Prefer #if defined to #ifdef
[ghc.git] / compiler / typecheck / TcSMonad.hs
1 {-# LANGUAGE CPP, TypeFamilies #-}
2
3 -- Type definitions for the constraint solver
4 module TcSMonad (
5
6 -- The work list
7 WorkList(..), isEmptyWorkList, emptyWorkList,
8 extendWorkListNonEq, extendWorkListCt, extendWorkListDerived,
9 extendWorkListCts, extendWorkListEq, extendWorkListFunEq,
10 appendWorkList,
11 selectNextWorkItem,
12 workListSize, workListWantedCount,
13 getWorkList, updWorkListTcS,
14
15 -- The TcS monad
16 TcS, runTcS, runTcSDeriveds, runTcSWithEvBinds,
17 failTcS, warnTcS, addErrTcS,
18 runTcSEqualities,
19 nestTcS, nestImplicTcS, setEvBindsTcS,
20
21 runTcPluginTcS, addUsedGRE, addUsedGREs, deferTcSForAllEq,
22
23 -- Tracing etc
24 panicTcS, traceTcS,
25 traceFireTcS, bumpStepCountTcS, csTraceTcS,
26 wrapErrTcS, wrapWarnTcS,
27
28 -- Evidence creation and transformation
29 MaybeNew(..), freshGoals, isFresh, getEvTerm,
30
31 newTcEvBinds,
32 newWantedEq, emitNewWantedEq,
33 newWanted, newWantedEvVar, newWantedNC, newWantedEvVarNC, newDerivedNC,
34 newBoundEvVarId,
35 unifyTyVar, unflattenFmv, reportUnifications,
36 setEvBind, setWantedEq, setEqIfWanted,
37 setWantedEvTerm, setWantedEvBind, setEvBindIfWanted,
38 newEvVar, newGivenEvVar, newGivenEvVars,
39 emitNewDerived, emitNewDeriveds, emitNewDerivedEq,
40 checkReductionDepth,
41
42 getInstEnvs, getFamInstEnvs, -- Getting the environments
43 getTopEnv, getGblEnv, getLclEnv,
44 getTcEvBindsVar, getTcLevel,
45 getTcEvBindsAndTCVs, getTcEvBindsMap,
46 tcLookupClass,
47 tcLookupId,
48
49 -- Inerts
50 InertSet(..), InertCans(..),
51 updInertTcS, updInertCans, updInertDicts, updInertIrreds,
52 getNoGivenEqs, setInertCans,
53 getInertEqs, getInertCans, getInertGivens,
54 getInertInsols,
55 emptyInert, getTcSInerts, setTcSInerts,
56 matchableGivens, prohibitedSuperClassSolve,
57 getUnsolvedInerts,
58 removeInertCts, getPendingScDicts,
59 addInertCan, addInertEq, insertFunEq,
60 emitInsoluble, emitWorkNC, emitWork,
61 isImprovable,
62
63 -- The Model
64 kickOutAfterUnification,
65
66 -- Inert Safe Haskell safe-overlap failures
67 addInertSafehask, insertSafeOverlapFailureTcS, updInertSafehask,
68 getSafeOverlapFailures,
69
70 -- Inert CDictCans
71 DictMap, emptyDictMap, lookupInertDict, findDictsByClass, addDict,
72 addDictsByClass, delDict, foldDicts, filterDicts, findDict,
73
74 -- Inert CTyEqCans
75 EqualCtList, findTyEqs, foldTyEqs, isInInertEqs,
76 lookupFlattenTyVar, lookupInertTyVar,
77
78 -- Inert solved dictionaries
79 addSolvedDict, lookupSolvedDict,
80
81 -- Irreds
82 foldIrreds,
83
84 -- The flattening cache
85 lookupFlatCache, extendFlatCache, newFlattenSkolem, -- Flatten skolems
86
87 -- Inert CFunEqCans
88 updInertFunEqs, findFunEq,
89 findFunEqsByTyCon,
90
91 instDFunType, -- Instantiation
92
93 -- MetaTyVars
94 newFlexiTcSTy, instFlexi, instFlexiX,
95 cloneMetaTyVar, demoteUnfilledFmv,
96 tcInstType,
97
98 TcLevel, isTouchableMetaTyVarTcS,
99 isFilledMetaTyVar_maybe, isFilledMetaTyVar,
100 zonkTyCoVarsAndFV, zonkTcType, zonkTcTypes, zonkTcTyVar, zonkCo,
101 zonkTyCoVarsAndFVList,
102 zonkSimples, zonkWC,
103
104 -- References
105 newTcRef, readTcRef, updTcRef,
106
107 -- Misc
108 getDefaultInfo, getDynFlags, getGlobalRdrEnvTcS,
109 matchFam, matchFamTcM,
110 checkWellStagedDFun,
111 pprEq -- Smaller utils, re-exported from TcM
112 -- TODO (DV): these are only really used in the
113 -- instance matcher in TcSimplify. I am wondering
114 -- if the whole instance matcher simply belongs
115 -- here
116 ) where
117
118 #include "HsVersions.h"
119
120 import HscTypes
121
122 import qualified Inst as TcM
123 import InstEnv
124 import FamInst
125 import FamInstEnv
126
127 import qualified TcRnMonad as TcM
128 import qualified TcMType as TcM
129 import qualified TcEnv as TcM
130 ( checkWellStaged, topIdLvl, tcGetDefaultTys, tcLookupClass, tcLookupId )
131 import PrelNames( heqTyConKey, eqTyConKey )
132 import Kind
133 import TcType
134 import DynFlags
135 import Type
136 import Coercion
137 import Unify
138
139 import TcEvidence
140 import Class
141 import TyCon
142 import TcErrors ( solverDepthErrorTcS )
143
144 import Name
145 import RdrName ( GlobalRdrEnv, GlobalRdrElt )
146 import qualified RnEnv as TcM
147 import Var
148 import VarEnv
149 import VarSet
150 import Outputable
151 import Bag
152 import UniqSupply
153 import Util
154 import TcRnTypes
155
156 import Unique
157 import UniqFM
158 import UniqDFM
159 import Maybes
160
161 import TrieMap
162 import Control.Monad
163 #if __GLASGOW_HASKELL__ > 710
164 import qualified Control.Monad.Fail as MonadFail
165 #endif
166 import MonadUtils
167 import Data.IORef
168 import Data.List ( foldl', partition )
169
170 #if defined(DEBUG)
171 import Digraph
172 import UniqSet
173 #endif
174
175 {-
176 ************************************************************************
177 * *
178 * Worklists *
179 * Canonical and non-canonical constraints that the simplifier has to *
180 * work on. Including their simplification depths. *
181 * *
182 * *
183 ************************************************************************
184
185 Note [WorkList priorities]
186 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
187 A WorkList contains canonical and non-canonical items (of all flavors).
188 Notice that each Ct now has a simplification depth. We may
189 consider using this depth for prioritization as well in the future.
190
191 As a simple form of priority queue, our worklist separates out
192 equalities (wl_eqs) from the rest of the canonical constraints,
193 so that it's easier to deal with them first, but the separation
194 is not strictly necessary. Notice that non-canonical constraints
195 are also parts of the worklist.
196
197 Note [Process derived items last]
198 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
199 We can often solve all goals without processing *any* derived constraints.
200 The derived constraints are just there to help us if we get stuck. So
201 we keep them in a separate list.
202
203 Note [Prioritise class equalities]
204 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
205 We prioritise equalities in the solver (see selectWorkItem). But class
206 constraints like (a ~ b) and (a ~~ b) are actually equalities too;
207 see Note [The equality types story] in TysPrim.
208
209 Failing to prioritise these is inefficient (more kick-outs etc).
210 But, worse, it can prevent us spotting a "recursive knot" among
211 Wanted constraints. See comment:10 of Trac #12734 for a worked-out
212 example.
213
214 So we arrange to put these particular class constraints in the wl_eqs.
215
216 NB: since we do not currently apply the substitution to the
217 inert_solved_dicts, the knot-tying still seems a bit fragile.
218 But this makes it better.
219 -}
220
221 -- See Note [WorkList priorities]
222 data WorkList
223 = WL { wl_eqs :: [Ct] -- Both equality constraints and their
224 -- class-level variants (a~b) and (a~~b);
225 -- See Note [Prioritise class equalities]
226
227 , wl_funeqs :: [Ct] -- LIFO stack of goals
228
229 , wl_rest :: [Ct]
230
231 , wl_deriv :: [CtEvidence] -- Implicitly non-canonical
232 -- See Note [Process derived items last]
233
234 , wl_implics :: Bag Implication -- See Note [Residual implications]
235 }
236
237 appendWorkList :: WorkList -> WorkList -> WorkList
238 appendWorkList
239 (WL { wl_eqs = eqs1, wl_funeqs = funeqs1, wl_rest = rest1
240 , wl_deriv = ders1, wl_implics = implics1 })
241 (WL { wl_eqs = eqs2, wl_funeqs = funeqs2, wl_rest = rest2
242 , wl_deriv = ders2, wl_implics = implics2 })
243 = WL { wl_eqs = eqs1 ++ eqs2
244 , wl_funeqs = funeqs1 ++ funeqs2
245 , wl_rest = rest1 ++ rest2
246 , wl_deriv = ders1 ++ ders2
247 , wl_implics = implics1 `unionBags` implics2 }
248
249 workListSize :: WorkList -> Int
250 workListSize (WL { wl_eqs = eqs, wl_funeqs = funeqs, wl_deriv = ders, wl_rest = rest })
251 = length eqs + length funeqs + length rest + length ders
252
253 workListWantedCount :: WorkList -> Int
254 workListWantedCount (WL { wl_eqs = eqs, wl_rest = rest })
255 = count isWantedCt eqs + count isWantedCt rest
256
257 extendWorkListEq :: Ct -> WorkList -> WorkList
258 extendWorkListEq ct wl = wl { wl_eqs = ct : wl_eqs wl }
259
260 extendWorkListEqs :: [Ct] -> WorkList -> WorkList
261 extendWorkListEqs cts wl = wl { wl_eqs = cts ++ wl_eqs wl }
262
263 extendWorkListFunEq :: Ct -> WorkList -> WorkList
264 extendWorkListFunEq ct wl = wl { wl_funeqs = ct : wl_funeqs wl }
265
266 extendWorkListNonEq :: Ct -> WorkList -> WorkList
267 -- Extension by non equality
268 extendWorkListNonEq ct wl = wl { wl_rest = ct : wl_rest wl }
269
270 extendWorkListDerived :: CtLoc -> CtEvidence -> WorkList -> WorkList
271 extendWorkListDerived loc ev wl
272 | isDroppableDerivedLoc loc = wl { wl_deriv = ev : wl_deriv wl }
273 | otherwise = extendWorkListEq (mkNonCanonical ev) wl
274
275 extendWorkListDeriveds :: CtLoc -> [CtEvidence] -> WorkList -> WorkList
276 extendWorkListDeriveds loc evs wl
277 | isDroppableDerivedLoc loc = wl { wl_deriv = evs ++ wl_deriv wl }
278 | otherwise = extendWorkListEqs (map mkNonCanonical evs) wl
279
280 extendWorkListImplic :: Implication -> WorkList -> WorkList
281 extendWorkListImplic implic wl = wl { wl_implics = implic `consBag` wl_implics wl }
282
283 extendWorkListCt :: Ct -> WorkList -> WorkList
284 -- Agnostic
285 extendWorkListCt ct wl
286 = case classifyPredType (ctPred ct) of
287 EqPred NomEq ty1 _
288 | Just tc <- tcTyConAppTyCon_maybe ty1
289 , isTypeFamilyTyCon tc
290 -> extendWorkListFunEq ct wl
291
292 EqPred {}
293 -> extendWorkListEq ct wl
294
295 ClassPred cls _ -- See Note [Prioritise class equalites]
296 | cls `hasKey` heqTyConKey
297 || cls `hasKey` eqTyConKey
298 -> extendWorkListEq ct wl
299
300 _ -> extendWorkListNonEq ct wl
301
302 extendWorkListCts :: [Ct] -> WorkList -> WorkList
303 -- Agnostic
304 extendWorkListCts cts wl = foldr extendWorkListCt wl cts
305
306 isEmptyWorkList :: WorkList -> Bool
307 isEmptyWorkList (WL { wl_eqs = eqs, wl_funeqs = funeqs
308 , wl_rest = rest, wl_deriv = ders, wl_implics = implics })
309 = null eqs && null rest && null funeqs && isEmptyBag implics && null ders
310
311 emptyWorkList :: WorkList
312 emptyWorkList = WL { wl_eqs = [], wl_rest = []
313 , wl_funeqs = [], wl_deriv = [], wl_implics = emptyBag }
314
315 selectWorkItem :: WorkList -> Maybe (Ct, WorkList)
316 selectWorkItem wl@(WL { wl_eqs = eqs, wl_funeqs = feqs
317 , wl_rest = rest })
318 | ct:cts <- eqs = Just (ct, wl { wl_eqs = cts })
319 | ct:fes <- feqs = Just (ct, wl { wl_funeqs = fes })
320 | ct:cts <- rest = Just (ct, wl { wl_rest = cts })
321 | otherwise = Nothing
322
323 getWorkList :: TcS WorkList
324 getWorkList = do { wl_var <- getTcSWorkListRef
325 ; wrapTcS (TcM.readTcRef wl_var) }
326
327 selectDerivedWorkItem :: WorkList -> Maybe (Ct, WorkList)
328 selectDerivedWorkItem wl@(WL { wl_deriv = ders })
329 | ev:evs <- ders = Just (mkNonCanonical ev, wl { wl_deriv = evs })
330 | otherwise = Nothing
331
332 selectNextWorkItem :: TcS (Maybe Ct)
333 selectNextWorkItem
334 = do { wl_var <- getTcSWorkListRef
335 ; wl <- wrapTcS (TcM.readTcRef wl_var)
336
337 ; let try :: Maybe (Ct,WorkList) -> TcS (Maybe Ct) -> TcS (Maybe Ct)
338 try mb_work do_this_if_fail
339 | Just (ct, new_wl) <- mb_work
340 = do { checkReductionDepth (ctLoc ct) (ctPred ct)
341 ; wrapTcS (TcM.writeTcRef wl_var new_wl)
342 ; return (Just ct) }
343 | otherwise
344 = do_this_if_fail
345
346 ; try (selectWorkItem wl) $
347
348 do { ics <- getInertCans
349 ; if inert_count ics == 0
350 then return Nothing
351 else try (selectDerivedWorkItem wl) (return Nothing) } }
352
353 -- Pretty printing
354 instance Outputable WorkList where
355 ppr (WL { wl_eqs = eqs, wl_funeqs = feqs
356 , wl_rest = rest, wl_implics = implics, wl_deriv = ders })
357 = text "WL" <+> (braces $
358 vcat [ ppUnless (null eqs) $
359 text "Eqs =" <+> vcat (map ppr eqs)
360 , ppUnless (null feqs) $
361 text "Funeqs =" <+> vcat (map ppr feqs)
362 , ppUnless (null rest) $
363 text "Non-eqs =" <+> vcat (map ppr rest)
364 , ppUnless (null ders) $
365 text "Derived =" <+> vcat (map ppr ders)
366 , ppUnless (isEmptyBag implics) $
367 sdocWithPprDebug $ \dbg ->
368 if dbg -- Typically we only want the work list for this level
369 then text "Implics =" <+> vcat (map ppr (bagToList implics))
370 else text "(Implics omitted)"
371 ])
372
373
374 {- *********************************************************************
375 * *
376 InertSet: the inert set
377 * *
378 * *
379 ********************************************************************* -}
380
381 data InertSet
382 = IS { inert_cans :: InertCans
383 -- Canonical Given, Wanted, Derived
384 -- Sometimes called "the inert set"
385
386 , inert_flat_cache :: ExactFunEqMap (TcCoercion, TcType, CtFlavour)
387 -- See Note [Type family equations]
388 -- If F tys :-> (co, rhs, flav),
389 -- then co :: F tys ~ rhs
390 -- flav is [G] or [WD]
391 --
392 -- Just a hash-cons cache for use when flattening only
393 -- These include entirely un-processed goals, so don't use
394 -- them to solve a top-level goal, else you may end up solving
395 -- (w:F ty ~ a) by setting w:=w! We just use the flat-cache
396 -- when allocating a new flatten-skolem.
397 -- Not necessarily inert wrt top-level equations (or inert_cans)
398
399 -- NB: An ExactFunEqMap -- this doesn't match via loose types!
400
401 , inert_solved_dicts :: DictMap CtEvidence
402 -- Of form ev :: C t1 .. tn
403 -- See Note [Solved dictionaries]
404 -- and Note [Do not add superclasses of solved dictionaries]
405 }
406
407 instance Outputable InertSet where
408 ppr is = vcat [ ppr $ inert_cans is
409 , ppUnless (null dicts) $
410 text "Solved dicts" <+> vcat (map ppr dicts) ]
411 where
412 dicts = bagToList (dictsToBag (inert_solved_dicts is))
413
414 emptyInert :: InertSet
415 emptyInert
416 = IS { inert_cans = IC { inert_count = 0
417 , inert_eqs = emptyDVarEnv
418 , inert_dicts = emptyDicts
419 , inert_safehask = emptyDicts
420 , inert_funeqs = emptyFunEqs
421 , inert_irreds = emptyCts
422 , inert_insols = emptyCts }
423 , inert_flat_cache = emptyExactFunEqs
424 , inert_solved_dicts = emptyDictMap }
425
426
427 {- Note [Solved dictionaries]
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
429 When we apply a top-level instance declaration, we add the "solved"
430 dictionary to the inert_solved_dicts. In general, we use it to avoid
431 creating a new EvVar when we have a new goal that we have solved in
432 the past.
