2 % (c) The University of Glasgow 2006
3 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
10 -- The above warning supression flag is a temporary kludge.
11 -- While working on this module you are encouraged to remove it and fix
12 -- any warnings in the module. See
13 -- http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
17 tcSimplifyInfer, tcSimplifyInferCheck,
18 tcSimplifyCheck, tcSimplifyRestricted,
19 tcSimplifyRuleLhs, tcSimplifyIPs,
20 tcSimplifySuperClasses,
21 tcSimplifyTop, tcSimplifyInteractive,
22 tcSimplifyBracket, tcSimplifyCheckPat,
24 tcSimplifyDeriv, tcSimplifyDefault,
25 bindInstsOfLocalFuns, bindIrreds,
30 #include "HsVersions.h"
32 import {-# SOURCE #-} TcUnify( unifyType )
73 %************************************************************************
77 %************************************************************************
79 --------------------------------------
80 Notes on functional dependencies (a bug)
81 --------------------------------------
88 instance D a b => C a b -- Undecidable
89 -- (Not sure if it's crucial to this eg)
90 f :: C a b => a -> Bool
93 g :: C a b => a -> Bool
96 Here f typechecks, but g does not!! Reason: before doing improvement,
97 we reduce the (C a b1) constraint from the call of f to (D a b1).
99 Here is a more complicated example:
101 | > class Foo a b | a->b
103 | > class Bar a b | a->b
107 | > instance Bar Obj Obj
109 | > instance (Bar a b) => Foo a b
111 | > foo:: (Foo a b) => a -> String
114 | > runFoo:: (forall a b. (Foo a b) => a -> w) -> w
120 | Could not deduce (Bar a b) from the context (Foo a b)
121 | arising from use of `foo' at <interactive>:1
123 | Add (Bar a b) to the expected type of an expression
124 | In the first argument of `runFoo', namely `foo'
125 | In the definition of `it': it = runFoo foo
127 | Why all of the sudden does GHC need the constraint Bar a b? The
128 | function foo didn't ask for that...
130 The trouble is that to type (runFoo foo), GHC has to solve the problem:
132 Given constraint Foo a b
133 Solve constraint Foo a b'
135 Notice that b and b' aren't the same. To solve this, just do
136 improvement and then they are the same. But GHC currently does
141 That is usually fine, but it isn't here, because it sees that Foo a b is
142 not the same as Foo a b', and so instead applies the instance decl for
143 instance Bar a b => Foo a b. And that's where the Bar constraint comes
146 The Right Thing is to improve whenever the constraint set changes at
147 all. Not hard in principle, but it'll take a bit of fiddling to do.
149 Note [Choosing which variables to quantify]
150 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
151 Suppose we are about to do a generalisation step. We have in our hand
154 T the type of the RHS
155 C the constraints from that RHS
157 The game is to figure out
159 Q the set of type variables over which to quantify
160 Ct the constraints we will *not* quantify over
161 Cq the constraints we will quantify over
163 So we're going to infer the type
167 and float the constraints Ct further outwards.
169 Here are the things that *must* be true:
171 (A) Q intersect fv(G) = EMPTY limits how big Q can be
172 (B) Q superset fv(Cq union T) \ oclose(fv(G),C) limits how small Q can be
174 (A) says we can't quantify over a variable that's free in the environment.
175 (B) says we must quantify over all the truly free variables in T, else
176 we won't get a sufficiently general type.
178 We do not *need* to quantify over any variable that is fixed by the
179 free vars of the environment G.
181 BETWEEN THESE TWO BOUNDS, ANY Q WILL DO!
183 Example: class H x y | x->y where ...
185 fv(G) = {a} C = {H a b, H c d}
188 (A) Q intersect {a} is empty
189 (B) Q superset {a,b,c,d} \ oclose({a}, C) = {a,b,c,d} \ {a,b} = {c,d}
191 So Q can be {c,d}, {b,c,d}
193 In particular, it's perfectly OK to quantify over more type variables
194 than strictly necessary; there is no need to quantify over 'b', since
195 it is determined by 'a' which is free in the envt, but it's perfectly
196 OK to do so. However we must not quantify over 'a' itself.
198 Other things being equal, however, we'd like to quantify over as few
199 variables as possible: smaller types, fewer type applications, more
200 constraints can get into Ct instead of Cq. Here's a good way to
203 Q = grow( fv(T), C ) \ oclose( fv(G), C )
205 That is, quantify over all variable that that MIGHT be fixed by the
206 call site (which influences T), but which aren't DEFINITELY fixed by
207 G. This choice definitely quantifies over enough type variables,
208 albeit perhaps too many.
210 Why grow( fv(T), C ) rather than fv(T)? Consider
212 class H x y | x->y where ...
217 If we used fv(T) = {c} we'd get the type
219 forall c. H c d => c -> b
221 And then if the fn was called at several different c's, each of
222 which fixed d differently, we'd get a unification error, because
223 d isn't quantified. Solution: quantify d. So we must quantify
224 everything that might be influenced by c.
226 Why not oclose( fv(T), C )? Because we might not be able to see
227 all the functional dependencies yet:
229 class H x y | x->y where ...
230 instance H x y => Eq (T x y) where ...
235 Now oclose(fv(T),C) = {c}, because the functional dependency isn't
236 apparent yet, and that's wrong. We must really quantify over d too.
238 There really isn't any point in quantifying over any more than
239 grow( fv(T), C ), because the call sites can't possibly influence
240 any other type variables.
244 -------------------------------------
246 -------------------------------------
248 It's very hard to be certain when a type is ambiguous. Consider
252 instance H x y => K (x,y)
254 Is this type ambiguous?
255 forall a b. (K (a,b), Eq b) => a -> a
257 Looks like it! But if we simplify (K (a,b)) we get (H a b) and
258 now we see that a fixes b. So we can't tell about ambiguity for sure
259 without doing a full simplification. And even that isn't possible if
260 the context has some free vars that may get unified. Urgle!
262 Here's another example: is this ambiguous?
263 forall a b. Eq (T b) => a -> a
264 Not if there's an insance decl (with no context)
265 instance Eq (T b) where ...
267 You may say of this example that we should use the instance decl right
268 away, but you can't always do that:
270 class J a b where ...
271 instance J Int b where ...
273 f :: forall a b. J a b => a -> a
275 (Notice: no functional dependency in J's class decl.)
276 Here f's type is perfectly fine, provided f is only called at Int.
277 It's premature to complain when meeting f's signature, or even
278 when inferring a type for f.
282 However, we don't *need* to report ambiguity right away. It'll always
283 show up at the call site.... and eventually at main, which needs special
284 treatment. Nevertheless, reporting ambiguity promptly is an excellent thing.
286 So here's the plan. We WARN about probable ambiguity if
288 fv(Cq) is not a subset of oclose(fv(T) union fv(G), C)
290 (all tested before quantification).
291 That is, all the type variables in Cq must be fixed by the the variables
292 in the environment, or by the variables in the type.
294 Notice that we union before calling oclose. Here's an example:
296 class J a b c | a b -> c
300 forall b c. (J a b c) => b -> b
302 Only if we union {a} from G with {b} from T before using oclose,
303 do we see that c is fixed.
305 It's a bit vague exactly which C we should use for this oclose call. If we
306 don't fix enough variables we might complain when we shouldn't (see
307 the above nasty example). Nothing will be perfect. That's why we can
308 only issue a warning.
311 Can we ever be *certain* about ambiguity? Yes: if there's a constraint
313 c in C such that fv(c) intersect (fv(G) union fv(T)) = EMPTY
315 then c is a "bubble"; there's no way it can ever improve, and it's
316 certainly ambiguous. UNLESS it is a constant (sigh). And what about
321 instance H x y => K (x,y)
323 Is this type ambiguous?
324 forall a b. (K (a,b), Eq b) => a -> a
326 Urk. The (Eq b) looks "definitely ambiguous" but it isn't. What we are after
327 is a "bubble" that's a set of constraints
329 Cq = Ca union Cq' st fv(Ca) intersect (fv(Cq') union fv(T) union fv(G)) = EMPTY
331 Hence another idea. To decide Q start with fv(T) and grow it
332 by transitive closure in Cq (no functional dependencies involved).
333 Now partition Cq using Q, leaving the definitely-ambiguous and probably-ok.
334 The definitely-ambiguous can then float out, and get smashed at top level
335 (which squashes out the constants, like Eq (T a) above)
338 --------------------------------------
339 Notes on principal types
340 --------------------------------------
345 f x = let g y = op (y::Int) in True
347 Here the principal type of f is (forall a. a->a)
348 but we'll produce the non-principal type
349 f :: forall a. C Int => a -> a
352 --------------------------------------
353 The need for forall's in constraints
354 --------------------------------------
356 [Exchange on Haskell Cafe 5/6 Dec 2000]
358 class C t where op :: t -> Bool
359 instance C [t] where op x = True
361 p y = (let f :: c -> Bool; f x = op (y >> return x) in f, y ++ [])
362 q y = (y ++ [], let f :: c -> Bool; f x = op (y >> return x) in f)
364 The definitions of p and q differ only in the order of the components in
365 the pair on their right-hand sides. And yet:
367 ghc and "Typing Haskell in Haskell" reject p, but accept q;
368 Hugs rejects q, but accepts p;
369 hbc rejects both p and q;
370 nhc98 ... (Malcolm, can you fill in the blank for us!).
372 The type signature for f forces context reduction to take place, and
373 the results of this depend on whether or not the type of y is known,
374 which in turn depends on which component of the pair the type checker
377 Solution: if y::m a, float out the constraints
378 Monad m, forall c. C (m c)
379 When m is later unified with [], we can solve both constraints.
382 --------------------------------------
383 Notes on implicit parameters
384 --------------------------------------
386 Note [Inheriting implicit parameters]
387 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
392 where f is *not* a top-level binding.
393 From the RHS of f we'll get the constraint (?y::Int).
394 There are two types we might infer for f:
398 (so we get ?y from the context of f's definition), or
400 f :: (?y::Int) => Int -> Int
402 At first you might think the first was better, becuase then
403 ?y behaves like a free variable of the definition, rather than
404 having to be passed at each call site. But of course, the WHOLE
405 IDEA is that ?y should be passed at each call site (that's what
406 dynamic binding means) so we'd better infer the second.
408 BOTTOM LINE: when *inferring types* you *must* quantify
409 over implicit parameters. See the predicate isFreeWhenInferring.
412 Note [Implicit parameters and ambiguity]
413 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
414 Only a *class* predicate can give rise to ambiguity
415 An *implicit parameter* cannot. For example:
416 foo :: (?x :: [a]) => Int
418 is fine. The call site will suppply a particular 'x'
420 Furthermore, the type variables fixed by an implicit parameter
421 propagate to the others. E.g.
422 foo :: (Show a, ?x::[a]) => Int
424 The type of foo looks ambiguous. But it isn't, because at a call site
426 let ?x = 5::Int in foo
427 and all is well. In effect, implicit parameters are, well, parameters,
428 so we can take their type variables into account as part of the
429 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
432 Question 2: type signatures
433 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
434 BUT WATCH OUT: When you supply a type signature, we can't force you
435 to quantify over implicit parameters. For example:
439 This is perfectly reasonable. We do not want to insist on
441 (?x + 1) :: (?x::Int => Int)
443 That would be silly. Here, the definition site *is* the occurrence site,
444 so the above strictures don't apply. Hence the difference between
445 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
446 and tcSimplifyCheckBind (which does not).
448 What about when you supply a type signature for a binding?
449 Is it legal to give the following explicit, user type
450 signature to f, thus:
455 At first sight this seems reasonable, but it has the nasty property
456 that adding a type signature changes the dynamic semantics.
459 (let f x = (x::Int) + ?y
460 in (f 3, f 3 with ?y=5)) with ?y = 6
466 in (f 3, f 3 with ?y=5)) with ?y = 6
470 Indeed, simply inlining f (at the Haskell source level) would change the
473 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
474 semantics for a Haskell program without knowing its typing, so if you
475 change the typing you may change the semantics.
477 To make things consistent in all cases where we are *checking* against
478 a supplied signature (as opposed to inferring a type), we adopt the
481 a signature does not need to quantify over implicit params.
