, avails) }
2175 -- ToDo: the superclasses of the original given all exist in Avails
2176 -- so we could really just cast them, but it's more awkward to do,
2177 -- and hopefully the optimiser will spot the duplicated work
2179 = return (refined_givens, avails)
2182 Note [ImplicInst rigidity]
2183 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2185 C :: forall ab. (Eq a, Ord b) => b -> T a
2187 ...(case x of C v -> <body>)...
2189 From the case (where x::T ty) we'll get an implication constraint
2190 forall b. (Eq ty, Ord b) => <body-constraints>
2191 Now suppose <body-constraints> itself has an implication constraint
2193 forall c. <reft> => <payload>
2194 Then, we can certainly apply the refinement <reft> to the Ord b, becuase it is
2195 existential, but we probably should not apply it to the (Eq ty) because it may
2196 be wobbly. Hence the isRigidInst
2198 @Insts@ are ordered by their class/type info, rather than by their
2199 unique. This allows the context-reduction mechanism to use standard finite
2200 maps to do their stuff. It's horrible that this code is here, rather
2201 than with the Avails handling stuff in TcSimplify
2204 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2205 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2206 addAvailAndSCs want_scs avails irred IsIrred
2208 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2209 addAvailAndSCs want_scs avails inst avail
2210 | not (isClassDict inst) = extendAvails avails inst avail
2211 | NoSCs <- want_scs = extendAvails avails inst avail
2212 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2213 ; avails' <- extendAvails avails inst avail
2214 ; addSCs is_loop avails' inst }
2216 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2217 -- Note: this compares by *type*, not by Unique
2218 deps = findAllDeps (unitVarSet (instToId inst)) avail
2219 dep_tys = map idType (varSetElems deps)
2221 findAllDeps :: IdSet -> AvailHow -> IdSet
2222 -- Find all the Insts that this one depends on
2223 -- See Note [SUPERCLASS-LOOP 2]
2224 -- Watch out, though. Since the avails may contain loops
2225 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2226 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2227 findAllDeps so_far other = so_far
2229 find_all :: IdSet -> Inst -> IdSet
2231 | kid_id `elemVarSet` so_far = so_far
2232 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2233 | otherwise = so_far'
2235 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2236 kid_id = instToId kid
2238 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2239 -- Add all the superclasses of the Inst to Avails
2240 -- The first param says "dont do this because the original thing
2241 -- depends on this one, so you'd build a loop"
2242 -- Invariant: the Inst is already in Avails.
2244 addSCs is_loop avails dict
2245 = ASSERT( isDict dict )
2246 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2247 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2249 (clas, tys) = getDictClassTys dict
2250 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2251 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2253 add_sc avails (sc_dict, sc_sel)
2254 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2255 | is_given sc_dict = return avails
2256 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2257 ; addSCs is_loop avails' sc_dict }
2259 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2260 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2262 is_given :: Inst -> Bool
2263 is_given sc_dict = case findAvail avails sc_dict of
2264 Just (Given _) -> True -- Given is cheaper than superclass selection
2268 %************************************************************************
2270 \section{tcSimplifyTop: defaulting}
2272 %************************************************************************
2275 @tcSimplifyTop@ is called once per module to simplify all the constant
2276 and ambiguous Insts.
2278 We need to be careful of one case. Suppose we have
2280 instance Num a => Num (Foo a b) where ...
2282 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2283 to (Num x), and default x to Int. But what about y??
2285 It's OK: the final zonking stage should zap y to (), which is fine.
2289 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2290 tcSimplifyTop wanteds
2291 = tc_simplify_top doc False wanteds
2293 doc = text "tcSimplifyTop"
2295 tcSimplifyInteractive wanteds
2296 = tc_simplify_top doc True wanteds
2298 doc = text "tcSimplifyInteractive"
2300 -- The TcLclEnv should be valid here, solely to improve
2301 -- error message generation for the monomorphism restriction
2302 tc_simplify_top doc interactive wanteds
2303 = do { dflags <- getDOpts
2304 ; wanteds <- mapM zonkInst wanteds
2305 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2307 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2308 ; (irreds2, binds2) <- approximateImplications doc2 (\d -> True) irreds1
2310 -- Use the defaulting rules to do extra unification
2311 -- NB: irreds2 are already zonked
2312 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2314 -- Deal with implicit parameters
2315 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2316 (ambigs, others) = partition isTyVarDict non_ips
2318 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2320 ; addNoInstanceErrs others
2321 ; addTopAmbigErrs ambigs
2323 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2325 doc1 = doc <+> ptext SLIT("(first round)")
2326 doc2 = doc <+> ptext SLIT("(approximate)")
2327 doc3 = doc <+> ptext SLIT("(disambiguate)")
2330 If a dictionary constrains a type variable which is
2331 * not mentioned in the environment
2332 * and not mentioned in the type of the expression
2333 then it is ambiguous. No further information will arise to instantiate
2334 the type variable; nor will it be generalised and turned into an extra
2335 parameter to a function.