433
434 But in particular, we can use it to create *recursive* dictionaries.
435 The simplest, degnerate case is
436 instance C [a] => C [a] where ...
437 If we have
438 [W] d1 :: C [x]
439 then we can apply the instance to get
440 d1 = $dfCList d
441 [W] d2 :: C [x]
442 Now 'd1' goes in inert_solved_dicts, and we can solve d2 directly from d1.
443 d1 = $dfCList d
444 d2 = d1
445
446 See Note [Example of recursive dictionaries]
447 Other notes about solved dictionaries
448
449 * See also Note [Do not add superclasses of solved dictionaries]
450
451 * The inert_solved_dicts field is not rewritten by equalities, so it may
452 get out of date.
453
454 Note [Do not add superclasses of solved dictionaries]
455 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
456 Every member of inert_solved_dicts is the result of applying a dictionary
457 function, NOT of applying superclass selection to anything.
458 Consider
459
460 class Ord a => C a where
461 instance Ord [a] => C [a] where ...
462
463 Suppose we are trying to solve
464 [G] d1 : Ord a
465 [W] d2 : C [a]
466
467 Then we'll use the instance decl to give
468
469 [G] d1 : Ord a Solved: d2 : C [a] = $dfCList d3
470 [W] d3 : Ord [a]
471
472 We must not add d4 : Ord [a] to the 'solved' set (by taking the
473 superclass of d2), otherwise we'll use it to solve d3, without ever
474 using d1, which would be a catastrophe.
475
476 Solution: when extending the solved dictionaries, do not add superclasses.
477 That's why each element of the inert_solved_dicts is the result of applying
478 a dictionary function.
479
480 Note [Example of recursive dictionaries]
481 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
482 --- Example 1
483
484 data D r = ZeroD | SuccD (r (D r));
485
486 instance (Eq (r (D r))) => Eq (D r) where
487 ZeroD == ZeroD = True
488 (SuccD a) == (SuccD b) = a == b
489 _ == _ = False;
490
491 equalDC :: D [] -> D [] -> Bool;
492 equalDC = (==);
493
494 We need to prove (Eq (D [])). Here's how we go:
495
496 [W] d1 : Eq (D [])
497 By instance decl of Eq (D r):
498 [W] d2 : Eq [D []] where d1 = dfEqD d2
499 By instance decl of Eq [a]:
500 [W] d3 : Eq (D []) where d2 = dfEqList d3
501 d1 = dfEqD d2
502 Now this wanted can interact with our "solved" d1 to get:
503 d3 = d1
504
505 -- Example 2:
506 This code arises in the context of "Scrap Your Boilerplate with Class"
507
508 class Sat a
509 class Data ctx a
510 instance Sat (ctx Char) => Data ctx Char -- dfunData1
511 instance (Sat (ctx [a]), Data ctx a) => Data ctx [a] -- dfunData2
512
513 class Data Maybe a => Foo a
514
515 instance Foo t => Sat (Maybe t) -- dfunSat
516
517 instance Data Maybe a => Foo a -- dfunFoo1
518 instance Foo a => Foo [a] -- dfunFoo2
519 instance Foo [Char] -- dfunFoo3
520
521 Consider generating the superclasses of the instance declaration
522 instance Foo a => Foo [a]
523
524 So our problem is this
525 [G] d0 : Foo t
526 [W] d1 : Data Maybe [t] -- Desired superclass
527
528 We may add the given in the inert set, along with its superclasses
529 Inert:
530 [G] d0 : Foo t
531 [G] d01 : Data Maybe t -- Superclass of d0
532 WorkList
533 [W] d1 : Data Maybe [t]
534
535 Solve d1 using instance dfunData2; d1 := dfunData2 d2 d3
536 Inert:
537 [G] d0 : Foo t
538 [G] d01 : Data Maybe t -- Superclass of d0
539 Solved:
540 d1 : Data Maybe [t]
541 WorkList:
542 [W] d2 : Sat (Maybe [t])
543 [W] d3 : Data Maybe t
544
545 Now, we may simplify d2 using dfunSat; d2 := dfunSat d4
546 Inert:
547 [G] d0 : Foo t
548 [G] d01 : Data Maybe t -- Superclass of d0
549 Solved:
550 d1 : Data Maybe [t]
551 d2 : Sat (Maybe [t])
552 WorkList:
553 [W] d3 : Data Maybe t
554 [W] d4 : Foo [t]
555
556 Now, we can just solve d3 from d01; d3 := d01
557 Inert
558 [G] d0 : Foo t
559 [G] d01 : Data Maybe t -- Superclass of d0
560 Solved:
561 d1 : Data Maybe [t]
562 d2 : Sat (Maybe [t])
563 WorkList
564 [W] d4 : Foo [t]
565
566 Now, solve d4 using dfunFoo2; d4 := dfunFoo2 d5
567 Inert
568 [G] d0 : Foo t
569 [G] d01 : Data Maybe t -- Superclass of d0
570 Solved:
571 d1 : Data Maybe [t]
572 d2 : Sat (Maybe [t])
573 d4 : Foo [t]
574 WorkList:
575 [W] d5 : Foo t
576
577 Now, d5 can be solved! d5 := d0
578
579 Result
580 d1 := dfunData2 d2 d3
581 d2 := dfunSat d4
582 d3 := d01
583 d4 := dfunFoo2 d5
584 d5 := d0
585 -}
586
587 {- *********************************************************************
588 * *
589 InertCans: the canonical inerts
590 * *
591 * *
592 ********************************************************************* -}
593
594 data InertCans -- See Note [Detailed InertCans Invariants] for more
595 = IC { inert_eqs :: InertEqs
596 -- See Note [inert_eqs: the inert equalities]
597 -- All CTyEqCans; index is the LHS tyvar
598 -- Domain = skolems and untouchables; a touchable would be unified
599
600 , inert_funeqs :: FunEqMap Ct
601 -- All CFunEqCans; index is the whole family head type.
602 -- All Nominal (that's an invarint of all CFunEqCans)
603 -- LHS is fully rewritten (modulo eqCanRewrite constraints)
604 -- wrt inert_eqs
605 -- Can include all flavours, [G], [W], [WD], [D]
606 -- See Note [Type family equations]
607
608 , inert_dicts :: DictMap Ct
609 -- Dictionaries only
610 -- All fully rewritten (modulo flavour constraints)
611 -- wrt inert_eqs
612
613 , inert_safehask :: DictMap Ct
614 -- Failed dictionary resolution due to Safe Haskell overlapping
615 -- instances restriction. We keep this separate from inert_dicts
616 -- as it doesn't cause compilation failure, just safe inference
617 -- failure.
618 --
619 -- ^ See Note [Safe Haskell Overlapping Instances Implementation]
620 -- in TcSimplify
621
622 , inert_irreds :: Cts
623 -- Irreducible predicates
624
625 , inert_insols :: Cts
626 -- Frozen errors (as non-canonicals)
627
628 , inert_count :: Int
629 -- Number of Wanted goals in
630 -- inert_eqs, inert_dicts, inert_safehask, inert_irreds
631 -- Does not include insolubles
632 -- When non-zero, keep trying to solved
633 }
634
635 type InertEqs = DTyVarEnv EqualCtList
636 type EqualCtList = [Ct] -- See Note [EqualCtList invariants]
637
638 {- Note [Detailed InertCans Invariants]
639 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
640 The InertCans represents a collection of constraints with the following properties:
641
642 * All canonical
643
644 * No two dictionaries with the same head
645 * No two CIrreds with the same type
646
647 * Family equations inert wrt top-level family axioms
648
649 * Dictionaries have no matching top-level instance
650
651 * Given family or dictionary constraints don't mention touchable
652 unification variables
653
654 * Non-CTyEqCan constraints are fully rewritten with respect
655 to the CTyEqCan equalities (modulo canRewrite of course;
656 eg a wanted cannot rewrite a given)
657
658 * CTyEqCan equalities: see Note [Applying the inert substitution]
659 in TcFlatten
660
661 Note [EqualCtList invariants]
662 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
663 * All are equalities
664 * All these equalities have the same LHS
665 * The list is never empty
666 * No element of the list can rewrite any other
667 * Derived before Wanted
668
669 From the fourth invariant it follows that the list is
670 - A single [G], or
671 - Zero or one [D] or [WD], followd by any number of [W]
672
673 The Wanteds can't rewrite anything which is why we put them last
674
675 Note [Type family equations]
676 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
677 Type-family equations, CFunEqCans, of form (ev : F tys ~ ty),
678 live in three places
679
680 * The work-list, of course
681
682 * The inert_funeqs are un-solved but fully processed, and in
683 the InertCans. They can be [G], [W], [WD], or [D].
684
685 * The inert_flat_cache. This is used when flattening, to get maximal
686 sharing. Everthing in the inert_flat_cache is [G] or [WD]
687
688 It contains lots of things that are still in the work-list.
689 E.g Suppose we have (w1: F (G a) ~ Int), and (w2: H (G a) ~ Int) in the
690 work list. Then we flatten w1, dumping (w3: G a ~ f1) in the work
691 list. Now if we flatten w2 before we get to w3, we still want to
692 share that (G a).
693 Because it contains work-list things, DO NOT use the flat cache to solve
694 a top-level goal. Eg in the above example we don't want to solve w3
695 using w3 itself!
696
697 The CFunEqCan Ownership Invariant:
698
699 * Each [G/W/WD] CFunEqCan has a distinct fsk or fmv
700 It "owns" that fsk/fmv, in the sense that:
701 - reducing a [W/WD] CFunEqCan fills in the fmv
702 - unflattening a [W/WD] CFunEqCan fills in the fmv
703 (in both cases unless an occurs-check would result)
704
705 * In contrast a [D] CFunEqCan does not "own" its fmv:
706 - reducing a [D] CFunEqCan does not fill in the fmv;
707 it just generates an equality
708 - unflattening ignores [D] CFunEqCans altogether
709
710
711 Note [inert_eqs: the inert equalities]
712 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
713 Definition [Can-rewrite relation]
714 A "can-rewrite" relation between flavours, written f1 >= f2, is a
715 binary relation with the following properties
716
717 (R1) >= is transitive
718 (R2) If f1 >= f, and f2 >= f,
719 then either f1 >= f2 or f2 >= f1
720
721 Lemma. If f1 >= f then f1 >= f1
722 Proof. By property (R2), with f1=f2
723
724 Definition [Generalised substitution]
725 A "generalised substitution" S is a set of triples (a -f-> t), where
726 a is a type variable
727 t is a type
728 f is a flavour
729 such that
730 (WF1) if (a -f1-> t1) in S
731 (a -f2-> t2) in S
732 then neither (f1 >= f2) nor (f2 >= f1) hold
733 (WF2) if (a -f-> t) is in S, then t /= a
734
735 Definition [Applying a generalised substitution]
736 If S is a generalised substitution
737 S(f,a) = t, if (a -fs-> t) in S, and fs >= f
738 = a, otherwise
739 Application extends naturally to types S(f,t), modulo roles.
740 See Note [Flavours with roles].
741
742 Theorem: S(f,a) is well defined as a function.
743 Proof: Suppose (a -f1-> t1) and (a -f2-> t2) are both in S,
744 and f1 >= f and f2 >= f
745 Then by (R2) f1 >= f2 or f2 >= f1, which contradicts (WF1)
746
747 Notation: repeated application.
748 S^0(f,t) = t
749 S^(n+1)(f,t) = S(f, S^n(t))
750
751 Definition: inert generalised substitution
752 A generalised substitution S is "inert" iff
753
754 (IG1) there is an n such that
755 for every f,t, S^n(f,t) = S^(n+1)(f,t)
756
757 By (IG1) we define S*(f,t) to be the result of exahaustively
758 applying S(f,_) to t.
759
760 ----------------------------------------------------------------
761 Our main invariant:
762 the inert CTyEqCans should be an inert generalised substitution
763 ----------------------------------------------------------------
764
765 Note that inertness is not the same as idempotence. To apply S to a
766 type, you may have to apply it recursive. But inertness does
767 guarantee that this recursive use will terminate.
768
769 Note [Extending the inert equalities]
770 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
771 Theorem [Stability under extension]
772 This is the main theorem!
773 Suppose we have a "work item"
774 a -fw-> t
775 and an inert generalised substitution S,
776 such that
777 (T1) S(fw,a) = a -- LHS of work-item is a fixpoint of S(fw,_)
778 (T2) S(fw,t) = t -- RHS of work-item is a fixpoint of S(fw,_)
779 (T3) a not in t -- No occurs check in the work item
780
781 (K1) for every (a -fs-> s) in S, then not (fw >= fs)
782 Reason: the work item is fully rewritten by S, hence not (fs >= fw)
783 but if (fw >= fs) then the work item could rewrite
784 the inert item
785
786 (K2) for every (b -fs-> s) in S, where b /= a, then
787 (K2a) not (fs >= fs)
788 or (K2b) fs >= fw
789 or (K2c) not (fw >= fs)
790 or (K2d) a not in s
791
792 (K3) See Note [K3: completeness of solving]
793 If (b -fs-> s) is in S with (fw >= fs), then
794 (K3a) If the role of fs is nominal: s /= a
795 (K3b) If the role of fs is representational: EITHER
796 a not in s, OR
797 the path from the top of s to a includes at least one non-newtype
798
799 then the extended substitution T = S+(a -fw-> t)
800 is an inert generalised substitution.
801
802 Conditions (T1-T3) are established by the canonicaliser
803 Conditions (K1-K3) are established by TcSMonad.kickOutRewritable
804
805 The idea is that
806 * (T1-2) are guaranteed by exhaustively rewriting the work-item
807 with S(fw,_).
808
809 * T3 is guaranteed by a simple occurs-check on the work item.
810 This is done during canonicalisation, in canEqTyVar;
811 (invariant: a CTyEqCan never has an occurs check).
812
813 * (K1-3) are the "kick-out" criteria. (As stated, they are really the
814 "keep" criteria.) If the current inert S contains a triple that does
815 not satisfy (K1-3), then we remove it from S by "kicking it out",
816 and re-processing it.
817
818 * Note that kicking out is a Bad Thing, because it means we have to
819 re-process a constraint. The less we kick out, the better.
820 TODO: Make sure that kicking out really *is* a Bad Thing. We've assumed
821 this but haven't done the empirical study to check.
822
823 * Assume we have G>=G, G>=W and that's all. Then, when performing
824 a unification we add a new given a -G-> ty. But doing so does NOT require
825 us to kick out an inert wanted that mentions a, because of (K2a). This
826 is a common case, hence good not to kick out.
827
828 * Lemma (L2): if not (fw >= fw), then K1-K3 all hold.
829 Proof: using Definition [Can-rewrite relation], fw can't rewrite anything
830 and so K1-K3 hold. Intuitively, since fw can't rewrite anything,
831 adding it cannot cause any loops
832 This is a common case, because Wanteds cannot rewrite Wanteds.
833
834 * Lemma (L1): The conditions of the Main Theorem imply that there is no
835 (a -fs-> t) in S, s.t. (fs >= fw).
836 Proof. Suppose the contrary (fs >= fw). Then because of (T1),
837 S(fw,a)=a. But since fs>=fw, S(fw,a) = s, hence s=a. But now we
838 have (a -fs-> a) in S, which contradicts (WF2).
839
840 * The extended substitution satisfies (WF1) and (WF2)
841 - (K1) plus (L1) guarantee that the extended substitution satisfies (WF1).
842 - (T3) guarantees (WF2).
843
844 * (K2) is about inertness. Intuitively, any infinite chain T^0(f,t),
845 T^1(f,t), T^2(f,T).... must pass through the new work item infnitely
846 often, since the substitution without the work item is inert; and must
847 pass through at least one of the triples in S infnitely often.
848
849 - (K2a): if not(fs>=fs) then there is no f that fs can rewrite (fs>=f),
850 and hence this triple never plays a role in application S(f,a).
851 It is always safe to extend S with such a triple.
852
853 (NB: we could strengten K1) in this way too, but see K3.
854
855 - (K2b): If this holds then, by (T2), b is not in t. So applying the
856 work item does not genenerate any new opportunities for applying S
857
858 - (K2c): If this holds, we can't pass through this triple infinitely
859 often, because if we did then fs>=f, fw>=f, hence by (R2)
860 * either fw>=fs, contradicting K2c
861 * or fs>=fw; so by the argument in K2b we can't have a loop
862
863 - (K2d): if a not in s, we hae no further opportunity to apply the
864 work item, similar to (K2b)
865
866 NB: Dimitrios has a PDF that does this in more detail
867
868 Key lemma to make it watertight.
869 Under the conditions of the Main Theorem,
870 forall f st fw >= f, a is not in S^k(f,t), for any k
871
872 Also, consider roles more carefully. See Note [Flavours with roles]
873
874 Note [K3: completeness of solving]
875 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
876 (K3) is not necessary for the extended substitution
877 to be inert. In fact K1 could be made stronger by saying
878 ... then (not (fw >= fs) or not (fs >= fs))
879 But it's not enough for S to be inert; we also want completeness.
880 That is, we want to be able to solve all soluble wanted equalities.
881 Suppose we have
882
883 work-item b -G-> a
884 inert-item a -W-> b
885
886 Assuming (G >= W) but not (W >= W), this fulfills all the conditions,
887 so we could extend the inerts, thus:
888
889 inert-items b -G-> a
890 a -W-> b
891
892 But if we kicked-out the inert item, we'd get
893
894 work-item a -W-> b
895 inert-item b -G-> a
896
897 Then rewrite the work-item gives us (a -W-> a), which is soluble via Refl.
898 So we add one more clause to the kick-out criteria
899
900 Another way to understand (K3) is that we treat an inert item
901 a -f-> b
902 in the same way as
903 b -f-> a
904 So if we kick out one, we should kick out the other. The orientation
905 is somewhat accidental.
