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 What type should we infer for this?
415 f x = (show ?y, x::Int)
416 Since we must quantify over the ?y, the most plausible type is
417 f :: (Show a, ?y::a) => Int -> (String, Int)
418 But notice that the type of the RHS is (String,Int), with no type
419 varibables mentioned at all! The type of f looks ambiguous. But
420 it isn't, because at a call site we might have
421 let ?y = 5::Int in f 7
422 and all is well. In effect, implicit parameters are, well, parameters,
423 so we can take their type variables into account as part of the
424 "tau-tvs" stuff. This is done in the function 'FunDeps.grow'.
427 Question 2: type signatures
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
429 BUT WATCH OUT: When you supply a type signature, we can't force you
430 to quantify over implicit parameters. For example:
434 This is perfectly reasonable. We do not want to insist on
436 (?x + 1) :: (?x::Int => Int)
438 That would be silly. Here, the definition site *is* the occurrence site,
439 so the above strictures don't apply. Hence the difference between
440 tcSimplifyCheck (which *does* allow implicit paramters to be inherited)
441 and tcSimplifyCheckBind (which does not).
443 What about when you supply a type signature for a binding?
444 Is it legal to give the following explicit, user type
445 signature to f, thus:
450 At first sight this seems reasonable, but it has the nasty property
451 that adding a type signature changes the dynamic semantics.
454 (let f x = (x::Int) + ?y
455 in (f 3, f 3 with ?y=5)) with ?y = 6
461 in (f 3, f 3 with ?y=5)) with ?y = 6
465 Indeed, simply inlining f (at the Haskell source level) would change the
468 Nevertheless, as Launchbury says (email Oct 01) we can't really give the
469 semantics for a Haskell program without knowing its typing, so if you
470 change the typing you may change the semantics.
472 To make things consistent in all cases where we are *checking* against
473 a supplied signature (as opposed to inferring a type), we adopt the
476 a signature does not need to quantify over implicit params.
478 [This represents a (rather marginal) change of policy since GHC 5.02,
479 which *required* an explicit signature to quantify over all implicit
480 params for the reasons mentioned above.]
482 But that raises a new question. Consider
484 Given (signature) ?x::Int
485 Wanted (inferred) ?x::Int, ?y::Bool
487 Clearly we want to discharge the ?x and float the ?y out. But
488 what is the criterion that distinguishes them? Clearly it isn't
489 what free type variables they have. The Right Thing seems to be
490 to float a constraint that
491 neither mentions any of the quantified type variables
492 nor any of the quantified implicit parameters
494 See the predicate isFreeWhenChecking.
497 Question 3: monomorphism
498 ~~~~~~~~~~~~~~~~~~~~~~~~
499 There's a nasty corner case when the monomorphism restriction bites:
503 The argument above suggests that we *must* generalise
504 over the ?y parameter, to get
505 z :: (?y::Int) => Int,
506 but the monomorphism restriction says that we *must not*, giving
508 Why does the momomorphism restriction say this? Because if you have
510 let z = x + ?y in z+z
512 you might not expect the addition to be done twice --- but it will if
513 we follow the argument of Question 2 and generalise over ?y.
516 Question 4: top level
517 ~~~~~~~~~~~~~~~~~~~~~
518 At the top level, monomorhism makes no sense at all.
521 main = let ?x = 5 in print foo
525 woggle :: (?x :: Int) => Int -> Int
528 We definitely don't want (foo :: Int) with a top-level implicit parameter
529 (?x::Int) becuase there is no way to bind it.
534 (A) Always generalise over implicit parameters
535 Bindings that fall under the monomorphism restriction can't
539 * Inlining remains valid
540 * No unexpected loss of sharing
541 * But simple bindings like
543 will be rejected, unless you add an explicit type signature
544 (to avoid the monomorphism restriction)
545 z :: (?y::Int) => Int
547 This seems unacceptable
549 (B) Monomorphism restriction "wins"
550 Bindings that fall under the monomorphism restriction can't
552 Always generalise over implicit parameters *except* for bindings
553 that fall under the monomorphism restriction
556 * Inlining isn't valid in general
557 * No unexpected loss of sharing
558 * Simple bindings like
560 accepted (get value of ?y from binding site)
562 (C) Always generalise over implicit parameters
563 Bindings that fall under the monomorphism restriction can't
564 be generalised, EXCEPT for implicit parameters
566 * Inlining remains valid
567 * Unexpected loss of sharing (from the extra generalisation)
568 * Simple bindings like
570 accepted (get value of ?y from occurrence sites)
575 None of these choices seems very satisfactory. But at least we should
576 decide which we want to do.
578 It's really not clear what is the Right Thing To Do. If you see
582 would you expect the value of ?y to be got from the *occurrence sites*
583 of 'z', or from the valuue of ?y at the *definition* of 'z'? In the
584 case of function definitions, the answer is clearly the former, but
585 less so in the case of non-fucntion definitions. On the other hand,
586 if we say that we get the value of ?y from the definition site of 'z',
587 then inlining 'z' might change the semantics of the program.
589 Choice (C) really says "the monomorphism restriction doesn't apply
590 to implicit parameters". Which is fine, but remember that every
591 innocent binding 'x = ...' that mentions an implicit parameter in
592 the RHS becomes a *function* of that parameter, called at each
593 use of 'x'. Now, the chances are that there are no intervening 'with'
594 clauses that bind ?y, so a decent compiler should common up all
595 those function calls. So I think I strongly favour (C). Indeed,
596 one could make a similar argument for abolishing the monomorphism
597 restriction altogether.
599 BOTTOM LINE: we choose (B) at present. See tcSimplifyRestricted
603 %************************************************************************
605 \subsection{tcSimplifyInfer}
607 %************************************************************************
609 tcSimplify is called when we *inferring* a type. Here's the overall game plan:
611 1. Compute Q = grow( fvs(T), C )
613 2. Partition C based on Q into Ct and Cq. Notice that ambiguous
614 predicates will end up in Ct; we deal with them at the top level
616 3. Try improvement, using functional dependencies
618 4. If Step 3 did any unification, repeat from step 1
619 (Unification can change the result of 'grow'.)
621 Note: we don't reduce dictionaries in step 2. For example, if we have
622 Eq (a,b), we don't simplify to (Eq a, Eq b). So Q won't be different
623 after step 2. However note that we may therefore quantify over more
624 type variables than we absolutely have to.
626 For the guts, we need a loop, that alternates context reduction and
627 improvement with unification. E.g. Suppose we have
629 class C x y | x->y where ...
631 and tcSimplify is called with:
633 Then improvement unifies a with b, giving
636 If we need to unify anything, we rattle round the whole thing all over
643 -> TcTyVarSet -- fv(T); type vars
645 -> TcM ([TcTyVar], -- Tyvars to quantify (zonked and quantified)
646 [Inst], -- Dict Ids that must be bound here (zonked)
647 TcDictBinds) -- Bindings
648 -- Any free (escaping) Insts are tossed into the environment
653 tcSimplifyInfer doc tau_tvs wanted
654 = do { tau_tvs1 <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
655 ; wanted' <- mappM zonkInst wanted -- Zonk before deciding quantified tyvars
656 ; gbl_tvs <- tcGetGlobalTyVars
657 ; let preds1 = fdPredsOfInsts wanted'
658 gbl_tvs1 = oclose preds1 gbl_tvs
659 qtvs = grow preds1 tau_tvs1 `minusVarSet` gbl_tvs1
660 -- See Note [Choosing which variables to quantify]
662 -- To maximise sharing, remove from consideration any
663 -- constraints that don't mention qtvs at all
664 ; let (free, bound) = partition (isFreeWhenInferring qtvs) wanted'
667 -- To make types simple, reduce as much as possible
668 ; traceTc (text "infer" <+> (ppr preds1 $$ ppr (grow preds1 tau_tvs1) $$ ppr gbl_tvs $$
669 ppr gbl_tvs1 $$ ppr free $$ ppr bound))
670 ; (irreds1, binds1) <- tryHardCheckLoop doc bound
672 -- Note [Inference and implication constraints]
673 ; let want_dict d = tyVarsOfInst d `intersectsVarSet` qtvs
674 ; (irreds2, binds2) <- approximateImplications doc want_dict irreds1
676 -- Now work out all over again which type variables to quantify,
677 -- exactly in the same way as before, but starting from irreds2. Why?
678 -- a) By now improvment may have taken place, and we must *not*
679 -- quantify over any variable free in the environment
680 -- tc137 (function h inside g) is an example
682 -- b) Do not quantify over constraints that *now* do not
683 -- mention quantified type variables, because they are
684 -- simply ambiguous (or might be bound further out). Example:
685 -- f :: Eq b => a -> (a, b)
687 -- From the RHS of g we get the MethodInst f77 :: alpha -> (alpha, beta)
688 -- We decide to quantify over 'alpha' alone, but free1 does not include f77
689 -- because f77 mentions 'alpha'. Then reducing leaves only the (ambiguous)
690 -- constraint (Eq beta), which we dump back into the free set
691 -- See test tcfail181
693 -- c) irreds may contain type variables not previously mentioned,
694 -- e.g. instance D a x => Foo [a]
696 -- Then after simplifying we'll get (D a x), and x is fresh
697 -- We must quantify over x else it'll be totally unbound
698 ; tau_tvs2 <- zonkTcTyVarsAndFV (varSetElems tau_tvs1)
699 ; gbl_tvs2 <- zonkTcTyVarsAndFV (varSetElems gbl_tvs1)
700 -- Note that we start from gbl_tvs1
701 -- We use tcGetGlobalTyVars, then oclose wrt preds2, because
702 -- we've already put some of the original preds1 into frees
703 -- E.g. wanteds = C a b (where a->b)
706 -- Then b is fixed by gbl_tvs, so (C a b) will be in free, and
707 -- irreds2 will be empty. But we don't want to generalise over b!
708 ; let preds2 = fdPredsOfInsts irreds2 -- irreds2 is zonked
709 qtvs = grow preds2 tau_tvs2 `minusVarSet` oclose preds2 gbl_tvs2
710 ; let (free, irreds3) = partition (isFreeWhenInferring qtvs) irreds2
713 -- Turn the quantified meta-type variables into real type variables
714 ; qtvs2 <- zonkQuantifiedTyVars (varSetElems qtvs)
716 -- We can't abstract over any remaining unsolved
717 -- implications so instead just float them outwards. Ugh.
718 ; let (q_dicts0, implics) = partition isAbstractableInst irreds3
719 ; loc <- getInstLoc (ImplicOrigin doc)
720 ; implic_bind <- bindIrreds loc qtvs2 q_dicts0 implics
722 -- Prepare equality instances for quantification
723 ; let (q_eqs0,q_dicts) = partition isEqInst q_dicts0
724 ; q_eqs <- mappM finalizeEqInst q_eqs0
726 ; return (qtvs2, q_eqs ++ q_dicts, binds1 `unionBags` binds2 `unionBags` implic_bind) }
727 -- NB: when we are done, we might have some bindings, but
728 -- the final qtvs might be empty. See Note [NO TYVARS] below.
730 approximateImplications :: SDoc -> (Inst -> Bool) -> [Inst] -> TcM ([Inst], TcDictBinds)
731 -- Note [Inference and implication constraints]
732 -- Given a bunch of Dict and ImplicInsts, try to approximate the implications by
733 -- - fetching any dicts inside them that are free
734 -- - using those dicts as cruder constraints, to solve the implications
735 -- - returning the extra ones too
737 approximateImplications doc want_dict irreds
739 = return (irreds, emptyBag)
741 = do { extra_dicts' <- mapM cloneDict extra_dicts
742 ; tryHardCheckLoop doc (extra_dicts' ++ irreds) }
743 -- By adding extra_dicts', we make them
744 -- available to solve the implication constraints
746 extra_dicts = get_dicts (filter isImplicInst irreds)
748 get_dicts :: [Inst] -> [Inst] -- Returns only Dicts
749 -- Find the wanted constraints in implication constraints that satisfy
750 -- want_dict, and are not bound by forall's in the constraint itself
751 get_dicts ds = concatMap get_dict ds
753 get_dict d@(Dict {}) | want_dict d = [d]
755 get_dict (ImplicInst {tci_tyvars = tvs, tci_wanted = wanteds})
756 = [ d | let tv_set = mkVarSet tvs
757 , d <- get_dicts wanteds
758 , not (tyVarsOfInst d `intersectsVarSet` tv_set)]
759 get_dict i@(EqInst {}) | want_dict i = [i]
761 get_dict other = pprPanic "approximateImplications" (ppr other)
764 Note [Inference and implication constraints]
765 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
766 Suppose we have a wanted implication constraint (perhaps arising from
767 a nested pattern match) like
769 and we are now trying to quantify over 'a' when inferring the type for
770 a function. In principle it's possible that there might be an instance
771 instance (C a, E a) => D [a]
772 so the context (E a) would suffice. The Right Thing is to abstract over
773 the implication constraint, but we don't do that (a) because it'll be
774 surprising to programmers and (b) because we don't have the machinery to deal
775 with 'given' implications.
777 So our best approximation is to make (D [a]) part of the inferred
778 context, so we can use that to discharge the implication. Hence
779 the strange function get_dictsin approximateImplications.
781 The common cases are more clear-cut, when we have things like
783 Here, abstracting over (C b) is not an approximation at all -- but see
784 Note [Freeness and implications].
786 See Trac #1430 and test tc228.
790 -----------------------------------------------------------
791 -- tcSimplifyInferCheck is used when we know the constraints we are to simplify
792 -- against, but we don't know the type variables over which we are going to quantify.
