TcInteract.lhs 88.5 KB
 simonpj@microsoft.com committed Sep 13, 2010 1 2 3 \begin{code} module TcInteract ( solveInteract, AtomicInert,  4  InertSet, emptyInert, updInertSet, extractUnsolved, solveOne  simonpj@microsoft.com committed Sep 13, 2010 5 6 7 8  ) where #include "HsVersions.h"  dimitris@microsoft.com committed Oct 04, 2010 9   simonpj@microsoft.com committed Sep 13, 2010 10 11 12 13 import BasicTypes import TcCanonical import VarSet import Type  dimitris@microsoft.com committed Oct 04, 2010 14 import TypeRep  simonpj@microsoft.com committed Sep 13, 2010 15 16  import Id  simonpj@microsoft.com committed Oct 07, 2010 17 import VarEnv  simonpj@microsoft.com committed Sep 13, 2010 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 import Var import TcType import HsBinds import InstEnv import Class import TyCon import Name import FunDeps import Control.Monad ( when ) import Coercion import Outputable import TcRnTypes import TcErrors import TcSMonad  simonpj@microsoft.com committed Oct 07, 2010 38 import Bag  dimitris@microsoft.com committed Oct 04, 2010 39 40 41 import qualified Data.Map as Map import Maybes  simonpj@microsoft.com committed Sep 13, 2010 42 43 44 45 46 import Control.Monad( zipWithM, unless ) import FastString ( sLit ) import DynFlags \end{code}  dimitris@microsoft.com committed Oct 06, 2010 47 Note [InertSet invariants]  simonpj@microsoft.com committed Sep 13, 2010 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 ~~~~~~~~~~~~~~~~~~~~~~~~~~~ An InertSet is a bag of canonical constraints, with the following invariants: 1 No two constraints react with each other. A tricky case is when there exists a given (solved) dictionary constraint and a wanted identical constraint in the inert set, but do not react because reaction would create loopy dictionary evidence for the wanted. See note [Recursive dictionaries] 2 Given equalities form an idempotent substitution [none of the given LHS's occur in any of the given RHS's or reactant parts] 3 Wanted equalities also form an idempotent substitution 4 The entire set of equalities is acyclic. 5 Wanted dictionaries are inert with the top-level axiom set 6 Equalities of the form tv1 ~ tv2 always have a touchable variable on the left (if possible). 7 No wanted constraints tv1 ~ tv2 with tv1 touchable. Such constraints will be marked as solved right before being pushed into the inert set. See note [Touchables and givens]. Note that 6 and 7 are /not/ enforced by canonicalization but rather by insertion in the inert list, ie by TcInteract. During the process of solving, the inert set will contain some previously given constraints, some wanted constraints, and some given constraints which have arisen from solving wanted constraints. For now we do not distinguish between given and solved constraints. Note that we must switch wanted inert items to given when going under an implication constraint (when in top-level inference mode).  