483 [This represents a (rather marginal) change of policy since GHC 5.02,
484 which *required* an explicit signature to quantify over all implicit
485 params for the reasons mentioned above.]
487 But that raises a new question. Consider
489 Given (signature) ?x::Int
490 Wanted (inferred) ?x::Int, ?y::Bool
492 Clearly we want to discharge the ?x and float the ?y out. But
493 what is the criterion that distinguishes them? Clearly it isn't
494 what free type variables they have. The Right Thing seems to be
495 to float a constraint that
496 neither mentions any of the quantified type variables
497 nor any of the quantified implicit parameters
499 See the predicate isFreeWhenChecking.
502 Question 3: monomorphism
503 ~~~~~~~~~~~~~~~~~~~~~~~~
504 There's a nasty corner case when the monomorphism restriction bites:
508 The argument above suggests that we *must* generalise
509 over the ?y parameter, to get
510 z :: (?y::Int) => Int,
511 but the monomorphism restriction says that we *must not*, giving
513 Why does the momomorphism restriction say this? Because if you have
515 let z = x + ?y in z+z
517 you might not expect the addition to be done twice --- but it will if
518 we follow the argument of Question 2 and generalise over ?y.
521 Question 4: top level
522 ~~~~~~~~~~~~~~~~~~~~~
523 At the top level, monomorhism makes no sense at all.
526 main = let ?x = 5 in print foo
530 woggle :: (?x :: Int) => Int -> Int
533 We definitely don't want (foo :: Int) with a top-level implicit parameter
534 (?x::Int) becuase there is no way to bind it.
539 (A) Always generalise over implicit parameters
540 Bindings that fall under the monomorphism restriction can't
544 * Inlining remains valid
545 * No unexpected loss of sharing
546 * But simple bindings like
548 will be rejected, unless you add an explicit type signature
549 (to avoid the monomorphism restriction)
550 z :: (?y::Int) => Int
552 This seems unacceptable
554 (B) Monomorphism restriction "wins"
555 Bindings that fall under the monomorphism restriction can't
557 Always generalise over implicit parameters *except* for bindings
558 that fall under the monomorphism restriction
561 * Inlining isn't valid in general
562 * No unexpected loss of sharing
563 * Simple bindings like
565 accepted (get value of ?y from binding site)
567 (C) Always generalise over implicit parameters
568 Bindings that fall under the monomorphism restriction can't
569 be generalised, EXCEPT for implicit parameters
571 * Inlining remains valid
572 * Unexpected loss of sharing (from the extra generalisation)
573 * Simple bindings like
575 accepted (get value of ?y from occurrence sites)
580 None of these choices seems very satisfactory. But at least we should
581 decide which we want to do.
583 It's really not clear what is the Right Thing To Do. If you see
587 would you expect the value of ?y to be got from the *occurrence sites*
588 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
589 case of function definitions, the answer is clearly the former, but
590 less so in the case of non-fucntion definitions. On the other hand,
591 if we say that we get the value of ?y from the definition site of 'z',
592 then inlining 'z' might change the semantics of the program.
594 Choice (C) really says "the monomorphism restriction doesn't apply
595 to implicit parameters". Which is fine, but remember that every
596 innocent binding 'x = ...' that mentions an implicit parameter in
597 the RHS becomes a *function* of that parameter, called at each
598 use of 'x'. Now, the chances are that there are no intervening 'with'
599 clauses that bind ?y, so a decent compiler should common up all
600 those function calls. So I think I strongly favour (C). Indeed,
601 one could make a similar argument for abolishing the monomorphism
602 restriction altogether.
604 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
608 %************************************************************************
610 \subsection{tcSimplifyInfer}
612 %************************************************************************
614 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
616 1. Compute Q = grow( fvs(T), C )
618 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
619 predicates will end up in Ct; we deal with them at the top level
621 3. Try improvement, using functional dependencies
623 4. If Step 3 did any unification, repeat from step 1
624 (Unification can change the result of 'grow'.)
626 Note: we don't reduce dictionaries in step 2. For example, if we have
627 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
628 after step 2. However note that we may therefore quantify over more
629 type variables than we absolutely have to.
631 For the guts, we need a loop, that alternates context reduction and
632 improvement with unification. E.g. Suppose we have
634 class C x y | x->y where ...
636 and tcSimplify is called with:
638 Then improvement unifies a with b, giving
641 If we need to unify anything, we rattle round the whole thing all over
648 -> TcTyVarSet -- fv(T); type vars
650 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
651 [Inst], -- Dict Ids that must be bound here (zonked)
652 TcDictBinds) -- Bindings
653 -- Any free (escaping) Insts are tossed into the environment
658 tcSimplifyInfer doc tau_tvs wanted
659 = do { tau_tvs1 <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
660 ; wanted' <- mappM zonkInst wanted -- Zonk before deciding quantified tyvars
661 ; gbl_tvs <- tcGetGlobalTyVars
662 ; let preds1 = fdPredsOfInsts wanted'
663 gbl_tvs1 = oclose preds1 gbl_tvs
664 qtvs = grow preds1 tau_tvs1 `minusVarSet` gbl_tvs1
665 -- See Note [Choosing which variables to quantify]
667 -- To maximise sharing, remove from consideration any
668 -- constraints that don't mention qtvs at all
669 ; let (free, bound) = partition (isFreeWhenInferring qtvs) wanted'
672 -- To make types simple, reduce as much as possible
673 ; traceTc (text "infer" <+> (ppr preds1 $$ ppr (grow preds1 tau_tvs1) $$ ppr gbl_tvs $$
674 ppr gbl_tvs1 $$ ppr free $$ ppr bound))
675 ; (irreds1, binds1) <- tryHardCheckLoop doc bound
677 -- Note [Inference and implication constraints]
678 ; let want_dict d = tyVarsOfInst d `intersectsVarSet` qtvs
679 ; (irreds2, binds2) <- approximateImplications doc want_dict irreds1
681 -- Now work out all over again which type variables to quantify,
682 -- exactly in the same way as before, but starting from irreds2. Why?
683 -- a) By now improvment may have taken place, and we must *not*
684 -- quantify over any variable free in the environment
685 -- tc137 (function h inside g) is an example
687 -- b) Do not quantify over constraints that *now* do not
688 -- mention quantified type variables, because they are
689 -- simply ambiguous (or might be bound further out). Example:
690 -- f :: Eq b => a -> (a, b)
692 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
693 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
694 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
695 -- constraint (Eq beta), which we dump back into the free set
696 -- See test tcfail181
698 -- c) irreds may contain type variables not previously mentioned,
699 -- e.g. instance D a x => Foo [a]
701 -- Then after simplifying we'll get (D a x), and x is fresh
702 -- We must quantify over x else it'll be totally unbound
703 ; tau_tvs2 <- zonkTcTyVarsAndFV (varSetElems tau_tvs1)
704 ; gbl_tvs2 <- zonkTcTyVarsAndFV (varSetElems gbl_tvs1)
705 -- Note that we start from gbl_tvs1
706 -- We use tcGetGlobalTyVars, then oclose wrt preds2, because
707 -- we've already put some of the original preds1 into frees
708 -- E.g. wanteds = C a b (where a->b)
711 -- Then b is fixed by gbl_tvs, so (C a b) will be in free, and
712 -- irreds2 will be empty. But we don't want to generalise over b!
713 ; let preds2 = fdPredsOfInsts irreds2 -- irreds2 is zonked
714 qtvs = grow preds2 tau_tvs2 `minusVarSet` oclose preds2 gbl_tvs2
715 ; let (free, irreds3) = partition (isFreeWhenInferring qtvs) irreds2
718 -- Turn the quantified meta-type variables into real type variables
719 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems qtvs)
721 -- We can't abstract over any remaining unsolved
722 -- implications so instead just float them outwards. Ugh.
723 ; let (q_dicts0, implics) = partition isAbstractableInst irreds3
724 ; loc <- getInstLoc (ImplicOrigin doc)
725 ; implic_bind <- bindIrreds loc qtvs2 q_dicts0 implics
727 -- Prepare equality instances for quantification
728 ; let (q_eqs0,q_dicts) = partition isEqInst q_dicts0
729 ; q_eqs <- mappM finalizeEqInst q_eqs0
731 ; return (qtvs2, q_eqs ++ q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
732 -- NB: when we are done, we might have some bindings, but
733 -- the final qtvs might be empty. See Note [NO TYVARS] below.
735 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
736 -- Note [Inference and implication constraints]
737 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
738 -- - fetching any dicts inside them that are free
739 -- - using those dicts as cruder constraints, to solve the implications
740 -- - returning the extra ones too
742 approximateImplications doc want_dict irreds
744 = return (irreds, emptyBag)
746 = do { extra_dicts' <- mapM cloneDict extra_dicts
747 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
748 -- By adding extra_dicts', we make them
749 -- available to solve the implication constraints
751 extra_dicts = get_dicts (filter isImplicInst irreds)
753 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
754 -- Find the wanted constraints in implication constraints that satisfy
755 -- want_dict, and are not bound by forall's in the constraint itself
756 get_dicts ds = concatMap get_dict ds
758 get_dict d@(Dict {}) | want_dict d = [d]
760 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
761 = [ d | let tv_set = mkVarSet tvs
762 , d <- get_dicts wanteds
763 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
764 get_dict i@(EqInst {}) | want_dict i = [i]
766 get_dict other = pprPanic "approximateImplications" (ppr other)
769 Note [Inference and implication constraints]
770 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
771 Suppose we have a wanted implication constraint (perhaps arising from
772 a nested pattern match) like
774 and we are now trying to quantify over 'a' when inferring the type for
775 a function. In principle it's possible that there might be an instance
776 instance (C a, E a) => D [a]
777 so the context (E a) would suffice. The Right Thing is to abstract over
778 the implication constraint, but we don't do that (a) because it'll be
779 surprising to programmers and (b) because we don't have the machinery to deal
780 with 'given' implications.
782 So our best approximation is to make (D [a]) part of the inferred
783 context, so we can use that to discharge the implication. Hence
784 the strange function get_dicts in approximateImplications.
786 The common cases are more clear-cut, when we have things like
788 Here, abstracting over (C b) is not an approximation at all -- but see
789 Note [Freeness and implications].
791 See Trac #1430 and test tc228.
795 -----------------------------------------------------------
796 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
797 -- against, but we don't know the type variables over which we are going to quantify.
798 -- This happens when we have a type signature for a mutually recursive group
801 -> TcTyVarSet -- fv(T)
804 -> TcM ([TyVar], -- Fully zonked, and quantified
805 TcDictBinds) -- Bindings
807 tcSimplifyInferCheck loc tau_tvs givens wanteds
808 = do { traceTc (text "tcSimplifyInferCheck <-" <+> ppr wanteds)
809 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
811 -- Figure out which type variables to quantify over
812 -- You might think it should just be the signature tyvars,
813 -- but in bizarre cases you can get extra ones
814 -- f :: forall a. Num a => a -> a
815 -- f x = fst (g (x, head [])) + 1
817 -- Here we infer g :: forall a b. a -> b -> (b,a)
818 -- We don't want g to be monomorphic in b just because
819 -- f isn't quantified over b.
820 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
821 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
822 ; gbl_tvs <- tcGetGlobalTyVars
823 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
824 -- We could close gbl_tvs, but its not necessary for
825 -- soundness, and it'll only affect which tyvars, not which
826 -- dictionaries, we quantify over
828 ; qtvs' <- zonkQuantifiedTyVars qtvs
830 -- Now we are back to normal (c.f. tcSimplCheck)
831 ; implic_bind <- bindIrreds loc qtvs' givens irreds
833 ; traceTc (text "tcSimplifyInferCheck ->" <+> ppr (implic_bind))
834 ; return (qtvs', binds `unionBags` implic_bind) }
837 Note [Squashing methods]
838 ~~~~~~~~~~~~~~~~~~~~~~~~~
839 Be careful if you want to float methods more:
840 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
841 From an application (truncate f i) we get
844 If we have also have a second occurrence of truncate, we get
847 When simplifying with i,f free, we might still notice that
848 t1=t3; but alas, the binding for t2 (which mentions t1)
849 may continue to float out!