793 -- This happens when we have a type signature for a mutually recursive group
796 -> TcTyVarSet -- fv(T)
799 -> TcM ([TyVar], -- Fully zonked, and quantified
800 TcDictBinds) -- Bindings
802 tcSimplifyInferCheck loc tau_tvs givens wanteds
803 = do { traceTc (text "tcSimplifyInferCheck <-" <+> ppr wanteds)
804 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
806 -- Figure out which type variables to quantify over
807 -- You might think it should just be the signature tyvars,
808 -- but in bizarre cases you can get extra ones
809 -- f :: forall a. Num a => a -> a
810 -- f x = fst (g (x, head [])) + 1
812 -- Here we infer g :: forall a b. a -> b -> (b,a)
813 -- We don't want g to be monomorphic in b just because
814 -- f isn't quantified over b.
815 ; let all_tvs = varSetElems (tau_tvs `unionVarSet` tyVarsOfInsts givens)
816 ; all_tvs <- zonkTcTyVarsAndFV all_tvs
817 ; gbl_tvs <- tcGetGlobalTyVars
818 ; let qtvs = varSetElems (all_tvs `minusVarSet` gbl_tvs)
819 -- We could close gbl_tvs, but its not necessary for
820 -- soundness, and it'll only affect which tyvars, not which
821 -- dictionaries, we quantify over
823 ; qtvs' <- zonkQuantifiedTyVars qtvs
825 -- Now we are back to normal (c.f. tcSimplCheck)
826 ; implic_bind <- bindIrreds loc qtvs' givens irreds
828 ; traceTc (text "tcSimplifyInferCheck ->" <+> ppr (implic_bind))
829 ; return (qtvs', binds `unionBags` implic_bind) }
832 Note [Squashing methods]
833 ~~~~~~~~~~~~~~~~~~~~~~~~~
834 Be careful if you want to float methods more:
835 truncate :: forall a. RealFrac a => forall b. Integral b => a -> b
836 From an application (truncate f i) we get
839 If we have also have a second occurrence of truncate, we get
842 When simplifying with i,f free, we might still notice that
843 t1=t3; but alas, the binding for t2 (which mentions t1)
844 may continue to float out!
849 class Y a b | a -> b where
852 instance Y [[a]] a where
855 k :: X a -> X a -> X a
857 g :: Num a => [X a] -> [X a]
860 h ys = ys ++ map (k (y [[0]])) xs
862 The excitement comes when simplifying the bindings for h. Initially
863 try to simplify {y @ [[t1]] t2, 0 @ t1}, with initial qtvs = {t2}.
864 From this we get t1:=:t2, but also various bindings. We can't forget
865 the bindings (because of [LOOP]), but in fact t1 is what g is
868 The net effect of [NO TYVARS]
871 isFreeWhenInferring :: TyVarSet -> Inst -> Bool
872 isFreeWhenInferring qtvs inst
873 = isFreeWrtTyVars qtvs inst -- Constrains no quantified vars
874 && isInheritableInst inst -- and no implicit parameter involved
875 -- see Note [Inheriting implicit parameters]
877 {- No longer used (with implication constraints)
878 isFreeWhenChecking :: TyVarSet -- Quantified tyvars
879 -> NameSet -- Quantified implicit parameters
881 isFreeWhenChecking qtvs ips inst
882 = isFreeWrtTyVars qtvs inst
883 && isFreeWrtIPs ips inst
886 isFreeWrtTyVars qtvs inst = tyVarsOfInst inst `disjointVarSet` qtvs
887 isFreeWrtIPs ips inst = not (any (`elemNameSet` ips) (ipNamesOfInst inst))
891 %************************************************************************
893 \subsection{tcSimplifyCheck}
895 %************************************************************************
897 @tcSimplifyCheck@ is used when we know exactly the set of variables
898 we are going to quantify over. For example, a class or instance declaration.
901 -----------------------------------------------------------
902 -- tcSimplifyCheck is used when checking expression type signatures,
903 -- class decls, instance decls etc.
904 tcSimplifyCheck :: InstLoc
905 -> [TcTyVar] -- Quantify over these
908 -> TcM TcDictBinds -- Bindings
909 tcSimplifyCheck loc qtvs givens wanteds
910 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
911 do { traceTc (text "tcSimplifyCheck")
912 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
913 ; implic_bind <- bindIrreds loc qtvs givens irreds
914 ; return (binds `unionBags` implic_bind) }
916 -----------------------------------------------------------
917 -- tcSimplifyCheckPat is used for existential pattern match
918 tcSimplifyCheckPat :: InstLoc
919 -> [CoVar] -> Refinement
920 -> [TcTyVar] -- Quantify over these
923 -> TcM TcDictBinds -- Bindings
924 tcSimplifyCheckPat loc co_vars reft qtvs givens wanteds
925 = ASSERT( all isTcTyVar qtvs && all isSkolemTyVar qtvs )
926 do { traceTc (text "tcSimplifyCheckPat")
927 ; (irreds, binds) <- gentleCheckLoop loc givens wanteds
928 ; implic_bind <- bindIrredsR loc qtvs co_vars reft
930 ; return (binds `unionBags` implic_bind) }
932 -----------------------------------------------------------
933 bindIrreds :: InstLoc -> [TcTyVar]
936 bindIrreds loc qtvs givens irreds
937 = bindIrredsR loc qtvs [] emptyRefinement givens irreds
939 bindIrredsR :: InstLoc -> [TcTyVar] -> [CoVar]
940 -> Refinement -> [Inst] -> [Inst]
942 -- Make a binding that binds 'irreds', by generating an implication
943 -- constraint for them, *and* throwing the constraint into the LIE
944 bindIrredsR loc qtvs co_vars reft givens irreds
948 = do { let givens' = filter isDict givens
949 -- The givens can include methods
950 -- See Note [Pruning the givens in an implication constraint]
952 -- If there are no 'givens' *and* the refinement is empty
953 -- (the refinement is like more givens), then it's safe to
954 -- partition the 'wanteds' by their qtvs, thereby trimming irreds
955 -- See Note [Freeness and implications]
956 ; irreds' <- if null givens' && isEmptyRefinement reft
958 { let qtv_set = mkVarSet qtvs
959 (frees, real_irreds) = partition (isFreeWrtTyVars qtv_set) irreds
961 ; return real_irreds }
964 ; let all_tvs = qtvs ++ co_vars -- Abstract over all these
965 ; (implics, bind) <- makeImplicationBind loc all_tvs reft givens' irreds'
966 -- This call does the real work
967 -- If irreds' is empty, it does something sensible
972 makeImplicationBind :: InstLoc -> [TcTyVar] -> Refinement
974 -> TcM ([Inst], TcDictBinds)
975 -- Make a binding that binds 'irreds', by generating an implication
976 -- constraint for them, *and* throwing the constraint into the LIE
977 -- The binding looks like
978 -- (ir1, .., irn) = f qtvs givens
979 -- where f is (evidence for) the new implication constraint
980 -- f :: forall qtvs. {reft} givens => (ir1, .., irn)
981 -- qtvs includes coercion variables
983 -- This binding must line up the 'rhs' in reduceImplication
984 makeImplicationBind loc all_tvs reft
985 givens -- Guaranteed all Dicts (TOMDO: true?)
987 | null irreds -- If there are no irreds, we are done
988 = return ([], emptyBag)
989 | otherwise -- Otherwise we must generate a binding
990 = do { uniq <- newUnique
991 ; span <- getSrcSpanM
992 ; let (eq_givens,dict_givens) = partitionGivenEqInsts givens
993 eq_tyvar_cos = map TyVarTy $ uniqSetToList $ tyVarsOfTypes $ map eqInstType eq_givens
994 ; let name = mkInternalName uniq (mkVarOcc "ic") span
995 implic_inst = ImplicInst { tci_name = name, tci_reft = reft,
996 tci_tyvars = all_tvs,
997 tci_given = (eq_givens ++ dict_givens),
998 tci_wanted = irreds, tci_loc = loc }
1000 -- only create binder for dict_irreds
1002 (eq_irreds,dict_irreds) = partitionWantedEqInsts irreds
1003 n_dict_irreds = length dict_irreds
1004 dict_irred_ids = map instToId dict_irreds
1005 tup_ty = mkTupleTy Boxed n_dict_irreds (map idType dict_irred_ids)
1006 pat = TuplePat (map nlVarPat dict_irred_ids) Boxed tup_ty
1007 rhs = L span (mkHsWrap co (HsVar (instToId implic_inst)))
1008 co = mkWpApps (map instToId dict_givens) <.> mkWpTyApps eq_tyvar_cos <.> mkWpTyApps (mkTyVarTys all_tvs)
1009 bind | [dict_irred_id] <- dict_irred_ids = VarBind dict_irred_id rhs
1010 | otherwise = PatBind { pat_lhs = L span pat,
1011 pat_rhs = unguardedGRHSs rhs,
1012 pat_rhs_ty = tup_ty,
1013 bind_fvs = placeHolderNames }
1014 ; -- pprTrace "Make implic inst" (ppr (implic_inst,irreds,dict_irreds,tup_ty)) $
1015 return ([implic_inst], unitBag (L span bind)) }
1017 -----------------------------------------------------------
1018 tryHardCheckLoop :: SDoc
1020 -> TcM ([Inst], TcDictBinds)
1022 tryHardCheckLoop doc wanteds
1023 = do { (irreds,binds,_) <- checkLoop (mkRedEnv doc try_me []) wanteds
1024 ; return (irreds,binds)
1027 try_me inst = ReduceMe AddSCs
1028 -- Here's the try-hard bit
1030 -----------------------------------------------------------
1031 gentleCheckLoop :: InstLoc
1034 -> TcM ([Inst], TcDictBinds)
1036 gentleCheckLoop inst_loc givens wanteds
1037 = do { (irreds,binds,_) <- checkLoop env wanteds
1038 ; return (irreds,binds)
1041 env = mkRedEnv (pprInstLoc inst_loc) try_me givens
1043 try_me inst | isMethodOrLit inst = ReduceMe AddSCs
1045 -- When checking against a given signature
1046 -- we MUST be very gentle: Note [Check gently]
1050 ~~~~~~~~~~~~~~~~~~~~
1051 We have to very careful about not simplifying too vigorously
1056 f :: Show b => T b -> b
1057 f (MkT x) = show [x]
1059 Inside the pattern match, which binds (a:*, x:a), we know that
1061 Hence we have a dictionary for Show [a] available; and indeed we
1062 need it. We are going to build an implication contraint
1063 forall a. (b~[a]) => Show [a]
1064 Later, we will solve this constraint using the knowledg e(Show b)
1066 But we MUST NOT reduce (Show [a]) to (Show a), else the whole
1067 thing becomes insoluble. So we simplify gently (get rid of literals
1068 and methods only, plus common up equal things), deferring the real
1069 work until top level, when we solve the implication constraint
1070 with tryHardCheckLooop.
1074 -----------------------------------------------------------
1077 -> TcM ([Inst], TcDictBinds,
1078 [Inst]) -- needed givens
1079 -- Precondition: givens are completely rigid
1080 -- Postcondition: returned Insts are zonked
1082 checkLoop env wanteds
1084 where go env wanteds needed_givens
1085 = do { -- Givens are skolems, so no need to zonk them
1086 wanteds' <- zonkInsts wanteds
1088 ; (improved, binds, irreds, more_needed_givens) <- reduceContext env wanteds'
1090 ; let all_needed_givens = needed_givens ++ more_needed_givens
1092 ; if not improved then
1093 return (irreds, binds, all_needed_givens)
1096 -- If improvement did some unification, we go round again.
1097 -- We start again with irreds, not wanteds
1098 -- Using an instance decl might have introduced a fresh type variable
1099 -- which might have been unified, so we'd get an infinite loop
1100 -- if we started again with wanteds! See Note [LOOP]
1101 { (irreds1, binds1, all_needed_givens1) <- go env irreds all_needed_givens
1102 ; return (irreds1, binds `unionBags` binds1, all_needed_givens1) } }
1107 class If b t e r | b t e -> r
1110 class Lte a b c | a b -> c where lte :: a -> b -> c
1112 instance (Lte a b l,If l b a c) => Max a b c
1114 Wanted: Max Z (S x) y
1116 Then we'll reduce using the Max instance to:
1117 (Lte Z (S x) l, If l (S x) Z y)
1118 and improve by binding l->T, after which we can do some reduction
1119 on both the Lte and If constraints. What we *can't* do is start again
1120 with (Max Z (S x) y)!
1124 %************************************************************************
1126 tcSimplifySuperClasses
1128 %************************************************************************
1130 Note [SUPERCLASS-LOOP 1]
1131 ~~~~~~~~~~~~~~~~~~~~~~~~
1132 We have to be very, very careful when generating superclasses, lest we
1133 accidentally build a loop. Here's an example:
1137 class S a => C a where { opc :: a -> a }
1138 class S b => D b where { opd :: b -> b }
1140 instance C Int where
1143 instance D Int where
1146 From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
1147 Simplifying, we may well get:
1148 $dfCInt = :C ds1 (opd dd)
1151 Notice that we spot that we can extract ds1 from dd.
1153 Alas! Alack! We can do the same for (instance D Int):
1155 $dfDInt = :D ds2 (opc dc)
1159 And now we've defined the superclass in terms of itself.
1161 Solution: never generate a superclass selectors at all when
1162 satisfying the superclass context of an instance declaration.