dimitris@microsoft.com committed Oct 06, 2010 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 Note [InertSet FlattenSkolemEqClass] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The inert_fsks field of the inert set contains an "inverse map" of all the flatten skolem equalities in the inert set. For instance, if inert_cts looks like this: fsk1 ~ fsk2 fsk3 ~ fsk2 fsk4 ~ fsk5 Then, the inert_fsks fields holds the following map: fsk2 |-> { fsk1, fsk3 } fsk5 |-> { fsk4 } Along with the necessary coercions to convert fsk1 and fsk3 back to fsk2 and fsk4 back to fsk5. Hence, the invariants of the inert_fsks field are: (a) All TcTyVars in the domain and range of inert_fsks are flatten skolems (b) All TcTyVars in the domain of inert_fsk occur naked as rhs in some equalities of inert_cts (c) For every mapping fsk1 |-> { (fsk2,co), ... } it must be: co : fsk2 ~ fsk1 The role of the inert_fsks is to make it easy to maintain the equivalence class of each flatten skolem, which is much needed to correctly do spontaneous solving. See Note [Loopy Spontaneous Solving]  simonpj@microsoft.com committed Sep 13, 2010 109 110 111 \begin{code} -- See Note [InertSet invariants]  dimitris@microsoft.com committed Oct 04, 2010 112 data InertSet  113  = IS { inert_eqs :: Bag.Bag CanonicalCt -- Equalities only **CTyEqCan**  114 115 116  , inert_cts :: Bag.Bag CanonicalCt -- Other constraints , inert_fds :: FDImprovements -- List of pairwise improvements that have kicked in already -- and reside either in the worklist or in the inerts  dimitris@microsoft.com committed Oct 04, 2010 117  , inert_fsks :: Map.Map TcTyVar [(TcTyVar,Coercion)] }  dimitris@microsoft.com committed Oct 06, 2010 118  -- See Note [InertSet FlattenSkolemEqClass]  dimitris@microsoft.com committed Oct 04, 2010 119   120 121 122 type FDImprovement = (PredType,PredType) type FDImprovements = [(PredType,PredType)]  simonpj@microsoft.com committed Sep 13, 2010 123 instance Outputable InertSet where  124 125  ppr is = vcat [ vcat (map ppr (Bag.bagToList $inert_eqs is)) , vcat (map ppr (Bag.bagToList$ inert_cts is))  dimitris@microsoft.com committed Oct 04, 2010 126 127 128 129 130  , vcat (map (\(v,rest) -> ppr v <+> text "|->" <+> hsep (map (ppr.fst) rest)) (Map.toList $inert_fsks is) ) ]  simonpj@microsoft.com committed Sep 13, 2010 131 emptyInert :: InertSet  132 emptyInert = IS { inert_eqs = Bag.emptyBag  133  , inert_cts = Bag.emptyBag, inert_fsks = Map.empty, inert_fds = [] }  dimitris@microsoft.com committed Oct 04, 2010 134 135 136  updInertSet :: InertSet -> AtomicInert -> InertSet -- Introduces an element in the inert set for the first time  137 updInertSet (IS { inert_eqs = eqs, inert_cts = cts, inert_fsks = fsks, inert_fds = fdis })  dimitris@microsoft.com committed Oct 04, 2010 138 139 140 141 142 143  item@(CTyEqCan { cc_id = cv , cc_tyvar = tv1 , cc_rhs = xi }) | Just tv2 <- tcGetTyVar_maybe xi, FlatSkol {} <- tcTyVarDetails tv1, FlatSkol {} <- tcTyVarDetails tv2  144  = let eqs' = eqs Bag.snocBag item  dimitris@microsoft.com committed Oct 04, 2010 145  fsks' = Map.insertWith (++) tv2 [(tv1, mkCoVarCoercion cv)] fsks  dimitris@microsoft.com committed Oct 06, 2010 146  -- See Note [InertSet FlattenSkolemEqClass]  147  in IS { inert_eqs = eqs', inert_cts = cts, inert_fsks = fsks', inert_fds = fdis }  148 updInertSet (IS { inert_eqs = eqs, inert_cts = cts  149  , inert_fsks = fsks, inert_fds = fdis }) item  150  | isTyEqCCan item  151  = let eqs' = eqs Bag.snocBag item  152  in IS { inert_eqs = eqs', inert_cts = cts, inert_fsks = fsks, inert_fds = fdis }  153  | otherwise  dimitris@microsoft.com committed Oct 04, 2010 154  = let cts' = cts Bag.snocBag item  155 156 157 158 159  in IS { inert_eqs = eqs, inert_cts = cts', inert_fsks = fsks, inert_fds = fdis } updInertSetFDImprs :: InertSet -> Maybe FDImprovement -> InertSet updInertSetFDImprs is (Just fdi) = is { inert_fds = fdi : inert_fds is } updInertSetFDImprs is Nothing = is  