854 class Y a b | a -> b where
857 instance Y [[a]] a where
860 k :: X a -> X a -> X a
862 g :: Num a => [X a] -> [X a]
865 h ys = ys ++ map (k (y [[0]])) xs
867 The excitement comes when simplifying the bindings for h. Initially
868 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
869 From this we get t1:=:t2, but also various bindings. We can't forget
870 the bindings (because of [LOOP]), but in fact t1 is what g is
873 The net effect of [NO TYVARS]
876 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
877 isFreeWhenInferring qtvs inst
878 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
879 && isInheritableInst inst -- and no implicit parameter involved
880 -- see Note [Inheriting implicit parameters]
882 {- No longer used (with implication constraints)
883 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
884 -> NameSet -- Quantified implicit parameters
886 isFreeWhenChecking qtvs ips inst
887 = isFreeWrtTyVars qtvs inst
888 && isFreeWrtIPs ips inst
891 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
892 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
896 %************************************************************************
898 \subsection{tcSimplifyCheck}
900 %************************************************************************
902 @tcSimplifyCheck@ is used when we know exactly the set of variables
903 we are going to quantify over. For example, a class or instance declaration.
906 -----------------------------------------------------------
907 -- tcSimplifyCheck is used when checking expression type signatures,
908 -- class decls, instance decls etc.
909 tcSimplifyCheck :: InstLoc
910 -> [TcTyVar] -- Quantify over these
913 -> TcM TcDictBinds -- Bindings
914 tcSimplifyCheck loc qtvs givens wanteds
915 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
916 do { traceTc (text "tcSimplifyCheck")
917 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
918 ; implic_bind <- bindIrreds loc qtvs givens irreds
919 ; return (binds `unionBags` implic_bind) }
921 -----------------------------------------------------------
922 -- tcSimplifyCheckPat is used for existential pattern match
923 tcSimplifyCheckPat :: InstLoc
924 -> [CoVar] -> Refinement
925 -> [TcTyVar] -- Quantify over these
928 -> TcM TcDictBinds -- Bindings
929 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
930 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
931 do { traceTc (text "tcSimplifyCheckPat")
932 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
933 ; implic_bind <- bindIrredsR loc qtvs co_vars reft
935 ; return (binds `unionBags` implic_bind) }
937 -----------------------------------------------------------
938 bindIrreds :: InstLoc -> [TcTyVar]
941 bindIrreds loc qtvs givens irreds
942 = bindIrredsR loc qtvs [] emptyRefinement givens irreds
944 bindIrredsR :: InstLoc -> [TcTyVar] -> [CoVar]
945 -> Refinement -> [Inst] -> [Inst]
947 -- Make a binding that binds 'irreds', by generating an implication
948 -- constraint for them, *and* throwing the constraint into the LIE
949 bindIrredsR loc qtvs co_vars reft givens irreds
953 = do { let givens' = filter isDict givens
954 -- The givens can include methods
955 -- See Note [Pruning the givens in an implication constraint]
957 -- If there are no 'givens' *and* the refinement is empty
958 -- (the refinement is like more givens), then it's safe to
959 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
960 -- See Note [Freeness and implications]
961 ; irreds' <- if null givens' && isEmptyRefinement reft
963 { let qtv_set = mkVarSet qtvs
964 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
966 ; return real_irreds }
969 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
970 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
971 -- This call does the real work
972 -- If irreds' is empty, it does something sensible
977 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
979 -> TcM ([Inst], TcDictBinds)
980 -- Make a binding that binds 'irreds', by generating an implication
981 -- constraint for them, *and* throwing the constraint into the LIE
982 -- The binding looks like
983 -- (ir1, .., irn) = f qtvs givens
984 -- where f is (evidence for) the new implication constraint
985 -- f :: forall qtvs. {reft} givens => (ir1, .., irn)
986 -- qtvs includes coercion variables
988 -- This binding must line up the 'rhs' in reduceImplication
989 makeImplicationBind loc all_tvs reft
990 givens -- Guaranteed all Dicts (TOMDO: true?)
992 | null irreds -- If there are no irreds, we are done
993 = return ([], emptyBag)
994 | otherwise -- Otherwise we must generate a binding
995 = do { uniq <- newUnique
996 ; span <- getSrcSpanM
997 ; let (eq_givens, dict_givens) = partition isEqInst givens
998 eq_tyvar_cos = map TyVarTy $ uniqSetToList $ tyVarsOfTypes $ map eqInstType eq_givens
999 ; let name = mkInternalName uniq (mkVarOcc "ic") span
1000 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
1001 tci_tyvars = all_tvs,
1002 tci_given = (eq_givens ++ dict_givens),
1003 tci_wanted = irreds, tci_loc = loc }
1004 ; let -- only create binder for dict_irreds
1005 (eq_irreds, dict_irreds) = partition isEqInst irreds
1006 n_dict_irreds = length dict_irreds
1007 dict_irred_ids = map instToId dict_irreds
1008 tup_ty = mkTupleTy Boxed n_dict_irreds (map idType dict_irred_ids)
1009 pat = TuplePat (map nlVarPat dict_irred_ids) Boxed tup_ty
1010 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
1011 co = mkWpApps (map instToId dict_givens) <.> mkWpTyApps eq_tyvar_cos <.> mkWpTyApps (mkTyVarTys all_tvs)
1012 bind | [dict_irred_id] <- dict_irred_ids = VarBind dict_irred_id rhs
1013 | otherwise = PatBind { pat_lhs = L span pat,
1014 pat_rhs = unguardedGRHSs rhs,
1015 pat_rhs_ty = tup_ty,
1016 bind_fvs = placeHolderNames }
1017 ; -- pprTrace "Make implic inst" (ppr (implic_inst,irreds,dict_irreds,tup_ty)) $
1018 return ([implic_inst], unitBag (L span bind)) }
1020 -----------------------------------------------------------
1021 tryHardCheckLoop :: SDoc
1023 -> TcM ([Inst], TcDictBinds)
1025 tryHardCheckLoop doc wanteds
1026 = do { (irreds,binds,_) <- checkLoop (mkRedEnv doc try_me []) wanteds
1027 ; return (irreds,binds)
1030 try_me inst = ReduceMe AddSCs
1031 -- Here's the try-hard bit
1033 -----------------------------------------------------------
1034 gentleCheckLoop :: InstLoc
1037 -> TcM ([Inst], TcDictBinds)
1039 gentleCheckLoop inst_loc givens wanteds
1040 = do { (irreds,binds,_) <- checkLoop env wanteds
1041 ; return (irreds,binds)
1044 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1046 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
1048 -- When checking against a given signature
1049 -- we MUST be very gentle: Note [Check gently]
1051 gentleInferLoop :: SDoc -> [Inst]
1052 -> TcM ([Inst], TcDictBinds)
1053 gentleInferLoop doc wanteds
1054 = do { (irreds, binds, _) <- checkLoop env wanteds
1055 ; return (irreds, binds) }
1057 env = mkRedEnv doc try_me []
1058 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
1063 ~~~~~~~~~~~~~~~~~~~~
1064 We have to very careful about not simplifying too vigorously
1069 f :: Show b => T b -> b
1070 f (MkT x) = show [x]
1072 Inside the pattern match, which binds (a:*, x:a), we know that
1074 Hence we have a dictionary for Show [a] available; and indeed we
1075 need it. We are going to build an implication contraint
1076 forall a. (b~[a]) => Show [a]
1077 Later, we will solve this constraint using the knowledge (Show b)
1079 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1080 thing becomes insoluble. So we simplify gently (get rid of literals
1081 and methods only, plus common up equal things), deferring the real
1082 work until top level, when we solve the implication constraint
1083 with tryHardCheckLooop.
1087 -----------------------------------------------------------
1090 -> TcM ([Inst], TcDictBinds,
1091 [Inst]) -- needed givens
1092 -- Precondition: givens are completely rigid
1093 -- Postcondition: returned Insts are zonked
1095 checkLoop env wanteds
1097 where go env wanteds needed_givens
1098 = do { -- We do need to zonk the givens; cf Note [Zonking RedEnv]
1099 ; env' <- zonkRedEnv env
1100 ; wanteds' <- zonkInsts wanteds
1102 ; (improved, binds, irreds, more_needed_givens) <- reduceContext env' wanteds'
1104 ; let all_needed_givens = needed_givens ++ more_needed_givens
1106 ; if not improved then
1107 return (irreds, binds, all_needed_givens)
1110 -- If improvement did some unification, we go round again.
1111 -- We start again with irreds, not wanteds
1112 -- Using an instance decl might have introduced a fresh type variable
1113 -- which might have been unified, so we'd get an infinite loop
1114 -- if we started again with wanteds! See Note [LOOP]
1115 { (irreds1, binds1, all_needed_givens1) <- go env' irreds all_needed_givens
1116 ; return (irreds1, binds `unionBags` binds1, all_needed_givens1) } }
1119 Note [Zonking RedEnv]
1120 ~~~~~~~~~~~~~~~~~~~~~
1121 It might appear as if the givens in RedEnv are always rigid, but that is not
1122 necessarily the case for programs involving higher-rank types that have class
1123 contexts constraining the higher-rank variables. An example from tc237 in the
1126 class Modular s a | s -> a
1128 wim :: forall a w. Integral a
1129 => a -> (forall s. Modular s a => M s w) -> w
1130 wim i k = error "urk"
1132 test5 :: (Modular s a, Integral a) => M s a
1135 test4 = wim 4 test4'
1137 Notice how the variable 'a' of (Modular s a) in the rank-2 type of wim is
1138 quantified further outside. When type checking test4, we have to check
1139 whether the signature of test5 is an instance of
1141 (forall s. Modular s a => M s w)
1143 Consequently, we will get (Modular s t_a), where t_a is a TauTv into the
1146 Given the FD of Modular in this example, class improvement will instantiate
1147 t_a to 'a', where 'a' is the skolem from test5's signatures (due to the
1148 Modular s a predicate in that signature). If we don't zonk (Modular s t_a) in
1149 the givens, we will get into a loop as improveOne uses the unification engine
1150 TcGadt.tcUnifyTys, which doesn't know about mutable type variables.
1155 class If b t e r | b t e -> r
1158 class Lte a b c | a b -> c where lte :: a -> b -> c
1160 instance (Lte a b l,If l b a c) => Max a b c
1162 Wanted: Max Z (S x) y
1164 Then we'll reduce using the Max instance to:
1165 (Lte Z (S x) l, If l (S x) Z y)
1166 and improve by binding l->T, after which we can do some reduction
1167 on both the Lte and If constraints. What we *can't* do is start again
1168 with (Max Z (S x) y)!
1172 %************************************************************************
1174 tcSimplifySuperClasses
1176 %************************************************************************
1178 Note [SUPERCLASS-LOOP 1]
1179 ~~~~~~~~~~~~~~~~~~~~~~~~
1180 We have to be very, very careful when generating superclasses, lest we
1181 accidentally build a loop. Here's an example:
1185 class S a => C a where { opc :: a -> a }
1186 class S b => D b where { opd :: b -> b }
1188 instance C Int where
1191 instance D Int where
1194 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1195 Simplifying, we may well get:
1196 $dfCInt = :C ds1 (opd dd)
1199 Notice that we spot that we can extract ds1 from dd.
1201 Alas! Alack! We can do the same for (instance D Int):
1203 $dfDInt = :D ds2 (opc dc)
1207 And now we've defined the superclass in terms of itself.
1209 Solution: never generate a superclass selectors at all when
1210 satisfying the superclass context of an instance declaration.
1212 Two more nasty cases are in
1217 tcSimplifySuperClasses
1222 tcSimplifySuperClasses loc givens sc_wanteds
1223 = do { traceTc (text "tcSimplifySuperClasses")
1224 ; (irreds,binds1,_) <- checkLoop env sc_wanteds
1225 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1226 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1229 env = mkRedEnv (pprInstLoc loc) try_me givens
1230 try_me inst = ReduceMe NoSCs
1231 -- Like tryHardCheckLoop, but with NoSCs
1235 %************************************************************************
1237 \subsection{tcSimplifyRestricted}
1239 %************************************************************************
1241 tcSimplifyRestricted infers which type variables to quantify for a
1242 group of restricted bindings. This isn't trivial.