1164 Two more nasty cases are in
1169 tcSimplifySuperClasses
1174 tcSimplifySuperClasses loc givens sc_wanteds
1175 = do { traceTc (text "tcSimplifySuperClasses")
1176 ; (irreds,binds1,_) <- checkLoop env sc_wanteds
1177 ; let (tidy_env, tidy_irreds) = tidyInsts irreds
1178 ; reportNoInstances tidy_env (Just (loc, givens)) tidy_irreds
1181 env = mkRedEnv (pprInstLoc loc) try_me givens
1182 try_me inst = ReduceMe NoSCs
1183 -- Like tryHardCheckLoop, but with NoSCs
1187 %************************************************************************
1189 \subsection{tcSimplifyRestricted}
1191 %************************************************************************
1193 tcSimplifyRestricted infers which type variables to quantify for a
1194 group of restricted bindings. This isn't trivial.
1197 We want to quantify over a to get id :: forall a. a->a
1200 We do not want to quantify over a, because there's an Eq a
1201 constraint, so we get eq :: a->a->Bool (notice no forall)
1204 RHS has type 'tau', whose free tyvars are tau_tvs
1205 RHS has constraints 'wanteds'
1208 Quantify over (tau_tvs \ ftvs(wanteds))
1209 This is bad. The constraints may contain (Monad (ST s))
1210 where we have instance Monad (ST s) where...
1211 so there's no need to be monomorphic in s!
1213 Also the constraint might be a method constraint,
1214 whose type mentions a perfectly innocent tyvar:
1215 op :: Num a => a -> b -> a
1216 Here, b is unconstrained. A good example would be
1218 We want to infer the polymorphic type
1219 foo :: forall b. b -> b
1222 Plan B (cunning, used for a long time up to and including GHC 6.2)
1223 Step 1: Simplify the constraints as much as possible (to deal
1224 with Plan A's problem). Then set
1225 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1227 Step 2: Now simplify again, treating the constraint as 'free' if
1228 it does not mention qtvs, and trying to reduce it otherwise.
1229 The reasons for this is to maximise sharing.
1231 This fails for a very subtle reason. Suppose that in the Step 2
1232 a constraint (Foo (Succ Zero) (Succ Zero) b) gets thrown upstairs as 'free'.
1233 In the Step 1 this constraint might have been simplified, perhaps to
1234 (Foo Zero Zero b), AND THEN THAT MIGHT BE IMPROVED, to bind 'b' to 'T'.
1235 This won't happen in Step 2... but that in turn might prevent some other
1236 constraint (Baz [a] b) being simplified (e.g. via instance Baz [a] T where {..})
1237 and that in turn breaks the invariant that no constraints are quantified over.
1239 Test typecheck/should_compile/tc177 (which failed in GHC 6.2) demonstrates
1244 Step 1: Simplify the constraints as much as possible (to deal
1245 with Plan A's problem). Then set
1246 qtvs = tau_tvs \ ftvs( simplify( wanteds ) )
1247 Return the bindings from Step 1.
1250 A note about Plan C (arising from "bug" reported by George Russel March 2004)
1253 instance (HasBinary ty IO) => HasCodedValue ty
1255 foo :: HasCodedValue a => String -> IO a
1257 doDecodeIO :: HasCodedValue a => () -> () -> IO a
1258 doDecodeIO codedValue view
1259 = let { act = foo "foo" } in act
1261 You might think this should work becuase the call to foo gives rise to a constraint
1262 (HasCodedValue t), which can be satisfied by the type sig for doDecodeIO. But the
1263 restricted binding act = ... calls tcSimplifyRestricted, and PlanC simplifies the
1264 constraint using the (rather bogus) instance declaration, and now we are stuffed.
1266 I claim this is not really a bug -- but it bit Sergey as well as George. So here's
1270 Plan D (a variant of plan B)
1271 Step 1: Simplify the constraints as much as possible (to deal
1272 with Plan A's problem), BUT DO NO IMPROVEMENT. 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.
1278 The point here is that it's generally OK to have too few qtvs; that is,
1279 to make the thing more monomorphic than it could be. We don't want to
1280 do that in the common cases, but in wierd cases it's ok: the programmer
1281 can always add a signature.
1283 Too few qtvs => too many wanteds, which is what happens if you do less
1288 tcSimplifyRestricted -- Used for restricted binding groups
1289 -- i.e. ones subject to the monomorphism restriction
1292 -> [Name] -- Things bound in this group
1293 -> TcTyVarSet -- Free in the type of the RHSs
1294 -> [Inst] -- Free in the RHSs
1295 -> TcM ([TyVar], -- Tyvars to quantify (zonked and quantified)
1296 TcDictBinds) -- Bindings
1297 -- tcSimpifyRestricted returns no constraints to
1298 -- quantify over; by definition there are none.
1299 -- They are all thrown back in the LIE
1301 tcSimplifyRestricted doc top_lvl bndrs tau_tvs wanteds
1302 -- Zonk everything in sight
1303 = do { traceTc (text "tcSimplifyRestricted")
1304 ; wanteds' <- zonkInsts wanteds
1306 -- 'ReduceMe': Reduce as far as we can. Don't stop at
1307 -- dicts; the idea is to get rid of as many type
1308 -- variables as possible, and we don't want to stop
1309 -- at (say) Monad (ST s), because that reduces
1310 -- immediately, with no constraint on s.
1312 -- BUT do no improvement! See Plan D above
1313 -- HOWEVER, some unification may take place, if we instantiate
1314 -- a method Inst with an equality constraint
1315 ; let env = mkNoImproveRedEnv doc (\i -> ReduceMe AddSCs)
1316 ; (_imp, _binds, constrained_dicts, _) <- reduceContext env wanteds'
1318 -- Next, figure out the tyvars we will quantify over
1319 ; tau_tvs' <- zonkTcTyVarsAndFV (varSetElems tau_tvs)
1320 ; gbl_tvs' <- tcGetGlobalTyVars
1321 ; constrained_dicts' <- zonkInsts constrained_dicts
1323 ; let qtvs1 = tau_tvs' `minusVarSet` oclose (fdPredsOfInsts constrained_dicts) gbl_tvs'
1324 -- As in tcSimplifyInfer
1326 -- Do not quantify over constrained type variables:
1327 -- this is the monomorphism restriction
1328 constrained_tvs' = tyVarsOfInsts constrained_dicts'
1329 qtvs = qtvs1 `minusVarSet` constrained_tvs'
1330 pp_bndrs = pprWithCommas (quotes . ppr) bndrs
1333 ; warn_mono <- doptM Opt_WarnMonomorphism
1334 ; warnTc (warn_mono && (constrained_tvs' `intersectsVarSet` qtvs1))
1335 (vcat[ ptext SLIT("the Monomorphism Restriction applies to the binding")
1336 <> plural bndrs <+> ptext SLIT("for") <+> pp_bndrs,
1337 ptext SLIT("Consider giving a type signature for") <+> pp_bndrs])
1339 ; traceTc (text "tcSimplifyRestricted" <+> vcat [
1340 pprInsts wanteds, pprInsts constrained_dicts',
1342 ppr constrained_tvs', ppr tau_tvs', ppr qtvs ])
1344 -- The first step may have squashed more methods than
1345 -- necessary, so try again, this time more gently, knowing the exact
1346 -- set of type variables to quantify over.
1348 -- We quantify only over constraints that are captured by qtvs;
1349 -- these will just be a subset of non-dicts. This in contrast
1350 -- to normal inference (using isFreeWhenInferring) in which we quantify over
1351 -- all *non-inheritable* constraints too. This implements choice
1352 -- (B) under "implicit parameter and monomorphism" above.
1354 -- Remember that we may need to do *some* simplification, to
1355 -- (for example) squash {Monad (ST s)} into {}. It's not enough
1356 -- just to float all constraints
1358 -- At top level, we *do* squash methods becuase we want to
1359 -- expose implicit parameters to the test that follows
1360 ; let is_nested_group = isNotTopLevel top_lvl
1361 try_me inst | isFreeWrtTyVars qtvs inst,
1362 (is_nested_group || isDict inst) = Stop
1363 | otherwise = ReduceMe AddSCs
1364 env = mkNoImproveRedEnv doc try_me
1365 ; (_imp, binds, irreds, _) <- reduceContext env wanteds'
1367 -- See "Notes on implicit parameters, Question 4: top level"
1368 ; ASSERT( all (isFreeWrtTyVars qtvs) irreds ) -- None should be captured
1369 if is_nested_group then
1371 else do { let (bad_ips, non_ips) = partition isIPDict irreds
1372 ; addTopIPErrs bndrs bad_ips
1373 ; extendLIEs non_ips }
1375 ; qtvs' <- zonkQuantifiedTyVars (varSetElems qtvs)
1376 ; return (qtvs', binds) }
1380 %************************************************************************
1384 %************************************************************************
1386 On the LHS of transformation rules we only simplify methods and constants,
1387 getting dictionaries. We want to keep all of them unsimplified, to serve
1388 as the available stuff for the RHS of the rule.
1390 Example. Consider the following left-hand side of a rule
1392 f (x == y) (y > z) = ...
1394 If we typecheck this expression we get constraints
1396 d1 :: Ord a, d2 :: Eq a
1398 We do NOT want to "simplify" to the LHS
1400 forall x::a, y::a, z::a, d1::Ord a.
1401 f ((==) (eqFromOrd d1) x y) ((>) d1 y z) = ...
1405 forall x::a, y::a, z::a, d1::Ord a, d2::Eq a.
1406 f ((==) d2 x y) ((>) d1 y z) = ...
1408 Here is another example:
1410 fromIntegral :: (Integral a, Num b) => a -> b
1411 {-# RULES "foo" fromIntegral = id :: Int -> Int #-}
1413 In the rule, a=b=Int, and Num Int is a superclass of Integral Int. But
1414 we *dont* want to get
1416 forall dIntegralInt.
1417 fromIntegral Int Int dIntegralInt (scsel dIntegralInt) = id Int
1419 because the scsel will mess up RULE matching. Instead we want
1421 forall dIntegralInt, dNumInt.
1422 fromIntegral Int Int dIntegralInt dNumInt = id Int
1426 g (x == y) (y == z) = ..
1428 where the two dictionaries are *identical*, we do NOT WANT
1430 forall x::a, y::a, z::a, d1::Eq a
1431 f ((==) d1 x y) ((>) d1 y z) = ...
1433 because that will only match if the dict args are (visibly) equal.
1434 Instead we want to quantify over the dictionaries separately.
1436 In short, tcSimplifyRuleLhs must *only* squash LitInst and MethInts, leaving
1437 all dicts unchanged, with absolutely no sharing. It's simpler to do this
1438 from scratch, rather than further parameterise simpleReduceLoop etc
1441 tcSimplifyRuleLhs :: [Inst] -> TcM ([Inst], TcDictBinds)
1442 tcSimplifyRuleLhs wanteds
1443 = go [] emptyBag wanteds
1446 = return (dicts, binds)
1447 go dicts binds (w:ws)
1449 = go (w:dicts) binds ws
1451 = do { w' <- zonkInst w -- So that (3::Int) does not generate a call
1452 -- to fromInteger; this looks fragile to me
1453 ; lookup_result <- lookupSimpleInst w'
1454 ; case lookup_result of
1455 GenInst ws' rhs -> go dicts (addBind binds w rhs) (ws' ++ ws)
1456 NoInstance -> pprPanic "tcSimplifyRuleLhs" (ppr w)
1460 tcSimplifyBracket is used when simplifying the constraints arising from
1461 a Template Haskell bracket [| ... |]. We want to check that there aren't
1462 any constraints that can't be satisfied (e.g. Show Foo, where Foo has no
1463 Show instance), but we aren't otherwise interested in the results.
1464 Nor do we care about ambiguous dictionaries etc. We will type check
1465 this bracket again at its usage site.
1468 tcSimplifyBracket :: [Inst] -> TcM ()
1469 tcSimplifyBracket wanteds
1470 = do { tryHardCheckLoop doc wanteds
1473 doc = text "tcSimplifyBracket"
1477 %************************************************************************
1479 \subsection{Filtering at a dynamic binding}
1481 %************************************************************************
1486 we must discharge all the ?x constraints from B. We also do an improvement
1487 step; if we have ?x::t1 and ?x::t2 we must unify t1, t2.
1489 Actually, the constraints from B might improve the types in ?x. For example
1491 f :: (?x::Int) => Char -> Char
1494 then the constraint (?x::Int) arising from the call to f will
1495 force the binding for ?x to be of type Int.
1498 tcSimplifyIPs :: [Inst] -- The implicit parameters bound here
1501 -- We need a loop so that we do improvement, and then
1502 -- (next time round) generate a binding to connect the two
1504 -- Here the two ?x's have different types, and improvement
1505 -- makes them the same.
