2 % (c) The University of Glasgow 2006
3 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
10 tcSimplifyInfer, tcSimplifyInferCheck,
11 tcSimplifyCheck, tcSimplifyRestricted,
12 tcSimplifyRuleLhs, tcSimplifyIPs,
13 tcSimplifySuperClasses,
14 tcSimplifyTop, tcSimplifyInteractive,
15 tcSimplifyBracket, tcSimplifyCheckPat,
17 tcSimplifyDeriv, tcSimplifyDefault,
18 bindInstsOfLocalFuns, bindIrreds,
21 #include "HsVersions.h"
23 import {-# SOURCE #-} TcUnify( unifyType )
59 %************************************************************************
63 %************************************************************************
65 --------------------------------------
66 Notes on functional dependencies (a bug)
67 --------------------------------------
74 instance D a b => C a b -- Undecidable
75 -- (Not sure if it's crucial to this eg)
76 f :: C a b => a -> Bool
79 g :: C a b => a -> Bool
82 Here f typechecks, but g does not!! Reason: before doing improvement,
83 we reduce the (C a b1) constraint from the call of f to (D a b1).
85 Here is a more complicated example:
87 | > class Foo a b | a->b
89 | > class Bar a b | a->b
93 | > instance Bar Obj Obj
95 | > instance (Bar a b) => Foo a b
97 | > foo:: (Foo a b) => a -> String
100 | > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
106 | Could not deduce (Bar a b) from the context (Foo a b)
107 | arising from use of `foo' at <interactive>:1
109 | Add (Bar a b) to the expected type of an expression
110 | In the first argument of `runFoo', namely `foo'
111 | In the definition of `it': it = runFoo foo
113 | Why all of the sudden does GHC need the constraint Bar a b? The
114 | function foo didn't ask for that...
116 The trouble is that to type (runFoo foo), GHC has to solve the problem:
118 Given constraint Foo a b
119 Solve constraint Foo a b'
121 Notice that b and b' aren't the same. To solve this, just do
122 improvement and then they are the same. But GHC currently does
127 That is usually fine, but it isn't here, because it sees that Foo a b is
128 not the same as Foo a b', and so instead applies the instance decl for
129 instance Bar a b => Foo a b. And that's where the Bar constraint comes
132 The Right Thing is to improve whenever the constraint set changes at
133 all. Not hard in principle, but it'll take a bit of fiddling to do.
137 --------------------------------------
138 Notes on quantification
139 --------------------------------------
141 Suppose we are about to do a generalisation step.
145 T the type of the RHS
146 C the constraints from that RHS
148 The game is to figure out
150 Q the set of type variables over which to quantify
151 Ct the constraints we will *not* quantify over
152 Cq the constraints we will quantify over
154 So we're going to infer the type
158 and float the constraints Ct further outwards.
160 Here are the things that *must* be true:
162 (A) Q intersect fv(G) = EMPTY limits how big Q can be
163 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
165 (A) says we can't quantify over a variable that's free in the
166 environment. (B) says we must quantify over all the truly free
167 variables in T, else we won't get a sufficiently general type. We do
168 not *need* to quantify over any variable that is fixed by the free
169 vars of the environment G.
171 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
173 Example: class H x y | x->y where ...
175 fv(G) = {a} C = {H a b, H c d}
178 (A) Q intersect {a} is empty
179 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
181 So Q can be {c,d}, {b,c,d}
183 Other things being equal, however, we'd like to quantify over as few
184 variables as possible: smaller types, fewer type applications, more
185 constraints can get into Ct instead of Cq.
188 -----------------------------------------
191 fv(T) the free type vars of T
193 oclose(vs,C) The result of extending the set of tyvars vs
194 using the functional dependencies from C
196 grow(vs,C) The result of extend the set of tyvars vs
197 using all conceivable links from C.
199 E.g. vs = {a}, C = {H [a] b, K (b,Int) c, Eq e}
200 Then grow(vs,C) = {a,b,c}
202 Note that grow(vs,C) `superset` grow(vs,simplify(C))
203 That is, simplfication can only shrink the result of grow.
206 oclose is conservative one way: v `elem` oclose(vs,C) => v is definitely fixed by vs
207 grow is conservative the other way: if v might be fixed by vs => v `elem` grow(vs,C)
210 -----------------------------------------
212 Note [Choosing which variables to quantify]
213 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
214 Here's a good way to choose Q:
216 Q = grow( fv(T), C ) \ oclose( fv(G), C )
218 That is, quantify over all variable that that MIGHT be fixed by the
219 call site (which influences T), but which aren't DEFINITELY fixed by
220 G. This choice definitely quantifies over enough type variables,
221 albeit perhaps too many.
223 Why grow( fv(T), C ) rather than fv(T)? Consider
225 class H x y | x->y where ...
230 If we used fv(T) = {c} we'd get the type
232 forall c. H c d => c -> b
234 And then if the fn was called at several different c's, each of
235 which fixed d differently, we'd get a unification error, because
236 d isn't quantified. Solution: quantify d. So we must quantify
237 everything that might be influenced by c.
239 Why not oclose( fv(T), C )? Because we might not be able to see
240 all the functional dependencies yet:
242 class H x y | x->y where ...
243 instance H x y => Eq (T x y) where ...
248 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
249 apparent yet, and that's wrong. We must really quantify over d too.
252 There really isn't any point in quantifying over any more than
253 grow( fv(T), C ), because the call sites can't possibly influence
254 any other type variables.
258 -------------------------------------
260 -------------------------------------
262 It's very hard to be certain when a type is ambiguous. Consider
266 instance H x y => K (x,y)
268 Is this type ambiguous?
269 forall a b. (K (a,b), Eq b) => a -> a
271 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
272 now we see that a fixes b. So we can't tell about ambiguity for sure
273 without doing a full simplification. And even that isn't possible if
274 the context has some free vars that may get unified. Urgle!
276 Here's another example: is this ambiguous?
277 forall a b. Eq (T b) => a -> a
278 Not if there's an insance decl (with no context)
279 instance Eq (T b) where ...
281 You may say of this example that we should use the instance decl right
282 away, but you can't always do that:
284 class J a b where ...
285 instance J Int b where ...
287 f :: forall a b. J a b => a -> a
289 (Notice: no functional dependency in J's class decl.)
290 Here f's type is perfectly fine, provided f is only called at Int.
291 It's premature to complain when meeting f's signature, or even
292 when inferring a type for f.
296 However, we don't *need* to report ambiguity right away. It'll always
297 show up at the call site.... and eventually at main, which needs special
298 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
300 So here's the plan. We WARN about probable ambiguity if
302 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
304 (all tested before quantification).
305 That is, all the type variables in Cq must be fixed by the the variables
306 in the environment, or by the variables in the type.
308 Notice that we union before calling oclose. Here's an example:
310 class J a b c | a b -> c
314 forall b c. (J a b c) => b -> b
316 Only if we union {a} from G with {b} from T before using oclose,
317 do we see that c is fixed.
319 It's a bit vague exactly which C we should use for this oclose call. If we
320 don't fix enough variables we might complain when we shouldn't (see
321 the above nasty example). Nothing will be perfect. That's why we can
322 only issue a warning.
325 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
327 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
329 then c is a "bubble"; there's no way it can ever improve, and it's
330 certainly ambiguous. UNLESS it is a constant (sigh). And what about
335 instance H x y => K (x,y)
337 Is this type ambiguous?
338 forall a b. (K (a,b), Eq b) => a -> a
340 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
341 is a "bubble" that's a set of constraints
343 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
345 Hence another idea. To decide Q start with fv(T) and grow it
346 by transitive closure in Cq (no functional dependencies involved).
347 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
348 The definitely-ambiguous can then float out, and get smashed at top level
349 (which squashes out the constants, like Eq (T a) above)
352 --------------------------------------
353 Notes on principal types
354 --------------------------------------
359 f x = let g y = op (y::Int) in True
361 Here the principal type of f is (forall a. a->a)
362 but we'll produce the non-principal type
363 f :: forall a. C Int => a -> a
366 --------------------------------------
367 The need for forall's in constraints
368 --------------------------------------
370 [Exchange on Haskell Cafe 5/6 Dec 2000]
372 class C t where op :: t -> Bool
373 instance C [t] where op x = True
375 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
376 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
378 The definitions of p and q differ only in the order of the components in
379 the pair on their right-hand sides. And yet:
381 ghc and "Typing Haskell in Haskell" reject p, but accept q;
382 Hugs rejects q, but accepts p;
383 hbc rejects both p and q;
384 nhc98 ... (Malcolm, can you fill in the blank for us!).
386 The type signature for f forces context reduction to take place, and
387 the results of this depend on whether or not the type of y is known,
388 which in turn depends on which component of the pair the type checker
391 Solution: if y::m a, float out the constraints
392 Monad m, forall c. C (m c)
393 When m is later unified with [], we can solve both constraints.
396 --------------------------------------
397 Notes on implicit parameters
398 --------------------------------------
400 Note [Inheriting implicit parameters]
401 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
406 where f is *not* a top-level binding.
407 From the RHS of f we'll get the constraint (?y::Int).
408 There are two types we might infer for f:
412 (so we get ?y from the context of f's definition), or
414 f :: (?y::Int) => Int -> Int
416 At first you might think the first was better, becuase then
417 ?y behaves like a free variable of the definition, rather than
418 having to be passed at each call site. But of course, the WHOLE
419 IDEA is that ?y should be passed at each call site (that's what
420 dynamic binding means) so we'd better infer the second.
422 BOTTOM LINE: when *inferring types* you *must* quantify
423 over implicit parameters. See the predicate isFreeWhenInferring.