dimitris@microsoft.com committed Oct 04, 2010 160   161 162 163 164 165 166 167 168 169 foldISEqCtsM :: Monad m => (a -> AtomicInert -> m a) -> a -> InertSet -> m a -- Fold over the equalities of the inerts foldISEqCtsM k z IS { inert_eqs = eqs } = Bag.foldlBagM k z eqs foldISOtherCtsM :: Monad m => (a -> AtomicInert -> m a) -> a -> InertSet -> m a -- Fold over other constraints in the inerts foldISOtherCtsM k z IS { inert_cts = cts } = Bag.foldlBagM k z cts  simonpj@microsoft.com committed Sep 13, 2010 170 171  extractUnsolved :: InertSet -> (InertSet, CanonicalCts)  172 extractUnsolved is@(IS {inert_eqs = eqs, inert_cts = cts, inert_fds = fdis })  173  = let is_init = is { inert_eqs = emptyCCan  174  , inert_cts = solved_cts, inert_fsks = Map.empty, inert_fds = fdis }  175 176 177 178 179  is_final = Bag.foldlBag updInertSet is_init solved_eqs -- Add equalities carefully in (is_final, unsolved) where (unsolved_cts, solved_cts) = Bag.partitionBag isWantedCt cts (unsolved_eqs, solved_eqs) = Bag.partitionBag isWantedCt eqs unsolved = unsolved_cts unionBags unsolved_eqs  simonpj@microsoft.com committed Sep 13, 2010 180   dimitris@microsoft.com committed Oct 04, 2010 181 182  getFskEqClass :: InertSet -> TcTyVar -> [(TcTyVar,Coercion)]  dimitris@microsoft.com committed Oct 06, 2010 183 -- Precondition: tv is a FlatSkol. See Note [InertSet FlattenSkolemEqClass]  dimitris@microsoft.com committed Oct 04, 2010 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 getFskEqClass (IS { inert_cts = cts, inert_fsks = fsks }) tv = case lkpTyEqCanByLhs of Nothing -> fromMaybe [] (Map.lookup tv fsks) Just ceq -> case tcGetTyVar_maybe (cc_rhs ceq) of Just tv_rhs | FlatSkol {} <- tcTyVarDetails tv_rhs -> let ceq_co = mkSymCoercion$ mkCoVarCoercion (cc_id ceq) mk_co (v,c) = (v, mkTransCoercion c ceq_co) in (tv_rhs, ceq_co): map mk_co (fromMaybe [] $Map.lookup tv fsks) _ -> [] where lkpTyEqCanByLhs = Bag.foldlBag lkp Nothing cts lkp :: Maybe CanonicalCt -> CanonicalCt -> Maybe CanonicalCt lkp Nothing ct@(CTyEqCan {cc_tyvar = tv'}) | tv' == tv = Just ct lkp other _ct = other  199 200 201 202 203 204 205 206 207 208 209 210 211 212 haveBeenImproved :: FDImprovements -> PredType -> PredType -> Bool haveBeenImproved [] _ _ = False haveBeenImproved ((pty1,pty2):fdimprs) pty1' pty2' | tcEqPred pty1 pty1' && tcEqPred pty2 pty2' = True | tcEqPred pty1 pty2' && tcEqPred pty2 pty1' = True | otherwise = haveBeenImproved fdimprs pty1' pty2' getFDImprovements :: InertSet -> FDImprovements -- Return a list of the improvements that have kicked in so far getFDImprovements = inert_fds  dimitris@microsoft.com committed Oct 04, 2010 213   simonpj@microsoft.com committed Sep 13, 2010 214 215 isWantedCt :: CanonicalCt -> Bool isWantedCt ct = isWanted (cc_flavor ct)  dimitris@microsoft.com committed Oct 04, 2010 216 217 218 219 220 221 222 223 224 225 226  {- TODO: Later ... data Inert = IS { class_inerts :: FiniteMap Class Atomics ip_inerts :: FiniteMap Class Atomics tyfun_inerts :: FiniteMap TyCon Atomics tyvar_inerts :: FiniteMap TyVar Atomics } Later should we also separate out givens and wanteds? -}  simonpj@microsoft.com committed Sep 13, 2010 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 \end{code} Note [Touchables and givens] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Touchable variables will never show up in givens which are inputs to the solver. However, touchables may show up in givens generated by the