1507 tcSimplifyIPs given_ips wanteds
1508 = do { wanteds' <- zonkInsts wanteds
1509 ; given_ips' <- zonkInsts given_ips
1510 -- Unusually for checking, we *must* zonk the given_ips
1512 ; let env = mkRedEnv doc try_me given_ips'
1513 ; (improved, binds, irreds, _) <- reduceContext env wanteds'
1515 ; if not improved then
1516 ASSERT( all is_free irreds )
1517 do { extendLIEs irreds
1520 tcSimplifyIPs given_ips wanteds }
1522 doc = text "tcSimplifyIPs" <+> ppr given_ips
1523 ip_set = mkNameSet (ipNamesOfInsts given_ips)
1524 is_free inst = isFreeWrtIPs ip_set inst
1526 -- Simplify any methods that mention the implicit parameter
1527 try_me inst | is_free inst = Stop
1528 | otherwise = ReduceMe NoSCs
1532 %************************************************************************
1534 \subsection[binds-for-local-funs]{@bindInstsOfLocalFuns@}
1536 %************************************************************************
1538 When doing a binding group, we may have @Insts@ of local functions.
1539 For example, we might have...
1541 let f x = x + 1 -- orig local function (overloaded)
1542 f.1 = f Int -- two instances of f
1547 The point is: we must drop the bindings for @f.1@ and @f.2@ here,
1548 where @f@ is in scope; those @Insts@ must certainly not be passed
1549 upwards towards the top-level. If the @Insts@ were binding-ified up
1550 there, they would have unresolvable references to @f@.
1552 We pass in an @init_lie@ of @Insts@ and a list of locally-bound @Ids@.
1553 For each method @Inst@ in the @init_lie@ that mentions one of the
1554 @Ids@, we create a binding. We return the remaining @Insts@ (in an
1555 @LIE@), as well as the @HsBinds@ generated.
1558 bindInstsOfLocalFuns :: [Inst] -> [TcId] -> TcM TcDictBinds
1559 -- Simlifies only MethodInsts, and generate only bindings of form
1561 -- We're careful not to even generate bindings of the form
1563 -- You'd think that'd be fine, but it interacts with what is
1564 -- arguably a bug in Match.tidyEqnInfo (see notes there)
1566 bindInstsOfLocalFuns wanteds local_ids
1567 | null overloaded_ids
1569 = extendLIEs wanteds `thenM_`
1570 returnM emptyLHsBinds
1573 = do { (irreds, binds,_) <- checkLoop env for_me
1574 ; extendLIEs not_for_me
1578 env = mkRedEnv doc try_me []
1579 doc = text "bindInsts" <+> ppr local_ids
1580 overloaded_ids = filter is_overloaded local_ids
1581 is_overloaded id = isOverloadedTy (idType id)
1582 (for_me, not_for_me) = partition (isMethodFor overloaded_set) wanteds
1584 overloaded_set = mkVarSet overloaded_ids -- There can occasionally be a lot of them
1585 -- so it's worth building a set, so that
1586 -- lookup (in isMethodFor) is faster
1587 try_me inst | isMethod inst = ReduceMe NoSCs
1592 %************************************************************************
1594 \subsection{Data types for the reduction mechanism}
1596 %************************************************************************
1598 The main control over context reduction is here
1602 = RedEnv { red_doc :: SDoc -- The context
1603 , red_try_me :: Inst -> WhatToDo
1604 , red_improve :: Bool -- True <=> do improvement
1605 , red_givens :: [Inst] -- All guaranteed rigid
1607 -- but see Note [Rigidity]
1608 , red_stack :: (Int, [Inst]) -- Recursion stack (for err msg)
1609 -- See Note [RedStack]
1613 -- The red_givens are rigid so far as cmpInst is concerned.
1614 -- There is one case where they are not totally rigid, namely in tcSimplifyIPs
1615 -- let ?x = e in ...
1616 -- Here, the given is (?x::a), where 'a' is not necy a rigid type
1617 -- But that doesn't affect the comparison, which is based only on mame.
1620 -- The red_stack pair (n,insts) pair is just used for error reporting.
1621 -- 'n' is always the depth of the stack.
1622 -- The 'insts' is the stack of Insts being reduced: to produce X
1623 -- I had to produce Y, to produce Y I had to produce Z, and so on.
1626 mkRedEnv :: SDoc -> (Inst -> WhatToDo) -> [Inst] -> RedEnv
1627 mkRedEnv doc try_me givens
1628 = RedEnv { red_doc = doc, red_try_me = try_me,
1629 red_givens = givens, red_stack = (0,[]),
1630 red_improve = True }
1632 mkNoImproveRedEnv :: SDoc -> (Inst -> WhatToDo) -> RedEnv
1633 -- Do not do improvement; no givens
1634 mkNoImproveRedEnv doc try_me
1635 = RedEnv { red_doc = doc, red_try_me = try_me,
1636 red_givens = [], red_stack = (0,[]),
1637 red_improve = True }
1640 = ReduceMe WantSCs -- Try to reduce this
1641 -- If there's no instance, add the inst to the
1642 -- irreductible ones, but don't produce an error
1643 -- message of any kind.
1644 -- It might be quite legitimate such as (Eq a)!
1646 | Stop -- Return as irreducible unless it can
1647 -- be reduced to a constant in one step
1648 -- Do not add superclasses; see
1650 data WantSCs = NoSCs | AddSCs -- Tells whether we should add the superclasses
1651 -- of a predicate when adding it to the avails
1652 -- The reason for this flag is entirely the super-class loop problem
1653 -- Note [SUPER-CLASS LOOP 1]
1656 %************************************************************************
1658 \subsection[reduce]{@reduce@}
1660 %************************************************************************
1664 reduceContext :: RedEnv
1666 -> TcM (ImprovementDone,
1667 TcDictBinds, -- Dictionary bindings
1668 [Inst], -- Irreducible
1669 [Inst]) -- Needed givens
1671 reduceContext env wanteds
1672 = do { traceTc (text "reduceContext" <+> (vcat [
1673 text "----------------------",
1675 text "given" <+> ppr (red_givens env),
1676 text "wanted" <+> ppr wanteds,
1677 text "----------------------"
1681 ; let givens = red_givens env
1682 (given_eqs0,given_dicts0) = partitionGivenEqInsts givens
1683 (wanted_eqs,wanted_dicts) = partitionWantedEqInsts wanteds
1685 ; -- 1. Normalise the *given* *equality* constraints
1686 (given_eqs,eliminate_skolems) <- normaliseGivens given_eqs0
1688 ; -- 2. Normalise the *given* *dictionary* constraints
1689 -- wrt. the toplevel and given equations
1690 (given_dicts,given_binds) <- normaliseGivenDicts given_eqs given_dicts0
1692 ; -- 3. Solve the *wanted* *equation* constraints
1693 eq_irreds0 <- solveWanteds given_eqs wanted_eqs
1695 -- 4. Normalise the *wanted* equality constraints with respect to each other
1696 ; eq_irreds <- normaliseWanteds eq_irreds0
1698 -- -- Do the real work
1699 -- -- Process non-implication constraints first, so that they are
1700 -- -- available to help solving the implication constraints
1701 -- -- ToDo: seems a bit inefficient and ad-hoc
1702 -- ; let (implics, rest) = partition isImplicInst wanteds
1703 -- ; avails <- reduceList env (rest ++ implics) init_state
1705 -- 5. Build the Avail mapping from "given_dicts"
1706 ; init_state <- foldlM addGiven emptyAvails given_dicts
1708 -- 6. Solve the *wanted* *dictionary* constraints
1709 -- This may expose some further equational constraints...
1710 ; wanted_dicts' <- zonkInsts wanted_dicts
1711 ; avails <- reduceList env wanted_dicts' init_state
1712 ; (binds, irreds0, needed_givens) <- extractResults avails wanted_dicts'
1713 ; traceTc (text "reduceContext extractresults" <+> vcat
1714 [ppr avails,ppr wanted_dicts',ppr binds,ppr needed_givens])
1716 ; -- 7. Normalise the *wanted* *dictionary* constraints
1717 -- wrt. the toplevel and given equations
1718 (irreds1,normalise_binds1) <- normaliseWantedDicts given_eqs irreds0
1720 ; -- 8. Substitute the wanted *equations* in the wanted *dictionaries*
1721 (irreds,normalise_binds2) <- substEqInDictInsts eq_irreds irreds1
1723 ; -- 9. eliminate the artificial skolem constants introduced in 1.
1726 -- If there was some FD improvement,
1727 -- or new wanted equations have been exposed,
1728 -- we should have another go at solving.
1729 ; let improved = availsImproved avails
1730 || (not $ isEmptyBag normalise_binds1)
1731 || (not $ isEmptyBag normalise_binds2)
1732 || (not $ null $ fst $ partitionGivenEqInsts irreds)
1734 ; traceTc (text "reduceContext end" <+> (vcat [
1735 text "----------------------",
1737 text "given" <+> ppr (red_givens env),
1738 text "wanted" <+> ppr wanteds,
1740 text "avails" <+> pprAvails avails,
1741 text "improved =" <+> ppr improved,
1742 text "irreds = " <+> ppr irreds,
1743 text "binds = " <+> ppr binds,
1744 text "needed givens = " <+> ppr needed_givens,
1745 text "----------------------"
1748 ; return (improved, given_binds `unionBags` normalise_binds1 `unionBags` normalise_binds2 `unionBags` binds, irreds ++ eq_irreds, needed_givens) }
1750 tcImproveOne :: Avails -> Inst -> TcM ImprovementDone
1751 tcImproveOne avails inst
1752 | not (isDict inst) = return False
1754 = do { inst_envs <- tcGetInstEnvs
1755 ; let eqns = improveOne (classInstances inst_envs)
1756 (dictPred inst, pprInstArising inst)
1757 [ (dictPred p, pprInstArising p)
1758 | p <- availsInsts avails, isDict p ]
1759 -- Avails has all the superclasses etc (good)
1760 -- It also has all the intermediates of the deduction (good)
1761 -- It does not have duplicates (good)
1762 -- NB that (?x::t1) and (?x::t2) will be held separately in avails
1763 -- so that improve will see them separate
1764 ; traceTc (text "improveOne" <+> ppr inst)
1767 unifyEqns :: [(Equation,(PredType,SDoc),(PredType,SDoc))]
1768 -> TcM ImprovementDone
1769 unifyEqns [] = return False
1771 = do { traceTc (ptext SLIT("Improve:") <+> vcat (map pprEquationDoc eqns))
1775 unify ((qtvs, pairs), what1, what2)
1776 = addErrCtxtM (mkEqnMsg what1 what2) $
1777 tcInstTyVars (varSetElems qtvs) `thenM` \ (_, _, tenv) ->
1778 mapM_ (unif_pr tenv) pairs
1779 unif_pr tenv (ty1,ty2) = unifyType (substTy tenv ty1) (substTy tenv ty2)
1781 pprEquationDoc (eqn, (p1,w1), (p2,w2)) = vcat [pprEquation eqn, nest 2 (ppr p1), nest 2 (ppr p2)]
1783 mkEqnMsg (pred1,from1) (pred2,from2) tidy_env
1784 = do { pred1' <- zonkTcPredType pred1; pred2' <- zonkTcPredType pred2
1785 ; let { pred1'' = tidyPred tidy_env pred1'; pred2'' = tidyPred tidy_env pred2' }
1786 ; let msg = vcat [ptext SLIT("When using functional dependencies to combine"),
1787 nest 2 (sep [ppr pred1'' <> comma, nest 2 from1]),
1788 nest 2 (sep [ppr pred2'' <> comma, nest 2 from2])]
1789 ; return (tidy_env, msg) }
1792 The main context-reduction function is @reduce@. Here's its game plan.
1795 reduceList :: RedEnv -> [Inst] -> Avails -> TcM Avails
1796 reduceList env@(RedEnv {red_stack = (n,stk)}) wanteds state
1797 = do { dopts <- getDOpts
1800 dumpTcRn (hang (ptext SLIT("Interesting! Context reduction stack depth") <+> int n)
1801 2 (ifPprDebug (nest 2 (pprStack stk))))
1804 ; if n >= ctxtStkDepth dopts then
1805 failWithTc (reduceDepthErr n stk)
1809 go [] state = return state
1810 go (w:ws) state = do { traceTc (text "reduceList " <+> ppr (w:ws) <+> ppr state)
1811 ; state' <- reduce (env {red_stack = (n+1, w:stk)}) w state
1814 -- Base case: we're done!
1815 reduce env wanted avails
1816 -- It's the same as an existing inst, or a superclass thereof
1817 | Just avail <- findAvail avails wanted
1818 = do { traceTc (text "reduce: found " <+> ppr wanted)
1823 = do { traceTc (text "reduce" <+> ppr avails <+> ppr wanted)
1824 ; case red_try_me env wanted of {
1825 Stop -> try_simple (addIrred NoSCs);
1826 -- See Note [No superclasses for Stop]
1828 ReduceMe want_scs -> do -- It should be reduced
1829 { (avails, lookup_result) <- reduceInst env avails wanted
1830 ; case lookup_result of
1831 NoInstance -> addIrred want_scs avails wanted
1832 -- Add it and its superclasses
1834 GenInst [] rhs -> addWanted want_scs avails wanted rhs []
1836 GenInst wanteds' rhs
1837 -> do { avails1 <- addIrred NoSCs avails wanted
1838 ; avails2 <- reduceList env wanteds' avails1
1839 ; addWanted want_scs avails2 wanted rhs wanteds' } }
1840 -- Temporarily do addIrred *before* the reduceList,
1841 -- which has the effect of adding the thing we are trying
1842 -- to prove to the database before trying to prove the things it
1843 -- needs. See note [RECURSIVE DICTIONARIES]
1844 -- NB: we must not do an addWanted before, because that adds the
1845 -- superclasses too, and that can lead to a spurious loop; see
1846 -- the examples in [SUPERCLASS-LOOP]
1847 -- So we do an addIrred before, and then overwrite it afterwards with addWanted
1850 -- First, see if the inst can be reduced to a constant in one step
1851 -- Works well for literals (1::Int) and constant dictionaries (d::Num Int)
1852 -- Don't bother for implication constraints, which take real work
1853 try_simple do_this_otherwise
1854 = do { res <- lookupSimpleInst wanted
1856 GenInst [] rhs -> addWanted AddSCs avails wanted rhs []
1857 other -> do_this_otherwise avails wanted }
1861 Note [SUPERCLASS-LOOP 2]
1862 ~~~~~~~~~~~~~~~~~~~~~~~~
1863 But the above isn't enough. Suppose we are *given* d1:Ord a,
1864 and want to deduce (d2:C [a]) where
1866 class Ord a => C a where
1867 instance Ord [a] => C [a] where ...