906
907 When considering roles, we also need the second clause (K3b). Consider
908
909 inert-item a -W/R-> b c
910 work-item c -G/N-> a
911
912 The work-item doesn't get rewritten by the inert, because (>=) doesn't hold.
913 We've satisfied conditions (T1)-(T3) and (K1) and (K2). If all we had were
914 condition (K3a), then we would keep the inert around and add the work item.
915 But then, consider if we hit the following:
916
917 work-item2 b -G/N-> Id
918
919 where
920
921 newtype Id x = Id x
922
923 For similar reasons, if we only had (K3a), we wouldn't kick the
924 representational inert out. And then, we'd miss solving the inert, which
925 now reduced to reflexivity. The solution here is to kick out representational
926 inerts whenever the tyvar of a work item is "exposed", where exposed means
927 not under some proper data-type constructor, like [] or Maybe. See
928 isTyVarExposed in TcType. This is encoded in (K3b).
929
930 Note [Flavours with roles]
931 ~~~~~~~~~~~~~~~~~~~~~~~~~~
932 The system described in Note [inert_eqs: the inert equalities]
933 discusses an abstract
934 set of flavours. In GHC, flavours have two components: the flavour proper,
935 taken from {Wanted, Derived, Given} and the equality relation (often called
936 role), taken from {NomEq, ReprEq}.
937 When substituting w.r.t. the inert set,
938 as described in Note [inert_eqs: the inert equalities],
939 we must be careful to respect all components of a flavour.
940 For example, if we have
941
942 inert set: a -G/R-> Int
943 b -G/R-> Bool
944
945 type role T nominal representational
946
947 and we wish to compute S(W/R, T a b), the correct answer is T a Bool, NOT
948 T Int Bool. The reason is that T's first parameter has a nominal role, and
949 thus rewriting a to Int in T a b is wrong. Indeed, this non-congruence of
950 substitution means that the proof in Note [The inert equalities] may need
951 to be revisited, but we don't think that the end conclusion is wrong.
952 -}
953
954 instance Outputable InertCans where
955 ppr (IC { inert_eqs = eqs
956 , inert_funeqs = funeqs, inert_dicts = dicts
957 , inert_safehask = safehask, inert_irreds = irreds
958 , inert_insols = insols, inert_count = count })
959 = braces $ vcat
960 [ ppUnless (isEmptyDVarEnv eqs) $
961 text "Equalities:"
962 <+> pprCts (foldDVarEnv (\eqs rest -> listToBag eqs `andCts` rest) emptyCts eqs)
963 , ppUnless (isEmptyTcAppMap funeqs) $
964 text "Type-function equalities =" <+> pprCts (funEqsToBag funeqs)
965 , ppUnless (isEmptyTcAppMap dicts) $
966 text "Dictionaries =" <+> pprCts (dictsToBag dicts)
967 , ppUnless (isEmptyTcAppMap safehask) $
968 text "Safe Haskell unsafe overlap =" <+> pprCts (dictsToBag safehask)
969 , ppUnless (isEmptyCts irreds) $
970 text "Irreds =" <+> pprCts irreds
971 , ppUnless (isEmptyCts insols) $
972 text "Insolubles =" <+> pprCts insols
973 , text "Unsolved goals =" <+> int count
974 ]
975
976 {- *********************************************************************
977 * *
978 Shadow constraints and improvement
979 * *
980 ************************************************************************
981
982 Note [The improvement story and derived shadows]
983 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
984 Because Wanteds cannot rewrite Wanteds (see Note [Wanteds do not
985 rewrite Wanteds] in TcRnTypes), we may miss some opportunities for
986 solving. Here's a classic example (indexed-types/should_fail/T4093a)
987
988 Ambiguity check for f: (Foo e ~ Maybe e) => Foo e
989
990 We get [G] Foo e ~ Maybe e
991 [W] Foo e ~ Foo ee -- ee is a unification variable
992 [W] Foo ee ~ Maybe ee
993
994 Flatten: [G] Foo e ~ fsk
995 [G] fsk ~ Maybe e -- (A)
996
997 [W] Foo ee ~ fmv
998 [W] fmv ~ fsk -- (B) From Foo e ~ Foo ee
999 [W] fmv ~ Maybe ee
1000
1001 --> rewrite (B) with (A)
1002 [W] Foo ee ~ fmv
1003 [W] fmv ~ Maybe e
1004 [W] fmv ~ Maybe ee
1005
1006 But now we appear to be stuck, since we don't rewrite Wanteds with
1007 Wanteds. This is silly because we can see that ee := e is the
1008 only solution.
1009
1010 The basic plan is
1011 * generate Derived constraints that shadow Wanted constraints
1012 * allow Derived to rewrite Derived
1013 * in order to cause some unifications to take place
1014 * that in turn solve the original Wanteds
1015
1016 The ONLY reason for all these Derived equalities is to tell us how to
1017 unify a variable: that is, what Mark Jones calls "improvement".
1018
1019 The same idea is sometimes also called "saturation"; find all the
1020 equalities that must hold in any solution.
1021
1022 Or, equivalently, you can think of the derived shadows as implementing
1023 the "model": an non-idempotent but no-occurs-check substitution,
1024 reflecting *all* *Nominal* equalities (a ~N ty) that are not
1025 immediately soluble by unification.
1026
1027 More specifically, here's how it works (Oct 16):
1028
1029 * Wanted constraints are born as [WD]; this behaves like a
1030 [W] and a [D] paired together.
1031
1032 * When we are about to add a [WD] to the inert set, if it can
1033 be rewritten by a [D] a ~ ty, then we split it into [W] and [D],
1034 putting the latter into the work list (see maybeEmitShadow).
1035
1036 In the example above, we get to the point where we are stuck:
1037 [WD] Foo ee ~ fmv
1038 [WD] fmv ~ Maybe e
1039 [WD] fmv ~ Maybe ee
1040
1041 But now when [WD] fmv ~ Maybe ee is about to be added, we'll
1042 split it into [W] and [D], since the inert [WD] fmv ~ Maybe e
1043 can rewrite it. Then:
1044 work item: [D] fmv ~ Maybe ee
1045 inert: [W] fmv ~ Maybe ee
1046 [WD] fmv ~ Maybe e -- (C)
1047 [WD] Foo ee ~ fmv
1048
1049 See Note [Splitting WD constraints]. Now the work item is rewritten
1050 by (C) and we soon get ee := e.
1051
1052 Additional notes:
1053
1054 * The derived shadow equalities live in inert_eqs, along with
1055 the Givens and Wanteds; see Note [EqualCtList invariants].
1056
1057 * We make Derived shadows only for Wanteds, not Givens. So we
1058 have only [G], not [GD] and [G] plus splitting. See
1059 Note [Add derived shadows only for Wanteds]
1060
1061 * We also get Derived equalities from functional dependencies
1062 and type-function injectivity; see calls to unifyDerived.
1063
1064 * This splitting business applies to CFunEqCans too; and then
1065 we do apply type-function reductions to the [D] CFunEqCan.
1066 See Note [Reduction for Derived CFunEqCans]
1067
1068 * It's worth having [WD] rather than just [W] and [D] because
1069 * efficiency: silly to process the same thing twice
1070 * inert_funeqs, inert_dicts is a finite map keyed by
1071 the type; it's inconvenient for it to map to TWO constraints
1072
1073 Note [Splitting WD constraints]
1074 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1075 We are about to add a [WD] constraint to the inert set; and we
1076 know that the inert set has fully rewritten it. Should we split
1077 it into [W] and [D], and put the [D] in the work list for further
1078 work?
1079
1080 * CDictCan (C tys) or CFunEqCan (F tys ~ fsk):
1081 Yes if the inert set could rewrite tys to make the class constraint,
1082 or type family, fire. That is, yes if the inert_eqs intersects
1083 with the free vars of tys. For this test we use
1084 (anyRewritableTyVar True) which ignores casts and coercions in tys,
1085 because rewriting the casts or coercions won't make the thing fire
1086 more often.
1087
1088 * CTyEqCan (a ~ ty): Yes if the inert set could rewrite 'a' or 'ty'.
1089 We need to check both 'a' and 'ty' against the inert set:
1090 - Inert set contains [D] a ~ ty2
1091 Then we want to put [D] a ~ ty in the worklist, so we'll
1092 get [D] ty ~ ty2 with consequent good things
1093
1094 - Inert set contains [D] b ~ a, where b is in ty.
1095 We can't just add [WD] a ~ ty[b] to the inert set, because
1096 that breaks the inert-set invariants. If we tried to
1097 canonicalise another [D] constraint mentioning 'a', we'd
1098 get an infinite loop
1099
1100 Moreover we must use (anyRewritableTyVar False) for the RHS,
1101 because even tyvars in the casts and coercions could give
1102 an infinite loop if we don't expose it
1103
1104 * Others: nothing is gained by splitting.
1105
1106 Note [Examples of how Derived shadows helps completeness]
1107 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1108 Trac #10009, a very nasty example:
1109
1110 f :: (UnF (F b) ~ b) => F b -> ()
1111
1112 g :: forall a. (UnF (F a) ~ a) => a -> ()
1113 g _ = f (undefined :: F a)
1114
1115 For g we get [G] UnF (F a) ~ a
1116 [WD] UnF (F beta) ~ beta
1117 [WD] F a ~ F beta
1118 Flatten:
1119 [G] g1: F a ~ fsk1 fsk1 := F a
1120 [G] g2: UnF fsk1 ~ fsk2 fsk2 := UnF fsk1
1121 [G] g3: fsk2 ~ a
1122
1123 [WD] w1: F beta ~ fmv1
1124 [WD] w2: UnF fmv1 ~ fmv2
1125 [WD] w3: fmv2 ~ beta
1126 [WD] w4: fmv1 ~ fsk1 -- From F a ~ F beta using flat-cache
1127 -- and re-orient to put meta-var on left
1128
1129 Rewrite w2 with w4: [D] d1: UnF fsk1 ~ fmv2
1130 React that with g2: [D] d2: fmv2 ~ fsk2
1131 React that with w3: [D] beta ~ fsk2
1132 and g3: [D] beta ~ a -- Hooray beta := a
1133 And that is enough to solve everything
1134
1135 Note [Add derived shadows only for Wanteds]
1136 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1137 We only add shadows for Wanted constraints. That is, we have
1138 [WD] but not [GD]; and maybeEmitShaodw looks only at [WD]
1139 constraints.
1140
1141 It does just possibly make sense ot add a derived shadow for a
1142 Given. If we created a Derived shadow of a Given, it could be
1143 rewritten by other Deriveds, and that could, conceivably, lead to a
1144 useful unification.
1145
1146 But (a) I have been unable to come up with an example of this
1147 happening
1148 (b) see Trac #12660 for how adding the derived shadows
1149 of a Given led to an infinite loop.
1150 (c) It's unlikely that rewriting derived Givens will lead
1151 to a unification because Givens don't mention touchable
1152 unification variables
1153
1154 For (b) there may be other ways to solve the loop, but simply
1155 reraining from adding derived shadows of Givens is particularly
1156 simple. And it's more efficient too!
1157
1158 Still, here's one possible reason for adding derived shadows
1159 for Givens. Consider
1160 work-item [G] a ~ [b], inerts has [D] b ~ a.
1161 If we added the derived shadow (into the work list)
1162 [D] a ~ [b]
1163 When we process it, we'll rewrite to a ~ [a] and get an
1164 occurs check. Without it we'll miss the occurs check (reporting
1165 inaccessible code); but that's probably OK.
1166
1167 Note [Keep CDictCan shadows as CDictCan]
1168 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1169 Suppose we have
1170 class C a => D a b
1171 and [G] D a b, [G] C a in the inert set. Now we insert
1172 [D] b ~ c. We want to kick out a derived shadow for [D] D a b,
1173 so we can rewrite it with the new constraint, and perhaps get
1174 instance reduction or other consequences.
1175
1176 BUT we do not want to kick out a *non-canonical* (D a b). If we
1177 did, we would do this:
1178 - rewrite it to [D] D a c, with pend_sc = True
1179 - use expandSuperClasses to add C a
1180 - go round again, which solves C a from the givens
1181 This loop goes on for ever and triggers the simpl_loop limit.
1182
1183 Solution: kick out the CDictCan which will have pend_sc = False,
1184 because we've already added its superclasses. So we won't re-add
1185 them. If we forget the pend_sc flag, our cunning scheme for avoiding
1186 generating superclasses repeatedly will fail.
1187
1188 See Trac #11379 for a case of this.
1189
1190 Note [Do not do improvement for WOnly]
1191 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1192 We do improvement between two constraints (e.g. for injectivity
1193 or functional dependencies) only if both are "improvable". And
1194 we improve a constraint wrt the top-level instances only if
1195 it is improvable.
1196
1197 Improvable: [G] [WD] [D}
1198 Not improvable: [W]
1199
1200 Reasons:
1201
1202 * It's less work: fewer pairs to compare
1203
1204 * Every [W] has a shadow [D] so nothing is lost
1205
1206 * Consider [WD] C Int b, where 'b' is a skolem, and
1207 class C a b | a -> b
1208 instance C Int Bool
1209 We'll do a fundep on it and emit [D] b ~ Bool
1210 That will kick out constraint [WD] C Int b
1211 Then we'll split it to [W] C Int b (keep in inert)
1212 and [D] C Int b (in work list)
1213 When processing the latter we'll rewrite it to
1214 [D] C Int Bool
1215 At that point it would be /stupid/ to interact it
1216 with the inert [W] C Int b in the inert set; after all,
1217 it's the very constraint from which the [D] C Int Bool
1218 was split! We can avoid this by not doing improvement
1219 on [W] constraints. This came up in Trac #12860.
1220 -}
1221
1222 maybeEmitShadow :: InertCans -> Ct -> TcS Ct
1223 -- See Note [The improvement story and derived shadows]
1224 maybeEmitShadow ics ct
1225 | let ev = ctEvidence ct
1226 , CtWanted { ctev_pred = pred, ctev_loc = loc
1227 , ctev_nosh = WDeriv } <- ev
1228 , shouldSplitWD (inert_eqs ics) ct
1229 = do { traceTcS "Emit derived shadow" (ppr ct)
1230 ; let derived_ev = CtDerived { ctev_pred = pred
1231 , ctev_loc = loc }
1232 shadow_ct = ct { cc_ev = derived_ev }
1233 -- Te shadow constraint keeps the canonical shape.
1234 -- This just saves work, but is sometimes important;
1235 -- see Note [Keep CDictCan shadows as CDictCan]
1236 ; emitWork [shadow_ct]
1237
1238 ; let ev' = ev { ctev_nosh = WOnly }
1239 ct' = ct { cc_ev = ev' }
1240 -- Record that it now has a shadow
1241 -- This is /the/ place we set the flag to WOnly
1242 ; return ct' }
1243
1244 | otherwise
1245 = return ct
1246
1247 shouldSplitWD :: InertEqs -> Ct -> Bool
1248 -- Precondition: 'ct' is [WD], and is inert
1249 -- True <=> we should split ct ito [W] and [D] because
1250 -- the inert_eqs can make progress on the [D]
1251 -- See Note [Splitting WD constraints]
1252
1253 shouldSplitWD inert_eqs (CFunEqCan { cc_tyargs = tys })
1254 = should_split_match_args inert_eqs tys
1255 -- We don't need to split if the tv is the RHS fsk
1256
1257 shouldSplitWD inert_eqs (CDictCan { cc_tyargs = tys })
1258 = should_split_match_args inert_eqs tys
1259 -- NB True: ignore coercions
1260 -- See Note [Splitting WD constraints]
1261
1262 shouldSplitWD inert_eqs (CTyEqCan { cc_tyvar = tv, cc_rhs = ty })
1263 = tv `elemDVarEnv` inert_eqs
1264 || anyRewritableTyVar False (`elemDVarEnv` inert_eqs) ty
1265 -- NB False: do not ignore casts and coercions
1266 -- See Note [Splitting WD constraints]
1267
1268 shouldSplitWD _ _ = False -- No point in splitting otherwise
1269
1270 should_split_match_args :: InertEqs -> [TcType] -> Bool
1271 -- True if the inert_eqs can rewrite anything in the argument
1272 -- types, ignoring casts and coercions
1273 should_split_match_args inert_eqs tys
1274 = any (anyRewritableTyVar True (`elemDVarEnv` inert_eqs)) tys
1275 -- NB True: ignore casts coercions
1276 -- See Note [Splitting WD constraints]
1277
1278 isImprovable :: CtEvidence -> Bool
1279 -- See Note [Do not do improvement for WOnly]
1280 isImprovable (CtWanted { ctev_nosh = WOnly }) = False
1281 isImprovable _ = True
1282
1283
1284 {- *********************************************************************
1285 * *
1286 Inert equalities
1287 * *
1288 ********************************************************************* -}
1289
1290 addTyEq :: InertEqs -> TcTyVar -> Ct -> InertEqs
1291 addTyEq old_eqs tv ct
1292 = extendDVarEnv_C add_eq old_eqs tv [ct]
1293 where
1294 add_eq old_eqs _
1295 | isWantedCt ct
1296 , (eq1 : eqs) <- old_eqs
1297 = eq1 : ct : eqs
1298 | otherwise
1299 = ct : old_eqs
1300
1301 foldTyEqs :: (Ct -> b -> b) -> InertEqs -> b -> b
1302 foldTyEqs k eqs z
1303 = foldDVarEnv (\cts z -> foldr k z cts) z eqs
1304
1305 findTyEqs :: InertCans -> TyVar -> EqualCtList
1306 findTyEqs icans tv = lookupDVarEnv (inert_eqs icans) tv `orElse` []
1307
1308 delTyEq :: InertEqs -> TcTyVar -> TcType -> InertEqs
1309 delTyEq m tv t = modifyDVarEnv (filter (not . isThisOne)) m tv
1310 where isThisOne (CTyEqCan { cc_rhs = t1 }) = eqType t t1
1311 isThisOne _ = False
1312
1313 lookupInertTyVar :: InertEqs -> TcTyVar -> Maybe TcType
1314 lookupInertTyVar ieqs tv
1315 = case lookupDVarEnv ieqs tv of
1316 Just (CTyEqCan { cc_rhs = rhs, cc_eq_rel = NomEq } : _ ) -> Just rhs
1317 _ -> Nothing
1318
1319 lookupFlattenTyVar :: InertEqs -> TcTyVar -> TcType
1320 -- See Note [lookupFlattenTyVar]
1321 lookupFlattenTyVar ieqs ftv
1322 = lookupInertTyVar ieqs ftv `orElse` mkTyVarTy ftv
1323
1324 {- Note [lookupFlattenTyVar]
1325 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1326 Suppose we have an injective function F and
1327 inert_funeqs: F t1 ~ fsk1
1328 F t2 ~ fsk2
1329 inert_eqs: fsk1 ~ fsk2
1330
1331 We never rewrite the RHS (cc_fsk) of a CFunEqCan. But we /do/ want to
1332 get the [D] t1 ~ t2 from the injectiveness of F. So we look up the
1333 cc_fsk of CFunEqCans in the inert_eqs when trying to find derived
1334 equalities arising from injectivity.