2337 It is an error for this to occur, except that Haskell provided for
2338 certain rules to be applied in the special case of numeric types.
2340 * at least one of its classes is a numeric class, and
2341 * all of its classes are numeric or standard
2342 then the type variable can be defaulted to the first type in the
2343 default-type list which is an instance of all the offending classes.
2345 So here is the function which does the work. It takes the ambiguous
2346 dictionaries and either resolves them (producing bindings) or
2347 complains. It works by splitting the dictionary list by type
2348 variable, and using @disambigOne@ to do the real business.
2350 @disambigOne@ assumes that its arguments dictionaries constrain all
2351 the same type variable.
2353 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2354 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2355 the most common use of defaulting is code like:
2357 _ccall_ foo `seqPrimIO` bar
2359 Since we're not using the result of @foo@, the result if (presumably)
2363 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2364 -- Just does unification to fix the default types
2365 -- The Insts are assumed to be pre-zonked
2366 disambiguate doc interactive dflags insts
2368 = return (insts, emptyBag)
2370 | null defaultable_groups
2371 = do { traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2372 ; return (insts, emptyBag) }
2375 = do { -- Figure out what default types to use
2376 default_tys <- getDefaultTys extended_defaulting ovl_strings
2378 ; traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2379 ; mapM_ (disambigGroup default_tys) defaultable_groups
2381 -- disambigGroup does unification, hence try again
2382 ; tryHardCheckLoop doc insts }
2385 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2386 ovl_strings = dopt Opt_OverloadedStrings dflags
2388 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2389 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2390 (unaries, bad_tvs_s) = partitionWith find_unary insts
2391 bad_tvs = unionVarSets bad_tvs_s
2393 -- Finds unary type-class constraints
2394 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2395 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2396 find_unary inst = Right (tyVarsOfInst inst)
2398 -- Group by type variable
2399 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2400 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2401 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2403 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2404 defaultable_group ds@((_,_,tv):_)
2405 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2406 && not (tv `elemVarSet` bad_tvs)
2407 && defaultable_classes [c | (_,c,_) <- ds]
2408 defaultable_group [] = panic "defaultable_group"
2410 defaultable_classes clss
2411 | extended_defaulting = any isInteractiveClass clss
2412 | otherwise = all is_std_class clss && (any is_num_class clss)
2414 -- In interactive mode, or with -fextended-default-rules,
2415 -- we default Show a to Show () to avoid graututious errors on "show []"
2416 isInteractiveClass cls
2417 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2419 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2420 -- is_num_class adds IsString to the standard numeric classes,
2421 -- when -foverloaded-strings is enabled
2423 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2424 -- Similarly is_std_class
2426 -----------------------
2427 disambigGroup :: [Type] -- The default types
2428 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2429 -> TcM () -- Just does unification, to fix the default types
2431 disambigGroup default_tys dicts
2432 = try_default default_tys
2434 (_,_,tyvar) = head dicts -- Should be non-empty
2435 classes = [c | (_,c,_) <- dicts]
2437 try_default [] = return ()
2438 try_default (default_ty : default_tys)
2439 = tryTcLIE_ (try_default default_tys) $
2440 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2441 -- This may fail; then the tryTcLIE_ kicks in
2442 -- Failure here is caused by there being no type in the
2443 -- default list which can satisfy all the ambiguous classes.
2444 -- For example, if Real a is reqd, but the only type in the
2445 -- default list is Int.