1869 Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
1870 superclasses of C [a] to avails. But we must not overwrite the binding
1871 for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
1874 Here's another variant, immortalised in tcrun020
1875 class Monad m => C1 m
1876 class C1 m => C2 m x
1877 instance C2 Maybe Bool
1878 For the instance decl we need to build (C1 Maybe), and it's no good if
1879 we run around and add (C2 Maybe Bool) and its superclasses to the avails
1880 before we search for C1 Maybe.
1882 Here's another example
1883 class Eq b => Foo a b
1884 instance Eq a => Foo [a] a
1888 we'll first deduce that it holds (via the instance decl). We must not
1889 then overwrite the Eq t constraint with a superclass selection!
1891 At first I had a gross hack, whereby I simply did not add superclass constraints
1892 in addWanted, though I did for addGiven and addIrred. This was sub-optimal,
1893 becuase it lost legitimate superclass sharing, and it still didn't do the job:
1894 I found a very obscure program (now tcrun021) in which improvement meant the
1895 simplifier got two bites a the cherry... so something seemed to be an Stop
1896 first time, but reducible next time.
1898 Now we implement the Right Solution, which is to check for loops directly
1899 when adding superclasses. It's a bit like the occurs check in unification.
1902 Note [RECURSIVE DICTIONARIES]
1903 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1905 data D r = ZeroD | SuccD (r (D r));
1907 instance (Eq (r (D r))) => Eq (D r) where
1908 ZeroD == ZeroD = True
1909 (SuccD a) == (SuccD b) = a == b
1912 equalDC :: D [] -> D [] -> Bool;
1915 We need to prove (Eq (D [])). Here's how we go:
1919 by instance decl, holds if
1923 by instance decl of Eq, holds if
1925 where d2 = dfEqList d3
1928 But now we can "tie the knot" to give
1934 and it'll even run! The trick is to put the thing we are trying to prove
1935 (in this case Eq (D []) into the database before trying to prove its
1936 contributing clauses.
1939 %************************************************************************
1941 Reducing a single constraint
1943 %************************************************************************
1946 ---------------------------------------------
1947 reduceInst :: RedEnv -> Avails -> Inst -> TcM (Avails, LookupInstResult)
1948 reduceInst env avails (ImplicInst { tci_tyvars = tvs, tci_reft = reft, tci_loc = loc,
1949 tci_given = extra_givens, tci_wanted = wanteds })
1950 = reduceImplication env avails reft tvs extra_givens wanteds loc
1952 reduceInst env avails other_inst
1953 = do { result <- lookupSimpleInst other_inst
1954 ; return (avails, result) }
1957 Note [Equational Constraints in Implication Constraints]
1958 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1960 An equational constraint is of the form
1962 where Given and Wanted may contain both equational and dictionary
1963 constraints. The delay and reduction of these two kinds of constraints
1966 -) In the generated code, wanted Dictionary constraints are wrapped up in an
1967 implication constraint that is created at the code site where the wanted
1968 dictionaries can be reduced via a let-binding. This let-bound implication
1969 constraint is deconstructed at the use-site of the wanted dictionaries.
1971 -) While the reduction of equational constraints is also delayed, the delay
1972 is not manifest in the generated code. The required evidence is generated
1973 in the code directly at the use-site. There is no let-binding and deconstruction
1974 necessary. The main disadvantage is that we cannot exploit sharing as the
1975 same evidence may be generated at multiple use-sites. However, this disadvantage
1976 is limited because it only concerns coercions which are erased.
1978 The different treatment is motivated by the different in representation. Dictionary
1979 constraints require manifest runtime dictionaries, while equations require coercions
1983 ---------------------------------------------
1984 reduceImplication :: RedEnv
1986 -> Refinement -- May refine the givens; often empty
1987 -> [TcTyVar] -- Quantified type variables; all skolems
1988 -> [Inst] -- Extra givens; all rigid
1991 -> TcM (Avails, LookupInstResult)
1994 Suppose we are simplifying the constraint
1995 forall bs. extras => wanted
1996 in the context of an overall simplification problem with givens 'givens',
1997 and refinment 'reft'.
2000 * The refinement is often empty
2002 * The 'extra givens' need not mention any of the quantified type variables
2003 e.g. forall {}. Eq a => Eq [a]
2004 forall {}. C Int => D (Tree Int)
2006 This happens when you have something like
2008 T1 :: Eq a => a -> T a
2011 f x = ...(case x of { T1 v -> v==v })...
2014 -- ToDo: should we instantiate tvs? I think it's not necessary
2016 -- Note on coercion variables:
2018 -- The extra given coercion variables are bound at two different sites:
2019 -- -) in the creation context of the implication constraint
2020 -- the solved equational constraints use these binders
2022 -- -) at the solving site of the implication constraint
2023 -- the solved dictionaries use these binders
2024 -- these binders are generated by reduceImplication
2026 reduceImplication env orig_avails reft tvs extra_givens wanteds inst_loc
2027 = do { -- Add refined givens, and the extra givens
2029 (refined_red_givens,refined_avails)
2030 <- if isEmptyRefinement reft then return (red_givens env,orig_avails)
2031 else foldlM (addRefinedGiven reft) ([],orig_avails) (red_givens env)
2033 -- Solve the sub-problem
2034 ; let try_me inst = ReduceMe AddSCs -- Note [Freeness and implications]
2035 env' = env { red_givens = refined_red_givens ++ extra_givens ++ availsInsts orig_avails
2036 , red_try_me = try_me }
2038 ; traceTc (text "reduceImplication" <+> vcat
2040 ppr (red_givens env), ppr extra_givens,
2041 ppr reft, ppr wanteds])
2042 ; (irreds,binds,needed_givens0) <- checkLoop env' wanteds
2043 ; let needed_givens1 = [ng | ng <- needed_givens0, notElem ng extra_givens]
2045 -- Note [Reducing implication constraints]
2046 -- Tom -- update note, put somewhere!
2048 ; traceTc (text "reduceImplication result" <+> vcat
2049 [ppr irreds, ppr binds, ppr needed_givens1])
2050 -- ; avails <- reduceList env' wanteds avails
2052 -- -- Extract the binding
2053 -- ; (binds, irreds) <- extractResults avails wanteds
2054 ; (refinement_binds,needed_givens) <- extractLocalResults refined_avails needed_givens1
2055 ; traceTc (text "reduceImplication local results" <+> vcat
2056 [ppr refinement_binds, ppr needed_givens])
2058 ; -- extract superclass binds
2059 -- (sc_binds,_) <- extractResults avails []
2060 -- ; traceTc (text "reduceImplication sc_binds" <+> vcat
2061 -- [ppr sc_binds, ppr avails])
2064 -- We always discard the extra avails we've generated;
2065 -- but we remember if we have done any (global) improvement
2066 -- ; let ret_avails = avails
2067 ; let ret_avails = orig_avails
2068 -- ; let ret_avails = updateImprovement orig_avails avails
2070 ; traceTc (text "reduceImplication condition" <+> ppr ((isEmptyLHsBinds binds) || (null irreds)))
2072 -- Porgress is no longer measered by the number of bindings
2073 -- ; if isEmptyLHsBinds binds then -- No progress
2074 ; if (isEmptyLHsBinds binds) && (not $ null irreds) then
2075 return (ret_avails, NoInstance)
2078 ; (implic_insts, bind) <- makeImplicationBind inst_loc tvs reft extra_givens irreds
2079 -- This binding is useless if the recursive simplification
2080 -- made no progress; but currently we don't try to optimise that
2081 -- case. After all, we only try hard to reduce at top level, or
2082 -- when inferring types.
2084 ; let dict_wanteds = filter (not . isEqInst) wanteds
2085 (extra_eq_givens,extra_dict_givens) = partitionGivenEqInsts extra_givens
2086 dict_ids = map instToId extra_dict_givens
2087 -- TOMDO: given equational constraints bug!
2088 -- we need a different evidence for given
2089 -- equations depending on whether we solve
2090 -- dictionary constraints or equational constraints
2091 eq_tyvars = uniqSetToList $ tyVarsOfTypes $ map eqInstType extra_eq_givens
2092 -- dict_ids = map instToId extra_givens
2093 co = mkWpTyLams tvs <.> mkWpTyLams eq_tyvars <.> mkWpLams dict_ids <.> WpLet (binds `unionBags` refinement_binds `unionBags` bind)
2094 rhs = mkHsWrap co payload
2095 loc = instLocSpan inst_loc
2096 payload | [dict_wanted] <- dict_wanteds = HsVar (instToId dict_wanted)
2097 | otherwise = ExplicitTuple (map (L loc . HsVar . instToId) dict_wanteds) Boxed
2100 ; traceTc (text "reduceImplication ->" <+> vcat
2103 -- If there are any irreds, we back off and return NoInstance
2104 ; return (ret_avails, GenInst (implic_insts ++ needed_givens) (L loc rhs))
2109 Note [Reducing implication constraints]
2110 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2111 Suppose we are trying to simplify
2112 (Ord a, forall b. C a b => (W [a] b, D c b))
2114 instance (C a b, Ord a) => W [a] b
2115 When solving the implication constraint, we'll start with
2117 in the Avails. Then we add (C a b -> Given) and solve. Extracting
2118 the results gives us a binding for the (W [a] b), with an Irred of
2119 (Ord a, D c b). Now, the (Ord a) comes from "outside" the implication,
2120 but the (D d b) is from "inside". So we want to generate a Rhs binding
2123 ic = /\b \dc:C a b). (df a b dc do, ic' b dc)
2126 ic' :: forall b. C a b => D c b
2128 The 'depending on' part of the Rhs is important, because it drives
2129 the extractResults code.
2131 The "inside" and "outside" distinction is what's going on with 'inner' and
2132 'outer' in reduceImplication
2135 Note [Freeness and implications]
2136 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2137 It's hard to say when an implication constraint can be floated out. Consider
2138 forall {} Eq a => Foo [a]
2139 The (Foo [a]) doesn't mention any of the quantified variables, but it
2140 still might be partially satisfied by the (Eq a).
2142 There is a useful special case when it *is* easy to partition the
2143 constraints, namely when there are no 'givens'. Consider
2144 forall {a}. () => Bar b
2145 There are no 'givens', and so there is no reason to capture (Bar b).
2146 We can let it float out. But if there is even one constraint we
2147 must be much more careful:
2148 forall {a}. C a b => Bar (m b)
2149 because (C a b) might have a superclass (D b), from which we might
2150 deduce (Bar [b]) when m later gets instantiated to []. Ha!
2152 Here is an even more exotic example
2154 Now consider the constraint
2155 forall b. D Int b => C Int
2156 We can satisfy the (C Int) from the superclass of D, so we don't want
2157 to float the (C Int) out, even though it mentions no type variable in
2160 Note [Pruning the givens in an implication constraint]
2161 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2162 Suppose we are about to form the implication constraint
2163 forall tvs. Eq a => Ord b
2164 The (Eq a) cannot contribute to the (Ord b), because it has no access to
2165 the type variable 'b'. So we could filter out the (Eq a) from the givens.
2167 Doing so would be a bit tidier, but all the implication constraints get
2168 simplified away by the optimiser, so it's no great win. So I don't take
2169 advantage of that at the moment.