1335 -}
1336
1337
1338 {- *********************************************************************
1339 * *
1340 Adding an inert
1341 * *
1342 ************************************************************************
1343
1344 Note [Adding an equality to the InertCans]
1345 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1346 When adding an equality to the inerts:
1347
1348 * Split [WD] into [W] and [D] if the inerts can rewrite the latter;
1349 done by maybeEmitShadow.
1350
1351 * Kick out any constraints that can be rewritten by the thing
1352 we are adding. Done by kickOutRewritable.
1353
1354 * Note that unifying a:=ty, is like adding [G] a~ty; just use
1355 kickOutRewritable with Nominal, Given. See kickOutAfterUnification.
1356
1357 Note [Kicking out CFunEqCan for fundeps]
1358 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1359 Consider:
1360 New: [D] fmv1 ~ fmv2
1361 Inert: [W] F alpha ~ fmv1
1362 [W] F beta ~ fmv2
1363
1364 where F is injective. The new (derived) equality certainly can't
1365 rewrite the inerts. But we *must* kick out the first one, to get:
1366
1367 New: [W] F alpha ~ fmv1
1368 Inert: [W] F beta ~ fmv2
1369 [D] fmv1 ~ fmv2
1370
1371 and now improvement will discover [D] alpha ~ beta. This is important;
1372 eg in Trac #9587.
1373
1374 So in kickOutRewritable we look at all the tyvars of the
1375 CFunEqCan, including the fsk.
1376 -}
1377
1378 addInertEq :: Ct -> TcS ()
1379 -- This is a key function, because of the kick-out stuff
1380 -- Precondition: item /is/ canonical
1381 -- See Note [Adding an equality to the InertCans]
1382 addInertEq ct
1383 = do { traceTcS "addInertEq {" $
1384 text "Adding new inert equality:" <+> ppr ct
1385 ; ics <- getInertCans
1386
1387 ; ct@(CTyEqCan { cc_tyvar = tv, cc_ev = ev }) <- maybeEmitShadow ics ct
1388
1389 ; (_, ics1) <- kickOutRewritable (ctEvFlavourRole ev) tv ics
1390
1391 ; let ics2 = ics1 { inert_eqs = addTyEq (inert_eqs ics1) tv ct
1392 , inert_count = bumpUnsolvedCount ev (inert_count ics1) }
1393 ; setInertCans ics2
1394
1395 ; traceTcS "addInertEq }" $ empty }
1396
1397 --------------
1398 addInertCan :: Ct -> TcS () -- Constraints *other than* equalities
1399 addInertCan ct
1400 = do { traceTcS "insertInertCan {" $
1401 text "Trying to insert new non-eq inert item:" <+> ppr ct
1402
1403 ; ics <- getInertCans
1404 ; ct <- maybeEmitShadow ics ct
1405 ; setInertCans (add_item ics ct)
1406
1407 ; traceTcS "addInertCan }" $ empty }
1408
1409 add_item :: InertCans -> Ct -> InertCans
1410 add_item ics item@(CFunEqCan { cc_fun = tc, cc_tyargs = tys })
1411 = ics { inert_funeqs = insertFunEq (inert_funeqs ics) tc tys item }
1412
1413 add_item ics item@(CIrredEvCan { cc_ev = ev })
1414 = ics { inert_irreds = inert_irreds ics `Bag.snocBag` item
1415 , inert_count = bumpUnsolvedCount ev (inert_count ics) }
1416 -- The 'False' is because the irreducible constraint might later instantiate
1417 -- to an equality.
1418 -- But since we try to simplify first, if there's a constraint function FC with
1419 -- type instance FC Int = Show
1420 -- we'll reduce a constraint (FC Int a) to Show a, and never add an inert irreducible
1421
1422 add_item ics item@(CDictCan { cc_ev = ev, cc_class = cls, cc_tyargs = tys })
1423 = ics { inert_dicts = addDict (inert_dicts ics) cls tys item
1424 , inert_count = bumpUnsolvedCount ev (inert_count ics) }
1425
1426 add_item _ item
1427 = pprPanic "upd_inert set: can't happen! Inserting " $
1428 ppr item -- CTyEqCan is dealt with by addInertEq
1429 -- Can't be CNonCanonical, CHoleCan,
1430 -- because they only land in inert_insols
1431
1432 bumpUnsolvedCount :: CtEvidence -> Int -> Int
1433 bumpUnsolvedCount ev n | isWanted ev = n+1
1434 | otherwise = n
1435
1436
1437 -----------------------------------------
1438 kickOutRewritable :: CtFlavourRole -- Flavour/role of the equality that
1439 -- is being added to the inert set
1440 -> TcTyVar -- The new equality is tv ~ ty
1441 -> InertCans
1442 -> TcS (Int, InertCans)
1443 kickOutRewritable new_fr new_tv ics
1444 = do { let (kicked_out, ics') = kick_out_rewritable new_fr new_tv ics
1445 n_kicked = workListSize kicked_out
1446
1447 ; unless (n_kicked == 0) $
1448 do { updWorkListTcS (appendWorkList kicked_out)
1449 ; csTraceTcS $
1450 hang (text "Kick out, tv =" <+> ppr new_tv)
1451 2 (vcat [ text "n-kicked =" <+> int n_kicked
1452 , text "kicked_out =" <+> ppr kicked_out
1453 , text "Residual inerts =" <+> ppr ics' ]) }
1454
1455 ; return (n_kicked, ics') }
1456
1457 kick_out_rewritable :: CtFlavourRole -- Flavour/role of the equality that
1458 -- is being added to the inert set
1459 -> TcTyVar -- The new equality is tv ~ ty
1460 -> InertCans
1461 -> (WorkList, InertCans)
1462 -- See Note [kickOutRewritable]
1463 kick_out_rewritable new_fr new_tv ics@(IC { inert_eqs = tv_eqs
1464 , inert_dicts = dictmap
1465 , inert_safehask = safehask
1466 , inert_funeqs = funeqmap
1467 , inert_irreds = irreds
1468 , inert_insols = insols
1469 , inert_count = n })
1470 | not (new_fr `eqMayRewriteFR` new_fr)
1471 = (emptyWorkList, ics)
1472 -- If new_fr can't rewrite itself, it can't rewrite
1473 -- anything else, so no need to kick out anything.
1474 -- (This is a common case: wanteds can't rewrite wanteds)
1475 -- Lemma (L2) in Note [Extending the inert equalities]
1476
1477 | otherwise
1478 = (kicked_out, inert_cans_in)
1479 where
1480 inert_cans_in = IC { inert_eqs = tv_eqs_in
1481 , inert_dicts = dicts_in
1482 , inert_safehask = safehask -- ??
1483 , inert_funeqs = feqs_in
1484 , inert_irreds = irs_in
1485 , inert_insols = insols_in
1486 , inert_count = n - workListWantedCount kicked_out }
1487
1488 kicked_out = WL { wl_eqs = tv_eqs_out
1489 , wl_funeqs = feqs_out
1490 , wl_deriv = []
1491 , wl_rest = bagToList (dicts_out `andCts` irs_out
1492 `andCts` insols_out)
1493 , wl_implics = emptyBag }
1494
1495 (tv_eqs_out, tv_eqs_in) = foldDVarEnv kick_out_eqs ([], emptyDVarEnv) tv_eqs
1496 (feqs_out, feqs_in) = partitionFunEqs kick_out_ct funeqmap
1497 -- See Note [Kicking out CFunEqCan for fundeps]
1498 (dicts_out, dicts_in) = partitionDicts kick_out_ct dictmap
1499 (irs_out, irs_in) = partitionBag kick_out_ct irreds
1500 (insols_out, insols_in) = partitionBag kick_out_ct insols
1501 -- Kick out even insolubles: See Note [Kick out insolubles]
1502
1503 fr_may_rewrite :: CtFlavourRole -> Bool
1504 fr_may_rewrite fs = new_fr `eqMayRewriteFR` fs
1505 -- Can the new item rewrite the inert item?
1506
1507 kick_out_ct :: Ct -> Bool
1508 -- Kick it out if the new CTyEqCan can rewrite the inert one
1509 -- See Note [kickOutRewritable]
1510 kick_out_ct ct | let ev = ctEvidence ct
1511 = fr_may_rewrite (ctEvFlavourRole ev)
1512 && anyRewritableTyVar False (== new_tv) (ctEvPred ev)
1513 -- False: ignore casts and coercions
1514 -- NB: this includes the fsk of a CFunEqCan. It can't
1515 -- actually be rewritten, but we need to kick it out
1516 -- so we get to take advantage of injectivity
1517 -- See Note [Kicking out CFunEqCan for fundeps]
1518
1519 kick_out_eqs :: EqualCtList -> ([Ct], DTyVarEnv EqualCtList)
1520 -> ([Ct], DTyVarEnv EqualCtList)
1521 kick_out_eqs eqs (acc_out, acc_in)
1522 = (eqs_out ++ acc_out, case eqs_in of
1523 [] -> acc_in
1524 (eq1:_) -> extendDVarEnv acc_in (cc_tyvar eq1) eqs_in)
1525 where
1526 (eqs_in, eqs_out) = partition keep_eq eqs
1527
1528 -- Implements criteria K1-K3 in Note [Extending the inert equalities]
1529 keep_eq (CTyEqCan { cc_tyvar = tv, cc_rhs = rhs_ty, cc_ev = ev
1530 , cc_eq_rel = eq_rel })
1531 | tv == new_tv
1532 = not (fr_may_rewrite fs) -- (K1)
1533
1534 | otherwise
1535 = check_k2 && check_k3
1536 where
1537 fs = ctEvFlavourRole ev
1538 check_k2 = not (fs `eqMayRewriteFR` fs) -- (K2a)
1539 || (fs `eqMayRewriteFR` new_fr) -- (K2b)
1540 || not (fr_may_rewrite fs) -- (K2c)
1541 || not (new_tv `elemVarSet` tyCoVarsOfType rhs_ty) -- (K2d)
1542
1543 check_k3
1544 | fr_may_rewrite fs
1545 = case eq_rel of
1546 NomEq -> not (rhs_ty `eqType` mkTyVarTy new_tv)
1547 ReprEq -> not (isTyVarExposed new_tv rhs_ty)
1548
1549 | otherwise
1550 = True
1551
1552 keep_eq ct = pprPanic "keep_eq" (ppr ct)
1553
1554 kickOutAfterUnification :: TcTyVar -> TcS Int
1555 kickOutAfterUnification new_tv
1556 = do { ics <- getInertCans
1557 ; (n_kicked, ics2) <- kickOutRewritable (Given,NomEq)
1558 new_tv ics
1559 -- Given because the tv := xi is given; NomEq because
1560 -- only nominal equalities are solved by unification
1561
1562 ; setInertCans ics2
1563 ; return n_kicked }
1564
1565 {- Note [kickOutRewritable]
1566 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1567 See also Note [inert_eqs: the inert equalities].
1568
1569 When we add a new inert equality (a ~N ty) to the inert set,
1570 we must kick out any inert items that could be rewritten by the
1571 new equality, to maintain the inert-set invariants.
1572
1573 - We want to kick out an existing inert constraint if
1574 a) the new constraint can rewrite the inert one
1575 b) 'a' is free in the inert constraint (so that it *will*)
1576 rewrite it if we kick it out.
1577
1578 For (b) we use tyCoVarsOfCt, which returns the type variables /and
1579 the kind variables/ that are directly visible in the type. Hence
1580 we will have exposed all the rewriting we care about to make the
1581 most precise kinds visible for matching classes etc. No need to
1582 kick out constraints that mention type variables whose kinds
1583 contain this variable!
1584
1585 - A Derived equality can kick out [D] constraints in inert_eqs,
1586 inert_dicts, inert_irreds etc.
1587
1588 - We don't kick out constraints from inert_solved_dicts, and
1589 inert_solved_funeqs optimistically. But when we lookup we have to
1590 take the substitution into account
1591
1592
1593 Note [Kick out insolubles]
1594 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1595 Suppose we have an insoluble alpha ~ [alpha], which is insoluble
1596 because an occurs check. And then we unify alpha := [Int].
1597 Then we really want to rewrite the insoluble to [Int] ~ [[Int]].
1598 Now it can be decomposed. Otherwise we end up with a "Can't match
1599 [Int] ~ [[Int]]" which is true, but a bit confusing because the
1600 outer type constructors match.
1601 -}
1602
1603
1604
1605 --------------
1606 addInertSafehask :: InertCans -> Ct -> InertCans
1607 addInertSafehask ics item@(CDictCan { cc_class = cls, cc_tyargs = tys })
1608 = ics { inert_safehask = addDict (inert_dicts ics) cls tys item }
1609
1610 addInertSafehask _ item
1611 = pprPanic "addInertSafehask: can't happen! Inserting " $ ppr item
1612
1613 insertSafeOverlapFailureTcS :: Ct -> TcS ()
1614 -- See Note [Safe Haskell Overlapping Instances Implementation] in TcSimplify
1615 insertSafeOverlapFailureTcS item
1616 = updInertCans (\ics -> addInertSafehask ics item)
1617
1618 getSafeOverlapFailures :: TcS Cts
1619 -- See Note [Safe Haskell Overlapping Instances Implementation] in TcSimplify
1620 getSafeOverlapFailures
1621 = do { IC { inert_safehask = safehask } <- getInertCans
1622 ; return $ foldDicts consCts safehask emptyCts }
1623
1624 --------------
1625 addSolvedDict :: CtEvidence -> Class -> [Type] -> TcS ()
1626 -- Add a new item in the solved set of the monad
1627 -- See Note [Solved dictionaries]
1628 addSolvedDict item cls tys
1629 | isIPPred (ctEvPred item) -- Never cache "solved" implicit parameters (not sure why!)