426 Note [Implicit parameters and ambiguity]
427 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
428 What type should we infer for this?
429 f x = (show ?y, x::Int)
430 Since we must quantify over the ?y, the most plausible type is
431 f :: (Show a, ?y::a) => Int -> (String, Int)
432 But notice that the type of the RHS is (String,Int), with no type
433 varibables mentioned at all! The type of f looks ambiguous. But
434 it isn't, because at a call site we might have
435 let ?y = 5::Int in f 7
436 and all is well. In effect, implicit parameters are, well, parameters,
437 so we can take their type variables into account as part of the
438 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
441 Question 2: type signatures
442 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
443 BUT WATCH OUT: When you supply a type signature, we can't force you
444 to quantify over implicit parameters. For example:
448 This is perfectly reasonable. We do not want to insist on
450 (?x + 1) :: (?x::Int => Int)
452 That would be silly. Here, the definition site *is* the occurrence site,
453 so the above strictures don't apply. Hence the difference between
454 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
455 and tcSimplifyCheckBind (which does not).
457 What about when you supply a type signature for a binding?
458 Is it legal to give the following explicit, user type
459 signature to f, thus:
464 At first sight this seems reasonable, but it has the nasty property
465 that adding a type signature changes the dynamic semantics.
468 (let f x = (x::Int) + ?y
469 in (f 3, f 3 with ?y=5)) with ?y = 6
475 in (f 3, f 3 with ?y=5)) with ?y = 6
479 Indeed, simply inlining f (at the Haskell source level) would change the
482 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
483 semantics for a Haskell program without knowing its typing, so if you
484 change the typing you may change the semantics.
486 To make things consistent in all cases where we are *checking* against
487 a supplied signature (as opposed to inferring a type), we adopt the
490 a signature does not need to quantify over implicit params.
492 [This represents a (rather marginal) change of policy since GHC 5.02,
493 which *required* an explicit signature to quantify over all implicit
494 params for the reasons mentioned above.]
496 But that raises a new question. Consider
498 Given (signature) ?x::Int
499 Wanted (inferred) ?x::Int, ?y::Bool
501 Clearly we want to discharge the ?x and float the ?y out. But
502 what is the criterion that distinguishes them? Clearly it isn't
503 what free type variables they have. The Right Thing seems to be
504 to float a constraint that
505 neither mentions any of the quantified type variables
506 nor any of the quantified implicit parameters
508 See the predicate isFreeWhenChecking.
511 Question 3: monomorphism
512 ~~~~~~~~~~~~~~~~~~~~~~~~
513 There's a nasty corner case when the monomorphism restriction bites:
517 The argument above suggests that we *must* generalise
518 over the ?y parameter, to get
519 z :: (?y::Int) => Int,
520 but the monomorphism restriction says that we *must not*, giving
522 Why does the momomorphism restriction say this? Because if you have
524 let z = x + ?y in z+z
526 you might not expect the addition to be done twice --- but it will if
527 we follow the argument of Question 2 and generalise over ?y.
530 Question 4: top level
531 ~~~~~~~~~~~~~~~~~~~~~
532 At the top level, monomorhism makes no sense at all.
535 main = let ?x = 5 in print foo
539 woggle :: (?x :: Int) => Int -> Int
542 We definitely don't want (foo :: Int) with a top-level implicit parameter
543 (?x::Int) becuase there is no way to bind it.
548 (A) Always generalise over implicit parameters
549 Bindings that fall under the monomorphism restriction can't
553 * Inlining remains valid
554 * No unexpected loss of sharing
555 * But simple bindings like
557 will be rejected, unless you add an explicit type signature
558 (to avoid the monomorphism restriction)
559 z :: (?y::Int) => Int
561 This seems unacceptable
563 (B) Monomorphism restriction "wins"
564 Bindings that fall under the monomorphism restriction can't
566 Always generalise over implicit parameters *except* for bindings
567 that fall under the monomorphism restriction
570 * Inlining isn't valid in general
571 * No unexpected loss of sharing
572 * Simple bindings like
574 accepted (get value of ?y from binding site)
576 (C) Always generalise over implicit parameters
577 Bindings that fall under the monomorphism restriction can't
578 be generalised, EXCEPT for implicit parameters
580 * Inlining remains valid
581 * Unexpected loss of sharing (from the extra generalisation)
582 * Simple bindings like
584 accepted (get value of ?y from occurrence sites)
589 None of these choices seems very satisfactory. But at least we should
590 decide which we want to do.
592 It's really not clear what is the Right Thing To Do. If you see
596 would you expect the value of ?y to be got from the *occurrence sites*
597 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
598 case of function definitions, the answer is clearly the former, but
599 less so in the case of non-fucntion definitions. On the other hand,
600 if we say that we get the value of ?y from the definition site of 'z',
601 then inlining 'z' might change the semantics of the program.
603 Choice (C) really says "the monomorphism restriction doesn't apply
604 to implicit parameters". Which is fine, but remember that every
605 innocent binding 'x = ...' that mentions an implicit parameter in
606 the RHS becomes a *function* of that parameter, called at each
607 use of 'x'. Now, the chances are that there are no intervening 'with'
608 clauses that bind ?y, so a decent compiler should common up all
609 those function calls. So I think I strongly favour (C). Indeed,
610 one could make a similar argument for abolishing the monomorphism
611 restriction altogether.
613 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
617 %************************************************************************
619 \subsection{tcSimplifyInfer}
621 %************************************************************************
623 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
625 1. Compute Q = grow( fvs(T), C )
627 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
628 predicates will end up in Ct; we deal with them at the top level
630 3. Try improvement, using functional dependencies
632 4. If Step 3 did any unification, repeat from step 1
633 (Unification can change the result of 'grow'.)
635 Note: we don't reduce dictionaries in step 2. For example, if we have
636 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
637 after step 2. However note that we may therefore quantify over more
638 type variables than we absolutely have to.
640 For the guts, we need a loop, that alternates context reduction and
641 improvement with unification. E.g. Suppose we have
643 class C x y | x->y where ...
645 and tcSimplify is called with:
647 Then improvement unifies a with b, giving
650 If we need to unify anything, we rattle round the whole thing all over
657 -> TcTyVarSet -- fv(T); type vars
659 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
660 [Inst], -- Dict Ids that must be bound here (zonked)
661 TcDictBinds) -- Bindings
662 -- Any free (escaping) Insts are tossed into the environment
667 tcSimplifyInfer doc tau_tvs wanted
668 = do { tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
669 ; wanted' <- mappM zonkInst wanted -- Zonk before deciding quantified tyvars
670 ; gbl_tvs <- tcGetGlobalTyVars
671 ; let preds = fdPredsOfInsts wanted'
672 qtvs = grow preds tau_tvs' `minusVarSet` oclose preds gbl_tvs
673 -- See Note [Choosing which variables to quantify]
675 -- To maximise sharing, remove from consideration any
676 -- constraints that don't mention qtvs at all
677 ; let (free1, bound) = partition (isFreeWhenInferring qtvs) wanted'
680 -- To make types simple, reduce as much as possible
681 ; traceTc (text "infer" <+> (ppr preds $$ ppr (grow preds tau_tvs') $$ ppr gbl_tvs $$
682 ppr (oclose preds gbl_tvs) $$ ppr free1 $$ ppr bound))
683 ; let try_me inst = ReduceMe AddSCs
684 red_env = mkRedEnv doc try_me []
685 ; (irreds1, binds1) <- checkLoop red_env bound
687 -- Note [Inference and implication constraints]
688 -- By putting extra_dicts first, we make them available
689 -- to solve the implication constraints
690 ; let extra_dicts = getImplicWanteds qtvs irreds1
691 ; (irreds2, binds2) <- if null extra_dicts
692 then return (irreds1, emptyBag)
693 else do { extra_dicts' <- mapM cloneDict extra_dicts
694 ; checkLoop red_env (extra_dicts' ++ irreds1) }
696 -- By now improvment may have taken place, and we must *not*
697 -- quantify over any variable free in the environment
698 -- tc137 (function h inside g) is an example
699 ; gbl_tvs <- tcGetGlobalTyVars
700 ; qtvs1 <- zonkTcTyVarsAndFV (varSetElems qtvs)
701 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems (qtvs1 `minusVarSet` gbl_tvs))
703 -- Do not quantify over constraints that *now* do not
704 -- mention quantified type variables, because they are
705 -- simply ambiguous (or might be bound further out). Example:
706 -- f :: Eq b => a -> (a, b)
708 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
709 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
710 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
711 -- constraint (Eq beta), which we dump back into the free set
712 -- See test tcfail181
713 ; let (free3, irreds3) = partition (isFreeWhenInferring (mkVarSet qtvs2)) irreds2
716 -- We can't abstract over any remaining unsolved
717 -- implications so instead just float them outwards. Ugh.
718 ; let (q_dicts, implics) = partition isDict irreds3
719 ; loc <- getInstLoc (ImplicOrigin doc)
720 ; implic_bind <- bindIrreds loc qtvs2 q_dicts implics
722 ; return (qtvs2, q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
723 -- NB: when we are done, we might have some bindings, but
724 -- the final qtvs might be empty. See Note [NO TYVARS] below.
726 getImplicWanteds :: TcTyVarSet -> [Inst] -> [Inst]
727 -- See Note [Inference and implication constraints]
728 -- Find the wanted constraints in implication constraints that mention the
729 -- quantified type variables, and are not bound by forall's in the constraint itself
730 -- Returns only Dicts
731 getImplicWanteds qtvs implics
732 = concatMap get implics
734 get d@(Dict {}) | tyVarsOfInst d `intersectsVarSet` qtvs = [d]
736 get (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
737 = [ d | let tv_set = mkVarSet tvs
738 , d <- getImplicWanteds qtvs wanteds
739 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
742 Note [Inference and implication constraints]
743 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
744 We can't (or at least don't) abstract over implications. But we might
745 have an implication constraint (perhaps arising from a nested pattern
748 when we are now trying to quantify over 'a'. Our best approximation
749 is to make (D a) part of the inferred context, so we can use that to
750 discharge the implication. Hence getImplicWanteds.
752 See Trac #1430 and test tc228.
756 -----------------------------------------------------------
757 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
758 -- against, but we don't know the type variables over which we are going to quantify.