2171 If you do, BE CAREFUL of wobbly type variables.
2174 %************************************************************************
2176 Avails and AvailHow: the pool of evidence
2178 %************************************************************************
2182 data Avails = Avails !ImprovementDone !AvailEnv
2184 type ImprovementDone = Bool -- True <=> some unification has happened
2185 -- so some Irreds might now be reducible
2186 -- keys that are now
2188 type AvailEnv = FiniteMap Inst AvailHow
2190 = IsIrred -- Used for irreducible dictionaries,
2191 -- which are going to be lambda bound
2193 | Given TcId -- Used for dictionaries for which we have a binding
2194 -- e.g. those "given" in a signature
2196 | Rhs -- Used when there is a RHS
2197 (LHsExpr TcId) -- The RHS
2198 [Inst] -- Insts free in the RHS; we need these too
2200 instance Outputable Avails where
2203 pprAvails (Avails imp avails)
2204 = vcat [ ptext SLIT("Avails") <> (if imp then ptext SLIT("[improved]") else empty)
2205 , nest 2 (vcat [sep [ppr inst, nest 2 (equals <+> ppr avail)]
2206 | (inst,avail) <- fmToList avails ])]
2208 instance Outputable AvailHow where
2211 -------------------------
2212 pprAvail :: AvailHow -> SDoc
2213 pprAvail IsIrred = text "Irred"
2214 pprAvail (Given x) = text "Given" <+> ppr x
2215 pprAvail (Rhs rhs bs) = text "Rhs" <+> sep [ppr rhs, braces (ppr bs)]
2217 -------------------------
2218 extendAvailEnv :: AvailEnv -> Inst -> AvailHow -> AvailEnv
2219 extendAvailEnv env inst avail = addToFM env inst avail
2221 findAvailEnv :: AvailEnv -> Inst -> Maybe AvailHow
2222 findAvailEnv env wanted = lookupFM env wanted
2223 -- NB 1: the Ord instance of Inst compares by the class/type info
2224 -- *not* by unique. So
2225 -- d1::C Int == d2::C Int
2227 emptyAvails :: Avails
2228 emptyAvails = Avails False emptyFM
2230 findAvail :: Avails -> Inst -> Maybe AvailHow
2231 findAvail (Avails _ avails) wanted = findAvailEnv avails wanted
2233 elemAvails :: Inst -> Avails -> Bool
2234 elemAvails wanted (Avails _ avails) = wanted `elemFM` avails
2236 extendAvails :: Avails -> Inst -> AvailHow -> TcM Avails
2238 extendAvails avails@(Avails imp env) inst avail
2239 = do { imp1 <- tcImproveOne avails inst -- Do any improvement
2240 ; return (Avails (imp || imp1) (extendAvailEnv env inst avail)) }
2242 availsInsts :: Avails -> [Inst]
2243 availsInsts (Avails _ avails) = keysFM avails
2245 availsImproved (Avails imp _) = imp
2247 updateImprovement :: Avails -> Avails -> Avails
2248 -- (updateImprovement a1 a2) sets a1's improvement flag from a2
2249 updateImprovement (Avails _ avails1) (Avails imp2 _) = Avails imp2 avails1
2252 Extracting the bindings from a bunch of Avails.
2253 The bindings do *not* come back sorted in dependency order.
2254 We assume that they'll be wrapped in a big Rec, so that the
2255 dependency analyser can sort them out later
2258 extractResults :: Avails
2260 -> TcM ( TcDictBinds, -- Bindings
2261 [Inst], -- Irreducible ones
2262 [Inst]) -- Needed givens, i.e. ones used in the bindings
2264 extractResults (Avails _ avails) wanteds
2265 = go avails emptyBag [] [] wanteds
2267 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst] -> [Inst]
2268 -> TcM (TcDictBinds, [Inst], [Inst])
2269 go avails binds irreds givens []
2270 = returnM (binds, irreds, givens)
2272 go avails binds irreds givens (w:ws)
2273 = case findAvailEnv avails w of
2274 Nothing -> pprTrace "Urk: extractResults" (ppr w) $
2275 go avails binds irreds givens ws
2278 | id == w_id -> go avails binds irreds (w:givens) ws
2279 | otherwise -> go avails (addBind binds w (nlHsVar id)) irreds (update_id w id:givens) ws
2280 -- The sought Id can be one of the givens, via a superclass chain
2281 -- and then we definitely don't want to generate an x=x binding!
2283 Just IsIrred -> go (add_given avails w) binds (w:irreds) givens ws
2284 -- The add_given handles the case where we want (Ord a, Eq a), and we
2285 -- don't want to emit *two* Irreds for Ord a, one via the superclass chain
2286 -- This showed up in a dupliated Ord constraint in the error message for
2289 Just (Rhs rhs ws') -> go (add_given avails w) new_binds irreds givens (ws' ++ ws)
2291 new_binds = addBind binds w rhs
2294 update_id m@(Method{}) id = m {tci_id = id}
2295 update_id w id = w {tci_name = idName id}
2297 add_given avails w = extendAvailEnv avails w (Given (instToId w))
2299 extractLocalResults :: Avails
2301 -> TcM ( TcDictBinds, -- Bindings
2302 [Inst]) -- Needed givens, i.e. ones used in the bindings
2304 extractLocalResults (Avails _ avails) wanteds
2305 = go avails emptyBag [] wanteds
2307 go :: AvailEnv -> TcDictBinds -> [Inst] -> [Inst]
2308 -> TcM (TcDictBinds, [Inst])
2309 go avails binds givens []
2310 = returnM (binds, givens)
2312 go avails binds givens (w:ws)
2313 = case findAvailEnv avails w of
2314 Nothing -> -- pprTrace "Urk: extractLocalResults" (ppr w) $
2315 go avails binds givens ws
2318 go avails binds givens ws
2321 | id == w_id -> go avails binds (w:givens) ws
2322 | otherwise -> go avails binds (w{tci_name=idName id}:givens) ws
2323 -- The sought Id can be one of the givens, via a superclass chain
2324 -- and then we definitely don't want to generate an x=x binding!
2326 Just (Rhs rhs ws') -> go (add_given avails w) new_binds givens (ws' ++ ws)
2328 new_binds = addBind binds w rhs
2332 add_given avails w = extendAvailEnv avails w (Given (instToId w))
2336 Note [No superclasses for Stop]
2337 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2338 When we decide not to reduce an Inst -- the 'WhatToDo' --- we still
2339 add it to avails, so that any other equal Insts will be commoned up
2340 right here. However, we do *not* add superclasses. If we have
2343 but a is not bound here, then we *don't* want to derive dn from df
2344 here lest we lose sharing.
2347 addWanted :: WantSCs -> Avails -> Inst -> LHsExpr TcId -> [Inst] -> TcM Avails
2348 addWanted want_scs avails wanted rhs_expr wanteds
2349 = addAvailAndSCs want_scs avails wanted avail
2351 avail = Rhs rhs_expr wanteds
2353 addGiven :: Avails -> Inst -> TcM Avails
2354 addGiven avails given = addAvailAndSCs AddSCs avails given (Given (instToId given))
2355 -- Always add superclasses for 'givens'
2357 -- No ASSERT( not (given `elemAvails` avails) ) because in an instance
2358 -- decl for Ord t we can add both Ord t and Eq t as 'givens',
2359 -- so the assert isn't true
2361 addRefinedGiven :: Refinement -> ([Inst], Avails) -> Inst -> TcM ([Inst], Avails)
2362 addRefinedGiven reft (refined_givens, avails) given
2363 | isDict given -- We sometimes have 'given' methods, but they
2364 -- are always optional, so we can drop them
2365 , let pred = dictPred given
2366 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2367 , Just (co, pred) <- refinePred reft pred
2368 = do { new_given <- newDictBndr (instLoc given) pred
2369 ; let rhs = L (instSpan given) $
2370 HsWrap (WpCo co) (HsVar (instToId given))
2371 ; avails <- addAvailAndSCs AddSCs avails new_given (Rhs rhs [given])
2372 ; return (new_given:refined_givens, avails) }
2373 -- ToDo: the superclasses of the original given all exist in Avails
2374 -- so we could really just cast them, but it's more awkward to do,
2375 -- and hopefully the optimiser will spot the duplicated work
2377 = return (refined_givens, avails)
2379 addRefinedGiven' :: Refinement -> [Inst] -> Inst -> TcM [Inst]
2380 addRefinedGiven' reft refined_givens given
2381 | isDict given -- We sometimes have 'given' methods, but they
2382 -- are always optional, so we can drop them
2383 , let pred = dictPred given
2384 , isRefineablePred pred -- See Note [ImplicInst rigidity]
2385 , Just (co, pred) <- refinePred reft pred
2386 = do { new_given <- newDictBndr (instLoc given) pred
2387 ; return (new_given:refined_givens) }
2388 -- ToDo: the superclasses of the original given all exist in Avails
2389 -- so we could really just cast them, but it's more awkward to do,
2390 -- and hopefully the optimiser will spot the duplicated work
2392 = return refined_givens
2395 Note [ImplicInst rigidity]
2396 ~~~~~~~~~~~~~~~~~~~~~~~~~~
2398 C :: forall ab. (Eq a, Ord b) => b -> T a
2400 ...(case x of C v -> <body>)...
2402 From the case (where x::T ty) we'll get an implication constraint
2403 forall b. (Eq ty, Ord b) => <body-constraints>
2404 Now suppose <body-constraints> itself has an implication constraint
2406 forall c. <reft> => <payload>
2407 Then, we can certainly apply the refinement <reft> to the Ord b, becuase it is
2408 existential, but we probably should not apply it to the (Eq ty) because it may
2409 be wobbly. Hence the isRigidInst
2411 @Insts@ are ordered by their class/type info, rather than by their
2412 unique. This allows the context-reduction mechanism to use standard finite
2413 maps to do their stuff. It's horrible that this code is here, rather
2414 than with the Avails handling stuff in TcSimplify
2417 addIrred :: WantSCs -> Avails -> Inst -> TcM Avails
2418 addIrred want_scs avails irred = ASSERT2( not (irred `elemAvails` avails), ppr irred $$ ppr avails )
2419 addAvailAndSCs want_scs avails irred IsIrred
2421 addAvailAndSCs :: WantSCs -> Avails -> Inst -> AvailHow -> TcM Avails
2422 addAvailAndSCs want_scs avails inst avail
2423 | not (isClassDict inst) = extendAvails avails inst avail
2424 | NoSCs <- want_scs = extendAvails avails inst avail
2425 | otherwise = do { traceTc (text "addAvailAndSCs" <+> vcat [ppr inst, ppr deps])
2426 ; avails' <- extendAvails avails inst avail
2427 ; addSCs is_loop avails' inst }
2429 is_loop pred = any (`tcEqType` mkPredTy pred) dep_tys
2430 -- Note: this compares by *type*, not by Unique
2431 deps = findAllDeps (unitVarSet (instToId inst)) avail
2432 dep_tys = map idType (varSetElems deps)
2434 findAllDeps :: IdSet -> AvailHow -> IdSet
2435 -- Find all the Insts that this one depends on
2436 -- See Note [SUPERCLASS-LOOP 2]
2437 -- Watch out, though. Since the avails may contain loops
2438 -- (see Note [RECURSIVE DICTIONARIES]), so we need to track the ones we've seen so far
2439 findAllDeps so_far (Rhs _ kids) = foldl find_all so_far kids
2440 findAllDeps so_far other = so_far
2442 find_all :: IdSet -> Inst -> IdSet
2444 | kid_id `elemVarSet` so_far = so_far
2445 | Just avail <- findAvail avails kid = findAllDeps so_far' avail
2446 | otherwise = so_far'
2448 so_far' = extendVarSet so_far kid_id -- Add the new kid to so_far
2449 kid_id = instToId kid
2451 addSCs :: (TcPredType -> Bool) -> Avails -> Inst -> TcM Avails
2452 -- Add all the superclasses of the Inst to Avails
2453 -- The first param says "dont do this because the original thing
2454 -- depends on this one, so you'd build a loop"
2455 -- Invariant: the Inst is already in Avails.
2457 addSCs is_loop avails dict
2458 = ASSERT( isDict dict )
2459 do { sc_dicts <- newDictBndrs (instLoc dict) sc_theta'
2460 ; foldlM add_sc avails (zipEqual "add_scs" sc_dicts sc_sels) }
2462 (clas, tys) = getDictClassTys dict
2463 (tyvars, sc_theta, sc_sels, _) = classBigSig clas
2464 sc_theta' = substTheta (zipTopTvSubst tyvars tys) sc_theta
2466 add_sc avails (sc_dict, sc_sel)
2467 | is_loop (dictPred sc_dict) = return avails -- See Note [SUPERCLASS-LOOP 2]
2468 | is_given sc_dict = return avails
2469 | otherwise = do { avails' <- extendAvails avails sc_dict (Rhs sc_sel_rhs [dict])
2470 ; addSCs is_loop avails' sc_dict }
2472 sc_sel_rhs = L (instSpan dict) (HsWrap co_fn (HsVar sc_sel))
2473 co_fn = WpApp (instToId dict) <.> mkWpTyApps tys
2475 is_given :: Inst -> Bool
2476 is_given sc_dict = case findAvail avails sc_dict of
2477 Just (Given _) -> True -- Given is cheaper than superclass selection
2481 %************************************************************************
2483 \section{tcSimplifyTop: defaulting}
2485 %************************************************************************
2488 @tcSimplifyTop@ is called once per module to simplify all the constant
2489 and ambiguous Insts.
2491 We need to be careful of one case. Suppose we have
2493 instance Num a => Num (Foo a b) where ...
2495 and @tcSimplifyTop@ is given a constraint (Num (Foo x y)). Then it'll simplify
2496 to (Num x), and default x to Int. But what about y??
2498 It's OK: the final zonking stage should zap y to (), which is fine.
2502 tcSimplifyTop, tcSimplifyInteractive :: [Inst] -> TcM TcDictBinds
2503 tcSimplifyTop wanteds
2504 = tc_simplify_top doc False wanteds
2506 doc = text "tcSimplifyTop"
2508 tcSimplifyInteractive wanteds
2509 = tc_simplify_top doc True wanteds
2511 doc = text "tcSimplifyInteractive"
2513 -- The TcLclEnv should be valid here, solely to improve
2514 -- error message generation for the monomorphism restriction
2515 tc_simplify_top doc interactive wanteds
2516 = do { dflags <- getDOpts
2517 ; wanteds <- zonkInsts wanteds
2518 ; mapM_ zonkTopTyVar (varSetElems (tyVarsOfInsts wanteds))
2520 ; traceTc (text "tc_simplify_top 0: " <+> ppr wanteds)
2521 ; (irreds1, binds1) <- tryHardCheckLoop doc1 wanteds
2522 ; traceTc (text "tc_simplify_top 1: " <+> ppr irreds1)
2523 ; (irreds2, binds2) <- approximateImplications doc2 (\d -> True) irreds1
2524 ; traceTc (text "tc_simplify_top 2: " <+> ppr irreds2)
2526 -- Use the defaulting rules to do extra unification
2527 -- NB: irreds2 are already zonked
2528 ; (irreds3, binds3) <- disambiguate doc3 interactive dflags irreds2
2530 -- Deal with implicit parameters
2531 ; let (bad_ips, non_ips) = partition isIPDict irreds3
2532 (ambigs, others) = partition isTyVarDict non_ips
2534 ; topIPErrs bad_ips -- Can arise from f :: Int -> Int
2536 ; addNoInstanceErrs others
2537 ; addTopAmbigErrs ambigs
2539 ; return (binds1 `unionBags` binds2 `unionBags` binds3) }
2541 doc1 = doc <+> ptext SLIT("(first round)")
2542 doc2 = doc <+> ptext SLIT("(approximate)")
2543 doc3 = doc <+> ptext SLIT("(disambiguate)")
2546 If a dictionary constrains a type variable which is
2547 * not mentioned in the environment
2548 * and not mentioned in the type of the expression
2549 then it is ambiguous. No further information will arise to instantiate
2550 the type variable; nor will it be generalised and turned into an extra
2551 parameter to a function.