flattener. For example, axioms: G Int ~ Char F Char ~ Int wanted: F (G alpha) ~w Int canonicalises to G alpha ~g b F b ~w Int which can be put in the inert set. Suppose we also have a wanted alpha ~w Int We cannot rewrite the given G alpha ~g b using the wanted alpha ~w Int. Instead, after reacting alpha ~w Int with the whole inert set, we observe that we can solve it by unifying alpha with Int, so we mark it as solved and put it back in the *work list*. [We also immediately unify alpha := Int, without telling anyone, see trySpontaneousSolve function, to avoid doing this in the end.] Later, because it is solved (given, in effect), we can use it to rewrite G alpha ~g b to G Int ~g b, which gets put back in the work list. Eventually, we will dispatch the remaining wanted constraints using the top-level axioms. Finally, note that after reacting a wanted equality with the entire inert set we may end up with something like b ~w alpha which we should flip around to generate the solved constraint alpha ~s b. %********************************************************************* %* * * Main Interaction Solver * * * ********************************************************************** Note [Basic plan] ~~~~~~~~~~~~~~~~~ 1. Canonicalise (unary) 2. Pairwise interaction (binary) * Take one from work list * Try all pair-wise interactions with each constraint in inert  281 282 283 284  As an optimisation, we prioritize the equalities both in the worklist and in the inerts.  simonpj@microsoft.com committed Sep 13, 2010 285 286 287 288 289 290 291 292 293 3. Try to solve spontaneously for equalities involving touchables 4. Top-level interaction (binary wrt top-level) Superclass decomposition belongs in (4), see note [Superclasses] \begin{code} type AtomicInert = CanonicalCt -- constraint pulled from InertSet type WorkItem = CanonicalCt -- constraint pulled from WorkList  294 295 296 297 298 299 -- A mixture of Given, Wanted, and Derived constraints. -- We split between equalities and the rest to process equalities first. data WorkList = WL { wl_eqs :: CanonicalCts -- Equalities (CTyEqCan, CFunEqCan) , wl_other :: CanonicalCts -- Other } type SWorkList = WorkList -- A worklist of solved  simonpj@microsoft.com committed Sep 13, 2010 300 301  unionWorkLists :: WorkList -> WorkList -> WorkList  302 303 304 unionWorkLists wl1 wl2 = WL { wl_eqs = andCCan (wl_eqs wl1) (wl_eqs wl2) , wl_other = andCCan (wl_other wl1) (wl_other wl2) }  simonpj@microsoft.com committed Sep 13, 2010 305   306 307 308 309 310 311 312 foldWorkListEqCtsM :: Monad m => (a -> WorkItem -> m a) -> a -> WorkList -> m a -- Fold over the equalities of a worklist foldWorkListEqCtsM f r wl = Bag.foldlBagM f r (wl_eqs wl) foldWorkListOtherCtsM :: Monad m => (a -> WorkItem -> m a) -> a -> WorkList -> m a -- Fold over non-equality constraints of a worklist foldWorkListOtherCtsM f r wl = Bag.foldlBagM f r (wl_other wl)  simonpj@microsoft.com committed Sep 13, 2010 313 314  isEmptyWorkList :: WorkList -> Bool  315 isEmptyWorkList wl = isEmptyCCan (wl_eqs wl) && isEmptyCCan (wl_other wl)  simonpj@microsoft.com committed Sep 13, 2010 316 317  emptyWorkList :: WorkList  318 319 320 321 322 emptyWorkList = WL { wl_eqs = emptyCCan, wl_other = emptyCCan } workListFromCCans :: CanonicalCts -> WorkList -- Generic, no precondition workListFromCCans cts = WL eqs others  323  where (eqs, others) = Bag.partitionBag isTyEqCCan cts  324   325 326 327 328 329 330 