1630 = return ()
1631 | otherwise
1632 = do { traceTcS "updSolvedSetTcs:" $ ppr item
1633 ; updInertTcS $ \ ics ->
1634 ics { inert_solved_dicts = addDict (inert_solved_dicts ics) cls tys item } }
1635
1636 {- *********************************************************************
1637 * *
1638 Other inert-set operations
1639 * *
1640 ********************************************************************* -}
1641
1642 updInertTcS :: (InertSet -> InertSet) -> TcS ()
1643 -- Modify the inert set with the supplied function
1644 updInertTcS upd_fn
1645 = do { is_var <- getTcSInertsRef
1646 ; wrapTcS (do { curr_inert <- TcM.readTcRef is_var
1647 ; TcM.writeTcRef is_var (upd_fn curr_inert) }) }
1648
1649 getInertCans :: TcS InertCans
1650 getInertCans = do { inerts <- getTcSInerts; return (inert_cans inerts) }
1651
1652 setInertCans :: InertCans -> TcS ()
1653 setInertCans ics = updInertTcS $ \ inerts -> inerts { inert_cans = ics }
1654
1655 updRetInertCans :: (InertCans -> (a, InertCans)) -> TcS a
1656 -- Modify the inert set with the supplied function
1657 updRetInertCans upd_fn
1658 = do { is_var <- getTcSInertsRef
1659 ; wrapTcS (do { inerts <- TcM.readTcRef is_var
1660 ; let (res, cans') = upd_fn (inert_cans inerts)
1661 ; TcM.writeTcRef is_var (inerts { inert_cans = cans' })
1662 ; return res }) }
1663
1664 updInertCans :: (InertCans -> InertCans) -> TcS ()
1665 -- Modify the inert set with the supplied function
1666 updInertCans upd_fn
1667 = updInertTcS $ \ inerts -> inerts { inert_cans = upd_fn (inert_cans inerts) }
1668
1669 updInertDicts :: (DictMap Ct -> DictMap Ct) -> TcS ()
1670 -- Modify the inert set with the supplied function
1671 updInertDicts upd_fn
1672 = updInertCans $ \ ics -> ics { inert_dicts = upd_fn (inert_dicts ics) }
1673
1674 updInertSafehask :: (DictMap Ct -> DictMap Ct) -> TcS ()
1675 -- Modify the inert set with the supplied function
1676 updInertSafehask upd_fn
1677 = updInertCans $ \ ics -> ics { inert_safehask = upd_fn (inert_safehask ics) }
1678
1679 updInertFunEqs :: (FunEqMap Ct -> FunEqMap Ct) -> TcS ()
1680 -- Modify the inert set with the supplied function
1681 updInertFunEqs upd_fn
1682 = updInertCans $ \ ics -> ics { inert_funeqs = upd_fn (inert_funeqs ics) }
1683
1684 updInertIrreds :: (Cts -> Cts) -> TcS ()
1685 -- Modify the inert set with the supplied function
1686 updInertIrreds upd_fn
1687 = updInertCans $ \ ics -> ics { inert_irreds = upd_fn (inert_irreds ics) }
1688
1689 getInertEqs :: TcS (DTyVarEnv EqualCtList)
1690 getInertEqs = do { inert <- getInertCans; return (inert_eqs inert) }
1691
1692 getInertInsols :: TcS Cts
1693 getInertInsols = do { inert <- getInertCans; return (inert_insols inert) }
1694
1695 getInertGivens :: TcS [Ct]
1696 -- Returns the Given constraints in the inert set,
1697 -- with type functions *not* unflattened
1698 getInertGivens
1699 = do { inerts <- getInertCans
1700 ; let all_cts = foldDicts (:) (inert_dicts inerts)
1701 $ foldFunEqs (:) (inert_funeqs inerts)
1702 $ concat (dVarEnvElts (inert_eqs inerts))
1703 ; return (filter isGivenCt all_cts) }
1704
1705 getPendingScDicts :: TcS [Ct]
1706 -- Find all inert Given dictionaries whose cc_pend_sc flag is True
1707 -- Set the flag to False in the inert set, and return that Ct
1708 getPendingScDicts = updRetInertCans get_sc_dicts
1709 where
1710 get_sc_dicts ic@(IC { inert_dicts = dicts })
1711 = (sc_pend_dicts, ic')
1712 where
1713 ic' = ic { inert_dicts = foldr add dicts sc_pend_dicts }
1714
1715 sc_pend_dicts :: [Ct]
1716 sc_pend_dicts = foldDicts get_pending dicts []
1717
1718 get_pending :: Ct -> [Ct] -> [Ct] -- Get dicts with cc_pend_sc = True
1719 -- but flipping the flag
1720 get_pending dict dicts
1721 | Just dict' <- isPendingScDict dict = dict' : dicts
1722 | otherwise = dicts
1723
1724 add :: Ct -> DictMap Ct -> DictMap Ct
1725 add ct@(CDictCan { cc_class = cls, cc_tyargs = tys }) dicts
1726 = addDict dicts cls tys ct
1727 add ct _ = pprPanic "getPendingScDicts" (ppr ct)
1728
1729 getUnsolvedInerts :: TcS ( Bag Implication
1730 , Cts -- Tyvar eqs: a ~ ty
1731 , Cts -- Fun eqs: F a ~ ty
1732 , Cts -- Insoluble
1733 , Cts ) -- All others
1734 -- Return all the unsolved [Wanted] or [Derived] constraints
1735 --
1736 -- Post-condition: the returned simple constraints are all fully zonked
1737 -- (because they come from the inert set)
1738 -- the unsolved implics may not be
1739 getUnsolvedInerts
1740 = do { IC { inert_eqs = tv_eqs
1741 , inert_funeqs = fun_eqs
1742 , inert_irreds = irreds
1743 , inert_dicts = idicts
1744 , inert_insols = insols
1745 } <- getInertCans
1746
1747 ; let unsolved_tv_eqs = foldTyEqs add_if_unsolved tv_eqs emptyCts
1748 unsolved_fun_eqs = foldFunEqs add_if_wanted fun_eqs emptyCts
1749 unsolved_irreds = Bag.filterBag is_unsolved irreds
1750 unsolved_dicts = foldDicts add_if_unsolved idicts emptyCts
1751 unsolved_others = unsolved_irreds `unionBags` unsolved_dicts
1752 unsolved_insols = filterBag is_unsolved insols
1753
1754 ; implics <- getWorkListImplics
1755
1756 ; traceTcS "getUnsolvedInerts" $
1757 vcat [ text " tv eqs =" <+> ppr unsolved_tv_eqs
1758 , text "fun eqs =" <+> ppr unsolved_fun_eqs
1759 , text "insols =" <+> ppr unsolved_insols
1760 , text "others =" <+> ppr unsolved_others
1761 , text "implics =" <+> ppr implics ]
1762
1763 ; return ( implics, unsolved_tv_eqs, unsolved_fun_eqs
1764 , unsolved_insols, unsolved_others) }
1765 where
1766 add_if_unsolved :: Ct -> Cts -> Cts
1767 add_if_unsolved ct cts | is_unsolved ct = ct `consCts` cts
1768 | otherwise = cts
1769
1770 is_unsolved ct = not (isGivenCt ct) -- Wanted or Derived
1771
1772 -- For CFunEqCans we ignore the Derived ones, and keep
1773 -- only the Wanteds for flattening. The Derived ones
1774 -- share a unification variable with the corresponding
1775 -- Wanted, so we definitely don't want to to participate
1776 -- in unflattening
1777 -- See Note [Type family equations]
1778 add_if_wanted ct cts | isWantedCt ct = ct `consCts` cts
1779 | otherwise = cts
1780
1781 isInInertEqs :: DTyVarEnv EqualCtList -> TcTyVar -> TcType -> Bool
1782 -- True if (a ~N ty) is in the inert set, in either Given or Wanted
1783 isInInertEqs eqs tv rhs
1784 = case lookupDVarEnv eqs tv of
1785 Nothing -> False
1786 Just cts -> any (same_pred rhs) cts
1787 where
1788 same_pred rhs ct
1789 | CTyEqCan { cc_rhs = rhs2, cc_eq_rel = eq_rel } <- ct
1790 , NomEq <- eq_rel
1791 , rhs `eqType` rhs2 = True
1792 | otherwise = False
1793
1794 getNoGivenEqs :: TcLevel -- TcLevel of this implication
1795 -> [TcTyVar] -- Skolems of this implication
1796 -> TcS ( Bool -- True <=> definitely no residual given equalities
1797 , Cts ) -- Insoluble constraints arising from givens
1798 -- See Note [When does an implication have given equalities?]
1799 getNoGivenEqs tclvl skol_tvs
1800 = do { inerts@(IC { inert_eqs = ieqs, inert_irreds = iirreds
1801 , inert_funeqs = funeqs
1802 , inert_insols = insols })
1803 <- getInertCans
1804 ; let local_fsks = foldFunEqs add_fsk funeqs emptyVarSet
1805
1806 has_given_eqs = foldrBag ((||) . ev_given_here . ctEvidence) False
1807 (iirreds `unionBags` insols)
1808 || anyDVarEnv (eqs_given_here local_fsks) ieqs
1809
1810 ; traceTcS "getNoGivenEqs" (vcat [ ppr has_given_eqs, ppr inerts
1811 , ppr insols])
1812 ; return (not has_given_eqs, insols) }
1813 where
1814 eqs_given_here :: VarSet -> EqualCtList -> Bool
1815 eqs_given_here local_fsks [CTyEqCan { cc_tyvar = tv, cc_ev = ev }]
1816 -- Givens are always a sigleton
1817 = not (skolem_bound_here local_fsks tv) && ev_given_here ev
1818 eqs_given_here _ _ = False
1819
1820 ev_given_here :: CtEvidence -> Bool
1821 -- True for a Given bound by the curent implication,
1822 -- i.e. the current level
1823 ev_given_here ev
1824 = isGiven ev
1825 && tclvl == ctLocLevel (ctEvLoc ev)
1826
1827 add_fsk :: Ct -> VarSet -> VarSet
1828 add_fsk ct fsks | CFunEqCan { cc_fsk = tv, cc_ev = ev } <- ct
1829 , isGiven ev = extendVarSet fsks tv
1830 | otherwise = fsks
1831
1832 skol_tv_set = mkVarSet skol_tvs
1833 skolem_bound_here local_fsks tv -- See Note [Let-bound skolems]
1834 = case tcTyVarDetails tv of
1835 SkolemTv {} -> tv `elemVarSet` skol_tv_set
1836 FlatSkol {} -> not (tv `elemVarSet` local_fsks)
1837 _ -> False
1838
1839 -- | Returns Given constraints that might,
1840 -- potentially, match the given pred. This is used when checking to see if a
1841 -- Given might overlap with an instance. See Note [Instance and Given overlap]
1842 -- in TcInteract.
1843 matchableGivens :: CtLoc -> PredType -> InertSet -> Cts
1844 matchableGivens loc_w pred (IS { inert_cans = inert_cans })
1845 = filterBag matchable_given all_relevant_givens
1846 where
1847 -- just look in class constraints and irreds. matchableGivens does get called
1848 -- for ~R constraints, but we don't need to look through equalities, because
1849 -- canonical equalities are used for rewriting. We'll only get caught by
1850 -- non-canonical -- that is, irreducible -- equalities.
1851 all_relevant_givens :: Cts
1852 all_relevant_givens
1853 | Just (clas, _) <- getClassPredTys_maybe pred
1854 = findDictsByClass (inert_dicts inert_cans) clas
1855 `unionBags` inert_irreds inert_cans
1856 | otherwise
1857 = inert_irreds inert_cans
1858
1859 matchable_given :: Ct -> Bool
1860 matchable_given ct
1861 | CtGiven { ctev_loc = loc_g } <- ctev
1862 , Just _ <- tcUnifyTys bind_meta_tv [ctEvPred ctev] [pred]
1863 , not (prohibitedSuperClassSolve loc_g loc_w)
1864 = True
1865
1866 | otherwise
1867 = False
1868 where
1869 ctev = cc_ev ct
1870
1871 bind_meta_tv :: TcTyVar -> BindFlag
1872 -- Any meta tyvar may be unified later, so we treat it as
1873 -- bindable when unifying with givens. That ensures that we
1874 -- conservatively assume that a meta tyvar might get unified with
1875 -- something that matches the 'given', until demonstrated
1876 -- otherwise.
1877 bind_meta_tv tv | isMetaTyVar tv = BindMe
1878 | otherwise = Skolem
1879
1880 prohibitedSuperClassSolve :: CtLoc -> CtLoc -> Bool
1881 -- See Note [Solving superclass constraints] in TcInstDcls
1882 prohibitedSuperClassSolve from_loc solve_loc
1883 | GivenOrigin (InstSC given_size) <- ctLocOrigin from_loc
1884 , ScOrigin wanted_size <- ctLocOrigin solve_loc
1885 = given_size >= wanted_size
1886 | otherwise
1887 = False
1888
1889 {- Note [Unsolved Derived equalities]
1890 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1891 In getUnsolvedInerts, we return a derived equality from the inert_eqs
1892 because it is a candidate for floating out of this implication. We
1893 only float equalities with a meta-tyvar on the left, so we only pull
1894 those out here.
1895
1896 Note [When does an implication have given equalities?]
1897 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1898 Consider an implication
1899 beta => alpha ~ Int
1900 where beta is a unification variable that has already been unified
1901 to () in an outer scope. Then we can float the (alpha ~ Int) out
1902 just fine. So when deciding whether the givens contain an equality,
1903 we should canonicalise first, rather than just looking at the original
1904 givens (Trac #8644).
1905
1906 So we simply look at the inert, canonical Givens and see if there are
1907 any equalities among them, the calculation of has_given_eqs. There
1908 are some wrinkles:
1909
1910 * We must know which ones are bound in *this* implication and which
1911 are bound further out. We can find that out from the TcLevel
1912 of the Given, which is itself recorded in the tcl_tclvl field
1913 of the TcLclEnv stored in the Given (ev_given_here).
1914
1915 What about interactions between inner and outer givens?
1916 - Outer given is rewritten by an inner given, then there must
1917 have been an inner given equality, hence the “given-eq” flag
1918 will be true anyway.
1919
1920 - Inner given rewritten by outer, retains its level (ie. The inner one)
1921
1922 * We must take account of *potential* equalities, like the one above:
1923 beta => ...blah...
1924 If we still don't know what beta is, we conservatively treat it as potentially
1925 becoming an equality. Hence including 'irreds' in the calculation or has_given_eqs.
1926
1927 * When flattening givens, we generate Given equalities like
1928 <F [a]> : F [a] ~ f,
1929 with Refl evidence, and we *don't* want those to count as an equality
1930 in the givens! After all, the entire flattening business is just an
1931 internal matter, and the evidence does not mention any of the 'givens'
1932 of this implication. So we do not treat inert_funeqs as a 'given equality'.
1933
1934 * See Note [Let-bound skolems] for another wrinkle
1935
1936 * We do *not* need to worry about representational equalities, because
1937 these do not affect the ability to float constraints.
1938
1939 Note [Let-bound skolems]
1940 ~~~~~~~~~~~~~~~~~~~~~~~~
1941 If * the inert set contains a canonical Given CTyEqCan (a ~ ty)
1942 and * 'a' is a skolem bound in this very implication, b
1943
1944 then:
1945 a) The Given is pretty much a let-binding, like
1946 f :: (a ~ b->c) => a -> a
1947 Here the equality constraint is like saying
1948 let a = b->c in ...
1949 It is not adding any new, local equality information,
1950 and hence can be ignored by has_given_eqs
1951
1952 b) 'a' will have been completely substituted out in the inert set,
1953 so we can safely discard it. Notably, it doesn't need to be
1954 returned as part of 'fsks'
1955
1956 For an example, see Trac #9211.
1957 -}
1958
1959 removeInertCts :: [Ct] -> InertCans -> InertCans
1960 -- ^ Remove inert constraints from the 'InertCans', for use when a
1961 -- typechecker plugin wishes to discard a given.
1962 removeInertCts cts icans = foldl' removeInertCt icans cts
1963
1964 removeInertCt :: InertCans -> Ct -> InertCans
1965 removeInertCt is ct =
1966 case ct of
1967
1968 CDictCan { cc_class = cl, cc_tyargs = tys } ->
1969 is { inert_dicts = delDict (inert_dicts is) cl tys }
1970
1971 CFunEqCan { cc_fun = tf, cc_tyargs = tys } ->
1972 is { inert_funeqs = delFunEq (inert_funeqs is) tf tys }
1973
1974 CTyEqCan { cc_tyvar = x, cc_rhs = ty } ->
1975 is { inert_eqs = delTyEq (inert_eqs is) x ty }
1976
1977 CIrredEvCan {} -> panic "removeInertCt: CIrredEvCan"
1978 CNonCanonical {} -> panic "removeInertCt: CNonCanonical"
1979 CHoleCan {} -> panic "removeInertCt: CHoleCan"
1980
1981
1982 lookupFlatCache :: TyCon -> [Type] -> TcS (Maybe (TcCoercion, TcType, CtFlavour))
1983 lookupFlatCache fam_tc tys
1984 = do { IS { inert_flat_cache = flat_cache
1985 , inert_cans = IC { inert_funeqs = inert_funeqs } } <- getTcSInerts
1986 ; return (firstJusts [lookup_inerts inert_funeqs,
1987 lookup_flats flat_cache]) }
1988 where
1989 lookup_inerts inert_funeqs
1990 | Just (CFunEqCan { cc_ev = ctev, cc_fsk = fsk, cc_tyargs = xis })
1991 <- findFunEq inert_funeqs fam_tc tys
1992 , tys `eqTypes` xis -- The lookup might find a near-match; see
1993 -- Note [Use loose types in inert set]
1994 = Just (ctEvCoercion ctev, mkTyVarTy fsk, ctEvFlavour ctev)
1995 | otherwise = Nothing
1996
1997 lookup_flats flat_cache = findExactFunEq flat_cache fam_tc tys
1998
1999
2000 lookupInInerts :: TcPredType -> TcS (Maybe CtEvidence)
2001 -- Is this exact predicate type cached in the solved or canonicals of the InertSet?
2002 lookupInInerts pty
2003 | ClassPred cls tys <- classifyPredType pty
2004 = do { inerts <- getTcSInerts
2005 ; return (lookupSolvedDict inerts cls tys `mplus`
2006 lookupInertDict (inert_cans inerts) cls tys) }
2007 | otherwise -- NB: No caching for equalities, IPs, holes, or errors
2008 = return Nothing
2009
2010 -- | Look up a dictionary inert. NB: the returned 'CtEvidence' might not
2011 -- match the input exactly. Note [Use loose types in inert set].
2012 lookupInertDict :: InertCans -> Class -> [Type] -> Maybe CtEvidence
2013 lookupInertDict (IC { inert_dicts = dicts }) cls tys
2014 = case findDict dicts cls tys of
2015 Just ct -> Just (ctEvidence ct)
2016 _ -> Nothing
2017
2018 -- | Look up a solved inert. NB: the returned 'CtEvidence' might not
2019 -- match the input exactly. See Note [Use loose types in inert set].
2020 lookupSolvedDict :: InertSet -> Class -> [Type] -> Maybe CtEvidence
2021 -- Returns just if exactly this predicate type exists in the solved.
2022 lookupSolvedDict (IS { inert_solved_dicts = solved }) cls tys
2023 = case findDict solved cls tys of
2024 Just ev -> Just ev
2025 _ -> Nothing
2026
2027 {- *********************************************************************
2028 * *
2029 Irreds
2030 * *
2031 ********************************************************************* -}
2032
2033 foldIrreds :: (Ct -> b -> b) -> Cts -> b -> b
2034 foldIrreds k irreds z = foldrBag k z irreds
2035
2036
2037 {- *********************************************************************
2038 * *
2039 TcAppMap
2040 * *
2041 ************************************************************************
2042
2043 Note [Use loose types in inert set]
2044 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2045 Say we know (Eq (a |> c1)) and we need (Eq (a |> c2)). One is clearly
2046 solvable from the other. So, we do lookup in the inert set using
2047 loose types, which omit the kind-check.