2553 It is an error for this to occur, except that Haskell provided for
2554 certain rules to be applied in the special case of numeric types.
2556 * at least one of its classes is a numeric class, and
2557 * all of its classes are numeric or standard
2558 then the type variable can be defaulted to the first type in the
2559 default-type list which is an instance of all the offending classes.
2561 So here is the function which does the work. It takes the ambiguous
2562 dictionaries and either resolves them (producing bindings) or
2563 complains. It works by splitting the dictionary list by type
2564 variable, and using @disambigOne@ to do the real business.
2566 @disambigOne@ assumes that its arguments dictionaries constrain all
2567 the same type variable.
2569 ADR Comment 20/6/94: I've changed the @CReturnable@ case to default to
2570 @()@ instead of @Int@. I reckon this is the Right Thing to do since
2571 the most common use of defaulting is code like:
2573 _ccall_ foo `seqPrimIO` bar
2575 Since we're not using the result of @foo@, the result if (presumably)
2579 disambiguate :: SDoc -> Bool -> DynFlags -> [Inst] -> TcM ([Inst], TcDictBinds)
2580 -- Just does unification to fix the default types
2581 -- The Insts are assumed to be pre-zonked
2582 disambiguate doc interactive dflags insts
2584 = return (insts, emptyBag)
2586 | null defaultable_groups
2587 = do { traceTc (text "disambigutate, no defaultable groups" <+> vcat [ppr unaries, ppr insts, ppr bad_tvs, ppr defaultable_groups])
2588 ; return (insts, emptyBag) }
2591 = do { -- Figure out what default types to use
2592 default_tys <- getDefaultTys extended_defaulting ovl_strings
2594 ; traceTc (text "disambiguate1" <+> vcat [ppr insts, ppr unaries, ppr bad_tvs, ppr defaultable_groups])
2595 ; mapM_ (disambigGroup default_tys) defaultable_groups
2597 -- disambigGroup does unification, hence try again
2598 ; tryHardCheckLoop doc insts }
2601 extended_defaulting = interactive || dopt Opt_ExtendedDefaultRules dflags
2602 ovl_strings = dopt Opt_OverloadedStrings dflags
2604 unaries :: [(Inst, Class, TcTyVar)] -- (C tv) constraints
2605 bad_tvs :: TcTyVarSet -- Tyvars mentioned by *other* constraints
2606 (unaries, bad_tvs_s) = partitionWith find_unary insts
2607 bad_tvs = unionVarSets bad_tvs_s
2609 -- Finds unary type-class constraints
2610 find_unary d@(Dict {tci_pred = ClassP cls [ty]})
2611 | Just tv <- tcGetTyVar_maybe ty = Left (d,cls,tv)
2612 find_unary inst = Right (tyVarsOfInst inst)
2614 -- Group by type variable
2615 defaultable_groups :: [[(Inst,Class,TcTyVar)]]
2616 defaultable_groups = filter defaultable_group (equivClasses cmp_tv unaries)
2617 cmp_tv (_,_,tv1) (_,_,tv2) = tv1 `compare` tv2
2619 defaultable_group :: [(Inst,Class,TcTyVar)] -> Bool
2620 defaultable_group ds@((_,_,tv):_)
2621 = isTyConableTyVar tv -- Note [Avoiding spurious errors]
2622 && not (tv `elemVarSet` bad_tvs)
2623 && defaultable_classes [c | (_,c,_) <- ds]
2624 defaultable_group [] = panic "defaultable_group"
2626 defaultable_classes clss
2627 | extended_defaulting = any isInteractiveClass clss
2628 | otherwise = all is_std_class clss && (any is_num_class clss)
2630 -- In interactive mode, or with -fextended-default-rules,
2631 -- we default Show a to Show () to avoid graututious errors on "show []"
2632 isInteractiveClass cls
2633 = is_num_class cls || (classKey cls `elem` [showClassKey, eqClassKey, ordClassKey])
2635 is_num_class cls = isNumericClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2636 -- is_num_class adds IsString to the standard numeric classes,
2637 -- when -foverloaded-strings is enabled
2639 is_std_class cls = isStandardClass cls || (ovl_strings && (cls `hasKey` isStringClassKey))
2640 -- Similarly is_std_class
2642 -----------------------
2643 disambigGroup :: [Type] -- The default types
2644 -> [(Inst,Class,TcTyVar)] -- All standard classes of form (C a)
2645 -> TcM () -- Just does unification, to fix the default types
2647 disambigGroup default_tys dicts
2648 = try_default default_tys
2650 (_,_,tyvar) = ASSERT(not (null dicts)) head dicts -- Should be non-empty
2651 classes = [c | (_,c,_) <- dicts]
2653 try_default [] = return ()
2654 try_default (default_ty : default_tys)
2655 = tryTcLIE_ (try_default default_tys) $
2656 do { tcSimplifyDefault [mkClassPred clas [default_ty] | clas <- classes]
2657 -- This may fail; then the tryTcLIE_ kicks in
2658 -- Failure here is caused by there being no type in the
2659 -- default list which can satisfy all the ambiguous classes.
2660 -- For example, if Real a is reqd, but the only type in the
2661 -- default list is Int.
2663 -- After this we can't fail
2664 ; warnDefault dicts default_ty
2665 ; unifyType default_ty (mkTyVarTy tyvar)
2666 ; return () -- TOMDO: do something with the coercion
2670 -----------------------
2671 getDefaultTys :: Bool -> Bool -> TcM [Type]
2672 getDefaultTys extended_deflts ovl_strings
2673 = do { mb_defaults <- getDeclaredDefaultTys
2674 ; case mb_defaults of {
2675 Just tys -> return tys ; -- User-supplied defaults
2678 -- No use-supplied default
2679 -- Use [Integer, Double], plus modifications
2680 { integer_ty <- tcMetaTy integerTyConName
2681 ; checkWiredInTyCon doubleTyCon
2682 ; string_ty <- tcMetaTy stringTyConName
2683 ; return (opt_deflt extended_deflts unitTy
2684 -- Note [Default unitTy]
2686 [integer_ty,doubleTy]
2688 opt_deflt ovl_strings string_ty) } } }
2690 opt_deflt True ty = [ty]
2691 opt_deflt False ty = []
2694 Note [Default unitTy]
2695 ~~~~~~~~~~~~~~~~~~~~~
2696 In interative mode (or with -fextended-default-rules) we add () as the first type we
2697 try when defaulting. This has very little real impact, except in the following case.
2699 Text.Printf.printf "hello"
2700 This has type (forall a. IO a); it prints "hello", and returns 'undefined'. We don't
2701 want the GHCi repl loop to try to print that 'undefined'. The neatest thing is to
2702 default the 'a' to (), rather than to Integer (which is what would otherwise happen;
2703 and then GHCi doesn't attempt to print the (). So in interactive mode, we add
2704 () to the list of defaulting types. See Trac #1200.
2706 Note [Avoiding spurious errors]
2707 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2708 When doing the unification for defaulting, we check for skolem
2709 type variables, and simply don't default them. For example:
2710 f = (*) -- Monomorphic
2711 g :: Num a => a -> a
2713 Here, we get a complaint when checking the type signature for g,
2714 that g isn't polymorphic enough; but then we get another one when
2715 dealing with the (Num a) context arising from f's definition;
2716 we try to unify a with Int (to default it), but find that it's
2717 already been unified with the rigid variable from g's type sig
2720 %************************************************************************
2722 \subsection[simple]{@Simple@ versions}
2724 %************************************************************************
2726 Much simpler versions when there are no bindings to make!
2728 @tcSimplifyThetas@ simplifies class-type constraints formed by
2729 @deriving@ declarations and when specialising instances. We are
2730 only interested in the simplified bunch of class/type constraints.
2732 It simplifies to constraints of the form (C a b c) where
2733 a,b,c are type variables. This is required for the context of
2734 instance declarations.
2737 tcSimplifyDeriv :: InstOrigin
2739 -> ThetaType -- Wanted
2740 -> TcM ThetaType -- Needed
2741 -- Given instance (wanted) => C inst_ty
2742 -- Simplify 'wanted' as much as possible
2743 -- The inst_ty is needed only for the termination check
2745 tcSimplifyDeriv orig tyvars theta
2746 = do { (tvs, _, tenv) <- tcInstTyVars tyvars
2747 -- The main loop may do unification, and that may crash if
2748 -- it doesn't see a TcTyVar, so we have to instantiate. Sigh
2749 -- ToDo: what if two of them do get unified?
2750 ; wanteds <- newDictBndrsO orig (substTheta tenv theta)
2751 ; (irreds, _) <- tryHardCheckLoop doc wanteds
2753 ; let (tv_dicts, others) = partition isTyVarDict irreds
2754 ; addNoInstanceErrs others
2756 ; let rev_env = zipTopTvSubst tvs (mkTyVarTys tyvars)
2757 simpl_theta = substTheta rev_env (map dictPred tv_dicts)
2758 -- This reverse-mapping is a pain, but the result
2759 -- should mention the original TyVars not TcTyVars
2761 ; return simpl_theta }
2763 doc = ptext SLIT("deriving classes for a data type")
2766 Note [Exotic derived instance contexts]
2767 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2769 data T a b c = MkT (Foo a b c) deriving( Eq )
2770 instance (C Int a, Eq b, Eq c) => Eq (Foo a b c)
2772 Notice that this instance (just) satisfies the Paterson termination
2773 conditions. Then we *could* derive an instance decl like this:
2775 instance (C Int a, Eq b, Eq c) => Eq (T a b c)
2777 even though there is no instance for (C Int a), because there just
2778 *might* be an instance for, say, (C Int Bool) at a site where we
2779 need the equality instance for T's.
2781 However, this seems pretty exotic, and it's quite tricky to allow
2782 this, and yet give sensible error messages in the (much more common)
2783 case where we really want that instance decl for C.
2785 So for now we simply require that the derived instance context
2786 should have only type-variable constraints.
2788 Here is another example:
2789 data Fix f = In (f (Fix f)) deriving( Eq )
2790 Here, if we are prepared to allow -fallow-undecidable-instances we
2791 could derive the instance
2792 instance Eq (f (Fix f)) => Eq (Fix f)
2793 but this is so delicate that I don't think it should happen inside
2794 'deriving'. If you want this, write it yourself!
2796 NB: if you want to lift this condition, make sure you still meet the
2797 termination conditions! If not, the deriving mechanism generates
2798 larger and larger constraints. Example:
2800 data Seq a = Cons a (Seq (Succ a)) | Nil deriving Show
2802 Note the lack of a Show instance for Succ. First we'll generate
2803 instance (Show (Succ a), Show a) => Show (Seq a)
2805 instance (Show (Succ (Succ a)), Show (Succ a), Show a) => Show (Seq a)
2806 and so on. Instead we want to complain of no instance for (Show (Succ a)).
2809 @tcSimplifyDefault@ just checks class-type constraints, essentially;
2810 used with \tr{default} declarations. We are only interested in
2811 whether it worked or not.
2814 tcSimplifyDefault :: ThetaType -- Wanted; has no type variables in it
2817 tcSimplifyDefault theta
2818 = newDictBndrsO DefaultOrigin theta `thenM` \ wanteds ->
2819 tryHardCheckLoop doc wanteds `thenM` \ (irreds, _) ->
2820 addNoInstanceErrs irreds `thenM_`
2826 doc = ptext SLIT("default declaration")
2830 %************************************************************************
2832 \section{Errors and contexts}
2834 %************************************************************************
2836 ToDo: for these error messages, should we note the location as coming
2837 from the insts, or just whatever seems to be around in the monad just
2841 groupErrs :: ([Inst] -> TcM ()) -- Deal with one group
2842 -> [Inst] -- The offending Insts
2844 -- Group together insts with the same origin
2845 -- We want to report them together in error messages
2847 groupErrs report_err []
2849 groupErrs report_err (inst:insts)
2850 = do_one (inst:friends) `thenM_`
2851 groupErrs report_err others
2854 -- (It may seem a bit crude to compare the error messages,
2855 -- but it makes sure that we combine just what the user sees,
2856 -- and it avoids need equality on InstLocs.)