759 -- This happens when we have a type signature for a mutually recursive group
762 -> TcTyVarSet -- fv(T)
765 -> TcM ([TyVar], -- Fully zonked, and quantified
766 TcDictBinds) -- Bindings
768 tcSimplifyInferCheck loc tau_tvs givens wanteds
769 = do { (irreds, binds) <- innerCheckLoop loc givens wanteds
771 -- Figure out which type variables to quantify over
772 -- You might think it should just be the signature tyvars,
773 -- but in bizarre cases you can get extra ones
774 -- f :: forall a. Num a => a -> a
775 -- f x = fst (g (x, head [])) + 1
777 -- Here we infer g :: forall a b. a -> b -> (b,a)
778 -- We don't want g to be monomorphic in b just because
779 -- f isn't quantified over b.
780 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
781 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
782 ; gbl_tvs <- tcGetGlobalTyVars
783 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
784 -- We could close gbl_tvs, but its not necessary for
785 -- soundness, and it'll only affect which tyvars, not which
786 -- dictionaries, we quantify over
788 ; qtvs' <- zonkQuantifiedTyVars qtvs
790 -- Now we are back to normal (c.f. tcSimplCheck)
791 ; implic_bind <- bindIrreds loc qtvs' givens irreds
793 ; return (qtvs', binds `unionBags` implic_bind) }
796 Note [Squashing methods]
797 ~~~~~~~~~~~~~~~~~~~~~~~~~
798 Be careful if you want to float methods more:
799 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
800 From an application (truncate f i) we get
803 If we have also have a second occurrence of truncate, we get
806 When simplifying with i,f free, we might still notice that
807 t1=t3; but alas, the binding for t2 (which mentions t1)
808 may continue to float out!
813 class Y a b | a -> b where
816 instance Y [[a]] a where
819 k :: X a -> X a -> X a
821 g :: Num a => [X a] -> [X a]
824 h ys = ys ++ map (k (y [[0]])) xs
826 The excitement comes when simplifying the bindings for h. Initially
827 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
828 From this we get t1:=:t2, but also various bindings. We can't forget
829 the bindings (because of [LOOP]), but in fact t1 is what g is
832 The net effect of [NO TYVARS]
835 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
836 isFreeWhenInferring qtvs inst
837 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
838 && isInheritableInst inst -- and no implicit parameter involved
839 -- see Note [Inheriting implicit parameters]
841 {- No longer used (with implication constraints)
842 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
843 -> NameSet -- Quantified implicit parameters
845 isFreeWhenChecking qtvs ips inst
846 = isFreeWrtTyVars qtvs inst
847 && isFreeWrtIPs ips inst
850 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
851 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
855 %************************************************************************
857 \subsection{tcSimplifyCheck}
859 %************************************************************************
861 @tcSimplifyCheck@ is used when we know exactly the set of variables
862 we are going to quantify over. For example, a class or instance declaration.
865 -----------------------------------------------------------
866 -- tcSimplifyCheck is used when checking expression type signatures,
867 -- class decls, instance decls etc.
868 tcSimplifyCheck :: InstLoc
869 -> [TcTyVar] -- Quantify over these
872 -> TcM TcDictBinds -- Bindings
873 tcSimplifyCheck loc qtvs givens wanteds
874 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
875 do { (irreds, binds) <- innerCheckLoop loc givens wanteds
876 ; implic_bind <- bindIrreds loc qtvs givens irreds
877 ; return (binds `unionBags` implic_bind) }
879 -----------------------------------------------------------
880 -- tcSimplifyCheckPat is used for existential pattern match
881 tcSimplifyCheckPat :: InstLoc
882 -> [CoVar] -> Refinement
883 -> [TcTyVar] -- Quantify over these
886 -> TcM TcDictBinds -- Bindings
887 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
888 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
889 do { (irreds, binds) <- innerCheckLoop loc givens wanteds
890 ; implic_bind <- bindIrredsR loc qtvs co_vars reft
892 ; return (binds `unionBags` implic_bind) }
894 -----------------------------------------------------------
895 bindIrreds :: InstLoc -> [TcTyVar]
898 bindIrreds loc qtvs givens irreds
899 = bindIrredsR loc qtvs [] emptyRefinement givens irreds
901 bindIrredsR :: InstLoc -> [TcTyVar] -> [CoVar]
902 -> Refinement -> [Inst] -> [Inst]
904 -- Make a binding that binds 'irreds', by generating an implication
905 -- constraint for them, *and* throwing the constraint into the LIE
906 bindIrredsR loc qtvs co_vars reft givens irreds
910 = do { let givens' = filter isDict givens
911 -- The givens can include methods
912 -- See Note [Pruning the givens in an implication constraint]
914 -- If there are no 'givens' *and* the refinement is empty
915 -- (the refinement is like more givens), then it's safe to
916 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
917 -- See Note [Freeness and implications]
918 ; irreds' <- if null givens' && isEmptyRefinement reft
920 { let qtv_set = mkVarSet qtvs
921 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
923 ; return real_irreds }
926 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
927 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
928 -- This call does the real work
929 -- If irreds' is empty, it does something sensible
934 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
936 -> TcM ([Inst], TcDictBinds)
937 -- Make a binding that binds 'irreds', by generating an implication
938 -- constraint for them, *and* throwing the constraint into the LIE
939 -- The binding looks like
940 -- (ir1, .., irn) = f qtvs givens
941 -- where f is (evidence for) the new implication constraint
942 -- f :: forall qtvs. {reft} givens => (ir1, .., irn)
943 -- qtvs includes coercion variables
945 -- This binding must line up the 'rhs' in reduceImplication
946 makeImplicationBind loc all_tvs reft
947 givens -- Guaranteed all Dicts
949 | null irreds -- If there are no irreds, we are done
950 = return ([], emptyBag)
951 | otherwise -- Otherwise we must generate a binding
952 = do { uniq <- newUnique
953 ; span <- getSrcSpanM
954 ; let name = mkInternalName uniq (mkVarOcc "ic") span
955 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
956 tci_tyvars = all_tvs,
958 tci_wanted = irreds, tci_loc = loc }
960 ; let n_irreds = length irreds
961 irred_ids = map instToId irreds
962 tup_ty = mkTupleTy Boxed n_irreds (map idType irred_ids)
963 pat = TuplePat (map nlVarPat irred_ids) Boxed tup_ty
964 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
965 co = mkWpApps (map instToId givens) <.> mkWpTyApps (mkTyVarTys all_tvs)
966 bind | n_irreds==1 = VarBind (head irred_ids) rhs
967 | otherwise = PatBind { pat_lhs = L span pat,
968 pat_rhs = unguardedGRHSs rhs,
970 bind_fvs = placeHolderNames }
971 ; -- pprTrace "Make implic inst" (ppr implic_inst) $
972 return ([implic_inst], unitBag (L span bind)) }
974 -----------------------------------------------------------
977 -> TcM ([Inst], TcDictBinds)
979 topCheckLoop doc wanteds
980 = checkLoop (mkRedEnv doc try_me []) wanteds
982 try_me inst = ReduceMe AddSCs
984 -----------------------------------------------------------
985 innerCheckLoop :: InstLoc
988 -> TcM ([Inst], TcDictBinds)
990 innerCheckLoop inst_loc givens wanteds
991 = checkLoop env wanteds
993 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
995 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
997 -- When checking against a given signature
998 -- we MUST be very gentle: Note [Check gently]
1002 ~~~~~~~~~~~~~~~~~~~~
1003 We have to very careful about not simplifying too vigorously
1008 f :: Show b => T b -> b
1009 f (MkT x) = show [x]
1011 Inside the pattern match, which binds (a:*, x:a), we know that
1013 Hence we have a dictionary for Show [a] available; and indeed we
1014 need it. We are going to build an implication contraint
1015 forall a. (b~[a]) => Show [a]
1016 Later, we will solve this constraint using the knowledg e(Show b)
1018 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1019 thing becomes insoluble. So we simplify gently (get rid of literals
1020 and methods only, plus common up equal things), deferring the real
1021 work until top level, when we solve the implication constraint
1026 -----------------------------------------------------------
1029 -> TcM ([Inst], TcDictBinds)
1030 -- Precondition: givens are completely rigid
1032 checkLoop env wanteds
1033 = do { -- Givens are skolems, so no need to zonk them
1034 wanteds' <- mappM zonkInst wanteds
1036 ; (improved, binds, irreds) <- reduceContext env wanteds'
1038 ; if not improved then
1039 return (irreds, binds)
1042 -- If improvement did some unification, we go round again.
1043 -- We start again with irreds, not wanteds
1044 -- Using an instance decl might have introduced a fresh type variable
1045 -- which might have been unified, so we'd get an infinite loop
1046 -- if we started again with wanteds! See Note [LOOP]
1047 { (irreds1, binds1) <- checkLoop env irreds
1048 ; return (irreds1, binds `unionBags` binds1) } }
1053 class If b t e r | b t e -> r
1056 class Lte a b c | a b -> c where lte :: a -> b -> c
1058 instance (Lte a b l,If l b a c) => Max a b c
1060 Wanted: Max Z (S x) y
1062 Then we'll reduce using the Max instance to:
1063 (Lte Z (S x) l, If l (S x) Z y)
1064 and improve by binding l->T, after which we can do some reduction
1065 on both the Lte and If constraints. What we *can't* do is start again
1066 with (Max Z (S x) y)!
1070 %************************************************************************
1072 tcSimplifySuperClasses
1074 %************************************************************************
1076 Note [SUPERCLASS-LOOP 1]
1077 ~~~~~~~~~~~~~~~~~~~~~~~~
1078 We have to be very, very careful when generating superclasses, lest we
1079 accidentally build a loop. Here's an example:
1083 class S a => C a where { opc :: a -> a }
1084 class S b => D b where { opd :: b -> b }
1086 instance C Int where
1089 instance D Int where
1092 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1093 Simplifying, we may well get:
1094 $dfCInt = :C ds1 (opd dd)
1097 Notice that we spot that we can extract ds1 from dd.
1099 Alas! Alack! We can do the same for (instance D Int):
1101 $dfDInt = :D ds2 (opc dc)
1105 And now we've defined the superclass in terms of itself.
1107 Solution: never generate a superclass selectors at all when
1108 satisfying the superclass context of an instance declaration.