2048
2049 We must be careful when using the result of a lookup because it may
2050 not match the requsted info exactly!
2051
2052 -}
2053
2054 type TcAppMap a = UniqDFM (ListMap LooseTypeMap a)
2055 -- Indexed by tycon then the arg types, using "loose" matching, where
2056 -- we don't require kind equality. This allows, for example, (a |> co)
2057 -- to match (a).
2058 -- See Note [Use loose types in inert set]
2059 -- Used for types and classes; hence UniqDFM
2060 -- See Note [foldTM determinism] for why we use UniqDFM here
2061
2062 isEmptyTcAppMap :: TcAppMap a -> Bool
2063 isEmptyTcAppMap m = isNullUDFM m
2064
2065 emptyTcAppMap :: TcAppMap a
2066 emptyTcAppMap = emptyUDFM
2067
2068 findTcApp :: TcAppMap a -> Unique -> [Type] -> Maybe a
2069 findTcApp m u tys = do { tys_map <- lookupUDFM m u
2070 ; lookupTM tys tys_map }
2071
2072 delTcApp :: TcAppMap a -> Unique -> [Type] -> TcAppMap a
2073 delTcApp m cls tys = adjustUDFM (deleteTM tys) m cls
2074
2075 insertTcApp :: TcAppMap a -> Unique -> [Type] -> a -> TcAppMap a
2076 insertTcApp m cls tys ct = alterUDFM alter_tm m cls
2077 where
2078 alter_tm mb_tm = Just (insertTM tys ct (mb_tm `orElse` emptyTM))
2079
2080 -- mapTcApp :: (a->b) -> TcAppMap a -> TcAppMap b
2081 -- mapTcApp f = mapUDFM (mapTM f)
2082
2083 filterTcAppMap :: (Ct -> Bool) -> TcAppMap Ct -> TcAppMap Ct
2084 filterTcAppMap f m
2085 = mapUDFM do_tm m
2086 where
2087 do_tm tm = foldTM insert_mb tm emptyTM
2088 insert_mb ct tm
2089 | f ct = insertTM tys ct tm
2090 | otherwise = tm
2091 where
2092 tys = case ct of
2093 CFunEqCan { cc_tyargs = tys } -> tys
2094 CDictCan { cc_tyargs = tys } -> tys
2095 _ -> pprPanic "filterTcAppMap" (ppr ct)
2096
2097 tcAppMapToBag :: TcAppMap a -> Bag a
2098 tcAppMapToBag m = foldTcAppMap consBag m emptyBag
2099
2100 foldTcAppMap :: (a -> b -> b) -> TcAppMap a -> b -> b
2101 foldTcAppMap k m z = foldUDFM (foldTM k) z m
2102
2103
2104 {- *********************************************************************
2105 * *
2106 DictMap
2107 * *
2108 ********************************************************************* -}
2109
2110 type DictMap a = TcAppMap a
2111
2112 emptyDictMap :: DictMap a
2113 emptyDictMap = emptyTcAppMap
2114
2115 -- sizeDictMap :: DictMap a -> Int
2116 -- sizeDictMap m = foldDicts (\ _ x -> x+1) m 0
2117
2118 findDict :: DictMap a -> Class -> [Type] -> Maybe a
2119 findDict m cls tys = findTcApp m (getUnique cls) tys
2120
2121 findDictsByClass :: DictMap a -> Class -> Bag a
2122 findDictsByClass m cls
2123 | Just tm <- lookupUDFM m cls = foldTM consBag tm emptyBag
2124 | otherwise = emptyBag
2125
2126 delDict :: DictMap a -> Class -> [Type] -> DictMap a
2127 delDict m cls tys = delTcApp m (getUnique cls) tys
2128
2129 addDict :: DictMap a -> Class -> [Type] -> a -> DictMap a
2130 addDict m cls tys item = insertTcApp m (getUnique cls) tys item
2131
2132 addDictsByClass :: DictMap Ct -> Class -> Bag Ct -> DictMap Ct
2133 addDictsByClass m cls items
2134 = addToUDFM m cls (foldrBag add emptyTM items)
2135 where
2136 add ct@(CDictCan { cc_tyargs = tys }) tm = insertTM tys ct tm
2137 add ct _ = pprPanic "addDictsByClass" (ppr ct)
2138
2139 filterDicts :: (Ct -> Bool) -> DictMap Ct -> DictMap Ct
2140 filterDicts f m = filterTcAppMap f m
2141
2142 partitionDicts :: (Ct -> Bool) -> DictMap Ct -> (Bag Ct, DictMap Ct)
2143 partitionDicts f m = foldTcAppMap k m (emptyBag, emptyDicts)
2144 where
2145 k ct (yeses, noes) | f ct = (ct `consBag` yeses, noes)
2146 | otherwise = (yeses, add ct noes)
2147 add ct@(CDictCan { cc_class = cls, cc_tyargs = tys }) m
2148 = addDict m cls tys ct
2149 add ct _ = pprPanic "partitionDicts" (ppr ct)
2150
2151 dictsToBag :: DictMap a -> Bag a
2152 dictsToBag = tcAppMapToBag
2153
2154 foldDicts :: (a -> b -> b) -> DictMap a -> b -> b
2155 foldDicts = foldTcAppMap
2156
2157 emptyDicts :: DictMap a
2158 emptyDicts = emptyTcAppMap
2159
2160
2161 {- *********************************************************************
2162 * *
2163 FunEqMap
2164 * *
2165 ********************************************************************* -}
2166
2167 type FunEqMap a = TcAppMap a -- A map whose key is a (TyCon, [Type]) pair
2168
2169 emptyFunEqs :: TcAppMap a
2170 emptyFunEqs = emptyTcAppMap
2171
2172 findFunEq :: FunEqMap a -> TyCon -> [Type] -> Maybe a
2173 findFunEq m tc tys = findTcApp m (getUnique tc) tys
2174
2175 funEqsToBag :: FunEqMap a -> Bag a
2176 funEqsToBag m = foldTcAppMap consBag m emptyBag
2177
2178 findFunEqsByTyCon :: FunEqMap a -> TyCon -> [a]
2179 -- Get inert function equation constraints that have the given tycon
2180 -- in their head. Not that the constraints remain in the inert set.
2181 -- We use this to check for derived interactions with built-in type-function
2182 -- constructors.
2183 findFunEqsByTyCon m tc
2184 | Just tm <- lookupUDFM m tc = foldTM (:) tm []
2185 | otherwise = []
2186
2187 foldFunEqs :: (a -> b -> b) -> FunEqMap a -> b -> b
2188 foldFunEqs = foldTcAppMap
2189
2190 -- mapFunEqs :: (a -> b) -> FunEqMap a -> FunEqMap b
2191 -- mapFunEqs = mapTcApp
2192
2193 -- filterFunEqs :: (Ct -> Bool) -> FunEqMap Ct -> FunEqMap Ct
2194 -- filterFunEqs = filterTcAppMap
2195
2196 insertFunEq :: FunEqMap a -> TyCon -> [Type] -> a -> FunEqMap a
2197 insertFunEq m tc tys val = insertTcApp m (getUnique tc) tys val
2198
2199 partitionFunEqs :: (Ct -> Bool) -> FunEqMap Ct -> ([Ct], FunEqMap Ct)
2200 -- Optimise for the case where the predicate is false
2201 -- partitionFunEqs is called only from kick-out, and kick-out usually
2202 -- kicks out very few equalities, so we want to optimise for that case
2203 partitionFunEqs f m = (yeses, foldr del m yeses)
2204 where
2205 yeses = foldTcAppMap k m []
2206 k ct yeses | f ct = ct : yeses
2207 | otherwise = yeses
2208 del (CFunEqCan { cc_fun = tc, cc_tyargs = tys }) m
2209 = delFunEq m tc tys
2210 del ct _ = pprPanic "partitionFunEqs" (ppr ct)
2211
2212 delFunEq :: FunEqMap a -> TyCon -> [Type] -> FunEqMap a
2213 delFunEq m tc tys = delTcApp m (getUnique tc) tys
2214
2215 ------------------------------
2216 type ExactFunEqMap a = UniqFM (ListMap TypeMap a)
2217
2218 emptyExactFunEqs :: ExactFunEqMap a
2219 emptyExactFunEqs = emptyUFM
2220
2221 findExactFunEq :: ExactFunEqMap a -> TyCon -> [Type] -> Maybe a
2222 findExactFunEq m tc tys = do { tys_map <- lookupUFM m (getUnique tc)
2223 ; lookupTM tys tys_map }
2224
2225 insertExactFunEq :: ExactFunEqMap a -> TyCon -> [Type] -> a -> ExactFunEqMap a
2226 insertExactFunEq m tc tys val = alterUFM alter_tm m (getUnique tc)
2227 where alter_tm mb_tm = Just (insertTM tys val (mb_tm `orElse` emptyTM))
2228
2229 {-
2230 ************************************************************************
2231 * *
2232 * The TcS solver monad *
2233 * *
2234 ************************************************************************
2235
2236 Note [The TcS monad]
2237 ~~~~~~~~~~~~~~~~~~~~
2238 The TcS monad is a weak form of the main Tc monad
2239
2240 All you can do is
2241 * fail
2242 * allocate new variables
2243 * fill in evidence variables
2244
2245 Filling in a dictionary evidence variable means to create a binding
2246 for it, so TcS carries a mutable location where the binding can be
2247 added. This is initialised from the innermost implication constraint.
2248 -}
2249
2250 data TcSEnv
2251 = TcSEnv {
2252 tcs_ev_binds :: EvBindsVar,
2253
2254 tcs_unified :: IORef Int,
2255 -- The number of unification variables we have filled
2256 -- The important thing is whether it is non-zero
2257
2258 tcs_count :: IORef Int, -- Global step count
2259
2260 tcs_inerts :: IORef InertSet, -- Current inert set
2261
2262 -- The main work-list and the flattening worklist
2263 -- See Note [Work list priorities] and
2264 tcs_worklist :: IORef WorkList -- Current worklist
2265 }
2266
2267 ---------------
2268 newtype TcS a = TcS { unTcS :: TcSEnv -> TcM a }
2269
2270 instance Functor TcS where
2271 fmap f m = TcS $ fmap f . unTcS m
2272
2273 instance Applicative TcS where
2274 pure x = TcS (\_ -> return x)
2275 (<*>) = ap
2276
2277 instance Monad TcS where
2278 fail err = TcS (\_ -> fail err)
2279 m >>= k = TcS (\ebs -> unTcS m ebs >>= \r -> unTcS (k r) ebs)
2280
2281 #if __GLASGOW_HASKELL__ > 710
2282 instance MonadFail.MonadFail TcS where
2283 fail err = TcS (\_ -> fail err)
2284 #endif
2285
2286 instance MonadUnique TcS where
2287 getUniqueSupplyM = wrapTcS getUniqueSupplyM
2288
2289 -- Basic functionality
2290 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2291 wrapTcS :: TcM a -> TcS a
2292 -- Do not export wrapTcS, because it promotes an arbitrary TcM to TcS,
2293 -- and TcS is supposed to have limited functionality
2294 wrapTcS = TcS . const -- a TcM action will not use the TcEvBinds
2295
2296 wrapErrTcS :: TcM a -> TcS a
2297 -- The thing wrapped should just fail
2298 -- There's no static check; it's up to the user
2299 -- Having a variant for each error message is too painful
2300 wrapErrTcS = wrapTcS
2301
2302 wrapWarnTcS :: TcM a -> TcS a
2303 -- The thing wrapped should just add a warning, or no-op
2304 -- There's no static check; it's up to the user
2305 wrapWarnTcS = wrapTcS
2306
2307 failTcS, panicTcS :: SDoc -> TcS a
2308 warnTcS :: WarningFlag -> SDoc -> TcS ()
2309 addErrTcS :: SDoc -> TcS ()
2310 failTcS = wrapTcS . TcM.failWith
2311 warnTcS flag = wrapTcS . TcM.addWarn (Reason flag)
2312 addErrTcS = wrapTcS . TcM.addErr
2313 panicTcS doc = pprPanic "TcCanonical" doc
2314
2315 traceTcS :: String -> SDoc -> TcS ()
2316 traceTcS herald doc = wrapTcS (TcM.traceTc herald doc)
2317
2318 runTcPluginTcS :: TcPluginM a -> TcS a
2319 runTcPluginTcS m = wrapTcS . runTcPluginM m =<< getTcEvBindsVar
2320
2321 instance HasDynFlags TcS where
2322 getDynFlags = wrapTcS getDynFlags
2323
2324 getGlobalRdrEnvTcS :: TcS GlobalRdrEnv
2325 getGlobalRdrEnvTcS = wrapTcS TcM.getGlobalRdrEnv
2326
2327 bumpStepCountTcS :: TcS ()
2328 bumpStepCountTcS = TcS $ \env -> do { let ref = tcs_count env
2329 ; n <- TcM.readTcRef ref
2330 ; TcM.writeTcRef ref (n+1) }
2331
2332 csTraceTcS :: SDoc -> TcS ()
2333 csTraceTcS doc
2334 = wrapTcS $ csTraceTcM (return doc)
2335
2336 traceFireTcS :: CtEvidence -> SDoc -> TcS ()
2337 -- Dump a rule-firing trace
2338 traceFireTcS ev doc
2339 = TcS $ \env -> csTraceTcM $
2340 do { n <- TcM.readTcRef (tcs_count env)
2341 ; tclvl <- TcM.getTcLevel
2342 ; return (hang (text "Step" <+> int n
2343 <> brackets (text "l:" <> ppr tclvl <> comma <>
2344 text "d:" <> ppr (ctLocDepth (ctEvLoc ev)))
2345 <+> doc <> colon)
2346 4 (ppr ev)) }
2347
2348 csTraceTcM :: TcM SDoc -> TcM ()
2349 -- Constraint-solver tracing, -ddump-cs-trace
2350 csTraceTcM mk_doc
2351 = do { dflags <- getDynFlags
2352 ; when ( dopt Opt_D_dump_cs_trace dflags
2353 || dopt Opt_D_dump_tc_trace dflags )
2354 ( do { msg <- mk_doc
2355 ; TcM.traceTcRn Opt_D_dump_cs_trace msg }) }
2356
2357 runTcS :: TcS a -- What to run
2358 -> TcM (a, EvBindMap)
2359 runTcS tcs
2360 = do { ev_binds_var <- TcM.newTcEvBinds
2361 ; res <- runTcSWithEvBinds ev_binds_var tcs
2362 ; ev_binds <- TcM.getTcEvBindsMap ev_binds_var
2363 ; return (res, ev_binds) }
2364
2365 -- | This variant of 'runTcS' will keep solving, even when only Deriveds
2366 -- are left around. It also doesn't return any evidence, as callers won't
2367 -- need it.
2368 runTcSDeriveds :: TcS a -> TcM a
2369 runTcSDeriveds tcs
2370 = do { ev_binds_var <- TcM.newTcEvBinds
2371 ; runTcSWithEvBinds ev_binds_var tcs }
2372
2373 -- | This can deal only with equality constraints.
2374 runTcSEqualities :: TcS a -> TcM a
2375 runTcSEqualities thing_inside
2376 = do { ev_binds_var <- TcM.newTcEvBinds
2377 ; runTcSWithEvBinds ev_binds_var thing_inside }
2378
2379 runTcSWithEvBinds :: EvBindsVar
2380 -> TcS a
2381 -> TcM a
2382 runTcSWithEvBinds ev_binds_var tcs
2383 = do { unified_var <- TcM.newTcRef 0
2384 ; step_count <- TcM.newTcRef 0
2385 ; inert_var <- TcM.newTcRef emptyInert
2386 ; wl_var <- TcM.newTcRef emptyWorkList
2387 ; let env = TcSEnv { tcs_ev_binds = ev_binds_var
2388 , tcs_unified = unified_var
2389 , tcs_count = step_count
2390 , tcs_inerts = inert_var
2391 , tcs_worklist = wl_var }
2392
2393 -- Run the computation
2394 ; res <- unTcS tcs env
2395
2396 ; count <- TcM.readTcRef step_count
2397 ; when (count > 0) $
2398 csTraceTcM $ return (text "Constraint solver steps =" <+> int count)
2399
2400 #if defined(DEBUG)
2401 ; ev_binds <- TcM.getTcEvBindsMap ev_binds_var
2402 ; checkForCyclicBinds ev_binds
2403 #endif
2404
2405 ; return res }
2406
2407 #if defined(DEBUG)
2408 checkForCyclicBinds :: EvBindMap -> TcM ()
2409 checkForCyclicBinds ev_binds_map
2410 | null cycles
2411 = return ()
2412 | null coercion_cycles
2413 = TcM.traceTc "Cycle in evidence binds" $ ppr cycles
2414 | otherwise
2415 = pprPanic "Cycle in coercion bindings" $ ppr coercion_cycles
2416 where
2417 ev_binds = evBindMapBinds ev_binds_map
2418
2419 cycles :: [[EvBind]]
2420 cycles = [c | CyclicSCC c <- stronglyConnCompFromEdgedVerticesUniq edges]
2421
2422 coercion_cycles = [c | c <- cycles, any is_co_bind c]
2423 is_co_bind (EvBind { eb_lhs = b }) = isEqPred (varType b)
2424
2425 edges :: [ Node EvVar EvBind ]
2426 edges = [ DigraphNode bind bndr (nonDetEltsUniqSet (evVarsOfTerm rhs))
2427 | bind@(EvBind { eb_lhs = bndr, eb_rhs = rhs}) <- bagToList ev_binds ]
2428 -- It's OK to use nonDetEltsUFM here as
2429 -- stronglyConnCompFromEdgedVertices is still deterministic even
2430 -- if the edges are in nondeterministic order as explained in
2431 -- Note [Deterministic SCC] in Digraph.