1245 We want to quantify over a to get id :: forall a. a->a
1248 We do not want to quantify over a, because there's an Eq a
1249 constraint, so we get eq :: a->a->Bool (notice no forall)
1252 RHS has type 'tau', whose free tyvars are tau_tvs
1253 RHS has constraints 'wanteds'
1256 Quantify over (tau_tvs \ ftvs(wanteds))
1257 This is bad. The constraints may contain (Monad (ST s))
1258 where we have instance Monad (ST s) where...
1259 so there's no need to be monomorphic in s!
1261 Also the constraint might be a method constraint,
1262 whose type mentions a perfectly innocent tyvar:
1263 op :: Num a => a -> b -> a
1264 Here, b is unconstrained. A good example would be
1266 We want to infer the polymorphic type
1267 foo :: forall b. b -> b
1270 Plan B (cunning, used for a long time up to and including GHC 6.2)
1271 Step 1: Simplify the constraints as much as possible (to deal
1272 with Plan A's problem). Then set
1273 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1275 Step 2: Now simplify again, treating the constraint as 'free' if
1276 it does not mention qtvs, and trying to reduce it otherwise.
1277 The reasons for this is to maximise sharing.
1279 This fails for a very subtle reason. Suppose that in the Step 2
1280 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1281 In the Step 1 this constraint might have been simplified, perhaps to
1282 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1283 This won't happen in Step 2... but that in turn might prevent some other
1284 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1285 and that in turn breaks the invariant that no constraints are quantified over.
1287 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1292 Step 1: Simplify the constraints as much as possible (to deal
1293 with Plan A's problem). Then set
1294 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1295 Return the bindings from Step 1.
1298 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1301 instance (HasBinary ty IO) => HasCodedValue ty
1303 foo :: HasCodedValue a => String -> IO a
1305 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1306 doDecodeIO codedValue view
1307 = let { act = foo "foo" } in act
1309 You might think this should work becuase the call to foo gives rise to a constraint
1310 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1311 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1312 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1314 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1318 Plan D (a variant of plan B)
1319 Step 1: Simplify the constraints as much as possible (to deal
1320 with Plan A's problem), BUT DO NO IMPROVEMENT. Then set
1321 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1323 Step 2: Now simplify again, treating the constraint as 'free' if
1324 it does not mention qtvs, and trying to reduce it otherwise.
1326 The point here is that it's generally OK to have too few qtvs; that is,
1327 to make the thing more monomorphic than it could be. We don't want to
1328 do that in the common cases, but in wierd cases it's ok: the programmer
1329 can always add a signature.
1331 Too few qtvs => too many wanteds, which is what happens if you do less
1336 tcSimplifyRestricted -- Used for restricted binding groups
1337 -- i.e. ones subject to the monomorphism restriction
1340 -> [Name] -- Things bound in this group
1341 -> TcTyVarSet -- Free in the type of the RHSs
1342 -> [Inst] -- Free in the RHSs
1343 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1344 TcDictBinds) -- Bindings
1345 -- tcSimpifyRestricted returns no constraints to
1346 -- quantify over; by definition there are none.
1347 -- They are all thrown back in the LIE
1349 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1350 -- Zonk everything in sight
1351 = do { traceTc (text "tcSimplifyRestricted")
1352 ; wanteds' <- zonkInsts wanteds
1354 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1355 -- dicts; the idea is to get rid of as many type
1356 -- variables as possible, and we don't want to stop
1357 -- at (say) Monad (ST s), because that reduces
1358 -- immediately, with no constraint on s.
1360 -- BUT do no improvement! See Plan D above
1361 -- HOWEVER, some unification may take place, if we instantiate
1362 -- a method Inst with an equality constraint
1363 ; let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1364 ; (_imp, _binds, constrained_dicts, _) <- reduceContext env wanteds'
1366 -- Next, figure out the tyvars we will quantify over
1367 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1368 ; gbl_tvs' <- tcGetGlobalTyVars
1369 ; constrained_dicts' <- zonkInsts constrained_dicts
1371 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1372 -- As in tcSimplifyInfer
1374 -- Do not quantify over constrained type variables:
1375 -- this is the monomorphism restriction
1376 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1377 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1378 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1381 ; warn_mono <- doptM Opt_WarnMonomorphism
1382 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1383 (vcat[ ptext SLIT("the Monomorphism Restriction applies to the binding")
1384 <> plural bndrs <+> ptext SLIT("for") <+> pp_bndrs,
1385 ptext SLIT("Consider giving a type signature for") <+> pp_bndrs])
1387 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1388 pprInsts wanteds, pprInsts constrained_dicts',
1390 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1392 -- The first step may have squashed more methods than
1393 -- necessary, so try again, this time more gently, knowing the exact
1394 -- set of type variables to quantify over.
1396 -- We quantify only over constraints that are captured by qtvs;
1397 -- these will just be a subset of non-dicts. This in contrast
1398 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1399 -- all *non-inheritable* constraints too. This implements choice
1400 -- (B) under "implicit parameter and monomorphism" above.
1402 -- Remember that we may need to do *some* simplification, to
1403 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1404 -- just to float all constraints
1406 -- At top level, we *do* squash methods becuase we want to
1407 -- expose implicit parameters to the test that follows
1408 ; let is_nested_group = isNotTopLevel top_lvl
1409 try_me inst | isFreeWrtTyVars qtvs inst,
1410 (is_nested_group || isDict inst) = Stop
1411 | otherwise = ReduceMe AddSCs
1412 env = mkNoImproveRedEnv doc try_me
1413 ; (_imp, binds, irreds, _) <- reduceContext env wanteds'
1415 -- See "Notes on implicit parameters, Question 4: top level"
1416 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1417 if is_nested_group then
1419 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1420 ; addTopIPErrs bndrs bad_ips
1421 ; extendLIEs non_ips }
1423 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1424 ; return (qtvs', binds) }
1428 %************************************************************************
1432 %************************************************************************
1434 On the LHS of transformation rules we only simplify methods and constants,
1435 getting dictionaries. We want to keep all of them unsimplified, to serve
1436 as the available stuff for the RHS of the rule.
1438 Example. Consider the following left-hand side of a rule
1440 f (x == y) (y > z) = ...
1442 If we typecheck this expression we get constraints
1444 d1 :: Ord a, d2 :: Eq a
1446 We do NOT want to "simplify" to the LHS
1448 forall x::a, y::a, z::a, d1::Ord a.
1449 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1453 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1454 f ((==) d2 x y) ((>) d1 y z) = ...
1456 Here is another example:
1458 fromIntegral :: (Integral a, Num b) => a -> b
1459 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1461 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1462 we *dont* want to get
1464 forall dIntegralInt.
1465 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1467 because the scsel will mess up RULE matching. Instead we want
1469 forall dIntegralInt, dNumInt.
1470 fromIntegral Int Int dIntegralInt dNumInt = id Int
1474 g (x == y) (y == z) = ..
1476 where the two dictionaries are *identical*, we do NOT WANT
1478 forall x::a, y::a, z::a, d1::Eq a
1479 f ((==) d1 x y) ((>) d1 y z) = ...
1481 because that will only match if the dict args are (visibly) equal.
1482 Instead we want to quantify over the dictionaries separately.
1484 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1485 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1486 from scratch, rather than further parameterise simpleReduceLoop etc
1489 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1490 tcSimplifyRuleLhs wanteds
1491 = go [] emptyBag wanteds
1494 = return (dicts, binds)
1495 go dicts binds (w:ws)
1497 = go (w:dicts) binds ws
1499 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1500 -- to fromInteger; this looks fragile to me
1501 ; lookup_result <- lookupSimpleInst w'
1502 ; case lookup_result of
1504 go dicts (addInstToDictBind binds w rhs) (ws' ++ ws)
1505 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1509 tcSimplifyBracket is used when simplifying the constraints arising from
1510 a Template Haskell bracket [| ... |]. We want to check that there aren't
1511 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1512 Show instance), but we aren't otherwise interested in the results.
1513 Nor do we care about ambiguous dictionaries etc. We will type check
1514 this bracket again at its usage site.
1517 tcSimplifyBracket :: [Inst] -> TcM ()
1518 tcSimplifyBracket wanteds
1519 = do { tryHardCheckLoop doc wanteds
1522 doc = text "tcSimplifyBracket"
1526 %************************************************************************
1528 \subsection{Filtering at a dynamic binding}
1530 %************************************************************************
1535 we must discharge all the ?x constraints from B. We also do an improvement
1536 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1538 Actually, the constraints from B might improve the types in ?x. For example
1540 f :: (?x::Int) => Char -> Char
1543 then the constraint (?x::Int) arising from the call to f will
1544 force the binding for ?x to be of type Int.
1547 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1550 -- We need a loop so that we do improvement, and then
1551 -- (next time round) generate a binding to connect the two
1553 -- Here the two ?x's have different types, and improvement
1554 -- makes them the same.
1556 tcSimplifyIPs given_ips wanteds
1557 = do { wanteds' <- zonkInsts wanteds
1558 ; given_ips' <- zonkInsts given_ips
1559 -- Unusually for checking, we *must* zonk the given_ips
1561 ; let env = mkRedEnv doc try_me given_ips'
1562 ; (improved, binds, irreds, _) <- reduceContext env wanteds'
1564 ; if not improved then
1565 ASSERT( all is_free irreds )
1566 do { extendLIEs irreds
1569 tcSimplifyIPs given_ips wanteds }
1571 doc = text "tcSimplifyIPs" <+> ppr given_ips
1572 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1573 is_free inst = isFreeWrtIPs ip_set inst
1575 -- Simplify any methods that mention the implicit parameter
1576 try_me inst | is_free inst = Stop
1577 | otherwise = ReduceMe NoSCs
1581 %************************************************************************
1583 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1585 %************************************************************************
1587 When doing a binding group, we may have @Insts@ of local functions.
1588 For example, we might have...
1590 let f x = x + 1 -- orig local function (overloaded)
1591 f.1 = f Int -- two instances of f
1596 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1597 where @f@ is in scope; those @Insts@ must certainly not be passed
1598 upwards towards the top-level. If the @Insts@ were binding-ified up
1599 there, they would have unresolvable references to @f@.
1601 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1602 For each method @Inst@ in the @init_lie@ that mentions one of the
1603 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1604 @LIE@), as well as the @HsBinds@ generated.
1607 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1608 -- Simlifies only MethodInsts, and generate only bindings of form
1610 -- We're careful not to even generate bindings of the form
1612 -- You'd think that'd be fine, but it interacts with what is
1613 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1615 bindInstsOfLocalFuns wanteds local_ids
1616 | null overloaded_ids
1618 = extendLIEs wanteds `thenM_`
1619 returnM emptyLHsBinds
1622 = do { (irreds, binds) <- gentleInferLoop doc for_me
1623 ; extendLIEs not_for_me
1627 doc = text "bindInsts" <+> ppr local_ids
1628 overloaded_ids = filter is_overloaded local_ids
1629 is_overloaded id = isOverloadedTy (idType id)
1630 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1632 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1633 -- so it's worth building a set, so that
1634 -- lookup (in isMethodFor) is faster
1638 %************************************************************************
1640 \subsection{Data types for the reduction mechanism}
1642 %************************************************************************
1644 The main control over context reduction is here
1648 = RedEnv { red_doc :: SDoc -- The context
1649 , red_try_me :: Inst -> WhatToDo
1650 , red_improve :: Bool -- True <=> do improvement
1651 , red_givens :: [Inst] -- All guaranteed rigid
1653 -- but see Note [Rigidity]
1654 , red_reft :: Refinement -- The refinement to apply to the 'givens'
1655 -- You should think of it as 'given equalities'
1656 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1657 -- See Note [RedStack]
1661 -- The red_givens are rigid so far as cmpInst is concerned.
1662 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1663 -- let ?x = e in ...
1664 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1665 -- But that doesn't affect the comparison, which is based only on mame.
1668 -- The red_stack pair (n,insts) pair is just used for error reporting.
1669 -- 'n' is always the depth of the stack.