2857 (friends, others) = partition is_friend insts
2858 loc_msg = showSDoc (pprInstLoc (instLoc inst))
2859 is_friend friend = showSDoc (pprInstLoc (instLoc friend)) == loc_msg
2860 do_one insts = addInstCtxt (instLoc (head insts)) (report_err insts)
2861 -- Add location and context information derived from the Insts
2863 -- Add the "arising from..." part to a message about bunch of dicts
2864 addInstLoc :: [Inst] -> Message -> Message
2865 addInstLoc insts msg = msg $$ nest 2 (pprInstArising (head insts))
2867 addTopIPErrs :: [Name] -> [Inst] -> TcM ()
2868 addTopIPErrs bndrs []
2870 addTopIPErrs bndrs ips
2871 = do { dflags <- getDOpts
2872 ; addErrTcM (tidy_env, mk_msg dflags tidy_ips) }
2874 (tidy_env, tidy_ips) = tidyInsts ips
2876 = vcat [sep [ptext SLIT("Implicit parameters escape from"),
2877 nest 2 (ptext SLIT("the monomorphic top-level binding")
2878 <> plural bndrs <+> ptext SLIT("of")
2879 <+> pprBinders bndrs <> colon)],
2880 nest 2 (vcat (map ppr_ip ips)),
2881 monomorphism_fix dflags]
2882 ppr_ip ip = pprPred (dictPred ip) <+> pprInstArising ip
2884 topIPErrs :: [Inst] -> TcM ()
2886 = groupErrs report tidy_dicts
2888 (tidy_env, tidy_dicts) = tidyInsts dicts
2889 report dicts = addErrTcM (tidy_env, mk_msg dicts)
2890 mk_msg dicts = addInstLoc dicts (ptext SLIT("Unbound implicit parameter") <>
2891 plural tidy_dicts <+> pprDictsTheta tidy_dicts)
2893 addNoInstanceErrs :: [Inst] -- Wanted (can include implications)
2895 addNoInstanceErrs insts
2896 = do { let (tidy_env, tidy_insts) = tidyInsts insts
2897 ; reportNoInstances tidy_env Nothing tidy_insts }
2901 -> Maybe (InstLoc, [Inst]) -- Context
2902 -- Nothing => top level
2903 -- Just (d,g) => d describes the construct
2905 -> [Inst] -- What is wanted (can include implications)
2908 reportNoInstances tidy_env mb_what insts
2909 = groupErrs (report_no_instances tidy_env mb_what) insts
2911 report_no_instances tidy_env mb_what insts
2912 = do { inst_envs <- tcGetInstEnvs
2913 ; let (implics, insts1) = partition isImplicInst insts
2914 (insts2, overlaps) = partitionWith (check_overlap inst_envs) insts1
2915 (eqInsts, insts3) = partition isEqInst insts2
2916 ; traceTc (text "reportNoInstances" <+> vcat
2917 [ppr implics, ppr insts1, ppr insts2])
2918 ; mapM_ complain_implic implics
2919 ; mapM_ (\doc -> addErrTcM (tidy_env, doc)) overlaps
2920 ; groupErrs complain_no_inst insts3
2921 ; mapM_ complain_eq eqInsts
2924 complain_no_inst insts = addErrTcM (tidy_env, mk_no_inst_err insts)
2926 complain_implic inst -- Recurse!
2927 = reportNoInstances tidy_env
2928 (Just (tci_loc inst, tci_given inst))
2931 complain_eq EqInst {tci_left = lty, tci_right = rty,
2932 tci_loc = InstLoc _ _ ctxt}
2933 = do { (env, msg) <- misMatchMsg lty rty
2935 failWithTcM (env, msg)
2938 check_overlap :: (InstEnv,InstEnv) -> Inst -> Either Inst SDoc
2939 -- Right msg => overlap message
2940 -- Left inst => no instance
2941 check_overlap inst_envs wanted
2942 | not (isClassDict wanted) = Left wanted
2944 = case lookupInstEnv inst_envs clas tys of
2945 -- The case of exactly one match and no unifiers means a
2946 -- successful lookup. That can't happen here, because dicts
2947 -- only end up here if they didn't match in Inst.lookupInst
2949 ([m],[]) -> pprPanic "reportNoInstance" (ppr wanted)
2951 ([], _) -> Left wanted -- No match
2952 res -> Right (mk_overlap_msg wanted res)
2954 (clas,tys) = getDictClassTys wanted
2956 mk_overlap_msg dict (matches, unifiers)
2957 = ASSERT( not (null matches) )
2958 vcat [ addInstLoc [dict] ((ptext SLIT("Overlapping instances for")
2959 <+> pprPred (dictPred dict))),
2960 sep [ptext SLIT("Matching instances") <> colon,
2961 nest 2 (vcat [pprInstances ispecs, pprInstances unifiers])],
2962 if not (isSingleton matches)
2963 then -- Two or more matches
2965 else -- One match, plus some unifiers
2966 ASSERT( not (null unifiers) )
2967 parens (vcat [ptext SLIT("The choice depends on the instantiation of") <+>
2968 quotes (pprWithCommas ppr (varSetElems (tyVarsOfInst dict))),
2969 ptext SLIT("To pick the first instance above, use -fallow-incoherent-instances"),
2970 ptext SLIT("when compiling the other instance declarations")])]
2972 ispecs = [ispec | (ispec, _) <- matches]
2974 mk_no_inst_err insts
2975 | null insts = empty
2977 | Just (loc, givens) <- mb_what, -- Nested (type signatures, instance decls)
2978 not (isEmptyVarSet (tyVarsOfInsts insts))
2979 = vcat [ addInstLoc insts $
2980 sep [ ptext SLIT("Could not deduce") <+> pprDictsTheta insts
2981 , nest 2 $ ptext SLIT("from the context") <+> pprDictsTheta givens]
2982 , show_fixes (fix1 loc : fixes2) ]
2984 | otherwise -- Top level
2985 = vcat [ addInstLoc insts $
2986 ptext SLIT("No instance") <> plural insts
2987 <+> ptext SLIT("for") <+> pprDictsTheta insts
2988 , show_fixes fixes2 ]
2991 fix1 loc = sep [ ptext SLIT("add") <+> pprDictsTheta insts
2992 <+> ptext SLIT("to the context of"),
2993 nest 2 (ppr (instLocOrigin loc)) ]
2994 -- I'm not sure it helps to add the location
2995 -- nest 2 (ptext SLIT("at") <+> ppr (instLocSpan loc)) ]
2997 fixes2 | null instance_dicts = []
2998 | otherwise = [sep [ptext SLIT("add an instance declaration for"),
2999 pprDictsTheta instance_dicts]]
3000 instance_dicts = [d | d <- insts, isClassDict d, not (isTyVarDict d)]
3001 -- Insts for which it is worth suggesting an adding an instance declaration
3002 -- Exclude implicit parameters, and tyvar dicts
3004 show_fixes :: [SDoc] -> SDoc
3005 show_fixes [] = empty
3006 show_fixes (f:fs) = sep [ptext SLIT("Possible fix:"),
3007 nest 2 (vcat (f : map (ptext SLIT("or") <+>) fs))]
3009 addTopAmbigErrs dicts
3010 -- Divide into groups that share a common set of ambiguous tyvars
3011 = ifErrsM (return ()) $ -- Only report ambiguity if no other errors happened
3012 -- See Note [Avoiding spurious errors]
3013 mapM_ report (equivClasses cmp [(d, tvs_of d) | d <- tidy_dicts])
3015 (tidy_env, tidy_dicts) = tidyInsts dicts
3017 tvs_of :: Inst -> [TcTyVar]
3018 tvs_of d = varSetElems (tyVarsOfInst d)
3019 cmp (_,tvs1) (_,tvs2) = tvs1 `compare` tvs2
3021 report :: [(Inst,[TcTyVar])] -> TcM ()
3022 report pairs@((inst,tvs) : _) -- The pairs share a common set of ambiguous tyvars
3023 = mkMonomorphismMsg tidy_env tvs `thenM` \ (tidy_env, mono_msg) ->
3024 setSrcSpan (instSpan inst) $
3025 -- the location of the first one will do for the err message
3026 addErrTcM (tidy_env, msg $$ mono_msg)
3028 dicts = map fst pairs
3029 msg = sep [text "Ambiguous type variable" <> plural tvs <+>
3030 pprQuotedList tvs <+> in_msg,
3031 nest 2 (pprDictsInFull dicts)]
3032 in_msg = text "in the constraint" <> plural dicts <> colon
3033 report [] = panic "addTopAmbigErrs"
3036 mkMonomorphismMsg :: TidyEnv -> [TcTyVar] -> TcM (TidyEnv, Message)
3037 -- There's an error with these Insts; if they have free type variables
3038 -- it's probably caused by the monomorphism restriction.
3039 -- Try to identify the offending variable
3040 -- ASSUMPTION: the Insts are fully zonked
3041 mkMonomorphismMsg tidy_env inst_tvs
3042 = do { dflags <- getDOpts
3043 ; (tidy_env, docs) <- findGlobals (mkVarSet inst_tvs) tidy_env
3044 ; return (tidy_env, mk_msg dflags docs) }
3046 mk_msg _ _ | any isRuntimeUnk inst_tvs
3047 = vcat [ptext SLIT("Cannot resolve unknown runtime types:") <+>
3048 (pprWithCommas ppr inst_tvs),
3049 ptext SLIT("Use :print or :force to determine these types")]
3050 mk_msg _ [] = ptext SLIT("Probable fix: add a type signature that fixes these type variable(s)")
3051 -- This happens in things like
3052 -- f x = show (read "foo")
3053 -- where monomorphism doesn't play any role
3055 = vcat [ptext SLIT("Possible cause: the monomorphism restriction applied to the following:"),
3057 monomorphism_fix dflags]
3059 isRuntimeUnk :: TcTyVar -> Bool
3060 isRuntimeUnk x | SkolemTv RuntimeUnkSkol <- tcTyVarDetails x = True
3063 monomorphism_fix :: DynFlags -> SDoc
3064 monomorphism_fix dflags
3065 = ptext SLIT("Probable fix:") <+> vcat
3066 [ptext SLIT("give these definition(s) an explicit type signature"),
3067 if dopt Opt_MonomorphismRestriction dflags
3068 then ptext SLIT("or use -fno-monomorphism-restriction")
3069 else empty] -- Only suggest adding "-fno-monomorphism-restriction"
3070 -- if it is not already set!
3072 warnDefault ups default_ty
3073 = doptM Opt_WarnTypeDefaults `thenM` \ warn_flag ->
3074 addInstCtxt (instLoc (head (dicts))) (warnTc warn_flag warn_msg)
3076 dicts = [d | (d,_,_) <- ups]
3079 (_, tidy_dicts) = tidyInsts dicts
3080 warn_msg = vcat [ptext SLIT("Defaulting the following constraint(s) to type") <+>
3081 quotes (ppr default_ty),
3082 pprDictsInFull tidy_dicts]
3084 reduceDepthErr n stack
3085 = vcat [ptext SLIT("Context reduction stack overflow; size =") <+> int n,
3086 ptext SLIT("Use -fcontext-stack=N to increase stack size to N"),
3087 nest 4 (pprStack stack)]
3089 pprStack stack = vcat (map pprInstInFull stack)
3091 -----------------------
3092 misMatchMsg :: TcType -> TcType -> TcM (TidyEnv, SDoc)
3093 -- Generate the message when two types fail to match,
3094 -- going to some trouble to make it helpful.
3095 -- The argument order is: actual type, expected type
3096 misMatchMsg ty_act ty_exp
3097 = do { env0 <- tcInitTidyEnv
3098 ; (env1, pp_exp, extra_exp) <- ppr_ty env0 ty_exp ty_act
3099 ; (env2, pp_act, extra_act) <- ppr_ty env1 ty_act ty_exp
3101 sep [sep [ptext SLIT("Couldn't match expected type") <+> pp_exp,
3103 ptext SLIT("against inferred type") <+> pp_act],
3104 nest 2 (extra_exp $$ extra_act)]) }
3106 ppr_ty :: TidyEnv -> TcType -> TcType -> TcM (TidyEnv, SDoc, SDoc)
3107 ppr_ty env ty other_ty
3108 = do { ty' <- zonkTcType ty
3109 ; let (env1, tidy_ty) = tidyOpenType env ty'
3110 ; (env2, extra) <- ppr_extra env1 tidy_ty other_ty
3111 ; return (env2, quotes (ppr tidy_ty), extra) }
3113 -- (ppr_extra env ty other_ty) shows extra info about 'ty'
3114 ppr_extra env (TyVarTy tv) other_ty
3115 | isSkolemTyVar tv || isSigTyVar tv
3116 = return (env1, pprSkolTvBinding tv1)
3118 (env1, tv1) = tidySkolemTyVar env tv
3120 ppr_extra env (TyConApp tc1 _) (TyConApp tc2 _)
3121 | getOccName tc1 == getOccName tc2
3122 = -- This case helps with messages that would otherwise say
3123 -- Could not match 'T' does not match 'M.T'
3124 -- which is not helpful
3125 do { this_mod <- getModule
3126 ; return (env, quotes (ppr tc1) <+> ptext SLIT("is defined") <+> mk_mod this_mod) }
3128 tc_mod = nameModule (getName tc1)
3129 tc_pkg = modulePackageId tc_mod
3130 tc2_pkg = modulePackageId (nameModule (getName tc2))
3132 | tc_mod == this_mod = ptext SLIT("in this module")
3134 | not home_pkg && tc2_pkg /= tc_pkg = pp_pkg
3135 -- Suppress the module name if (a) it's from another package
3136 -- (b) other_ty isn't from that same package
3138 | otherwise = ptext SLIT("in module") <+> quotes (ppr tc_mod) <+> pp_pkg
3140 home_pkg = tc_pkg == modulePackageId this_mod
3141 pp_pkg | home_pkg = empty
3142 | otherwise = ptext SLIT("in package") <+> quotes (ppr tc_pkg)
3144 ppr_extra env ty other_ty = return (env, empty) -- Normal case