2432 #endif
2433
2434 setEvBindsTcS :: EvBindsVar -> TcS a -> TcS a
2435 setEvBindsTcS ref (TcS thing_inside)
2436 = TcS $ \ env -> thing_inside (env { tcs_ev_binds = ref })
2437
2438 nestImplicTcS :: EvBindsVar
2439 -> TcLevel -> TcS a
2440 -> TcS a
2441 nestImplicTcS ref inner_tclvl (TcS thing_inside)
2442 = TcS $ \ TcSEnv { tcs_unified = unified_var
2443 , tcs_inerts = old_inert_var
2444 , tcs_count = count
2445 } ->
2446 do { inerts <- TcM.readTcRef old_inert_var
2447 ; let nest_inert = inerts { inert_flat_cache = emptyExactFunEqs }
2448 -- See Note [Do not inherit the flat cache]
2449 ; new_inert_var <- TcM.newTcRef nest_inert
2450 ; new_wl_var <- TcM.newTcRef emptyWorkList
2451 ; let nest_env = TcSEnv { tcs_ev_binds = ref
2452 , tcs_unified = unified_var
2453 , tcs_count = count
2454 , tcs_inerts = new_inert_var
2455 , tcs_worklist = new_wl_var }
2456 ; res <- TcM.setTcLevel inner_tclvl $
2457 thing_inside nest_env
2458
2459 #if defined(DEBUG)
2460 -- Perform a check that the thing_inside did not cause cycles
2461 ; ev_binds <- TcM.getTcEvBindsMap ref
2462 ; checkForCyclicBinds ev_binds
2463 #endif
2464 ; return res }
2465
2466 {- Note [Do not inherit the flat cache]
2467 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2468 We do not want to inherit the flat cache when processing nested
2469 implications. Consider
2470 a ~ F b, forall c. b~Int => blah
2471 If we have F b ~ fsk in the flat-cache, and we push that into the
2472 nested implication, we might miss that F b can be rewritten to F Int,
2473 and hence perhpas solve it. Moreover, the fsk from outside is
2474 flattened out after solving the outer level, but and we don't
2475 do that flattening recursively.
2476 -}
2477
2478 nestTcS :: TcS a -> TcS a
2479 -- Use the current untouchables, augmenting the current
2480 -- evidence bindings, and solved dictionaries
2481 -- But have no effect on the InertCans, or on the inert_flat_cache
2482 -- (we want to inherit the latter from processing the Givens)
2483 nestTcS (TcS thing_inside)
2484 = TcS $ \ env@(TcSEnv { tcs_inerts = inerts_var }) ->
2485 do { inerts <- TcM.readTcRef inerts_var
2486 ; new_inert_var <- TcM.newTcRef inerts
2487 ; new_wl_var <- TcM.newTcRef emptyWorkList
2488 ; let nest_env = env { tcs_inerts = new_inert_var
2489 , tcs_worklist = new_wl_var }
2490
2491 ; res <- thing_inside nest_env
2492
2493 ; new_inerts <- TcM.readTcRef new_inert_var
2494
2495 -- we want to propogate the safe haskell failures
2496 ; let old_ic = inert_cans inerts
2497 new_ic = inert_cans new_inerts
2498 nxt_ic = old_ic { inert_safehask = inert_safehask new_ic }
2499
2500 ; TcM.writeTcRef inerts_var -- See Note [Propagate the solved dictionaries]
2501 (inerts { inert_solved_dicts = inert_solved_dicts new_inerts
2502 , inert_cans = nxt_ic })
2503
2504 ; return res }
2505
2506 {-
2507 Note [Propagate the solved dictionaries]
2508 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2509 It's really quite important that nestTcS does not discard the solved
2510 dictionaries from the thing_inside.
2511 Consider
2512 Eq [a]
2513 forall b. empty => Eq [a]
2514 We solve the simple (Eq [a]), under nestTcS, and then turn our attention to
2515 the implications. It's definitely fine to use the solved dictionaries on
2516 the inner implications, and it can make a signficant performance difference
2517 if you do so.
2518 -}
2519
2520 -- Getters and setters of TcEnv fields
2521 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2522
2523 -- Getter of inerts and worklist
2524 getTcSInertsRef :: TcS (IORef InertSet)
2525 getTcSInertsRef = TcS (return . tcs_inerts)
2526
2527 getTcSWorkListRef :: TcS (IORef WorkList)
2528 getTcSWorkListRef = TcS (return . tcs_worklist)
2529
2530 getTcSInerts :: TcS InertSet
2531 getTcSInerts = getTcSInertsRef >>= wrapTcS . (TcM.readTcRef)
2532
2533 setTcSInerts :: InertSet -> TcS ()
2534 setTcSInerts ics = do { r <- getTcSInertsRef; wrapTcS (TcM.writeTcRef r ics) }
2535
2536 getWorkListImplics :: TcS (Bag Implication)
2537 getWorkListImplics
2538 = do { wl_var <- getTcSWorkListRef
2539 ; wl_curr <- wrapTcS (TcM.readTcRef wl_var)
2540 ; return (wl_implics wl_curr) }
2541
2542 updWorkListTcS :: (WorkList -> WorkList) -> TcS ()
2543 updWorkListTcS f
2544 = do { wl_var <- getTcSWorkListRef
2545 ; wl_curr <- wrapTcS (TcM.readTcRef wl_var)
2546 ; let new_work = f wl_curr
2547 ; wrapTcS (TcM.writeTcRef wl_var new_work) }
2548
2549 emitWorkNC :: [CtEvidence] -> TcS ()
2550 emitWorkNC evs
2551 | null evs
2552 = return ()
2553 | otherwise
2554 = emitWork (map mkNonCanonical evs)
2555
2556 emitWork :: [Ct] -> TcS ()
2557 emitWork cts
2558 = do { traceTcS "Emitting fresh work" (vcat (map ppr cts))
2559 ; updWorkListTcS (extendWorkListCts cts) }
2560
2561 emitInsoluble :: Ct -> TcS ()
2562 -- Emits a non-canonical constraint that will stand for a frozen error in the inerts.
2563 emitInsoluble ct
2564 = do { traceTcS "Emit insoluble" (ppr ct $$ pprCtLoc (ctLoc ct))
2565 ; updInertTcS add_insol }
2566 where
2567 this_pred = ctPred ct
2568 add_insol is@(IS { inert_cans = ics@(IC { inert_insols = old_insols }) })
2569 | already_there = is
2570 | otherwise = is { inert_cans = ics { inert_insols = old_insols `snocCts` ct } }
2571 where
2572 already_there = not (isWantedCt ct) && anyBag (tcEqType this_pred . ctPred) old_insols
2573 -- See Note [Do not add duplicate derived insolubles]
2574
2575 newTcRef :: a -> TcS (TcRef a)
2576 newTcRef x = wrapTcS (TcM.newTcRef x)
2577
2578 readTcRef :: TcRef a -> TcS a
2579 readTcRef ref = wrapTcS (TcM.readTcRef ref)
2580
2581 updTcRef :: TcRef a -> (a->a) -> TcS ()
2582 updTcRef ref upd_fn = wrapTcS (TcM.updTcRef ref upd_fn)
2583
2584 getTcEvBindsVar :: TcS EvBindsVar
2585 getTcEvBindsVar = TcS (return . tcs_ev_binds)
2586
2587 getTcLevel :: TcS TcLevel
2588 getTcLevel = wrapTcS TcM.getTcLevel
2589
2590 getTcEvBindsAndTCVs :: EvBindsVar -> TcS (EvBindMap, TyCoVarSet)
2591 getTcEvBindsAndTCVs ev_binds_var
2592 = wrapTcS $ do { bnds <- TcM.getTcEvBindsMap ev_binds_var
2593 ; tcvs <- TcM.getTcEvTyCoVars ev_binds_var
2594 ; return (bnds, tcvs) }
2595
2596 getTcEvBindsMap :: TcS EvBindMap
2597 getTcEvBindsMap
2598 = do { ev_binds_var <- getTcEvBindsVar
2599 ; wrapTcS $ TcM.getTcEvBindsMap ev_binds_var }
2600
2601 unifyTyVar :: TcTyVar -> TcType -> TcS ()
2602 -- Unify a meta-tyvar with a type
2603 -- We keep track of how many unifications have happened in tcs_unified,
2604 --
2605 -- We should never unify the same variable twice!
2606 unifyTyVar tv ty
2607 = ASSERT2( isMetaTyVar tv, ppr tv )
2608 TcS $ \ env ->
2609 do { TcM.traceTc "unifyTyVar" (ppr tv <+> text ":=" <+> ppr ty)
2610 ; TcM.writeMetaTyVar tv ty
2611 ; TcM.updTcRef (tcs_unified env) (+1) }
2612
2613 unflattenFmv :: TcTyVar -> TcType -> TcS ()
2614 -- Fill a flatten-meta-var, simply by unifying it.
2615 -- This does NOT count as a unification in tcs_unified.
2616 unflattenFmv tv ty
2617 = ASSERT2( isMetaTyVar tv, ppr tv )
2618 TcS $ \ _ ->
2619 do { TcM.traceTc "unflattenFmv" (ppr tv <+> text ":=" <+> ppr ty)
2620 ; TcM.writeMetaTyVar tv ty }
2621
2622 reportUnifications :: TcS a -> TcS (Int, a)
2623 reportUnifications (TcS thing_inside)
2624 = TcS $ \ env ->
2625 do { inner_unified <- TcM.newTcRef 0
2626 ; res <- thing_inside (env { tcs_unified = inner_unified })
2627 ; n_unifs <- TcM.readTcRef inner_unified
2628 ; TcM.updTcRef (tcs_unified env) (+ n_unifs)
2629 ; return (n_unifs, res) }
2630
2631 getDefaultInfo :: TcS ([Type], (Bool, Bool))
2632 getDefaultInfo = wrapTcS TcM.tcGetDefaultTys
2633
2634 -- Just get some environments needed for instance looking up and matching
2635 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2636
2637 getInstEnvs :: TcS InstEnvs
2638 getInstEnvs = wrapTcS $ TcM.tcGetInstEnvs
2639
2640 getFamInstEnvs :: TcS (FamInstEnv, FamInstEnv)
2641 getFamInstEnvs = wrapTcS $ FamInst.tcGetFamInstEnvs
2642
2643 getTopEnv :: TcS HscEnv
2644 getTopEnv = wrapTcS $ TcM.getTopEnv
2645
2646 getGblEnv :: TcS TcGblEnv
2647 getGblEnv = wrapTcS $ TcM.getGblEnv
2648
2649 getLclEnv :: TcS TcLclEnv
2650 getLclEnv = wrapTcS $ TcM.getLclEnv
2651
2652 tcLookupClass :: Name -> TcS Class
2653 tcLookupClass c = wrapTcS $ TcM.tcLookupClass c
2654
2655 tcLookupId :: Name -> TcS Id
2656 tcLookupId n = wrapTcS $ TcM.tcLookupId n
2657
2658 -- Setting names as used (used in the deriving of Coercible evidence)
2659 -- Too hackish to expose it to TcS? In that case somehow extract the used
2660 -- constructors from the result of solveInteract
2661 addUsedGREs :: [GlobalRdrElt] -> TcS ()
2662 addUsedGREs gres = wrapTcS $ TcM.addUsedGREs gres
2663
2664 addUsedGRE :: Bool -> GlobalRdrElt -> TcS ()
2665 addUsedGRE warn_if_deprec gre = wrapTcS $ TcM.addUsedGRE warn_if_deprec gre
2666
2667
2668 -- Various smaller utilities [TODO, maybe will be absorbed in the instance matcher]
2669 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2670
2671 checkWellStagedDFun :: PredType -> DFunId -> CtLoc -> TcS ()
2672 checkWellStagedDFun pred dfun_id loc
2673 = wrapTcS $ TcM.setCtLocM loc $
2674 do { use_stage <- TcM.getStage
2675 ; TcM.checkWellStaged pp_thing bind_lvl (thLevel use_stage) }
2676 where
2677 pp_thing = text "instance for" <+> quotes (ppr pred)
2678 bind_lvl = TcM.topIdLvl dfun_id
2679
2680 pprEq :: TcType -> TcType -> SDoc
2681 pprEq ty1 ty2 = pprParendType ty1 <+> char '~' <+> pprParendType ty2
2682
2683 isTouchableMetaTyVarTcS :: TcTyVar -> TcS Bool
2684 isTouchableMetaTyVarTcS tv
2685 = do { tclvl <- getTcLevel
2686 ; return $ isTouchableMetaTyVar tclvl tv }
2687
2688 isFilledMetaTyVar_maybe :: TcTyVar -> TcS (Maybe Type)
2689 isFilledMetaTyVar_maybe tv
2690 = case tcTyVarDetails tv of
2691 MetaTv { mtv_ref = ref }
2692 -> do { cts <- wrapTcS (TcM.readTcRef ref)
2693 ; case cts of
2694 Indirect ty -> return (Just ty)
2695 Flexi -> return Nothing }
2696 _ -> return Nothing
2697
2698 isFilledMetaTyVar :: TcTyVar -> TcS Bool
2699 isFilledMetaTyVar tv = wrapTcS (TcM.isFilledMetaTyVar tv)
2700
2701 zonkTyCoVarsAndFV :: TcTyCoVarSet -> TcS TcTyCoVarSet
2702 zonkTyCoVarsAndFV tvs = wrapTcS (TcM.zonkTyCoVarsAndFV tvs)
2703
2704 zonkTyCoVarsAndFVList :: [TcTyCoVar] -> TcS [TcTyCoVar]
2705 zonkTyCoVarsAndFVList tvs = wrapTcS (TcM.zonkTyCoVarsAndFVList tvs)
2706
2707 zonkCo :: Coercion -> TcS Coercion
2708 zonkCo = wrapTcS . TcM.zonkCo
2709
2710 zonkTcType :: TcType -> TcS TcType
2711 zonkTcType ty = wrapTcS (TcM.zonkTcType ty)
2712
2713 zonkTcTypes :: [TcType] -> TcS [TcType]
2714 zonkTcTypes tys = wrapTcS (TcM.zonkTcTypes tys)
2715
2716 zonkTcTyVar :: TcTyVar -> TcS TcType
2717 zonkTcTyVar tv = wrapTcS (TcM.zonkTcTyVar tv)
2718
2719 zonkSimples :: Cts -> TcS Cts
2720 zonkSimples cts = wrapTcS (TcM.zonkSimples cts)
2721
2722 zonkWC :: WantedConstraints -> TcS WantedConstraints
2723 zonkWC wc = wrapTcS (TcM.zonkWC wc)
2724
2725 {-
2726 Note [Do not add duplicate derived insolubles]
2727 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2728 In general we *must* add an insoluble (Int ~ Bool) even if there is
2729 one such there already, because they may come from distinct call
2730 sites. Not only do we want an error message for each, but with
2731 -fdefer-type-errors we must generate evidence for each. But for
2732 *derived* insolubles, we only want to report each one once. Why?
2733
2734 (a) A constraint (C r s t) where r -> s, say, may generate the same fundep
2735 equality many times, as the original constraint is successively rewritten.
2736
2737 (b) Ditto the successive iterations of the main solver itself, as it traverses
2738 the constraint tree. See example below.
2739
2740 Also for *given* insolubles we may get repeated errors, as we
2741 repeatedly traverse the constraint tree. These are relatively rare
2742 anyway, so removing duplicates seems ok. (Alternatively we could take
2743 the SrcLoc into account.)
2744
2745 Note that the test does not need to be particularly efficient because
2746 it is only used if the program has a type error anyway.