1110 Two more nasty cases are in
1115 tcSimplifySuperClasses
1120 tcSimplifySuperClasses loc givens sc_wanteds
1121 = do { (irreds, binds1) <- checkLoop env sc_wanteds
1122 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1123 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1126 env = mkRedEnv (pprInstLoc loc) try_me givens
1127 try_me inst = ReduceMe NoSCs
1128 -- Like topCheckLoop, but with NoSCs
1132 %************************************************************************
1134 \subsection{tcSimplifyRestricted}
1136 %************************************************************************
1138 tcSimplifyRestricted infers which type variables to quantify for a
1139 group of restricted bindings. This isn't trivial.
1142 We want to quantify over a to get id :: forall a. a->a
1145 We do not want to quantify over a, because there's an Eq a
1146 constraint, so we get eq :: a->a->Bool (notice no forall)
1149 RHS has type 'tau', whose free tyvars are tau_tvs
1150 RHS has constraints 'wanteds'
1153 Quantify over (tau_tvs \ ftvs(wanteds))
1154 This is bad. The constraints may contain (Monad (ST s))
1155 where we have instance Monad (ST s) where...
1156 so there's no need to be monomorphic in s!
1158 Also the constraint might be a method constraint,
1159 whose type mentions a perfectly innocent tyvar:
1160 op :: Num a => a -> b -> a
1161 Here, b is unconstrained. A good example would be
1163 We want to infer the polymorphic type
1164 foo :: forall b. b -> b
1167 Plan B (cunning, used for a long time up to and including GHC 6.2)
1168 Step 1: Simplify the constraints as much as possible (to deal
1169 with Plan A's problem). Then set
1170 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1172 Step 2: Now simplify again, treating the constraint as 'free' if
1173 it does not mention qtvs, and trying to reduce it otherwise.
1174 The reasons for this is to maximise sharing.
1176 This fails for a very subtle reason. Suppose that in the Step 2
1177 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1178 In the Step 1 this constraint might have been simplified, perhaps to
1179 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1180 This won't happen in Step 2... but that in turn might prevent some other
1181 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1182 and that in turn breaks the invariant that no constraints are quantified over.
1184 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1189 Step 1: Simplify the constraints as much as possible (to deal
1190 with Plan A's problem). Then set
1191 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1192 Return the bindings from Step 1.
1195 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1198 instance (HasBinary ty IO) => HasCodedValue ty
1200 foo :: HasCodedValue a => String -> IO a
1202 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1203 doDecodeIO codedValue view
1204 = let { act = foo "foo" } in act
1206 You might think this should work becuase the call to foo gives rise to a constraint
1207 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1208 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1209 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1211 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1215 Plan D (a variant of plan B)
1216 Step 1: Simplify the constraints as much as possible (to deal
1217 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1218 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1220 Step 2: Now simplify again, treating the constraint as 'free' if
1221 it does not mention qtvs, and trying to reduce it otherwise.
1223 The point here is that it's generally OK to have too few qtvs; that is,
1224 to make the thing more monomorphic than it could be. We don't want to
1225 do that in the common cases, but in wierd cases it's ok: the programmer
1226 can always add a signature.
1228 Too few qtvs => too many wanteds, which is what happens if you do less
1233 tcSimplifyRestricted -- Used for restricted binding groups
1234 -- i.e. ones subject to the monomorphism restriction
1237 -> [Name] -- Things bound in this group
1238 -> TcTyVarSet -- Free in the type of the RHSs
1239 -> [Inst] -- Free in the RHSs
1240 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1241 TcDictBinds) -- Bindings
1242 -- tcSimpifyRestricted returns no constraints to
1243 -- quantify over; by definition there are none.
1244 -- They are all thrown back in the LIE
1246 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1247 -- Zonk everything in sight
1248 = do { wanteds' <- mappM zonkInst wanteds
1250 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1251 -- dicts; the idea is to get rid of as many type
1252 -- variables as possible, and we don't want to stop
1253 -- at (say) Monad (ST s), because that reduces
1254 -- immediately, with no constraint on s.
1256 -- BUT do no improvement! See Plan D above
1257 -- HOWEVER, some unification may take place, if we instantiate
1258 -- a method Inst with an equality constraint
1259 ; let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1260 ; (_imp, _binds, constrained_dicts) <- reduceContext env wanteds'
1262 -- Next, figure out the tyvars we will quantify over
1263 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1264 ; gbl_tvs' <- tcGetGlobalTyVars
1265 ; constrained_dicts' <- mappM zonkInst constrained_dicts
1267 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1268 -- As in tcSimplifyInfer
1270 -- Do not quantify over constrained type variables:
1271 -- this is the monomorphism restriction
1272 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1273 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1274 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1277 ; warn_mono <- doptM Opt_WarnMonomorphism
1278 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1279 (vcat[ ptext SLIT("the Monomorphism Restriction applies to the binding")
1280 <> plural bndrs <+> ptext SLIT("for") <+> pp_bndrs,
1281 ptext SLIT("Consider giving a type signature for") <+> pp_bndrs])
1283 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1284 pprInsts wanteds, pprInsts constrained_dicts',
1286 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1288 -- The first step may have squashed more methods than
1289 -- necessary, so try again, this time more gently, knowing the exact
1290 -- set of type variables to quantify over.
1292 -- We quantify only over constraints that are captured by qtvs;
1293 -- these will just be a subset of non-dicts. This in contrast
1294 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1295 -- all *non-inheritable* constraints too. This implements choice
1296 -- (B) under "implicit parameter and monomorphism" above.
1298 -- Remember that we may need to do *some* simplification, to
1299 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1300 -- just to float all constraints
1302 -- At top level, we *do* squash methods becuase we want to
1303 -- expose implicit parameters to the test that follows
1304 ; let is_nested_group = isNotTopLevel top_lvl
1305 try_me inst | isFreeWrtTyVars qtvs inst,
1306 (is_nested_group || isDict inst) = Stop
1307 | otherwise = ReduceMe AddSCs
1308 env = mkNoImproveRedEnv doc try_me
1309 ; (_imp, binds, irreds) <- reduceContext env wanteds'
1311 -- See "Notes on implicit parameters, Question 4: top level"
1312 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1313 if is_nested_group then
1315 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1316 ; addTopIPErrs bndrs bad_ips
1317 ; extendLIEs non_ips }
1319 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1320 ; return (qtvs', binds) }
1324 %************************************************************************
1328 %************************************************************************
1330 On the LHS of transformation rules we only simplify methods and constants,
1331 getting dictionaries. We want to keep all of them unsimplified, to serve
1332 as the available stuff for the RHS of the rule.
1334 Example. Consider the following left-hand side of a rule
1336 f (x == y) (y > z) = ...
1338 If we typecheck this expression we get constraints
1340 d1 :: Ord a, d2 :: Eq a
1342 We do NOT want to "simplify" to the LHS
1344 forall x::a, y::a, z::a, d1::Ord a.
1345 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1349 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1350 f ((==) d2 x y) ((>) d1 y z) = ...
1352 Here is another example:
1354 fromIntegral :: (Integral a, Num b) => a -> b
1355 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1357 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1358 we *dont* want to get
1360 forall dIntegralInt.
1361 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1363 because the scsel will mess up RULE matching. Instead we want
1365 forall dIntegralInt, dNumInt.
1366 fromIntegral Int Int dIntegralInt dNumInt = id Int
1370 g (x == y) (y == z) = ..
1372 where the two dictionaries are *identical*, we do NOT WANT
1374 forall x::a, y::a, z::a, d1::Eq a
1375 f ((==) d1 x y) ((>) d1 y z) = ...
1377 because that will only match if the dict args are (visibly) equal.
1378 Instead we want to quantify over the dictionaries separately.
1380 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1381 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1382 from scratch, rather than further parameterise simpleReduceLoop etc
1385 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1386 tcSimplifyRuleLhs wanteds
1387 = go [] emptyBag wanteds
1390 = return (dicts, binds)
1391 go dicts binds (w:ws)
1393 = go (w:dicts) binds ws
1395 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1396 -- to fromInteger; this looks fragile to me
1397 ; lookup_result <- lookupSimpleInst w'
1398 ; case lookup_result of
1399 GenInst ws' rhs -> go dicts (addBind binds (instToId w) rhs) (ws' ++ ws)
1400 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1404 tcSimplifyBracket is used when simplifying the constraints arising from
1405 a Template Haskell bracket [| ... |]. We want to check that there aren't
1406 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1407 Show instance), but we aren't otherwise interested in the results.
1408 Nor do we care about ambiguous dictionaries etc. We will type check
1409 this bracket again at its usage site.
1412 tcSimplifyBracket :: [Inst] -> TcM ()
1413 tcSimplifyBracket wanteds
1414 = do { topCheckLoop doc wanteds
1417 doc = text "tcSimplifyBracket"
1421 %************************************************************************
1423 \subsection{Filtering at a dynamic binding}
1425 %************************************************************************
1430 we must discharge all the ?x constraints from B. We also do an improvement
1431 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1433 Actually, the constraints from B might improve the types in ?x. For example
1435 f :: (?x::Int) => Char -> Char
1438 then the constraint (?x::Int) arising from the call to f will
1439 force the binding for ?x to be of type Int.
1442 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1445 -- We need a loop so that we do improvement, and then
1446 -- (next time round) generate a binding to connect the two
1448 -- Here the two ?x's have different types, and improvement
1449 -- makes them the same.
1451 tcSimplifyIPs given_ips wanteds
1452 = do { wanteds' <- mappM zonkInst wanteds
1453 ; given_ips' <- mappM zonkInst given_ips
1454 -- Unusually for checking, we *must* zonk the given_ips
1456 ; let env = mkRedEnv doc try_me given_ips'
1457 ; (improved, binds, irreds) <- reduceContext env wanteds'
1459 ; if not improved then
1460 ASSERT( all is_free irreds )
1461 do { extendLIEs irreds
1464 tcSimplifyIPs given_ips wanteds }
1466 doc = text "tcSimplifyIPs" <+> ppr given_ips
1467 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1468 is_free inst = isFreeWrtIPs ip_set inst
1470 -- Simplify any methods that mention the implicit parameter
1471 try_me inst | is_free inst = Stop
1472 | otherwise = ReduceMe NoSCs
1476 %************************************************************************
1478 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1480 %************************************************************************
1482 When doing a binding group, we may have @Insts@ of local functions.
1483 For example, we might have...
1485 let f x = x + 1 -- orig local function (overloaded)
1486 f.1 = f Int -- two instances of f
1491 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1492 where @f@ is in scope; those @Insts@ must certainly not be passed
1493 upwards towards the top-level. If the @Insts@ were binding-ified up
1494 there, they would have unresolvable references to @f@.