2447 -- After this we can't fail
2448 ; warnDefault dicts default_ty
2449 ; unifyType default_ty (mkTyVarTy tyvar) }
2452 -----------------------
2453 getDefaultTys :: Bool -> Bool -> TcM [Type]
2454 getDefaultTys extended_deflts ovl_strings
2455 = do { mb_defaults <- getDeclaredDefaultTys
2456 ; case mb_defaults of {
2457 Just tys -> return tys ; -- User-supplied defaults
2460 -- No use-supplied default
2461 -- Use [Integer, Double], plus modifications
2462 { integer_ty <- tcMetaTy integerTyConName
2463 ; checkWiredInTyCon doubleTyCon
2464 ; string_ty <- tcMetaTy stringTyConName
2465 ; return (opt_deflt extended_deflts unitTy
2466 -- Note [Default unitTy]
2468 [integer_ty,doubleTy]
2470 opt_deflt ovl_strings string_ty) } } }
2472 opt_deflt True ty = [ty]
2473 opt_deflt False ty = []
2476 Note [Default unitTy]
2477 ~~~~~~~~~~~~~~~~~~~~~
2478 In interative mode (or with -fextended-default-rules) we add () as the first type we
2479 try when defaulting. This has very little real impact, except in the following case.
2481 Text.Printf.printf "hello"
2482 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2483 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2484 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2485 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2486 () to the list of defaulting types. See Trac #1200.
2488 Note [Avoiding spurious errors]
2489 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2490 When doing the unification for defaulting, we check for skolem
2491 type variables, and simply don't default them. For example:
2492 f = (*) -- Monomorphic
2493 g :: Num a => a -> a
2495 Here, we get a complaint when checking the type signature for g,
2496 that g isn't polymorphic enough; but then we get another one when
2497 dealing with the (Num a) context arising from f's definition;
2498 we try to unify a with Int (to default it), but find that it's
2499 already been unified with the rigid variable from g's type sig
2502 %************************************************************************
2504 \subsection[simple]{@Simple@ versions}
2506 %************************************************************************
2508 Much simpler versions when there are no bindings to make!
2510 @tcSimplifyThetas@ simplifies class-type constraints formed by
2511 @deriving@ declarations and when specialising instances. We are
2512 only interested in the simplified bunch of class/type constraints.
2514 It simplifies to constraints of the form (C a b c) where
2515 a,b,c are type variables. This is required for the context of
2516 instance declarations.
2519 tcSimplifyDeriv :: InstOrigin
2521 -> ThetaType -- Wanted
2522 -> TcM ThetaType -- Needed
2523 -- Given instance (wanted) => C inst_ty
2524 -- Simplify 'wanted' as much as possible
2525 -- The inst_ty is needed only for the termination check
2527 tcSimplifyDeriv orig tyvars theta
2528 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2529 -- The main loop may do unification, and that may crash if
2530 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2531 -- ToDo: what if two of them do get unified?
2532 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2533 ; (irreds, _) <- tryHardCheckLoop doc wanteds
2535 ; let (tv_dicts, others) = partition isTyVarDict irreds
2536 ; addNoInstanceErrs others
2538 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2539 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
2540 -- This reverse-mapping is a pain, but the result
2541 -- should mention the original TyVars not TcTyVars
2543 ; return simpl_theta }
2545 doc = ptext SLIT("deriving classes for a data type")
2548 Note [Exotic derived instance contexts]
2549 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2551 data T a b c = MkT (Foo a b c) deriving( Eq )
2552 instance (C Int a, Eq b, Eq c) => Eq (Foo a b c)
2554 Notice that this instance (just) satisfies the Paterson termination
2555 conditions. Then we *could* derive an instance decl like this:
2557 instance (C Int a, Eq b, Eq c) => Eq (T a b c)
2559 even though there is no instance for (C Int a), because there just
2560 *might* be an instance for, say, (C Int Bool) at a site where we
2561 need the equality instance for T's.
2563 However, this seems pretty exotic, and it's quite tricky to allow
2564 this, and yet give sensible error messages in the (much more common)
2565 case where we really want that instance decl for C.
2567 So for now we simply require that the derived instance context
2568 should have only type-variable constraints.
2570 Here is another example:
2571 data Fix f = In (f (Fix f)) deriving( Eq )
2572 Here, if we are prepared to allow -fallow-undecidable-instances we
2573 could derive the instance
2574 instance Eq (f (Fix f)) => Eq (Fix f)
2575 but this is so delicate that I don't think it should happen inside
2576 'deriving'. If you want this, write it yourself!