1670 -- The 'insts' is the stack of Insts being reduced: to produce X
1671 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1674 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1675 mkRedEnv doc try_me givens
1676 = RedEnv { red_doc = doc, red_try_me = try_me,
1677 red_givens = givens,
1678 red_reft = emptyRefinement,
1680 red_improve = True }
1682 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1683 -- Do not do improvement; no givens
1684 mkNoImproveRedEnv doc try_me
1685 = RedEnv { red_doc = doc, red_try_me = try_me,
1686 red_givens = [], red_reft = emptyRefinement,
1688 red_improve = True }
1691 = ReduceMe WantSCs -- Try to reduce this
1692 -- If there's no instance, add the inst to the
1693 -- irreductible ones, but don't produce an error
1694 -- message of any kind.
1695 -- It might be quite legitimate such as (Eq a)!
1697 | Stop -- Return as irreducible unless it can
1698 -- be reduced to a constant in one step
1699 -- Do not add superclasses; see
1701 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1702 -- of a predicate when adding it to the avails
1703 -- The reason for this flag is entirely the super-class loop problem
1704 -- Note [SUPER-CLASS LOOP 1]
1706 zonkRedEnv :: RedEnv -> TcM RedEnv
1708 = do { givens' <- mappM zonkInst (red_givens env)
1709 ; return $ env {red_givens = givens'}
1714 %************************************************************************
1716 \subsection[reduce]{@reduce@}
1718 %************************************************************************
1720 Note [Ancestor Equalities]
1721 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1722 During context reduction, we add to the wanted equalities also those
1723 equalities that (transitively) occur in superclass contexts of wanted
1724 class constraints. Consider the following code
1726 class a ~ Int => C a
1729 If (C a) is wanted, we want to add (a ~ Int), which will be discharged by
1730 substituting Int for a. Hence, we ultimately want (C Int), which we
1731 discharge with the explicit instance.
1734 reduceContext :: RedEnv
1736 -> TcM (ImprovementDone,
1737 TcDictBinds, -- Dictionary bindings
1738 [Inst], -- Irreducible
1739 [Inst]) -- Needed givens
1741 reduceContext env wanteds
1742 = do { traceTc (text "reduceContext" <+> (vcat [
1743 text "----------------------",
1745 text "given" <+> ppr (red_givens env),
1746 text "wanted" <+> ppr wanteds,
1747 text "----------------------"
1751 ; let givens = red_givens env
1752 (given_eqs0, given_dicts0) = partition isEqInst givens
1753 (wanted_eqs0, wanted_dicts) = partition isEqInst wanteds
1755 -- We want to add as wanted equalities those that (transitively)
1756 -- occur in superclass contexts of wanted class constraints.
1757 -- See Note [Ancestor Equalities]
1758 ; ancestor_eqs <- ancestorEqualities wanted_dicts
1759 ; let wanted_eqs = wanted_eqs0 ++ ancestor_eqs
1760 ; traceTc $ text "reduceContext: ancestor eqs" <+> ppr ancestor_eqs
1762 -- 1. Normalise the *given* *equality* constraints
1763 ; (given_eqs, eliminate_skolems) <- normaliseGivenEqs given_eqs0
1765 -- 2. Normalise the *given* *dictionary* constraints
1766 -- wrt. the toplevel and given equations
1767 ; (given_dicts, given_binds) <- normaliseGivenDicts given_eqs
1770 -- 3. Solve the *wanted* *equation* constraints
1771 ; eq_irreds0 <- solveWantedEqs given_eqs wanted_eqs
1773 -- 4. Normalise the *wanted* equality constraints with respect to
1775 ; eq_irreds <- normaliseWantedEqs eq_irreds0
1777 -- 5. Build the Avail mapping from "given_dicts"
1778 -- Add dicts refined by the current type refinement
1779 ; init_state <- foldlM addGiven emptyAvails given_dicts
1780 ; let reft = red_reft env
1781 ; init_state <- if isEmptyRefinement reft then return init_state
1782 else foldlM (addRefinedGiven reft)
1783 init_state given_dicts
1785 -- 6. Solve the *wanted* *dictionary* constraints
1786 -- This may expose some further equational constraints...
1787 ; wanted_dicts' <- zonkInsts wanted_dicts
1788 ; avails <- reduceList env wanted_dicts' init_state
1789 ; (binds, irreds0, needed_givens) <- extractResults avails wanted_dicts'
1790 ; traceTc $ text "reduceContext extractresults" <+> vcat
1791 [ppr avails,ppr wanted_dicts',ppr binds,ppr needed_givens]
1793 -- 7. Normalise the *wanted* *dictionary* constraints
1794 -- wrt. the toplevel and given equations
1795 ; (irreds1,normalise_binds1) <- normaliseWantedDicts given_eqs irreds0
1797 -- 8. Substitute the wanted *equations* in the wanted *dictionaries*
1798 ; (irreds,normalise_binds2) <- substEqInDictInsts eq_irreds irreds1
1800 -- 9. eliminate the artificial skolem constants introduced in 1.
1803 -- If there was some FD improvement,
1804 -- or new wanted equations have been exposed,
1805 -- we should have another go at solving.
1806 ; let improved = availsImproved avails
1807 || (not $ isEmptyBag normalise_binds1)
1808 || (not $ isEmptyBag normalise_binds2)
1809 || (any isEqInst irreds)
1811 ; traceTc (text "reduceContext end" <+> (vcat [
1812 text "----------------------",
1814 text "given" <+> ppr (red_givens env),
1815 text "wanted" <+> ppr wanteds,
1817 text "avails" <+> pprAvails avails,
1818 text "improved =" <+> ppr improved,
1819 text "irreds = " <+> ppr irreds,
1820 text "binds = " <+> ppr binds,
1821 text "needed givens = " <+> ppr needed_givens,
1822 text "----------------------"
1826 given_binds `unionBags` normalise_binds1
1827 `unionBags` normalise_binds2
1829 irreds ++ eq_irreds,
1833 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1834 tcImproveOne avails inst
1835 | not (isDict inst) = return False
1837 = do { inst_envs <- tcGetInstEnvs
1838 ; let eqns = improveOne (classInstances inst_envs)
1839 (dictPred inst, pprInstArising inst)
1840 [ (dictPred p, pprInstArising p)
1841 | p <- availsInsts avails, isDict p ]
1842 -- Avails has all the superclasses etc (good)
1843 -- It also has all the intermediates of the deduction (good)
1844 -- It does not have duplicates (good)
1845 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1846 -- so that improve will see them separate
1847 ; traceTc (text "improveOne" <+> ppr inst)
1850 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1851 -> TcM ImprovementDone
1852 unifyEqns [] = return False
1854 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1858 unify ((qtvs, pairs), what1, what2)
1859 = addErrCtxtM (mkEqnMsg what1 what2) $
1860 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1861 mapM_ (unif_pr tenv) pairs
1862 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1864 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1866 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1867 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1868 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1869 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1870 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1871 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1872 ; return (tidy_env, msg) }
1875 The main context-reduction function is @reduce@. Here's its game plan.
1878 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1879 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1880 = do { traceTc (text "reduceList " <+> (ppr wanteds $$ ppr state))
1884 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1885 2 (ifPprDebug (nest 2 (pprStack stk))))
1888 ; if n >= ctxtStkDepth dopts then
1889 failWithTc (reduceDepthErr n stk)
1893 go [] state = return state
1894 go (w:ws) state = do { state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1897 -- Base case: we're done!
1898 reduce env wanted avails
1899 -- It's the same as an existing inst, or a superclass thereof
1900 | Just avail <- findAvail avails wanted
1901 = do { traceTc (text "reduce: found " <+> ppr wanted)
1906 = do { traceTc (text "reduce" <+> ppr wanted $$ ppr avails)
1907 ; case red_try_me env wanted of {
1908 Stop -> try_simple (addIrred NoSCs);
1909 -- See Note [No superclasses for Stop]
1911 ReduceMe want_scs -> do -- It should be reduced
1912 { (avails, lookup_result) <- reduceInst env avails wanted
1913 ; case lookup_result of
1914 NoInstance -> addIrred want_scs avails wanted
1915 -- Add it and its superclasses
1917 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1919 GenInst wanteds' rhs
1920 -> do { avails1 <- addIrred NoSCs avails wanted
1921 ; avails2 <- reduceList env wanteds' avails1
1922 ; addWanted want_scs avails2 wanted rhs wanteds' } }
1923 -- Temporarily do addIrred *before* the reduceList,
1924 -- which has the effect of adding the thing we are trying
1925 -- to prove to the database before trying to prove the things it
1926 -- needs. See note [RECURSIVE DICTIONARIES]
1927 -- NB: we must not do an addWanted before, because that adds the
1928 -- superclasses too, and that can lead to a spurious loop; see
1929 -- the examples in [SUPERCLASS-LOOP]
1930 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1933 -- First, see if the inst can be reduced to a constant in one step
1934 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1935 -- Don't bother for implication constraints, which take real work
1936 try_simple do_this_otherwise
1937 = do { res <- lookupSimpleInst wanted
1939 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1940 other -> do_this_otherwise avails wanted }
1944 Note [SUPERCLASS-LOOP 2]
1945 ~~~~~~~~~~~~~~~~~~~~~~~~
1946 But the above isn't enough. Suppose we are *given* d1:Ord a,
1947 and want to deduce (d2:C [a]) where
1949 class Ord a => C a where
1950 instance Ord [a] => C [a] where ...
1952 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1953 superclasses of C [a] to avails. But we must not overwrite the binding
1954 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1957 Here's another variant, immortalised in tcrun020
1958 class Monad m => C1 m
1959 class C1 m => C2 m x
1960 instance C2 Maybe Bool
1961 For the instance decl we need to build (C1 Maybe), and it's no good if
1962 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1963 before we search for C1 Maybe.
1965 Here's another example
1966 class Eq b => Foo a b
1967 instance Eq a => Foo [a] a
1971 we'll first deduce that it holds (via the instance decl). We must not
1972 then overwrite the Eq t constraint with a superclass selection!
1974 At first I had a gross hack, whereby I simply did not add superclass constraints
1975 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1976 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1977 I found a very obscure program (now tcrun021) in which improvement meant the
1978 simplifier got two bites a the cherry... so something seemed to be an Stop
1979 first time, but reducible next time.
1981 Now we implement the Right Solution, which is to check for loops directly
1982 when adding superclasses. It's a bit like the occurs check in unification.
1985 Note [RECURSIVE DICTIONARIES]
1986 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1988 data D r = ZeroD | SuccD (r (D r));
1990 instance (Eq (r (D r))) => Eq (D r) where
1991 ZeroD == ZeroD = True
1992 (SuccD a) == (SuccD b) = a == b
1995 equalDC :: D [] -> D [] -> Bool;
1998 We need to prove (Eq (D [])). Here's how we go:
2002 by instance decl, holds if
2006 by instance decl of Eq, holds if
2008 where d2 = dfEqList d3
2011 But now we can "tie the knot" to give
2017 and it'll even run! The trick is to put the thing we are trying to prove
2018 (in this case Eq (D []) into the database before trying to prove its
2019 contributing clauses.
2022 %************************************************************************
2024 Reducing a single constraint
2026 %************************************************************************
2029 ---------------------------------------------
2030 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
2031 reduceInst env avails (ImplicInst { tci_name = name,
2032 tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
2033 tci_given = extra_givens, tci_wanted = wanteds })
2034 = reduceImplication env avails name reft tvs extra_givens wanteds loc
2036 reduceInst env avails other_inst
2037 = do { result <- lookupSimpleInst other_inst
2038 ; return (avails, result) }
2041 Note [Equational Constraints in Implication Constraints]
2042 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2044 An implication constraint is of the form
2046 where Given and Wanted may contain both equational and dictionary
2047 constraints. The delay and reduction of these two kinds of constraints
2050 -) In the generated code, wanted Dictionary constraints are wrapped up in an
2051 implication constraint that is created at the code site where the wanted
2052 dictionaries can be reduced via a let-binding. This let-bound implication
2053 constraint is deconstructed at the use-site of the wanted dictionaries.
2055 -) While the reduction of equational constraints is also delayed, the delay
2056 is not manifest in the generated code. The required evidence is generated
2057 in the code directly at the use-site. There is no let-binding and deconstruction
2058 necessary. The main disadvantage is that we cannot exploit sharing as the
2059 same evidence may be generated at multiple use-sites. However, this disadvantage
2060 is limited because it only concerns coercions which are erased.