1496 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1497 For each method @Inst@ in the @init_lie@ that mentions one of the
1498 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1499 @LIE@), as well as the @HsBinds@ generated.
1502 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1503 -- Simlifies only MethodInsts, and generate only bindings of form
1505 -- We're careful not to even generate bindings of the form
1507 -- You'd think that'd be fine, but it interacts with what is
1508 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1510 bindInstsOfLocalFuns wanteds local_ids
1511 | null overloaded_ids
1513 = extendLIEs wanteds `thenM_`
1514 returnM emptyLHsBinds
1517 = do { (irreds, binds) <- checkLoop env for_me
1518 ; extendLIEs not_for_me
1522 env = mkRedEnv doc try_me []
1523 doc = text "bindInsts" <+> ppr local_ids
1524 overloaded_ids = filter is_overloaded local_ids
1525 is_overloaded id = isOverloadedTy (idType id)
1526 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1528 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1529 -- so it's worth building a set, so that
1530 -- lookup (in isMethodFor) is faster
1531 try_me inst | isMethod inst = ReduceMe NoSCs
1536 %************************************************************************
1538 \subsection{Data types for the reduction mechanism}
1540 %************************************************************************
1542 The main control over context reduction is here
1546 = RedEnv { red_doc :: SDoc -- The context
1547 , red_try_me :: Inst -> WhatToDo
1548 , red_improve :: Bool -- True <=> do improvement
1549 , red_givens :: [Inst] -- All guaranteed rigid
1551 -- but see Note [Rigidity]
1552 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1553 -- See Note [RedStack]
1557 -- The red_givens are rigid so far as cmpInst is concerned.
1558 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1559 -- let ?x = e in ...
1560 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1561 -- But that doesn't affect the comparison, which is based only on mame.
1564 -- The red_stack pair (n,insts) pair is just used for error reporting.
1565 -- 'n' is always the depth of the stack.
1566 -- The 'insts' is the stack of Insts being reduced: to produce X
1567 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1570 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1571 mkRedEnv doc try_me givens
1572 = RedEnv { red_doc = doc, red_try_me = try_me,
1573 red_givens = givens, red_stack = (0,[]),
1574 red_improve = True }
1576 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1577 -- Do not do improvement; no givens
1578 mkNoImproveRedEnv doc try_me
1579 = RedEnv { red_doc = doc, red_try_me = try_me,
1580 red_givens = [], red_stack = (0,[]),
1581 red_improve = True }
1584 = ReduceMe WantSCs -- Try to reduce this
1585 -- If there's no instance, add the inst to the
1586 -- irreductible ones, but don't produce an error
1587 -- message of any kind.
1588 -- It might be quite legitimate such as (Eq a)!
1590 | Stop -- Return as irreducible unless it can
1591 -- be reduced to a constant in one step
1592 -- Do not add superclasses; see
1594 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1595 -- of a predicate when adding it to the avails
1596 -- The reason for this flag is entirely the super-class loop problem
1597 -- Note [SUPER-CLASS LOOP 1]
1600 %************************************************************************
1602 \subsection[reduce]{@reduce@}
1604 %************************************************************************
1608 reduceContext :: RedEnv
1610 -> TcM (ImprovementDone,
1611 TcDictBinds, -- Dictionary bindings
1612 [Inst]) -- Irreducible
1614 reduceContext env wanteds
1615 = do { traceTc (text "reduceContext" <+> (vcat [
1616 text "----------------------",
1618 text "given" <+> ppr (red_givens env),
1619 text "wanted" <+> ppr wanteds,
1620 text "----------------------"
1623 -- Build the Avail mapping from "givens"
1624 ; init_state <- foldlM addGiven emptyAvails (red_givens env)
1627 ; avails <- reduceList env wanteds init_state
1629 ; let improved = availsImproved avails
1630 ; (binds, irreds) <- extractResults avails wanteds
1632 ; traceTc (text "reduceContext end" <+> (vcat [
1633 text "----------------------",
1635 text "given" <+> ppr (red_givens env),
1636 text "wanted" <+> ppr wanteds,
1638 text "avails" <+> pprAvails avails,
1639 text "improved =" <+> ppr improved,
1640 text "irreds = " <+> ppr irreds,
1641 text "----------------------"
1644 ; return (improved, binds, irreds) }
1646 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1647 tcImproveOne avails inst
1648 | not (isDict inst) = return False
1650 = do { inst_envs <- tcGetInstEnvs
1651 ; let eqns = improveOne (classInstances inst_envs)
1652 (dictPred inst, pprInstArising inst)
1653 [ (dictPred p, pprInstArising p)
1654 | p <- availsInsts avails, isDict p ]
1655 -- Avails has all the superclasses etc (good)
1656 -- It also has all the intermediates of the deduction (good)
1657 -- It does not have duplicates (good)
1658 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1659 -- so that improve will see them separate
1660 ; traceTc (text "improveOne" <+> ppr inst)
1663 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1664 -> TcM ImprovementDone
1665 unifyEqns [] = return False
1667 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1671 unify ((qtvs, pairs), what1, what2)
1672 = addErrCtxtM (mkEqnMsg what1 what2) $
1673 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1674 mapM_ (unif_pr tenv) pairs
1675 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1677 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1679 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1680 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1681 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1682 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1683 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1684 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1685 ; return (tidy_env, msg) }
1688 The main context-reduction function is @reduce@. Here's its game plan.
1691 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1692 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1693 = do { dopts <- getDOpts
1696 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1697 2 (ifPprDebug (nest 2 (pprStack stk))))
1700 ; if n >= ctxtStkDepth dopts then
1701 failWithTc (reduceDepthErr n stk)
1705 go [] state = return state
1706 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1709 -- Base case: we're done!
1710 reduce env wanted avails
1711 -- It's the same as an existing inst, or a superclass thereof
1712 | Just avail <- findAvail avails wanted
1716 = case red_try_me env wanted of {
1717 ; Stop -> try_simple (addIrred NoSCs) -- See Note [No superclasses for Stop]
1719 ; ReduceMe want_scs -> -- It should be reduced
1720 reduceInst env avails wanted `thenM` \ (avails, lookup_result) ->
1721 case lookup_result of
1722 NoInstance -> -- No such instance!
1723 -- Add it and its superclasses
1724 addIrred want_scs avails wanted
1726 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1728 GenInst wanteds' rhs -> do { avails1 <- addIrred NoSCs avails wanted
1729 ; avails2 <- reduceList env wanteds' avails1
1730 ; addWanted want_scs avails2 wanted rhs wanteds' }
1731 -- Temporarily do addIrred *before* the reduceList,
1732 -- which has the effect of adding the thing we are trying
1733 -- to prove to the database before trying to prove the things it
1734 -- needs. See note [RECURSIVE DICTIONARIES]
1735 -- NB: we must not do an addWanted before, because that adds the
1736 -- superclasses too, and thaat can lead to a spurious loop; see
1737 -- the examples in [SUPERCLASS-LOOP]
1738 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1742 -- First, see if the inst can be reduced to a constant in one step
1743 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1744 -- Don't bother for implication constraints, which take real work
1745 try_simple do_this_otherwise
1746 = do { res <- lookupSimpleInst wanted
1748 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1749 other -> do_this_otherwise avails wanted }
1753 Note [SUPERCLASS-LOOP 2]
1754 ~~~~~~~~~~~~~~~~~~~~~~~~
1755 But the above isn't enough. Suppose we are *given* d1:Ord a,
1756 and want to deduce (d2:C [a]) where
1758 class Ord a => C a where
1759 instance Ord [a] => C [a] where ...
1761 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1762 superclasses of C [a] to avails. But we must not overwrite the binding
1763 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1766 Here's another variant, immortalised in tcrun020
1767 class Monad m => C1 m
1768 class C1 m => C2 m x
1769 instance C2 Maybe Bool
1770 For the instance decl we need to build (C1 Maybe), and it's no good if
1771 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1772 before we search for C1 Maybe.
1774 Here's another example
1775 class Eq b => Foo a b
1776 instance Eq a => Foo [a] a
1780 we'll first deduce that it holds (via the instance decl). We must not
1781 then overwrite the Eq t constraint with a superclass selection!
1783 At first I had a gross hack, whereby I simply did not add superclass constraints
1784 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1785 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1786 I found a very obscure program (now tcrun021) in which improvement meant the
1787 simplifier got two bites a the cherry... so something seemed to be an Stop
1788 first time, but reducible next time.
1790 Now we implement the Right Solution, which is to check for loops directly
1791 when adding superclasses. It's a bit like the occurs check in unification.
1794 Note [RECURSIVE DICTIONARIES]
1795 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1797 data D r = ZeroD | SuccD (r (D r));
1799 instance (Eq (r (D r))) => Eq (D r) where
1800 ZeroD == ZeroD = True
1801 (SuccD a) == (SuccD b) = a == b
1804 equalDC :: D [] -> D [] -> Bool;
1807 We need to prove (Eq (D [])). Here's how we go:
1811 by instance decl, holds if
1815 by instance decl of Eq, holds if
1817 where d2 = dfEqList d3
1820 But now we can "tie the knot" to give
1826 and it'll even run! The trick is to put the thing we are trying to prove
1827 (in this case Eq (D []) into the database before trying to prove its
1828 contributing clauses.