2747
2748 Example of (b): assume a top-level class and instance declaration:
2749
2750 class D a b | a -> b
2751 instance D [a] [a]
2752
2753 Assume we have started with an implication:
2754
2755 forall c. Eq c => { wc_simple = D [c] c [W] }
2756
2757 which we have simplified to:
2758
2759 forall c. Eq c => { wc_simple = D [c] c [W]
2760 , wc_insols = (c ~ [c]) [D] }
2761
2762 For some reason, e.g. because we floated an equality somewhere else,
2763 we might try to re-solve this implication. If we do not do a
2764 dropDerivedWC, then we will end up trying to solve the following
2765 constraints the second time:
2766
2767 (D [c] c) [W]
2768 (c ~ [c]) [D]
2769
2770 which will result in two Deriveds to end up in the insoluble set:
2771
2772 wc_simple = D [c] c [W]
2773 wc_insols = (c ~ [c]) [D], (c ~ [c]) [D]
2774 -}
2775
2776 -- Flatten skolems
2777 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2778 newFlattenSkolem :: CtFlavour -> CtLoc
2779 -> TyCon -> [TcType] -- F xis
2780 -> TcS (CtEvidence, Coercion, TcTyVar) -- [G/WD] x:: F xis ~ fsk
2781 newFlattenSkolem flav loc tc xis
2782 = do { stuff@(ev, co, fsk) <- new_skolem
2783 ; let fsk_ty = mkTyVarTy fsk
2784 ; extendFlatCache tc xis (co, fsk_ty, ctEvFlavour ev)
2785 ; return stuff }
2786 where
2787 fam_ty = mkTyConApp tc xis
2788
2789 new_skolem
2790 | Given <- flav
2791 = do { fsk <- wrapTcS (TcM.newFskTyVar fam_ty)
2792 ; let co = mkNomReflCo fam_ty
2793 ; ev <- newGivenEvVar loc (mkPrimEqPred fam_ty (mkTyVarTy fsk),
2794 EvCoercion co)
2795 ; return (ev, co, fsk) }
2796
2797 | otherwise -- Generate a [WD] for both Wanted and Derived
2798 -- See Note [No Derived CFunEqCans]
2799 = do { fmv <- wrapTcS (TcM.newFmvTyVar fam_ty)
2800 ; (ev, hole_co) <- newWantedEq loc Nominal fam_ty (mkTyVarTy fmv)
2801 ; return (ev, hole_co, fmv) }
2802
2803 extendFlatCache :: TyCon -> [Type] -> (TcCoercion, TcType, CtFlavour) -> TcS ()
2804 extendFlatCache tc xi_args stuff@(_, ty, fl)
2805 | isGivenOrWDeriv fl -- Maintain the invariant that inert_flat_cache
2806 -- only has [G] and [WD] CFunEqCans
2807 = do { dflags <- getDynFlags
2808 ; when (gopt Opt_FlatCache dflags) $
2809 do { traceTcS "extendFlatCache" (vcat [ ppr tc <+> ppr xi_args
2810 , ppr fl, ppr ty ])
2811 -- 'co' can be bottom, in the case of derived items
2812 ; updInertTcS $ \ is@(IS { inert_flat_cache = fc }) ->
2813 is { inert_flat_cache = insertExactFunEq fc tc xi_args stuff } } }
2814
2815 | otherwise
2816 = return ()
2817
2818 -- Instantiations
2819 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2820
2821 instDFunType :: DFunId -> [DFunInstType] -> TcS ([TcType], TcThetaType)
2822 instDFunType dfun_id inst_tys
2823 = wrapTcS $ TcM.instDFunType dfun_id inst_tys
2824
2825 newFlexiTcSTy :: Kind -> TcS TcType
2826 newFlexiTcSTy knd = wrapTcS (TcM.newFlexiTyVarTy knd)
2827
2828 cloneMetaTyVar :: TcTyVar -> TcS TcTyVar
2829 cloneMetaTyVar tv = wrapTcS (TcM.cloneMetaTyVar tv)
2830
2831 demoteUnfilledFmv :: TcTyVar -> TcS ()
2832 -- If a flatten-meta-var is still un-filled,
2833 -- turn it into an ordinary meta-var
2834 demoteUnfilledFmv fmv
2835 = wrapTcS $ do { is_filled <- TcM.isFilledMetaTyVar fmv
2836 ; unless is_filled $
2837 do { tv_ty <- TcM.newFlexiTyVarTy (tyVarKind fmv)
2838 ; TcM.writeMetaTyVar fmv tv_ty } }
2839
2840 instFlexi :: [TKVar] -> TcS TCvSubst
2841 instFlexi = instFlexiX emptyTCvSubst
2842
2843 instFlexiX :: TCvSubst -> [TKVar] -> TcS TCvSubst
2844 instFlexiX subst tvs
2845 = wrapTcS (foldlM instFlexiHelper subst tvs)
2846
2847 instFlexiHelper :: TCvSubst -> TKVar -> TcM TCvSubst
2848 instFlexiHelper subst tv
2849 = do { uniq <- TcM.newUnique
2850 ; details <- TcM.newMetaDetails TauTv
2851 ; let name = setNameUnique (tyVarName tv) uniq
2852 kind = substTyUnchecked subst (tyVarKind tv)
2853 ty' = mkTyVarTy (mkTcTyVar name kind details)
2854 ; return (extendTvSubst subst tv ty') }
2855
2856 tcInstType :: ([TyVar] -> TcM (TCvSubst, [TcTyVar]))
2857 -- ^ How to instantiate the type variables
2858 -> Id -- ^ Type to instantiate
2859 -> TcS ([(Name, TcTyVar)], TcThetaType, TcType) -- ^ Result
2860 -- (type vars, preds (incl equalities), rho)
2861 tcInstType inst_tyvars id = wrapTcS (TcM.tcInstType inst_tyvars id)
2862
2863
2864
2865 -- Creating and setting evidence variables and CtFlavors
2866 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2867
2868 data MaybeNew = Fresh CtEvidence | Cached EvTerm
2869
2870 isFresh :: MaybeNew -> Bool
2871 isFresh (Fresh {}) = True
2872 isFresh (Cached {}) = False
2873
2874 freshGoals :: [MaybeNew] -> [CtEvidence]
2875 freshGoals mns = [ ctev | Fresh ctev <- mns ]
2876
2877 getEvTerm :: MaybeNew -> EvTerm
2878 getEvTerm (Fresh ctev) = ctEvTerm ctev
2879 getEvTerm (Cached evt) = evt
2880
2881 setEvBind :: EvBind -> TcS ()
2882 setEvBind ev_bind
2883 = do { evb <- getTcEvBindsVar
2884 ; wrapTcS $ TcM.addTcEvBind evb ev_bind }
2885
2886 -- | Mark variables as used filling a coercion hole
2887 useVars :: CoVarSet -> TcS ()
2888 useVars vars
2889 = do { EvBindsVar { ebv_tcvs = ref } <- getTcEvBindsVar
2890 ; wrapTcS $
2891 do { tcvs <- TcM.readTcRef ref
2892 ; let tcvs' = tcvs `unionVarSet` vars
2893 ; TcM.writeTcRef ref tcvs' } }
2894
2895 -- | Equalities only
2896 setWantedEq :: TcEvDest -> Coercion -> TcS ()
2897 setWantedEq (HoleDest hole) co
2898 = do { useVars (coVarsOfCo co)
2899 ; wrapTcS $ TcM.fillCoercionHole hole co }
2900 setWantedEq (EvVarDest ev) _ = pprPanic "setWantedEq" (ppr ev)
2901
2902 -- | Equalities only
2903 setEqIfWanted :: CtEvidence -> Coercion -> TcS ()
2904 setEqIfWanted (CtWanted { ctev_dest = dest }) co = setWantedEq dest co
2905 setEqIfWanted _ _ = return ()
2906
2907 -- | Good for equalities and non-equalities
2908 setWantedEvTerm :: TcEvDest -> EvTerm -> TcS ()
2909 setWantedEvTerm (HoleDest hole) tm
2910 = do { let co = evTermCoercion tm
2911 ; useVars (coVarsOfCo co)
2912 ; wrapTcS $ TcM.fillCoercionHole hole co }
2913 setWantedEvTerm (EvVarDest ev) tm = setWantedEvBind ev tm
2914
2915 setWantedEvBind :: EvVar -> EvTerm -> TcS ()
2916 setWantedEvBind ev_id tm = setEvBind (mkWantedEvBind ev_id tm)
2917
2918 setEvBindIfWanted :: CtEvidence -> EvTerm -> TcS ()
2919 setEvBindIfWanted ev tm
2920 = case ev of
2921 CtWanted { ctev_dest = dest }
2922 -> setWantedEvTerm dest tm
2923 _ -> return ()
2924
2925 newTcEvBinds :: TcS EvBindsVar
2926 newTcEvBinds = wrapTcS TcM.newTcEvBinds
2927
2928 newEvVar :: TcPredType -> TcS EvVar
2929 newEvVar pred = wrapTcS (TcM.newEvVar pred)
2930
2931 newGivenEvVar :: CtLoc -> (TcPredType, EvTerm) -> TcS CtEvidence
2932 -- Make a new variable of the given PredType,
2933 -- immediately bind it to the given term
2934 -- and return its CtEvidence
2935 -- See Note [Bind new Givens immediately] in TcRnTypes
2936 newGivenEvVar loc (pred, rhs)
2937 = do { new_ev <- newBoundEvVarId pred rhs
2938 ; return (CtGiven { ctev_pred = pred, ctev_evar = new_ev, ctev_loc = loc }) }
2939
2940 -- | Make a new 'Id' of the given type, bound (in the monad's EvBinds) to the
2941 -- given term
2942 newBoundEvVarId :: TcPredType -> EvTerm -> TcS EvVar
2943 newBoundEvVarId pred rhs
2944 = do { new_ev <- newEvVar pred
2945 ; setEvBind (mkGivenEvBind new_ev rhs)
2946 ; return new_ev }
2947
2948 newGivenEvVars :: CtLoc -> [(TcPredType, EvTerm)] -> TcS [CtEvidence]
2949 newGivenEvVars loc pts = mapM (newGivenEvVar loc) pts
2950
2951 emitNewWantedEq :: CtLoc -> Role -> TcType -> TcType -> TcS Coercion
2952 -- | Emit a new Wanted equality into the work-list
2953 emitNewWantedEq loc role ty1 ty2
2954 | otherwise
2955 = do { (ev, co) <- newWantedEq loc role ty1 ty2
2956 ; updWorkListTcS $
2957 extendWorkListEq (mkNonCanonical ev)
2958 ; return co }
2959
2960 -- | Make a new equality CtEvidence
2961 newWantedEq :: CtLoc -> Role -> TcType -> TcType -> TcS (CtEvidence, Coercion)
2962 newWantedEq loc role ty1 ty2
2963 = do { hole <- wrapTcS $ TcM.newCoercionHole
2964 ; traceTcS "Emitting new coercion hole" (ppr hole <+> dcolon <+> ppr pty)
2965 ; return ( CtWanted { ctev_pred = pty, ctev_dest = HoleDest hole
2966 , ctev_nosh = WDeriv
2967 , ctev_loc = loc}
2968 , mkHoleCo hole role ty1 ty2 ) }
2969 where
2970 pty = mkPrimEqPredRole role ty1 ty2
2971
2972 -- no equalities here. Use newWantedEq instead
2973 newWantedEvVarNC :: CtLoc -> TcPredType -> TcS CtEvidence
2974 -- Don't look up in the solved/inerts; we know it's not there
2975 newWantedEvVarNC loc pty
2976 = do { new_ev <- newEvVar pty
2977 ; traceTcS "Emitting new wanted" (ppr new_ev <+> dcolon <+> ppr pty $$
2978 pprCtLoc loc)
2979 ; return (CtWanted { ctev_pred = pty, ctev_dest = EvVarDest new_ev
2980 , ctev_nosh = WDeriv
2981 , ctev_loc = loc })}
2982
2983 newWantedEvVar :: CtLoc -> TcPredType -> TcS MaybeNew
2984 -- For anything except ClassPred, this is the same as newWantedEvVarNC
2985 newWantedEvVar loc pty
2986 = do { mb_ct <- lookupInInerts pty
2987 ; case mb_ct of
2988 Just ctev
2989 | not (isDerived ctev)
2990 -> do { traceTcS "newWantedEvVar/cache hit" $ ppr ctev
2991 ; return $ Cached (ctEvTerm ctev) }
2992 _ -> do { ctev <- newWantedEvVarNC loc pty
2993 ; return (Fresh ctev) } }
2994
2995 -- deals with both equalities and non equalities. Tries to look
2996 -- up non-equalities in the cache
2997 newWanted :: CtLoc -> PredType -> TcS MaybeNew
2998 newWanted loc pty
2999 | Just (role, ty1, ty2) <- getEqPredTys_maybe pty
3000 = Fresh . fst <$> newWantedEq loc role ty1 ty2
3001 | otherwise
3002 = newWantedEvVar loc pty
3003
3004 -- deals with both equalities and non equalities. Doesn't do any cache lookups.
3005 newWantedNC :: CtLoc -> PredType -> TcS CtEvidence
3006 newWantedNC loc pty
3007 | Just (role, ty1, ty2) <- getEqPredTys_maybe pty
3008 = fst <$> newWantedEq loc role ty1 ty2
3009 | otherwise
3010 = newWantedEvVarNC loc pty
3011
3012 emitNewDerived :: CtLoc -> TcPredType -> TcS ()
3013 emitNewDerived loc pred
3014 = do { ev <- newDerivedNC loc pred
3015 ; traceTcS "Emitting new derived" (ppr ev)
3016 ; updWorkListTcS (extendWorkListDerived loc ev) }
3017
3018 emitNewDeriveds :: CtLoc -> [TcPredType] -> TcS ()
3019 emitNewDeriveds loc preds
3020 | null preds
3021 = return ()
3022 | otherwise
3023 = do { evs <- mapM (newDerivedNC loc) preds
3024 ; traceTcS "Emitting new deriveds" (ppr evs)
3025 ; updWorkListTcS (extendWorkListDeriveds loc evs) }
3026
3027 emitNewDerivedEq :: CtLoc -> Role -> TcType -> TcType -> TcS ()
3028 -- Create new equality Derived and put it in the work list
3029 -- There's no caching, no lookupInInerts
3030 emitNewDerivedEq loc role ty1 ty2
3031 = do { ev <- newDerivedNC loc (mkPrimEqPredRole role ty1 ty2)
3032 ; traceTcS "Emitting new derived equality" (ppr ev $$ pprCtLoc loc)
3033 ; updWorkListTcS (extendWorkListDerived loc ev) }
3034
3035 newDerivedNC :: CtLoc -> TcPredType -> TcS CtEvidence
3036 newDerivedNC loc pred
3037 = do { -- checkReductionDepth loc pred
3038 ; return (CtDerived { ctev_pred = pred, ctev_loc = loc }) }
3039
3040 -- --------- Check done in TcInteract.selectNewWorkItem???? ---------
3041 -- | Checks if the depth of the given location is too much. Fails if
3042 -- it's too big, with an appropriate error message.
3043 checkReductionDepth :: CtLoc -> TcType -- ^ type being reduced
3044 -> TcS ()
3045 checkReductionDepth loc ty
3046 = do { dflags <- getDynFlags
3047 ; when (subGoalDepthExceeded dflags (ctLocDepth loc)) $
3048 wrapErrTcS $
3049 solverDepthErrorTcS loc ty }
3050
3051 matchFam :: TyCon -> [Type] -> TcS (Maybe (Coercion, TcType))
3052 matchFam tycon args = wrapTcS $ matchFamTcM tycon args
3053
3054 matchFamTcM :: TyCon -> [Type] -> TcM (Maybe (Coercion, TcType))
3055 -- Given (F tys) return (ty, co), where co :: F tys ~ ty
3056 matchFamTcM tycon args
3057 = do { fam_envs <- FamInst.tcGetFamInstEnvs
3058 ; let match_fam_result
3059 = reduceTyFamApp_maybe fam_envs Nominal tycon args
3060 ; TcM.traceTc "matchFamTcM" $
3061 vcat [ text "Matching:" <+> ppr (mkTyConApp tycon args)
3062 , ppr_res match_fam_result ]
3063 ; return match_fam_result }
3064 where
3065 ppr_res Nothing = text "Match failed"
3066 ppr_res (Just (co,ty)) = hang (text "Match succeeded:")
3067 2 (vcat [ text "Rewrites to:" <+> ppr ty
3068 , text "Coercion:" <+> ppr co ])
3069
3070 {-
3071 Note [Residual implications]
3072 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3073 The wl_implics in the WorkList are the residual implication
3074 constraints that are generated while solving or canonicalising the
3075 current worklist. Specifically, when canonicalising
3076 (forall a. t1 ~ forall a. t2)
3077 from which we get the implication
3078 (forall a. t1 ~ t2)
3079 See TcSMonad.deferTcSForAllEq
3080 -}
3081
3082 -- Deferring forall equalities as implications
3083 -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3084
3085 deferTcSForAllEq :: Role -- Nominal or Representational
3086 -> CtLoc -- Original wanted equality flavor
3087 -> [Coercion] -- among the kinds of the binders
3088 -> ([TyVarBinder],TcType) -- ForAll tvs1 body1
3089 -> ([TyVarBinder],TcType) -- ForAll tvs2 body2
3090 -> TcS Coercion
3091 deferTcSForAllEq role loc kind_cos (bndrs1,body1) (bndrs2,body2)
3092 = do { let tvs1' = zipWithEqual "deferTcSForAllEq"
3093 mkCastTy (mkTyVarTys tvs1) kind_cos
3094 body2' = substTyWithUnchecked tvs2 tvs1' body2
3095 ; (subst, skol_tvs) <- wrapTcS $ TcM.tcInstSkolTyVars tvs1
3096 ; let phi1 = Type.substTyUnchecked subst body1
3097 phi2 = Type.substTyUnchecked subst body2'
3098 skol_info = UnifyForAllSkol phi1
3099
3100 ; (ctev, hole_co) <- newWantedEq loc role phi1 phi2
3101 ; env <- getLclEnv
3102 ; ev_binds <- newTcEvBinds
3103 -- We have nowhere to put these bindings
3104 -- but TcSimplify.setImplicationStatus
3105 -- checks that we don't actually use them
3106 ; let new_tclvl = pushTcLevel (tcl_tclvl env)
3107 wc = WC { wc_simple = singleCt (mkNonCanonical ctev)
3108 , wc_impl = emptyBag
3109 , wc_insol = emptyCts }
3110 imp = Implic { ic_tclvl = new_tclvl
3111 , ic_skols = skol_tvs
3112 , ic_no_eqs = True
3113 , ic_given = []
3114 , ic_wanted = wc
3115 , ic_status = IC_Unsolved
3116 , ic_binds = ev_binds
3117 , ic_env = env
3118 , ic_needed = emptyVarSet
3119 , ic_info = skol_info }
3120 ; updWorkListTcS (extendWorkListImplic imp)
3121 ; let cobndrs = zip skol_tvs kind_cos
3122 ; return $ mkForAllCos cobndrs hole_co }
3123 where
3124 tvs1 = binderVars bndrs1
3125 tvs2 = binderVars bndrs2