2062 The different treatment is motivated by the different in representation. Dictionary
2063 constraints require manifest runtime dictionaries, while equations require coercions
2067 ---------------------------------------------
2068 reduceImplication :: RedEnv
2071 -> Refinement -- May refine the givens; often empty
2072 -> [TcTyVar] -- Quantified type variables; all skolems
2073 -> [Inst] -- Extra givens; all rigid
2076 -> TcM (Avails, LookupInstResult)
2079 Suppose we are simplifying the constraint
2080 forall bs. extras => wanted
2081 in the context of an overall simplification problem with givens 'givens',
2082 and refinment 'reft'.
2085 * The refinement is often empty
2087 * The 'extra givens' need not mention any of the quantified type variables
2088 e.g. forall {}. Eq a => Eq [a]
2089 forall {}. C Int => D (Tree Int)
2091 This happens when you have something like
2093 T1 :: Eq a => a -> T a
2096 f x = ...(case x of { T1 v -> v==v })...
2099 -- ToDo: should we instantiate tvs? I think it's not necessary
2101 -- Note on coercion variables:
2103 -- The extra given coercion variables are bound at two different sites:
2104 -- -) in the creation context of the implication constraint
2105 -- the solved equational constraints use these binders
2107 -- -) at the solving site of the implication constraint
2108 -- the solved dictionaries use these binders
2109 -- these binders are generated by reduceImplication
2111 reduceImplication env orig_avails name reft tvs extra_givens wanteds inst_loc
2112 = do { -- Add refined givens, and the extra givens
2114 -- (refined_red_givens,refined_avails)
2115 -- <- if isEmptyRefinement reft then return (red_givens env,orig_avails)
2116 -- else foldlM (addRefinedGiven reft) ([],orig_avails) (red_givens env)
2117 -- Commented out SLPJ Sept 07; see comment with extractLocalResults below
2118 let refined_red_givens = []
2120 -- Solve the sub-problem
2121 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
2122 env' = env { red_givens = extra_givens ++ availsInsts orig_avails
2124 , red_doc = sep [ptext SLIT("reduceImplication for") <+> ppr name,
2125 nest 2 (parens $ ptext SLIT("within") <+> red_doc env)]
2126 , red_try_me = try_me }
2128 ; traceTc (text "reduceImplication" <+> vcat
2130 ppr (red_givens env), ppr extra_givens,
2131 ppr reft, ppr wanteds])
2132 ; (irreds,binds,needed_givens0) <- checkLoop env' wanteds
2133 ; let (extra_eq_givens, extra_dict_givens) = partition isEqInst extra_givens
2134 -- SLPJ Sept 07: I think this is bogus; currently
2135 -- there are no Eqinsts in extra_givens
2136 dict_ids = map instToId extra_dict_givens
2138 -- needed_givens0 is the free vars of the bindings
2139 -- Remove the ones we are going to lambda-bind
2140 -- Use the actual dictionary identity *not* equality on Insts
2141 -- (Mind you, it should make no difference here.)
2142 ; let needed_givens = [ng | ng <- needed_givens0
2143 , instToVar ng `notElem` dict_ids]
2145 -- Note [Reducing implication constraints]
2146 -- Tom -- update note, put somewhere!
2148 ; traceTc (text "reduceImplication result" <+> vcat
2149 [ppr irreds, ppr binds, ppr needed_givens])
2151 ; -- extract superclass binds
2152 -- (sc_binds,_) <- extractResults avails []
2153 -- ; traceTc (text "reduceImplication sc_binds" <+> vcat
2154 -- [ppr sc_binds, ppr avails])
2157 -- We always discard the extra avails we've generated;
2158 -- but we remember if we have done any (global) improvement
2159 -- ; let ret_avails = avails
2160 ; let ret_avails = orig_avails
2161 -- ; let ret_avails = updateImprovement orig_avails avails
2163 -- SLPJ Sept 07: what if improvement happened inside the checkLoop?
2164 -- Then we must iterate the outer loop too!
2166 ; traceTc (text "reduceImplication condition" <+> ppr ((isEmptyLHsBinds binds) || (null irreds)))
2168 -- Progress is no longer measered by the number of bindings
2169 -- ; if isEmptyLHsBinds binds then -- No progress
2170 ; if (isEmptyLHsBinds binds) && (not $ null irreds) then
2171 return (ret_avails, NoInstance)
2174 ; (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
2175 -- This binding is useless if the recursive simplification
2176 -- made no progress; but currently we don't try to optimise that
2177 -- case. After all, we only try hard to reduce at top level, or
2178 -- when inferring types.
2180 ; let dict_wanteds = filter (not . isEqInst) wanteds
2181 -- TOMDO: given equational constraints bug!
2182 -- we need a different evidence for given
2183 -- equations depending on whether we solve
2184 -- dictionary constraints or equational constraints
2186 eq_tyvars = uniqSetToList $ tyVarsOfTypes $ map eqInstType extra_eq_givens
2187 -- SLPJ Sept07: this looks Utterly Wrong to me, but I think
2188 -- that current extra_givens has no EqInsts, so
2189 -- it makes no difference
2190 -- dict_ids = map instToId extra_givens
2191 co = mkWpTyLams tvs <.> mkWpTyLams eq_tyvars <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` bind)
2192 rhs = mkHsWrap co payload
2193 loc = instLocSpan inst_loc
2194 payload | [dict_wanted] <- dict_wanteds = HsVar (instToId dict_wanted)
2195 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) dict_wanteds) Boxed
2198 ; traceTc (vcat [text "reduceImplication" <+> ppr name,
2200 text "->" <+> sep [ppr needed_givens, ppr rhs]])
2201 -- If there are any irreds, we back off and return NoInstance
2202 ; return (ret_avails, GenInst (implic_insts ++ needed_givens) (L loc rhs))
2207 Note [Reducing implication constraints]
2208 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2209 Suppose we are trying to simplify
2211 ic: (forall b. C a b => (W [a] b, D c b)) )
2213 instance (C a b, Ord a) => W [a] b
2214 When solving the implication constraint, we'll start with
2216 in the Avails. Then we add (C a b -> Given) and solve. Extracting
2217 the results gives us a binding for the (W [a] b), with an Irred of
2218 (Ord a, D c b). Now, the (Ord a) comes from "outside" the implication,
2219 but the (D d b) is from "inside". So we want to generate a GenInst
2224 ic' :: forall b. C a b => D c b]
2225 (/\b \(dc:C a b). (df a b dc do, ic' b dc))
2227 The first arg of GenInst gives the free dictionary variables of the
2228 second argument -- the "needed givens". And that list in turn is
2229 vital because it's used to determine what other dicts must be solved.
2230 This very list ends up in the second field of the Rhs, and drives
2233 The need for this field is why we have to return "needed givens"
2234 from extractResults, reduceContext, checkLoop, and so on.
2236 NB: the "needed givens" in a GenInst or Rhs, may contain two dicts
2237 with the same type but different Ids, e.g. [d12 :: Eq a, d81 :: Eq a]
2238 That says we must generate a binding for both d12 and d81.
2240 The "inside" and "outside" distinction is what's going on with 'inner' and
2241 'outer' in reduceImplication
2244 Note [Freeness and implications]
2245 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2246 It's hard to say when an implication constraint can be floated out. Consider
2247 forall {} Eq a => Foo [a]
2248 The (Foo [a]) doesn't mention any of the quantified variables, but it
2249 still might be partially satisfied by the (Eq a).
2251 There is a useful special case when it *is* easy to partition the
2252 constraints, namely when there are no 'givens'. Consider
2253 forall {a}. () => Bar b
2254 There are no 'givens', and so there is no reason to capture (Bar b).
2255 We can let it float out. But if there is even one constraint we
2256 must be much more careful:
2257 forall {a}. C a b => Bar (m b)
2258 because (C a b) might have a superclass (D b), from which we might
2259 deduce (Bar [b]) when m later gets instantiated to []. Ha!
2261 Here is an even more exotic example
2263 Now consider the constraint
2264 forall b. D Int b => C Int
2265 We can satisfy the (C Int) from the superclass of D, so we don't want
2266 to float the (C Int) out, even though it mentions no type variable in
2269 Note [Pruning the givens in an implication constraint]
2270 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2271 Suppose we are about to form the implication constraint
2272 forall tvs. Eq a => Ord b
2273 The (Eq a) cannot contribute to the (Ord b), because it has no access to
2274 the type variable 'b'. So we could filter out the (Eq a) from the givens.
2276 Doing so would be a bit tidier, but all the implication constraints get
2277 simplified away by the optimiser, so it's no great win. So I don't take
2278 advantage of that at the moment.
2280 If you do, BE CAREFUL of wobbly type variables.
2283 %************************************************************************
2285 Avails and AvailHow: the pool of evidence
2287 %************************************************************************
2291 data Avails = Avails !ImprovementDone !AvailEnv
2293 type ImprovementDone = Bool -- True <=> some unification has happened
2294 -- so some Irreds might now be reducible
2295 -- keys that are now
2297 type AvailEnv = FiniteMap Inst AvailHow
2299 = IsIrred -- Used for irreducible dictionaries,
2300 -- which are going to be lambda bound
2302 | Given Inst -- Used for dictionaries for which we have a binding
2303 -- e.g. those "given" in a signature
2305 | Rhs -- Used when there is a RHS
2306 (LHsExpr TcId) -- The RHS
2307 [Inst] -- Insts free in the RHS; we need these too
2309 instance Outputable Avails where
2312 pprAvails (Avails imp avails)
2313 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
2315 vcat [ sep [ppr inst, nest 2 (equals <+> ppr avail)]
2316 | (inst,avail) <- fmToList avails ]]
2318 instance Outputable AvailHow where
2321 -------------------------
2322 pprAvail :: AvailHow -> SDoc
2323 pprAvail IsIrred = text "Irred"
2324 pprAvail (Given x) = text "Given" <+> ppr x
2325 pprAvail (Rhs rhs bs) = sep [text "Rhs" <+> ppr bs,
2328 -------------------------
2329 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2330 extendAvailEnv env inst avail = addToFM env inst avail
2332 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2333 findAvailEnv env wanted = lookupFM env wanted
2334 -- NB 1: the Ord instance of Inst compares by the class/type info
2335 -- *not* by unique. So
2336 -- d1::C Int == d2::C Int
2338 emptyAvails :: Avails
2339 emptyAvails = Avails False emptyFM
2341 findAvail :: Avails -> Inst -> Maybe AvailHow
2342 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2344 elemAvails :: Inst -> Avails -> Bool
2345 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2347 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2349 extendAvails avails@(Avails imp env) inst avail
2350 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2351 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2353 availsInsts :: Avails -> [Inst]
2354 availsInsts (Avails _ avails) = keysFM avails
2356 availsImproved (Avails imp _) = imp
2358 updateImprovement :: Avails -> Avails -> Avails
2359 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
2360 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
2363 Extracting the bindings from a bunch of Avails.
2364 The bindings do *not* come back sorted in dependency order.
2365 We assume that they'll be wrapped in a big Rec, so that the
2366 dependency analyser can sort them out later
2369 extractResults :: Avails
2371 -> TcM ( TcDictBinds, -- Bindings
2372 [Inst], -- Irreducible ones
2373 [Inst]) -- Needed givens, i.e. ones used in the bindings
2374 -- Postcondition: needed-givens = free vars( binds ) \ irreds
2375 -- needed-gives is subset of Givens in incoming Avails
2376 -- Note [Reducing implication constraints]
2378 extractResults (Avails _ avails) wanteds
2379 = go avails emptyBag [] [] wanteds
2381 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst] -> [Inst]
2382 -> TcM (TcDictBinds, [Inst], [Inst])
2383 go avails binds irreds givens []
2384 = returnM (binds, irreds, givens)
2386 go avails binds irreds givens (w:ws)
2387 = case findAvailEnv avails w of
2388 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2389 go avails binds irreds givens ws
2391 Just (Given g) -> go (avails_with g g_id)
2392 (add_triv_bind g_id)
2393 irreds (g:givens) ws
2394 -- avail_with g ensures that we don't emit the
2395 -- same given twice into needed-givens
2399 Just IsIrred -> go (avails_with w w_id) binds (w:irreds) givens ws
2401 -- The avails_with_w handles the case where we want (Ord a, Eq a), and we
2402 -- don't want to emit *two* Irreds for Ord a, one via the superclass chain
2403 -- This showed up in a dupliated Ord constraint in the error message for
2405 -- More generally, we don't want to emit two irreds with
2408 Just (Rhs rhs@(L _ (HsVar g_id)) ws')
2409 -> go avails (add_triv_bind g_id) irreds givens (ws' ++ ws)
2412 -> go (avails_with w w_id) (add_bind rhs)
2413 irreds givens (ws' ++ ws)
2414 -- The avails-with w replaces a complex RHS with a simple one
2415 -- for the benefit of subsequent lookups
2419 add_triv_bind rhs_id | rhs_id == w_id = binds
2420 | otherwise = add_bind (nlHsVar rhs_id)
2421 -- The sought Id can be one of the givens, via a
2422 -- superclass chain and then we definitely don't
2423 -- want to generate an x=x binding!