2578 NB: if you want to lift this condition, make sure you still meet the
2579 termination conditions! If not, the deriving mechanism generates
2580 larger and larger constraints. Example:
2582 data Seq a = Cons a (Seq (Succ a)) | Nil deriving Show
2584 Note the lack of a Show instance for Succ. First we'll generate
2585 instance (Show (Succ a), Show a) => Show (Seq a)
2587 instance (Show (Succ (Succ a)), Show (Succ a), Show a) => Show (Seq a)
2588 and so on. Instead we want to complain of no instance for (Show (Succ a)).
2591 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2592 used with \tr{default} declarations. We are only interested in
2593 whether it worked or not.
2596 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2599 tcSimplifyDefault theta
2600 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2601 tryHardCheckLoop doc wanteds `thenM` \ (irreds, _) ->
2602 addNoInstanceErrs irreds `thenM_`
2608 doc = ptext SLIT("default declaration")
2612 %************************************************************************
2614 \section{Errors and contexts}
2616 %************************************************************************
2618 ToDo: for these error messages, should we note the location as coming
2619 from the insts, or just whatever seems to be around in the monad just
2623 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2624 -> [Inst] -- The offending Insts
2626 -- Group together insts with the same origin
2627 -- We want to report them together in error messages
2629 groupErrs report_err []
2631 groupErrs report_err (inst:insts)
2632 = do_one (inst:friends) `thenM_`
2633 groupErrs report_err others
2636 -- (It may seem a bit crude to compare the error messages,
2637 -- but it makes sure that we combine just what the user sees,
2638 -- and it avoids need equality on InstLocs.)
2639 (friends, others) = partition is_friend insts
2640 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2641 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2642 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2643 -- Add location and context information derived from the Insts
2645 -- Add the "arising from..." part to a message about bunch of dicts
2646 addInstLoc :: [Inst] -> Message -> Message
2647 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2649 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2650 addTopIPErrs bndrs []
2652 addTopIPErrs bndrs ips
2653 = do { dflags <- getDOpts
2654 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
2656 (tidy_env, tidy_ips) = tidyInsts ips
2658 = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2659 nest 2 (ptext SLIT("the monomorphic top-level binding")
2660 <> plural bndrs <+> ptext SLIT("of")
2661 <+> pprBinders bndrs <> colon)],
2662 nest 2 (vcat (map ppr_ip ips)),
2663 monomorphism_fix dflags]
2664 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2666 topIPErrs :: [Inst] -> TcM ()
2668 = groupErrs report tidy_dicts
2670 (tidy_env, tidy_dicts) = tidyInsts dicts
2671 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2672 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2673 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2675 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2677 addNoInstanceErrs insts
2678 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2679 ; reportNoInstances tidy_env Nothing tidy_insts }
2683 -> Maybe (InstLoc, [Inst]) -- Context
2684 -- Nothing => top level
2685 -- Just (d,g) => d describes the construct
2687 -> [Inst] -- What is wanted (can include implications)
2690 reportNoInstances tidy_env mb_what insts
2691 = groupErrs (report_no_instances tidy_env mb_what) insts
2693 report_no_instances tidy_env mb_what insts
2694 = do { inst_envs <- tcGetInstEnvs
2695 ; let (implics, insts1) = partition isImplicInst insts
2696 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2697 ; traceTc (text "reportNoInstnces" <+> vcat
2698 [ppr implics, ppr insts1, ppr insts2])
2699 ; mapM_ complain_implic implics
2700 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2701 ; groupErrs complain_no_inst insts2 }
2703 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2705 complain_implic inst -- Recurse!