1831 %************************************************************************
1833 Reducing a single constraint
1835 %************************************************************************
1838 ---------------------------------------------
1839 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1840 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1841 tci_given = extra_givens, tci_wanted = wanteds })
1842 = reduceImplication env avails reft tvs extra_givens wanteds loc
1844 reduceInst env avails other_inst
1845 = do { result <- lookupSimpleInst other_inst
1846 ; return (avails, result) }
1850 ---------------------------------------------
1851 reduceImplication :: RedEnv
1853 -> Refinement -- May refine the givens; often empty
1854 -> [TcTyVar] -- Quantified type variables; all skolems
1855 -> [Inst] -- Extra givens; all rigid
1858 -> TcM (Avails, LookupInstResult)
1861 Suppose we are simplifying the constraint
1862 forall bs. extras => wanted
1863 in the context of an overall simplification problem with givens 'givens',
1864 and refinment 'reft'.
1867 * The refinement is often empty
1869 * The 'extra givens' need not mention any of the quantified type variables
1870 e.g. forall {}. Eq a => Eq [a]
1871 forall {}. C Int => D (Tree Int)
1873 This happens when you have something like
1875 T1 :: Eq a => a -> T a
1878 f x = ...(case x of { T1 v -> v==v })...
1881 -- ToDo: should we instantiate tvs? I think it's not necessary
1883 -- ToDo: what about improvement? There may be some improvement
1884 -- exposed as a result of the simplifications done by reduceList
1885 -- which are discarded if we back off.
1886 -- This is almost certainly Wrong, but we'll fix it when dealing
1887 -- better with equality constraints
1888 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
1889 = do { -- Add refined givens, and the extra givens
1890 (refined_red_givens, avails)
1891 <- if isEmptyRefinement reft then return (red_givens env, orig_avails)
1892 else foldlM (addRefinedGiven reft) ([], orig_avails) (red_givens env)
1893 ; avails <- foldlM addGiven avails extra_givens
1895 -- Solve the sub-problem
1896 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
1897 env' = env { red_givens = refined_red_givens ++ extra_givens
1898 , red_try_me = try_me }
1900 ; traceTc (text "reduceImplication" <+> vcat
1902 ppr (red_givens env), ppr extra_givens,
1903 ppr reft, ppr wanteds, ppr avails ])
1904 ; avails <- reduceList env' wanteds avails
1906 -- Extract the binding
1907 ; (binds, irreds) <- extractResults avails wanteds
1909 ; traceTc (text "reduceImplication result" <+> vcat
1910 [ ppr irreds, ppr binds])
1912 -- We always discard the extra avails we've generated;
1913 -- but we remember if we have done any (global) improvement
1914 ; let ret_avails = updateImprovement orig_avails avails
1916 ; if isEmptyLHsBinds binds then -- No progress
1917 return (ret_avails, NoInstance)
1919 { (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
1921 ; let dict_ids = map instToId extra_givens
1922 co = mkWpTyLams tvs <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
1923 rhs = mkHsWrap co payload
1924 loc = instLocSpan inst_loc
1925 payload | [wanted] <- wanteds = HsVar (instToId wanted)
1926 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) wanteds) Boxed
1928 -- If there are any irreds, we back off and return NoInstance
1929 ; return (ret_avails, GenInst implic_insts (L loc rhs))
1933 Note [Freeness and implications]
1934 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1935 It's hard to say when an implication constraint can be floated out. Consider
1936 forall {} Eq a => Foo [a]
1937 The (Foo [a]) doesn't mention any of the quantified variables, but it
1938 still might be partially satisfied by the (Eq a).
1940 There is a useful special case when it *is* easy to partition the
1941 constraints, namely when there are no 'givens'. Consider
1942 forall {a}. () => Bar b
1943 There are no 'givens', and so there is no reason to capture (Bar b).
1944 We can let it float out. But if there is even one constraint we
1945 must be much more careful:
1946 forall {a}. C a b => Bar (m b)
1947 because (C a b) might have a superclass (D b), from which we might
1948 deduce (Bar [b]) when m later gets instantiated to []. Ha!
1950 Here is an even more exotic example
1952 Now consider the constraint
1953 forall b. D Int b => C Int
1954 We can satisfy the (C Int) from the superclass of D, so we don't want
1955 to float the (C Int) out, even though it mentions no type variable in
1958 Note [Pruning the givens in an implication constraint]
1959 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1960 Suppose we are about to form the implication constraint
1961 forall tvs. Eq a => Ord b
1962 The (Eq a) cannot contribute to the (Ord b), because it has no access to
1963 the type variable 'b'. So we could filter out the (Eq a) from the givens.
1965 Doing so would be a bit tidier, but all the implication constraints get
1966 simplified away by the optimiser, so it's no great win. So I don't take
1967 advantage of that at the moment.
1969 If you do, BE CAREFUL of wobbly type variables.
1972 %************************************************************************
1974 Avails and AvailHow: the pool of evidence
1976 %************************************************************************
1980 data Avails = Avails !ImprovementDone !AvailEnv
1982 type ImprovementDone = Bool -- True <=> some unification has happened
1983 -- so some Irreds might now be reducible
1984 -- keys that are now
1986 type AvailEnv = FiniteMap Inst AvailHow
1988 = IsIrred TcId -- Used for irreducible dictionaries,
1989 -- which are going to be lambda bound
1991 | Given TcId -- Used for dictionaries for which we have a binding
1992 -- e.g. those "given" in a signature
1994 | Rhs -- Used when there is a RHS
1995 (LHsExpr TcId) -- The RHS
1996 [Inst] -- Insts free in the RHS; we need these too
1998 instance Outputable Avails where
2001 pprAvails (Avails imp avails)
2002 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
2003 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
2004 | (inst,avail) <- fmToList avails ])]
2006 instance Outputable AvailHow where
2009 -------------------------
2010 pprAvail :: AvailHow -> SDoc
2011 pprAvail (IsIrred x) = text "Irred" <+> ppr x
2012 pprAvail (Given x) = text "Given" <+> ppr x
2013 pprAvail (Rhs rhs bs) = text "Rhs" <+> ppr rhs <+> braces (ppr bs)
2015 -------------------------
2016 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2017 extendAvailEnv env inst avail = addToFM env inst avail
2019 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2020 findAvailEnv env wanted = lookupFM env wanted
2021 -- NB 1: the Ord instance of Inst compares by the class/type info
2022 -- *not* by unique. So
2023 -- d1::C Int == d2::C Int
2025 emptyAvails :: Avails
2026 emptyAvails = Avails False emptyFM
2028 findAvail :: Avails -> Inst -> Maybe AvailHow
2029 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2031 elemAvails :: Inst -> Avails -> Bool
2032 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2034 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2036 extendAvails avails@(Avails imp env) inst avail
2037 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2038 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2040 availsInsts :: Avails -> [Inst]
2041 availsInsts (Avails _ avails) = keysFM avails
2043 availsImproved (Avails imp _) = imp
2045 updateImprovement :: Avails -> Avails -> Avails
2046 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
2047 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
2050 Extracting the bindings from a bunch of Avails.
2051 The bindings do *not* come back sorted in dependency order.
2052 We assume that they'll be wrapped in a big Rec, so that the
2053 dependency analyser can sort them out later
2056 extractResults :: Avails
2058 -> TcM ( TcDictBinds, -- Bindings
2059 [Inst]) -- Irreducible ones
2061 extractResults (Avails _ avails) wanteds
2062 = go avails emptyBag [] wanteds
2064 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
2065 -> TcM (TcDictBinds, [Inst])
2066 go avails binds irreds []
2067 = returnM (binds, irreds)
2069 go avails binds irreds (w:ws)
2070 = case findAvailEnv avails w of
2071 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2072 go avails binds irreds ws
2075 | id == w_id -> go avails binds irreds ws
2076 | otherwise -> go avails (addBind binds w_id (nlHsVar id)) irreds ws
2077 -- The sought Id can be one of the givens, via a superclass chain
2078 -- and then we definitely don't want to generate an x=x binding!
2081 | id == w_id -> go (add_given avails w) binds (w:irreds) ws
2082 | otherwise -> go avails (addBind binds w_id (nlHsVar id)) irreds ws
2083 -- The add_given handles the case where we want (Ord a, Eq a), and we
2084 -- don't want to emit *two* Irreds for Ord a, one via the superclass chain
2085 -- This showed up in a dupliated Ord constraint in the error message for
2088 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds (ws' ++ ws)
2090 new_binds = addBind binds w_id rhs
2094 add_given avails w = extendAvailEnv avails w (Given (instToId w))
2095 -- Don't add the same binding twice
2097 addBind binds id rhs = binds `unionBags` unitBag (L (getSrcSpan id) (VarBind id rhs))
2101 Note [No superclasses for Stop]
2102 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2103 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2104 add it to avails, so that any other equal Insts will be commoned up
2105 right here. However, we do *not* add superclasses. If we have
2108 but a is not bound here, then we *don't* want to derive dn from df
2109 here lest we lose sharing.
2112 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2113 addWanted want_scs avails wanted rhs_expr wanteds
2114 = addAvailAndSCs want_scs avails wanted avail
2116 avail = Rhs rhs_expr wanteds
2118 addGiven :: Avails -> Inst -> TcM Avails
2119 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2120 -- Always add superclasses for 'givens'
2122 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2123 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2124 -- so the assert isn't true
2126 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2127 addRefinedGiven reft (refined_givens, avails) given
2128 | isDict given -- We sometimes have 'given' methods, but they
2129 -- are always optional, so we can drop them
2130 , let pred = dictPred given
2131 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2132 , Just (co, pred) <- refinePred reft pred
2133 = do { new_given <- newDictBndr (instLoc given) pred
2134 ; let rhs = L (instSpan given) $
2135 HsWrap (WpCo co) (HsVar (instToId given))
2136 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2137 ; return (new_given:refined_givens, avails) }
2138 -- ToDo: the superclasses of the original given all exist in Avails
2139 -- so we could really just cast them, but it's more awkward to do,
2140 -- and hopefully the optimiser will spot the duplicated work
2142 = return (refined_givens, avails)
2145 Note [ImplicInst rigidity]
2146 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2148 C :: forall ab. (Eq a, Ord b) => b -> T a
2150 ...(case x of C v -> <body>)...