2425 add_bind rhs = addInstToDictBind binds w rhs
2426 avails_with w w_id = extendAvailEnv avails w (Rhs (nlHsVar w_id) [])
2430 Note [No superclasses for Stop]
2431 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2432 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2433 add it to avails, so that any other equal Insts will be commoned up
2434 right here. However, we do *not* add superclasses. If we have
2437 but a is not bound here, then we *don't* want to derive dn from df
2438 here lest we lose sharing.
2441 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2442 addWanted want_scs avails wanted rhs_expr wanteds
2443 = addAvailAndSCs want_scs avails wanted avail
2445 avail = Rhs rhs_expr wanteds
2447 addGiven :: Avails -> Inst -> TcM Avails
2448 addGiven avails given = addAvailAndSCs AddSCs avails given (Given given)
2449 -- Always add superclasses for 'givens'
2451 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2452 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2453 -- so the assert isn't true
2455 addRefinedGiven :: Refinement -> Avails -> Inst -> TcM Avails
2456 addRefinedGiven reft avails given
2457 | isDict given -- We sometimes have 'given' methods, but they
2458 -- are always optional, so we can drop them
2459 , let pred = dictPred given
2460 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2461 , Just (co, pred) <- refinePred reft pred
2462 = do { new_given <- newDictBndr (instLoc given) pred
2463 ; let rhs = L (instSpan given) $
2464 HsWrap (WpCo co) (HsVar (instToId given))
2465 ; addAvailAndSCs AddSCs avails new_given (Rhs rhs [given]) }
2466 -- ToDo: the superclasses of the original given all exist in Avails
2467 -- so we could really just cast them, but it's more awkward to do,
2468 -- and hopefully the optimiser will spot the duplicated work
2473 Note [ImplicInst rigidity]
2474 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2476 C :: forall ab. (Eq a, Ord b) => b -> T a
2478 ...(case x of C v -> <body>)...
2480 From the case (where x::T ty) we'll get an implication constraint
2481 forall b. (Eq ty, Ord b) => <body-constraints>
2482 Now suppose <body-constraints> itself has an implication constraint
2484 forall c. <reft> => <payload>
2485 Then, we can certainly apply the refinement <reft> to the Ord b, becuase it is
2486 existential, but we probably should not apply it to the (Eq ty) because it may
2487 be wobbly. Hence the isRigidInst
2489 @Insts@ are ordered by their class/type info, rather than by their
2490 unique. This allows the context-reduction mechanism to use standard finite
2491 maps to do their stuff. It's horrible that this code is here, rather
2492 than with the Avails handling stuff in TcSimplify
2495 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2496 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2497 addAvailAndSCs want_scs avails irred IsIrred
2499 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2500 addAvailAndSCs want_scs avails inst avail
2501 | not (isClassDict inst) = extendAvails avails inst avail
2502 | NoSCs <- want_scs = extendAvails avails inst avail
2503 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2504 ; avails' <- extendAvails avails inst avail
2505 ; addSCs is_loop avails' inst }
2507 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2508 -- Note: this compares by *type*, not by Unique
2509 deps = findAllDeps (unitVarSet (instToVar inst)) avail
2510 dep_tys = map idType (varSetElems deps)
2512 findAllDeps :: IdSet -> AvailHow -> IdSet
2513 -- Find all the Insts that this one depends on
2514 -- See Note [SUPERCLASS-LOOP 2]
2515 -- Watch out, though. Since the avails may contain loops
2516 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2517 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2518 findAllDeps so_far other = so_far
2520 find_all :: IdSet -> Inst -> IdSet
2522 | isEqInst kid = so_far
2523 | kid_id `elemVarSet` so_far = so_far
2524 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2525 | otherwise = so_far'
2527 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2528 kid_id = instToId kid
2530 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2531 -- Add all the superclasses of the Inst to Avails
2532 -- The first param says "don't do this because the original thing
2533 -- depends on this one, so you'd build a loop"
2534 -- Invariant: the Inst is already in Avails.
2536 addSCs is_loop avails dict
2537 = ASSERT( isDict dict )
2538 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2539 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2541 (clas, tys) = getDictClassTys dict
2542 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2543 sc_theta' = filter (not . isEqPred) $
2544 substTheta (zipTopTvSubst tyvars tys) sc_theta
2546 add_sc avails (sc_dict, sc_sel)
2547 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2548 | is_given sc_dict = return avails
2549 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2550 ; addSCs is_loop avails' sc_dict }
2552 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2553 co_fn = WpApp (instToVar dict) <.> mkWpTyApps tys
2555 is_given :: Inst -> Bool
2556 is_given sc_dict = case findAvail avails sc_dict of
2557 Just (Given _) -> True -- Given is cheaper than superclass selection
2560 -- From the a set of insts obtain all equalities that (transitively) occur in
2561 -- superclass contexts of class constraints (aka the ancestor equalities).
2563 ancestorEqualities :: [Inst] -> TcM [Inst]
2565 = mapM mkWantedEqInst -- turn only equality predicates..
2566 . filter isEqPred -- ..into wanted equality insts
2568 . addAEsToBag emptyBag -- collect the superclass constraints..
2569 . map dictPred -- ..of all predicates in a bag
2570 . filter isClassDict
2572 addAEsToBag :: Bag PredType -> [PredType] -> Bag PredType
2573 addAEsToBag bag [] = bag
2574 addAEsToBag bag (pred:preds)
2575 | pred `elemBag` bag = addAEsToBag bag preds
2576 | isEqPred pred = addAEsToBag bagWithPred preds
2577 | isClassPred pred = addAEsToBag bagWithPred predsWithSCs
2578 | otherwise = addAEsToBag bag preds
2580 bagWithPred = bag `snocBag` pred
2581 predsWithSCs = preds ++ substTheta (zipTopTvSubst tyvars tys) sc_theta
2583 (tyvars, sc_theta, _, _) = classBigSig clas
2584 (clas, tys) = getClassPredTys pred
2588 %************************************************************************
2590 \section{tcSimplifyTop: defaulting}
2592 %************************************************************************
2595 @tcSimplifyTop@ is called once per module to simplify all the constant
2596 and ambiguous Insts.
2598 We need to be careful of one case. Suppose we have
2600 instance Num a => Num (Foo a b) where ...
2602 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2603 to (Num x), and default x to Int. But what about y??
2605 It's OK: the final zonking stage should zap y to (), which is fine.
2609 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2610 tcSimplifyTop wanteds
2611 = tc_simplify_top doc False wanteds
2613 doc = text "tcSimplifyTop"
2615 tcSimplifyInteractive wanteds
2616 = tc_simplify_top doc True wanteds
2618 doc = text "tcSimplifyInteractive"
2620 -- The TcLclEnv should be valid here, solely to improve
2621 -- error message generation for the monomorphism restriction
2622 tc_simplify_top doc interactive wanteds
2623 = do { dflags <- getDOpts
2624 ; wanteds <- zonkInsts wanteds
2625 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2627 ; traceTc (text "tc_simplify_top 0: " <+> ppr wanteds)
2628 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2629 -- ; (irreds1, binds1) <- gentleInferLoop doc1 wanteds
2630 ; traceTc (text "tc_simplify_top 1: " <+> ppr irreds1)
2631 ; (irreds2, binds2) <- approximateImplications doc2 (\d -> True) irreds1
2632 ; traceTc (text "tc_simplify_top 2: " <+> ppr irreds2)
2634 -- Use the defaulting rules to do extra unification
2635 -- NB: irreds2 are already zonked
2636 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2638 -- Deal with implicit parameters
2639 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2640 (ambigs, others) = partition isTyVarDict non_ips
2642 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2644 ; addNoInstanceErrs others
2645 ; addTopAmbigErrs ambigs
2647 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2649 doc1 = doc <+> ptext SLIT("(first round)")
2650 doc2 = doc <+> ptext SLIT("(approximate)")
2651 doc3 = doc <+> ptext SLIT("(disambiguate)")
2654 If a dictionary constrains a type variable which is
2655 * not mentioned in the environment
2656 * and not mentioned in the type of the expression
2657 then it is ambiguous. No further information will arise to instantiate
2658 the type variable; nor will it be generalised and turned into an extra
2659 parameter to a function.
2661 It is an error for this to occur, except that Haskell provided for
2662 certain rules to be applied in the special case of numeric types.
2664 * at least one of its classes is a numeric class, and
2665 * all of its classes are numeric or standard
2666 then the type variable can be defaulted to the first type in the
2667 default-type list which is an instance of all the offending classes.
2669 So here is the function which does the work. It takes the ambiguous
2670 dictionaries and either resolves them (producing bindings) or
2671 complains. It works by splitting the dictionary list by type
2672 variable, and using @disambigOne@ to do the real business.
2674 @disambigOne@ assumes that its arguments dictionaries constrain all
2675 the same type variable.
2677 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2678 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2679 the most common use of defaulting is code like:
2681 _ccall_ foo `seqPrimIO` bar
2683 Since we're not using the result of @foo@, the result if (presumably)
2687 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2688 -- Just does unification to fix the default types
2689 -- The Insts are assumed to be pre-zonked
2690 disambiguate doc interactive dflags insts
2692 = return (insts, emptyBag)
2694 | null defaultable_groups
2695 = do { traceTc (text "disambigutate, no defaultable groups" <+> vcat [ppr unaries, ppr insts, ppr bad_tvs, ppr defaultable_groups])
2696 ; return (insts, emptyBag) }
2699 = do { -- Figure out what default types to use
2700 default_tys <- getDefaultTys extended_defaulting ovl_strings
2702 ; traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2703 ; mapM_ (disambigGroup default_tys) defaultable_groups
2705 -- disambigGroup does unification, hence try again
2706 ; tryHardCheckLoop doc insts }
2709 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2710 ovl_strings = dopt Opt_OverloadedStrings dflags
2712 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2713 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2714 (unaries, bad_tvs_s) = partitionWith find_unary insts
2715 bad_tvs = unionVarSets bad_tvs_s
2717 -- Finds unary type-class constraints
2718 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2719 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2720 find_unary inst = Right (tyVarsOfInst inst)
2722 -- Group by type variable
2723 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2724 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2725 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2727 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2728 defaultable_group ds@((_,_,tv):_)
2729 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2730 && not (tv `elemVarSet` bad_tvs)
2731 && defaultable_classes [c | (_,c,_) <- ds]
2732 defaultable_group [] = panic "defaultable_group"
2734 defaultable_classes clss
2735 | extended_defaulting = any isInteractiveClass clss
2736 | otherwise = all is_std_class clss && (any is_num_class clss)
2738 -- In interactive mode, or with -fextended-default-rules,
2739 -- we default Show a to Show () to avoid graututious errors on "show []"
2740 isInteractiveClass cls
2741 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2743 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2744 -- is_num_class adds IsString to the standard numeric classes,
2745 -- when -foverloaded-strings is enabled
2747 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2748 -- Similarly is_std_class
2750 -----------------------
2751 disambigGroup :: [Type] -- The default types
2752 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2753 -> TcM () -- Just does unification, to fix the default types
2755 disambigGroup default_tys dicts
2756 = try_default default_tys
2758 (_,_,tyvar) = ASSERT(not (null dicts)) head dicts -- Should be non-empty
2759 classes = [c | (_,c,_) <- dicts]
2761 try_default [] = return ()
2762 try_default (default_ty : default_tys)
2763 = tryTcLIE_ (try_default default_tys) $
2764 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2765 -- This may fail; then the tryTcLIE_ kicks in
2766 -- Failure here is caused by there being no type in the
2767 -- default list which can satisfy all the ambiguous classes.