2706 = reportNoInstances tidy_env
2707 (Just (tci_loc inst, tci_given inst))
2710 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2711 -- Right msg => overlap message
2712 -- Left inst => no instance
2713 check_overlap inst_envs wanted
2714 | not (isClassDict wanted) = Left wanted
2716 = case lookupInstEnv inst_envs clas tys of
2717 -- The case of exactly one match and no unifiers means
2718 -- a successful lookup. That can't happen here, becuase
2719 -- dicts only end up here if they didn't match in Inst.lookupInst
2721 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2723 ([], _) -> Left wanted -- No match
2724 res -> Right (mk_overlap_msg wanted res)
2726 (clas,tys) = getDictClassTys wanted
2728 mk_overlap_msg dict (matches, unifiers)
2729 = ASSERT( not (null matches) )
2730 vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2731 <+> pprPred (dictPred dict))),
2732 sep [ptext SLIT("Matching instances") <> colon,
2733 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2734 if not (isSingleton matches)
2735 then -- Two or more matches
2737 else -- One match, plus some unifiers
2738 ASSERT( not (null unifiers) )
2739 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2740 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2741 ptext SLIT("To pick the first instance above, use -fallow-incoherent-instances"),
2742 ptext SLIT("when compiling the other instances")])]
2744 ispecs = [ispec | (ispec, _) <- matches]
2746 mk_no_inst_err insts
2747 | null insts = empty
2749 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2750 not (isEmptyVarSet (tyVarsOfInsts insts))
2751 = vcat [ addInstLoc insts $
2752 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2753 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2754 , show_fixes (fix1 loc : fixes2) ]
2756 | otherwise -- Top level
2757 = vcat [ addInstLoc insts $
2758 ptext SLIT("No instance") <> plural insts
2759 <+> ptext SLIT("for") <+> pprDictsTheta insts
2760 , show_fixes fixes2 ]
2763 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2764 <+> ptext SLIT("to the context of"),
2765 nest 2 (ppr (instLocOrigin loc)) ]
2766 -- I'm not sure it helps to add the location
2767 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2769 fixes2 | null instance_dicts = []
2770 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2771 pprDictsTheta instance_dicts]]
2772 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2773 -- Insts for which it is worth suggesting an adding an instance declaration
2774 -- Exclude implicit parameters, and tyvar dicts
2776 show_fixes :: [SDoc] -> SDoc
2777 show_fixes [] = empty
2778 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2779 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2781 addTopAmbigErrs dicts
2782 -- Divide into groups that share a common set of ambiguous tyvars
2783 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2784 -- See Note [Avoiding spurious errors]
2785 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2787 (tidy_env, tidy_dicts) = tidyInsts dicts
2789 tvs_of :: Inst -> [TcTyVar]
2790 tvs_of d = varSetElems (tyVarsOfInst d)
2791 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2793 report :: [(Inst,[TcTyVar])] -> TcM ()
2794 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2795 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2796 setSrcSpan (instSpan inst) $
2797 -- the location of the first one will do for the err message
2798 addErrTcM (tidy_env, msg $$ mono_msg)
2800 dicts = map fst pairs
2801 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2802 pprQuotedList tvs <+> in_msg,
2803 nest 2 (pprDictsInFull dicts)]
2804 in_msg = text "in the constraint" <> plural dicts <> colon
2805 report [] = panic "addTopAmbigErrs"
2808 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2809 -- There's an error with these Insts; if they have free type variables
2810 -- it's probably caused by the monomorphism restriction.
2811 -- Try to identify the offending variable
2812 -- ASSUMPTION: the Insts are fully zonked
2813 mkMonomorphismMsg tidy_env inst_tvs
2814 = do { dflags <- getDOpts
2815 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
2816 ; return (tidy_env, mk_msg dflags docs) }
2818 mk_msg _ _ | any isRuntimeUnk inst_tvs
2819 = vcat [ptext SLIT("Cannot resolve unknown runtime types:") <+>
2820 (pprWithCommas ppr inst_tvs),
2821 ptext SLIT("Use :print or :force to determine these types")]
2822 mk_msg _ [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2823 -- This happens in things like
2824 -- f x = show (read "foo")
2825 -- where monomorphism doesn't play any role
2827 = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2829 monomorphism_fix dflags]
2831 isRuntimeUnk :: TcTyVar -> Bool
2832 isRuntimeUnk x | SkolemTv RuntimeUnkSkol <- tcTyVarDetails x = True
2835 monomorphism_fix :: DynFlags -> SDoc
2836 monomorphism_fix dflags
2837 = ptext SLIT("Probable fix:") <+> vcat
2838 [ptext SLIT("give these definition(s) an explicit type signature"),
2839 if dopt Opt_MonomorphismRestriction dflags
2840 then ptext SLIT("or use -fno-monomorphism-restriction")
2841 else empty] -- Only suggest adding "-fno-monomorphism-restriction"
2842 -- if it is not already set!
2844 warnDefault ups default_ty
2845 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2846 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2848 dicts = [d | (d,_,_) <- ups]
2851 (_, tidy_dicts) = tidyInsts dicts
2852 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2853 quotes (ppr default_ty),
2854 pprDictsInFull tidy_dicts]
2856 reduceDepthErr n stack
2857 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2858 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2859 nest 4 (pprStack stack)]
2861 pprStack stack = vcat (map pprInstInFull stack)