2152 From the case (where x::T ty) we'll get an implication constraint
2153 forall b. (Eq ty, Ord b) => <body-constraints>
2154 Now suppose <body-constraints> itself has an implication constraint
2156 forall c. <reft> => <payload>
2157 Then, we can certainly apply the refinement <reft> to the Ord b, becuase it is
2158 existential, but we probably should not apply it to the (Eq ty) because it may
2159 be wobbly. Hence the isRigidInst
2161 @Insts@ are ordered by their class/type info, rather than by their
2162 unique. This allows the context-reduction mechanism to use standard finite
2163 maps to do their stuff. It's horrible that this code is here, rather
2164 than with the Avails handling stuff in TcSimplify
2167 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2168 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2169 addAvailAndSCs want_scs avails irred (IsIrred (instToId irred))
2171 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2172 addAvailAndSCs want_scs avails inst avail
2173 | not (isClassDict inst) = extendAvails avails inst avail
2174 | NoSCs <- want_scs = extendAvails avails inst avail
2175 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2176 ; avails' <- extendAvails avails inst avail
2177 ; addSCs is_loop avails' inst }
2179 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2180 -- Note: this compares by *type*, not by Unique
2181 deps = findAllDeps (unitVarSet (instToId inst)) avail
2182 dep_tys = map idType (varSetElems deps)
2184 findAllDeps :: IdSet -> AvailHow -> IdSet
2185 -- Find all the Insts that this one depends on
2186 -- See Note [SUPERCLASS-LOOP 2]
2187 -- Watch out, though. Since the avails may contain loops
2188 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2189 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2190 findAllDeps so_far other = so_far
2192 find_all :: IdSet -> Inst -> IdSet
2194 | kid_id `elemVarSet` so_far = so_far
2195 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2196 | otherwise = so_far'
2198 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2199 kid_id = instToId kid
2201 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2202 -- Add all the superclasses of the Inst to Avails
2203 -- The first param says "dont do this because the original thing
2204 -- depends on this one, so you'd build a loop"
2205 -- Invariant: the Inst is already in Avails.
2207 addSCs is_loop avails dict
2208 = ASSERT( isDict dict )
2209 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2210 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2212 (clas, tys) = getDictClassTys dict
2213 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2214 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2216 add_sc avails (sc_dict, sc_sel)
2217 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2218 | is_given sc_dict = return avails
2219 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2220 ; addSCs is_loop avails' sc_dict }
2222 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2223 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2225 is_given :: Inst -> Bool
2226 is_given sc_dict = case findAvail avails sc_dict of
2227 Just (Given _) -> True -- Given is cheaper than superclass selection
2231 %************************************************************************
2233 \section{tcSimplifyTop: defaulting}
2235 %************************************************************************
2238 @tcSimplifyTop@ is called once per module to simplify all the constant
2239 and ambiguous Insts.
2241 We need to be careful of one case. Suppose we have
2243 instance Num a => Num (Foo a b) where ...
2245 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2246 to (Num x), and default x to Int. But what about y??
2248 It's OK: the final zonking stage should zap y to (), which is fine.
2252 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2253 tcSimplifyTop wanteds
2254 = tc_simplify_top doc False wanteds
2256 doc = text "tcSimplifyTop"
2258 tcSimplifyInteractive wanteds
2259 = tc_simplify_top doc True wanteds
2261 doc = text "tcSimplifyInteractive"
2263 -- The TcLclEnv should be valid here, solely to improve
2264 -- error message generation for the monomorphism restriction
2265 tc_simplify_top doc interactive wanteds
2266 = do { wanteds <- mapM zonkInst wanteds
2267 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2269 ; (irreds1, binds1) <- topCheckLoop doc wanteds
2271 ; if null irreds1 then
2274 -- OK, so there are some errors
2275 { -- Use the defaulting rules to do extra unification
2276 -- NB: irreds are already zonked
2277 ; dflags <- getDOpts
2278 ; disambiguate interactive dflags irreds1 -- Does unification
2281 -- Deal with implicit parameters
2282 ; (irreds2, binds2) <- topCheckLoop doc irreds1
2283 ; let (bad_ips, non_ips) = partition isIPDict irreds2
2284 (ambigs, others) = partition isTyVarDict non_ips
2286 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2288 ; addNoInstanceErrs others
2289 ; addTopAmbigErrs ambigs
2291 ; return (binds1 `unionBags` binds2) }}
2294 If a dictionary constrains a type variable which is
2295 * not mentioned in the environment
2296 * and not mentioned in the type of the expression
2297 then it is ambiguous. No further information will arise to instantiate
2298 the type variable; nor will it be generalised and turned into an extra
2299 parameter to a function.
2301 It is an error for this to occur, except that Haskell provided for
2302 certain rules to be applied in the special case of numeric types.
2304 * at least one of its classes is a numeric class, and
2305 * all of its classes are numeric or standard
2306 then the type variable can be defaulted to the first type in the
2307 default-type list which is an instance of all the offending classes.
2309 So here is the function which does the work. It takes the ambiguous
2310 dictionaries and either resolves them (producing bindings) or
2311 complains. It works by splitting the dictionary list by type
2312 variable, and using @disambigOne@ to do the real business.
2314 @disambigOne@ assumes that its arguments dictionaries constrain all
2315 the same type variable.
2317 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2318 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2319 the most common use of defaulting is code like:
2321 _ccall_ foo `seqPrimIO` bar
2323 Since we're not using the result of @foo@, the result if (presumably)
2327 disambiguate :: Bool -> DynFlags -> [Inst] -> TcM ()
2328 -- Just does unification to fix the default types
2329 -- The Insts are assumed to be pre-zonked
2330 disambiguate interactive dflags insts
2331 | null defaultable_groups
2332 = do { traceTc (text "disambigutate" <+> vcat [ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2335 = do { -- Figure out what default types to use
2336 ; default_tys <- getDefaultTys extended_defaulting ovl_strings
2338 ; traceTc (text "disambigutate" <+> vcat [ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2339 ; mapM_ (disambigGroup default_tys) defaultable_groups }
2341 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2342 ovl_strings = dopt Opt_OverloadedStrings dflags
2344 unaries :: [(Inst,Class, TcTyVar)] -- (C tv) constraints
2345 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2346 (unaries, bad_tvs) = getDefaultableDicts insts
2348 -- Group by type variable
2349 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2350 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2351 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2353 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2354 defaultable_group ds@((_,_,tv):_)
2355 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2356 && not (tv `elemVarSet` bad_tvs)
2357 && defaultable_classes [c | (_,c,_) <- ds]
2358 defaultable_group [] = panic "defaultable_group"
2360 defaultable_classes clss
2361 | extended_defaulting = any isInteractiveClass clss
2362 | otherwise = all is_std_class clss && (any is_num_class clss)
2364 -- In interactive mode, or with -fextended-default-rules,
2365 -- we default Show a to Show () to avoid graututious errors on "show []"
2366 isInteractiveClass cls
2367 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2369 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2370 -- is_num_class adds IsString to the standard numeric classes,
2371 -- when -foverloaded-strings is enabled
2373 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2374 -- Similarly is_std_class
2376 -----------------------
2377 disambigGroup :: [Type] -- The default types
2378 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2379 -> TcM () -- Just does unification, to fix the default types
2381 disambigGroup default_tys dicts
2382 = try_default default_tys
2384 (_,_,tyvar) = head dicts -- Should be non-empty
2385 classes = [c | (_,c,_) <- dicts]
2387 try_default [] = return ()
2388 try_default (default_ty : default_tys)
2389 = tryTcLIE_ (try_default default_tys) $
2390 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2391 -- This may fail; then the tryTcLIE_ kicks in
2392 -- Failure here is caused by there being no type in the
2393 -- default list which can satisfy all the ambiguous classes.
2394 -- For example, if Real a is reqd, but the only type in the
2395 -- default list is Int.
2397 -- After this we can't fail
2398 ; warnDefault dicts default_ty
2399 ; unifyType default_ty (mkTyVarTy tyvar) }
2402 -----------------------
2403 getDefaultTys :: Bool -> Bool -> TcM [Type]
2404 getDefaultTys extended_deflts ovl_strings
2405 = do { mb_defaults <- getDeclaredDefaultTys
2406 ; case mb_defaults of {
2407 Just tys -> return tys ; -- User-supplied defaults
2410 -- No use-supplied default
2411 -- Use [Integer, Double], plus modifications
2412 { integer_ty <- tcMetaTy integerTyConName
2413 ; checkWiredInTyCon doubleTyCon
2414 ; string_ty <- tcMetaTy stringTyConName
2415 ; return (opt_deflt extended_deflts unitTy
2416 -- Note [Default unitTy]
2418 [integer_ty,doubleTy]
2420 opt_deflt ovl_strings string_ty) } } }
2422 opt_deflt True ty = [ty]
2423 opt_deflt False ty = []
2425 -----------------------
2426 getDefaultableDicts :: [Inst] -> ([(Inst, Class, TcTyVar)], TcTyVarSet)
2427 -- Look for free dicts of the form (C tv), even inside implications
2428 -- *and* the set of tyvars mentioned by all *other* constaints
2429 -- This disgustingly ad-hoc function is solely to support defaulting
2430 getDefaultableDicts insts
2431 = (concat ps, unionVarSets tvs)
2433 (ps, tvs) = mapAndUnzip get insts
2434 get d@(Dict {tci_pred = ClassP cls [ty]})
2435 | Just tv <- tcGetTyVar_maybe ty = ([(d,cls,tv)], emptyVarSet)
2436 | otherwise = ([], tyVarsOfType ty)
2437 get (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
2438 = ([ up | up@(_,_,tv) <- ups, not (tv `elemVarSet` tv_set)],
2439 ftvs `minusVarSet` tv_set)
2441 tv_set = mkVarSet tvs
2442 (ups, ftvs) = getDefaultableDicts wanteds
2443 get inst = ([], tyVarsOfInst inst)
2446 Note [Default unitTy]
2447 ~~~~~~~~~~~~~~~~~~~~~
2448 In interative mode (or with -fextended-default-rules) we add () as the first type we
2449 try when defaulting. This has very little real impact, except in the following case.
2451 Text.Printf.printf "hello"
2452 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2453 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2454 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2455 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2456 () to the list of defaulting types. See Trac #1200.