2768 -- For example, if Real a is reqd, but the only type in the
2769 -- default list is Int.
2771 -- After this we can't fail
2772 ; warnDefault dicts default_ty
2773 ; unifyType default_ty (mkTyVarTy tyvar)
2774 ; return () -- TOMDO: do something with the coercion
2778 -----------------------
2779 getDefaultTys :: Bool -> Bool -> TcM [Type]
2780 getDefaultTys extended_deflts ovl_strings
2781 = do { mb_defaults <- getDeclaredDefaultTys
2782 ; case mb_defaults of {
2783 Just tys -> return tys ; -- User-supplied defaults
2786 -- No use-supplied default
2787 -- Use [Integer, Double], plus modifications
2788 { integer_ty <- tcMetaTy integerTyConName
2789 ; checkWiredInTyCon doubleTyCon
2790 ; string_ty <- tcMetaTy stringTyConName
2791 ; return (opt_deflt extended_deflts unitTy
2792 -- Note [Default unitTy]
2794 [integer_ty,doubleTy]
2796 opt_deflt ovl_strings string_ty) } } }
2798 opt_deflt True ty = [ty]
2799 opt_deflt False ty = []
2802 Note [Default unitTy]
2803 ~~~~~~~~~~~~~~~~~~~~~
2804 In interative mode (or with -fextended-default-rules) we add () as the first type we
2805 try when defaulting. This has very little real impact, except in the following case.
2807 Text.Printf.printf "hello"
2808 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2809 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2810 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2811 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2812 () to the list of defaulting types. See Trac #1200.
2814 Note [Avoiding spurious errors]
2815 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2816 When doing the unification for defaulting, we check for skolem
2817 type variables, and simply don't default them. For example:
2818 f = (*) -- Monomorphic
2819 g :: Num a => a -> a
2821 Here, we get a complaint when checking the type signature for g,
2822 that g isn't polymorphic enough; but then we get another one when
2823 dealing with the (Num a) context arising from f's definition;
2824 we try to unify a with Int (to default it), but find that it's
2825 already been unified with the rigid variable from g's type sig
2828 %************************************************************************
2830 \subsection[simple]{@Simple@ versions}
2832 %************************************************************************
2834 Much simpler versions when there are no bindings to make!
2836 @tcSimplifyThetas@ simplifies class-type constraints formed by
2837 @deriving@ declarations and when specialising instances. We are
2838 only interested in the simplified bunch of class/type constraints.
2840 It simplifies to constraints of the form (C a b c) where
2841 a,b,c are type variables. This is required for the context of
2842 instance declarations.
2845 tcSimplifyDeriv :: InstOrigin
2847 -> ThetaType -- Wanted
2848 -> TcM ThetaType -- Needed
2849 -- Given instance (wanted) => C inst_ty
2850 -- Simplify 'wanted' as much as possible
2852 tcSimplifyDeriv orig tyvars theta
2853 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2854 -- The main loop may do unification, and that may crash if
2855 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2856 -- ToDo: what if two of them do get unified?
2857 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2858 ; (irreds, _) <- tryHardCheckLoop doc wanteds
2860 ; let (tv_dicts, others) = partition ok irreds
2861 ; addNoInstanceErrs others
2862 -- See Note [Exotic derived instance contexts] in TcMType
2864 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2865 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
2866 -- This reverse-mapping is a pain, but the result
2867 -- should mention the original TyVars not TcTyVars
2869 ; return simpl_theta }
2871 doc = ptext SLIT("deriving classes for a data type")
2873 ok dict | isDict dict = validDerivPred (dictPred dict)
2878 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2879 used with \tr{default} declarations. We are only interested in
2880 whether it worked or not.
2883 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2886 tcSimplifyDefault theta
2887 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2888 tryHardCheckLoop doc wanteds `thenM` \ (irreds, _) ->
2889 addNoInstanceErrs irreds `thenM_`
2895 doc = ptext SLIT("default declaration")
2899 %************************************************************************
2901 \section{Errors and contexts}
2903 %************************************************************************
2905 ToDo: for these error messages, should we note the location as coming
2906 from the insts, or just whatever seems to be around in the monad just
2910 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2911 -> [Inst] -- The offending Insts
2913 -- Group together insts with the same origin
2914 -- We want to report them together in error messages
2916 groupErrs report_err []
2918 groupErrs report_err (inst:insts)
2919 = do_one (inst:friends) `thenM_`
2920 groupErrs report_err others
2923 -- (It may seem a bit crude to compare the error messages,
2924 -- but it makes sure that we combine just what the user sees,
2925 -- and it avoids need equality on InstLocs.)
2926 (friends, others) = partition is_friend insts
2927 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2928 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2929 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2930 -- Add location and context information derived from the Insts
2932 -- Add the "arising from..." part to a message about bunch of dicts
2933 addInstLoc :: [Inst] -> Message -> Message
2934 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2936 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2937 addTopIPErrs bndrs []
2939 addTopIPErrs bndrs ips
2940 = do { dflags <- getDOpts
2941 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
2943 (tidy_env, tidy_ips) = tidyInsts ips
2945 = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2946 nest 2 (ptext SLIT("the monomorphic top-level binding")
2947 <> plural bndrs <+> ptext SLIT("of")
2948 <+> pprBinders bndrs <> colon)],
2949 nest 2 (vcat (map ppr_ip ips)),
2950 monomorphism_fix dflags]
2951 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2953 topIPErrs :: [Inst] -> TcM ()
2955 = groupErrs report tidy_dicts
2957 (tidy_env, tidy_dicts) = tidyInsts dicts
2958 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2959 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2960 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2962 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2964 addNoInstanceErrs insts
2965 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2966 ; reportNoInstances tidy_env Nothing tidy_insts }
2970 -> Maybe (InstLoc, [Inst]) -- Context
2971 -- Nothing => top level
2972 -- Just (d,g) => d describes the construct
2974 -> [Inst] -- What is wanted (can include implications)
2977 reportNoInstances tidy_env mb_what insts
2978 = groupErrs (report_no_instances tidy_env mb_what) insts
2980 report_no_instances tidy_env mb_what insts
2981 = do { inst_envs <- tcGetInstEnvs
2982 ; let (implics, insts1) = partition isImplicInst insts
2983 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2984 (eqInsts, insts3) = partition isEqInst insts2
2985 ; traceTc (text "reportNoInstances" <+> vcat
2986 [ppr implics, ppr insts1, ppr insts2])
2987 ; mapM_ complain_implic implics
2988 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2989 ; groupErrs complain_no_inst insts3
2990 ; mapM_ eqInstMisMatch eqInsts
2993 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2995 complain_implic inst -- Recurse!
2996 = reportNoInstances tidy_env
2997 (Just (tci_loc inst, tci_given inst))
3000 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
3001 -- Right msg => overlap message
3002 -- Left inst => no instance
3003 check_overlap inst_envs wanted
3004 | not (isClassDict wanted) = Left wanted
3006 = case lookupInstEnv inst_envs clas tys of
3007 -- The case of exactly one match and no unifiers means a
3008 -- successful lookup. That can't happen here, because dicts
3009 -- only end up here if they didn't match in Inst.lookupInst
3011 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
3013 ([], _) -> Left wanted -- No match
3014 res -> Right (mk_overlap_msg wanted res)
3016 (clas,tys) = getDictClassTys wanted
3018 mk_overlap_msg dict (matches, unifiers)
3019 = ASSERT( not (null matches) )
3020 vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
3021 <+> pprPred (dictPred dict))),
3022 sep [ptext SLIT("Matching instances") <> colon,
3023 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
3024 if not (isSingleton matches)
3025 then -- Two or more matches
3027 else -- One match, plus some unifiers
3028 ASSERT( not (null unifiers) )
3029 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
3030 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
3031 ptext SLIT("To pick the first instance above, use -fallow-incoherent-instances"),
3032 ptext SLIT("when compiling the other instance declarations")])]
3034 ispecs = [ispec | (ispec, _) <- matches]
3036 mk_no_inst_err insts
3037 | null insts = empty
3039 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
3040 not (isEmptyVarSet (tyVarsOfInsts insts))
3041 = vcat [ addInstLoc insts $
3042 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
3043 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
3044 , show_fixes (fix1 loc : fixes2) ]
3046 | otherwise -- Top level
3047 = vcat [ addInstLoc insts $
3048 ptext SLIT("No instance") <> plural insts
3049 <+> ptext SLIT("for") <+> pprDictsTheta insts
3050 , show_fixes fixes2 ]
3053 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
3054 <+> ptext SLIT("to the context of"),
3055 nest 2 (ppr (instLocOrigin loc)) ]
3056 -- I'm not sure it helps to add the location
3057 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
3059 fixes2 | null instance_dicts = []
3060 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
3061 pprDictsTheta instance_dicts]]
3062 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
3063 -- Insts for which it is worth suggesting an adding an instance declaration
3064 -- Exclude implicit parameters, and tyvar dicts
3066 show_fixes :: [SDoc] -> SDoc
3067 show_fixes [] = empty
3068 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
3069 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
3071 addTopAmbigErrs dicts
3072 -- Divide into groups that share a common set of ambiguous tyvars
3073 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
3074 -- See Note [Avoiding spurious errors]
3075 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
3077 (tidy_env, tidy_dicts) = tidyInsts dicts
3079 tvs_of :: Inst -> [TcTyVar]
3080 tvs_of d = varSetElems (tyVarsOfInst d)
3081 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
3083 report :: [(Inst,[TcTyVar])] -> TcM ()
3084 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
3085 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
3086 setSrcSpan (instSpan inst) $
3087 -- the location of the first one will do for the err message
3088 addErrTcM (tidy_env, msg $$ mono_msg)
3090 dicts = map fst pairs
3091 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
3092 pprQuotedList tvs <+> in_msg,
3093 nest 2 (pprDictsInFull dicts)]
3094 in_msg = text "in the constraint" <> plural dicts <> colon
3095 report [] = panic "addTopAmbigErrs"
3098 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
3099 -- There's an error with these Insts; if they have free type variables
3100 -- it's probably caused by the monomorphism restriction.
3101 -- Try to identify the offending variable
3102 -- ASSUMPTION: the Insts are fully zonked
3103 mkMonomorphismMsg tidy_env inst_tvs
3104 = do { dflags <- getDOpts
3105 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
3106 ; return (tidy_env, mk_msg dflags docs) }
3108 mk_msg _ _ | any isRuntimeUnk inst_tvs
3109 = vcat [ptext SLIT("Cannot resolve unknown runtime types:") <+>
3110 (pprWithCommas ppr inst_tvs),
3111 ptext SLIT("Use :print or :force to determine these types")]
3112 mk_msg _ [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
3113 -- This happens in things like
3114 -- f x = show (read "foo")
3115 -- where monomorphism doesn't play any role
3117 = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
3119 monomorphism_fix dflags]
3121 isRuntimeUnk :: TcTyVar -> Bool
3122 isRuntimeUnk x | SkolemTv RuntimeUnkSkol <- tcTyVarDetails x = True
3125 monomorphism_fix :: DynFlags -> SDoc
3126 monomorphism_fix dflags
3127 = ptext SLIT("Probable fix:") <+> vcat
3128 [ptext SLIT("give these definition(s) an explicit type signature"),
3129 if dopt Opt_MonomorphismRestriction dflags
3130 then ptext SLIT("or use -fno-monomorphism-restriction")
3131 else empty] -- Only suggest adding "-fno-monomorphism-restriction"
3132 -- if it is not already set!
3134 warnDefault ups default_ty
3135 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
3136 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
3138 dicts = [d | (d,_,_) <- ups]
3141 (_, tidy_dicts) = tidyInsts dicts
3142 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
3143 quotes (ppr default_ty),
3144 pprDictsInFull tidy_dicts]
3146 reduceDepthErr n stack
3147 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
3148 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
3149 nest 4 (pprStack stack)]
3151 pprStack stack = vcat (map pprInstInFull stack)