2458 Note [Avoiding spurious errors]
2459 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2460 When doing the unification for defaulting, we check for skolem
2461 type variables, and simply don't default them. For example:
2462 f = (*) -- Monomorphic
2463 g :: Num a => a -> a
2465 Here, we get a complaint when checking the type signature for g,
2466 that g isn't polymorphic enough; but then we get another one when
2467 dealing with the (Num a) context arising from f's definition;
2468 we try to unify a with Int (to default it), but find that it's
2469 already been unified with the rigid variable from g's type sig
2472 %************************************************************************
2474 \subsection[simple]{@Simple@ versions}
2476 %************************************************************************
2478 Much simpler versions when there are no bindings to make!
2480 @tcSimplifyThetas@ simplifies class-type constraints formed by
2481 @deriving@ declarations and when specialising instances. We are
2482 only interested in the simplified bunch of class/type constraints.
2484 It simplifies to constraints of the form (C a b c) where
2485 a,b,c are type variables. This is required for the context of
2486 instance declarations.
2489 tcSimplifyDeriv :: InstOrigin
2491 -> ThetaType -- Wanted
2492 -> TcM ThetaType -- Needed
2493 -- Given instance (wanted) => C inst_ty
2494 -- Simplify 'wanted' as much as possible
2495 -- The inst_ty is needed only for the termination check
2497 tcSimplifyDeriv orig tyvars theta
2498 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2499 -- The main loop may do unification, and that may crash if
2500 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2501 -- ToDo: what if two of them do get unified?
2502 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2503 ; (irreds, _) <- topCheckLoop doc wanteds
2505 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2506 simpl_theta = substTheta rev_env (map dictPred irreds)
2507 -- This reverse-mapping is a pain, but the result
2508 -- should mention the original TyVars not TcTyVars
2510 -- NB: the caller will further check the tv_dicts for
2511 -- legal instance-declaration form
2513 ; return simpl_theta }
2515 doc = ptext SLIT("deriving classes for a data type")
2520 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2521 used with \tr{default} declarations. We are only interested in
2522 whether it worked or not.
2525 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2528 tcSimplifyDefault theta
2529 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2530 topCheckLoop doc wanteds `thenM` \ (irreds, _) ->
2531 addNoInstanceErrs irreds `thenM_`
2537 doc = ptext SLIT("default declaration")
2541 %************************************************************************
2543 \section{Errors and contexts}
2545 %************************************************************************
2547 ToDo: for these error messages, should we note the location as coming
2548 from the insts, or just whatever seems to be around in the monad just
2552 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2553 -> [Inst] -- The offending Insts
2555 -- Group together insts with the same origin
2556 -- We want to report them together in error messages
2558 groupErrs report_err []
2560 groupErrs report_err (inst:insts)
2561 = do_one (inst:friends) `thenM_`
2562 groupErrs report_err others
2565 -- (It may seem a bit crude to compare the error messages,
2566 -- but it makes sure that we combine just what the user sees,
2567 -- and it avoids need equality on InstLocs.)
2568 (friends, others) = partition is_friend insts
2569 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2570 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2571 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2572 -- Add location and context information derived from the Insts
2574 -- Add the "arising from..." part to a message about bunch of dicts
2575 addInstLoc :: [Inst] -> Message -> Message
2576 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2578 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2579 addTopIPErrs bndrs []
2581 addTopIPErrs bndrs ips
2582 = do { dflags <- getDOpts
2583 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
2585 (tidy_env, tidy_ips) = tidyInsts ips
2587 = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2588 nest 2 (ptext SLIT("the monomorphic top-level binding")
2589 <> plural bndrs <+> ptext SLIT("of")
2590 <+> pprBinders bndrs <> colon)],
2591 nest 2 (vcat (map ppr_ip ips)),
2592 monomorphism_fix dflags]
2593 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2595 topIPErrs :: [Inst] -> TcM ()
2597 = groupErrs report tidy_dicts
2599 (tidy_env, tidy_dicts) = tidyInsts dicts
2600 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2601 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2602 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2604 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2606 addNoInstanceErrs insts
2607 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2608 ; reportNoInstances tidy_env Nothing tidy_insts }
2612 -> Maybe (InstLoc, [Inst]) -- Context
2613 -- Nothing => top level
2614 -- Just (d,g) => d describes the construct
2616 -> [Inst] -- What is wanted (can include implications)
2619 reportNoInstances tidy_env mb_what insts
2620 = groupErrs (report_no_instances tidy_env mb_what) insts
2622 report_no_instances tidy_env mb_what insts
2623 = do { inst_envs <- tcGetInstEnvs
2624 ; let (implics, insts1) = partition isImplicInst insts
2625 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2626 ; traceTc (text "reportNoInstnces" <+> vcat
2627 [ppr implics, ppr insts1, ppr insts2])
2628 ; mapM_ complain_implic implics
2629 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2630 ; groupErrs complain_no_inst insts2 }
2632 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2634 complain_implic inst -- Recurse!
2635 = reportNoInstances tidy_env
2636 (Just (tci_loc inst, tci_given inst))
2639 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2640 -- Right msg => overlap message
2641 -- Left inst => no instance
2642 check_overlap inst_envs wanted
2643 | not (isClassDict wanted) = Left wanted
2645 = case lookupInstEnv inst_envs clas tys of
2646 -- The case of exactly one match and no unifiers means
2647 -- a successful lookup. That can't happen here, becuase
2648 -- dicts only end up here if they didn't match in Inst.lookupInst
2650 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2652 ([], _) -> Left wanted -- No match
2653 res -> Right (mk_overlap_msg wanted res)
2655 (clas,tys) = getDictClassTys wanted
2657 mk_overlap_msg dict (matches, unifiers)
2658 = vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2659 <+> pprPred (dictPred dict))),
2660 sep [ptext SLIT("Matching instances") <> colon,
2661 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2662 ASSERT( not (null matches) )
2663 if not (isSingleton matches)
2664 then -- Two or more matches
2666 else -- One match, plus some unifiers
2667 ASSERT( not (null unifiers) )
2668 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2669 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2670 ptext SLIT("Use -fallow-incoherent-instances to use the first choice above")])]
2672 ispecs = [ispec | (ispec, _) <- matches]
2674 mk_no_inst_err insts
2675 | null insts = empty
2677 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2678 not (isEmptyVarSet (tyVarsOfInsts insts))
2679 = vcat [ addInstLoc insts $
2680 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2681 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2682 , show_fixes (fix1 loc : fixes2) ]
2684 | otherwise -- Top level
2685 = vcat [ addInstLoc insts $
2686 ptext SLIT("No instance") <> plural insts
2687 <+> ptext SLIT("for") <+> pprDictsTheta insts
2688 , show_fixes fixes2 ]
2691 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2692 <+> ptext SLIT("to the context of"),
2693 nest 2 (ppr (instLocOrigin loc)) ]
2694 -- I'm not sure it helps to add the location
2695 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2697 fixes2 | null instance_dicts = []
2698 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2699 pprDictsTheta instance_dicts]]
2700 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
2701 -- Insts for which it is worth suggesting an adding an instance declaration
2702 -- Exclude implicit parameters, and tyvar dicts
2704 show_fixes :: [SDoc] -> SDoc
2705 show_fixes [] = empty
2706 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
2707 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
2709 addTopAmbigErrs dicts
2710 -- Divide into groups that share a common set of ambiguous tyvars
2711 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
2712 -- See Note [Avoiding spurious errors]
2713 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
2715 (tidy_env, tidy_dicts) = tidyInsts dicts
2717 tvs_of :: Inst -> [TcTyVar]
2718 tvs_of d = varSetElems (tyVarsOfInst d)
2719 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
2721 report :: [(Inst,[TcTyVar])] -> TcM ()
2722 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
2723 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
2724 setSrcSpan (instSpan inst) $
2725 -- the location of the first one will do for the err message
2726 addErrTcM (tidy_env, msg $$ mono_msg)
2728 dicts = map fst pairs
2729 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
2730 pprQuotedList tvs <+> in_msg,
2731 nest 2 (pprDictsInFull dicts)]
2732 in_msg = text "in the constraint" <> plural dicts <> colon
2733 report [] = panic "addTopAmbigErrs"
2736 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
2737 -- There's an error with these Insts; if they have free type variables
2738 -- it's probably caused by the monomorphism restriction.
2739 -- Try to identify the offending variable
2740 -- ASSUMPTION: the Insts are fully zonked
2741 mkMonomorphismMsg tidy_env inst_tvs
2742 = do { dflags <- getDOpts
2743 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
2744 ; return (tidy_env, mk_msg dflags docs) }
2746 mk_msg _ _ | any isRuntimeUnk inst_tvs
2747 = vcat [ptext SLIT("Cannot resolve unknown runtime types:") <+>
2748 (pprWithCommas ppr inst_tvs),
2749 ptext SLIT("Use :print or :force to determine these types")]
2750 mk_msg _ [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
2751 -- This happens in things like
2752 -- f x = show (read "foo")
2753 -- where monomorphism doesn't play any role
2755 = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
2757 monomorphism_fix dflags]
2759 isRuntimeUnk :: TcTyVar -> Bool
2760 isRuntimeUnk x | SkolemTv RuntimeUnkSkol <- tcTyVarDetails x = True
2763 monomorphism_fix :: DynFlags -> SDoc
2764 monomorphism_fix dflags
2765 = ptext SLIT("Probable fix:") <+> vcat
2766 [ptext SLIT("give these definition(s) an explicit type signature"),
2767 if dopt Opt_MonomorphismRestriction dflags
2768 then ptext SLIT("or use -fno-monomorphism-restriction")
2769 else empty] -- Only suggest adding "-fno-monomorphism-restriction"
2770 -- if it is not already set!
2772 warnDefault ups default_ty
2773 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
2774 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
2776 dicts = [d | (d,_,_) <- ups]
2779 (_, tidy_dicts) = tidyInsts dicts
2780 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
2781 quotes (ppr default_ty),
2782 pprDictsInFull tidy_dicts]
2784 reduceDepthErr n stack
2785 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
2786 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
2787 nest 4 (pprStack stack)]
2789 pprStack stack = vcat (map pprInstInFull stack)