331 workListFromCCan :: CanonicalCt -> WorkList workListFromCCan ct | isTyEqCCan ct = WL (singleCCan ct) emptyCCan | otherwise = WL emptyCCan (singleCCan ct) -- TODO: -- At the call sites of workListFromCCan(s), sometimes we know whether the new work -- involves equalities or not. It's probably a good idea to add specialized calls for -- those, to avoid asking whether 'isTyEqCCan' all the time.  simonpj@microsoft.com committed Sep 13, 2010 332   dimitris@microsoft.com committed Oct 04, 2010 333   simonpj@microsoft.com committed Sep 13, 2010 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 data StopOrContinue = Stop -- Work item is consumed | ContinueWith WorkItem -- Not consumed instance Outputable StopOrContinue where ppr Stop = ptext (sLit "Stop") ppr (ContinueWith w) = ptext (sLit "ContinueWith") <+> ppr w -- Results after interacting a WorkItem as far as possible with an InertSet data StageResult = SR { sr_inerts :: InertSet -- The new InertSet to use (REPLACES the old InertSet) , sr_new_work :: WorkList -- Any new work items generated (should be ADDED to the old WorkList) -- Invariant: -- sr_stop = Just workitem => workitem is *not* in sr_inerts and -- workitem is inert wrt to sr_inerts , sr_stop :: StopOrContinue } instance Outputable StageResult where ppr (SR { sr_inerts = inerts, sr_new_work = work, sr_stop = stop }) = ptext (sLit "SR") <+> braces (sep [ ptext (sLit "inerts =") <+> ppr inerts <> comma , ptext (sLit "new work =") <+> ppr work <> comma , ptext (sLit "stop =") <+> ppr stop])  361 362 363 instance Outputable WorkList where ppr (WL eqcts othercts) = vcat [ppr eqcts, ppr othercts]  simonpj@microsoft.com committed Sep 13, 2010 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 type SimplifierStage = WorkItem -> InertSet -> TcS StageResult -- Combine a sequence of simplifier 'stages' to create a pipeline runSolverPipeline :: [(String, SimplifierStage)] -> InertSet -> WorkItem -> TcS (InertSet, WorkList) -- Precondition: non-empty list of stages runSolverPipeline pipeline inerts workItem = do { traceTcS "Start solver pipeline"$ vcat [ ptext (sLit "work item =") <+> ppr workItem , ptext (sLit "inerts =") <+> ppr inerts] ; let itr_in = SR { sr_inerts = inerts , sr_new_work = emptyWorkList , sr_stop = ContinueWith workItem } ; itr_out <- run_pipeline pipeline itr_in ; let new_inert = case sr_stop itr_out of Stop -> sr_inerts itr_out  dimitris@microsoft.com committed Oct 04, 2010 383  ContinueWith item -> sr_inerts itr_out updInertSet item  simonpj@microsoft.com committed Sep 13, 2010 384 385 386 387 388 389 390 391 392 393 394 395 396  ; return (new_inert, sr_new_work itr_out) } where run_pipeline :: [(String, SimplifierStage)] -> StageResult -> TcS StageResult run_pipeline [] itr = return itr run_pipeline _ itr@(SR { sr_stop = Stop }) = return itr run_pipeline ((name,stage):stages) (SR { sr_new_work = accum_work , sr_inerts = inerts , sr_stop = ContinueWith work_item }) = do { itr <- stage work_item inerts ; traceTcS ("Stage result (" ++ name ++ ")") (ppr itr)  397  ; let itr' = itr { sr_new_work = accum_work unionWorkLists sr_new_work itr }  simonpj@microsoft.com committed Sep 13, 2010 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428  ; run_pipeline stages itr' } \end{code} Example 1: Inert: {c ~ d, F a ~ t, b ~ Int, a ~ ty} (all given) Reagent: a ~ [b] (given) React with (c~d) ==> IR (ContinueWith (a~[b])) True [] React with (F a ~ t) ==> IR (ContinueWith (a~[b])) False [F [b] ~ t] React with (b ~ Int) ==> IR (ContinueWith (a~[Int]) True [] Example 2: Inert: {c ~w d, F a ~g t, b ~w Int, a ~w ty} Reagent: a ~w [b] React with (c ~w d) ==> IR (ContinueWith (a~[b])) True [] React with (F a ~g t) ==> IR (ContinueWith (a~[b])) True [] (can't rewrite given with wanted!) etc. Example 3: Inert: {a ~ Int, F Int ~ b} (given) Reagent: F a ~ b (wanted) React with (a ~ Int) ==> IR (ContinueWith (F Int ~ b)) True [] React with (F Int ~ b) ==> IR Stop True [] -- after substituting we re-canonicalize and get nothing \begin{code} -- Main interaction solver: we fully solve the worklist 'in one go', -- returning an extended inert set. -- -- See Note [Touchables and givens].  429 solveInteract :: InertSet -> CanonicalCts -> TcS InertSet  simonpj@microsoft.com committed Sep 13, 2010 430 431 solveInteract inert ws = do { dyn_flags <- getDynFlags  432 433  ; let worklist = workListFromCCans ws ; solveInteractWithDepth (ctxtStkDepth dyn_flags,0,[]) inert worklist  simonpj@microsoft.com committed Sep 13, 2010 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455  } solveOne :: InertSet -> WorkItem -> TcS InertSet solveOne inerts workItem = do { dyn_flags <- getDynFlags ; solveOneWithDepth (ctxtStkDepth dyn_flags,0,[]) inerts workItem } ----------------- solveInteractWithDepth :: (Int, Int, [WorkItem]) -> InertSet -> WorkList -> TcS InertSet solveInteractWithDepth ctxt@(max_depth,n,stack) inert ws | isEmptyWorkList ws = return inert | n > max_depth = solverDepthErrorTcS n stack | otherwise = do { traceTcS "solveInteractWithDepth" $vcat [ text "Current depth =" <+> ppr n , text "Max depth =" <+> ppr max_depth ]  456 457 458  ; is_from_eqs <- foldWorkListEqCtsM (solveOneWithDepth ctxt) inert ws ; foldWorkListOtherCtsM (solveOneWithDepth ctxt) is_from_eqs ws }  simonpj@microsoft.com committed Sep 13, 2010 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481  ------------------ -- Fully interact the given work item with an inert set, and return a -- new inert set which has assimilated the new information. solveOneWithDepth :: (Int, Int, [WorkItem]) -> InertSet -> WorkItem -> TcS InertSet solveOneWithDepth (max_depth, n, stack) inert work = do { traceTcS0 (indent ++ "Solving {") (ppr work) ; (new_inert, new_work) <- runSolverPipeline thePipeline inert work ; traceTcS0 (indent ++ "Subgoals:") (ppr new_work) -- Recursively solve the new work generated -- from workItem, with a greater depth ; res_inert <- solveInteractWithDepth (max_depth, n+1, work:stack) new_inert new_work ; traceTcS0 (indent ++ "Done }") (ppr work) ; return res_inert } where indent = replicate (2*n) ' ' thePipeline :: [(String,SimplifierStage)]  482 483 484 485 thePipeline = [ ("interact with inert eqs", interactWithInertEqsStage) , ("interact with inerts", interactWithInertsStage) , ("spontaneous solve", spontaneousSolveStage) , ("top-level reactions", topReactionsStage) ]  simonpj@microsoft.com committed Sep 13, 2010 486 487 488 489 490 491 492 493 494 495 496 \end{code} ********************************************************************************* * * The spontaneous-solve Stage * * ********************************************************************************* \begin{code} spontaneousSolveStage :: SimplifierStage spontaneousSolveStage workItem inerts  497  = do { mSolve <- trySpontaneousSolve workItem inerts  simonpj@microsoft.com committed Sep 13, 2010 498 499  ; case mSolve of Nothing -> -- no spontaneous solution for him, keep going  500 501  return$ SR { sr_new_work = emptyWorkList , sr_inerts = inerts  simonpj@microsoft.com committed Sep 13, 2010 502 503  , sr_stop = ContinueWith workItem }  504  Just workList' -> -- He has been solved; workList' are all givens  dimitris@microsoft.com committed Oct 04, 2010 505 506  return $SR { sr_new_work = workList' , sr_inerts = inerts  507  , sr_stop = Stop }  dimitris@microsoft.com committed Oct 04, 2010 508  }  dimitris@microsoft.com committed Oct 06, 2010 509   simonpj@microsoft.com committed Sep 13, 2010 510 511 512 513 514 -- @trySpontaneousSolve wi@ solves equalities where one side is a -- touchable unification variable. Returns: -- * Nothing if we were not able to solve it -- * Just wi' if we solved it, wi' (now a "given") should be put in the work list. -- See Note [Touchables and givens]  515 -- NB: just passing the inerts through for the skolem equivalence classes  dimitris@microsoft.com committed Oct 04, 2010 516 517 trySpontaneousSolve :: WorkItem -> InertSet -> TcS (Maybe SWorkList) trySpontaneousSolve (CTyEqCan { cc_id = cv, cc_flavor = gw, cc_tyvar = tv1, cc_rhs = xi }) inerts  simonpj@microsoft.com committed Oct 07, 2010 518 519  | isGiven gw = return Nothing  simonpj@microsoft.com committed Sep 13, 2010 520 521 522 523  | Just tv2 <- tcGetTyVar_maybe xi = do { tch1 <- isTouchableMetaTyVar tv1 ; tch2 <- isTouchableMetaTyVar tv2 ; case (tch1, tch2) of  dimitris@microsoft.com committed Oct 04, 2010 524 525  (True, True) -> trySpontaneousEqTwoWay inerts cv gw tv1 tv2 (True, False) -> trySpontaneousEqOneWay inerts cv gw tv1 xi  526  (False, True) -> trySpontaneousEqOneWay inerts cv gw tv2 (mkTyVarTy tv1)  simonpj@microsoft.com committed Sep 13, 2010 527 528 529  _ -> return Nothing } | otherwise = do { tch1 <- isTouchableMetaTyVar tv1  dimitris@microsoft.com committed Oct 04, 2010 530  ; if tch1 then trySpontaneousEqOneWay inerts cv gw tv1 xi  simonpj@microsoft.com committed Sep 13, 2010 531 532 533 534 535  else return Nothing } -- No need for -- trySpontaneousSolve (CFunEqCan ...) = ... -- See Note [No touchables as FunEq RHS] in TcSMonad  dimitris@microsoft.com committed Oct 04, 2010 536 trySpontaneousSolve _ _ = return Nothing  simonpj@microsoft.com committed Sep 13, 2010 537 538  ----------------  dimitris@microsoft.com committed Oct 04, 2010 539 540 trySpontaneousEqOneWay :: InertSet -> CoVar -> CtFlavor -> TcTyVar -> Xi -> TcS (Maybe SWorkList)  simonpj@microsoft.com committed Sep 13, 2010 541 -- tv is a MetaTyVar, not untouchable  dimitris@microsoft.com committed Oct 04, 2010 542 trySpontaneousEqOneWay inerts cv gw tv xi  543 544 545 546 547 548 549 550 551 552 553 554  | not (isSigTyVar tv) || isTyVarTy xi = if typeKind xi isSubKind tyVarKind tv then solveWithIdentity inerts cv gw tv xi else if tyVarKind tv isSubKind typeKind xi then return Nothing -- kinds are compatible but we can't solveWithIdentity this way -- This case covers the a_touchable :: * ~ b_untouchable :: ?? -- which has to be deferred or floated out for someone else to solve -- it in a scope where 'b' is no longer untouchable. else kindErrorTcS gw (mkTyVarTy tv) xi -- See Note [Kind errors] | otherwise -- Still can't solve, sig tyvar and non-variable rhs = return Nothing  simonpj@microsoft.com committed Sep 13, 2010 555 556  ----------------  dimitris@microsoft.com committed Oct 04, 2010 557 558 trySpontaneousEqTwoWay :: InertSet -> CoVar -> CtFlavor -> TcTyVar -> TcTyVar -> TcS (Maybe SWorkList)  559 -- Both tyvars are *touchable* MetaTyvars so there is only a chance for kind error here  dimitris@microsoft.com committed Oct 04, 2010 560 trySpontaneousEqTwoWay inerts cv gw tv1 tv2  561  | k1 isSubKind k2  dimitris@microsoft.com committed Oct 04, 2010 562  , nicer_to_update_tv2 = solveWithIdentity inerts cv gw tv2 (mkTyVarTy tv1)  563 564  | k2 isSubKind k1 = solveWithIdentity inerts cv gw tv1 (mkTyVarTy tv2)  565 566  | otherwise -- None is a subkind of the other, but they are both touchable! = kindErrorTcS gw (mkTyVarTy tv1) (mkTyVarTy tv2) -- See Note [Kind errors]  simonpj@microsoft.com committed Sep 13, 2010 567 568 569 570 571 572  where k1 = tyVarKind tv1 k2 = tyVarKind tv2 nicer_to_update_tv2 = isSigTyVar tv1 || isSystemName (Var.varName tv2) \end{code}  573 574 575 576 577 578 579 580 581 582 583 Note [Kind errors] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Consider the wanted problem: alpha ~ (# Int, Int #) where alpha :: ?? and (# Int, Int #) :: (#). We can't spontaneously solve this constraint, but we should rather reject the program that give rise to it. If 'trySpontaneousEqTwoWay' simply returns @Nothing@ then that wanted constraint is going to propagate all the way and get quantified over in inference mode. That's bad because we do know at this point that the constraint is insoluble. Instead, we call 'kindErrorTcS' here, which immediately fails. The same applies in canonicalization code in case of kind errors in the givens.  584 585 586 587 588 589 590 591 592 593 594  Note [Spontaneous solving and kind compatibility] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Note that our canonical constraints insist that only *given* equalities (tv ~ xi) or (F xis ~ rhs) require the LHS and the RHS to have exactly the same kinds. - We have to require this because: Given equalities can be freely used to rewrite inside other types or constraints. - We do not have to do the same for wanteds because:  595 596 597 598 599 600 601 602  First, wanted equations (tv ~ xi) where tv is a touchable unification variable may have kinds that do not agree (the kind of xi must be a sub kind of the kind of tv). Second, any potential kind mismatch will result in the constraint not being soluble, which will be reported anyway. This is the reason that @trySpontaneousOneWay@ and @trySpontaneousTwoWay@ will perform a kind compatibility check, and only then will they proceed to @solveWithIdentity@.  603 604 605 606 607 608 609 610 611 612 613  Caveat: - Givens from higher-rank, such as: type family T b :: * -> * -> * type instance T Bool = (->) f :: forall a. ((T a ~ (->)) => ...) -> a -> ... flop = f (...) True Whereas we would be able to apply the type instance, we would not be able to use the given (T Bool ~ (->)) in the body of 'flop'  simonpj@microsoft.com committed Sep 13, 2010 614 615 616 617 618 619 620 621 622 623 624 625 626 Note [Loopy spontaneous solving] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Consider the original wanted: wanted : Maybe (E alpha) ~ alpha where E is a type family, such that E (T x) = x. After canonicalization, as a result of flattening, we will get: given : E alpha ~ fsk wanted : alpha ~ Maybe fsk where (fsk := E alpha, on the side). Now, if we spontaneously *solve* (alpha := Maybe fsk) we are in trouble! Instead, we should refrain from solving it and keep it as wanted. In inference mode we'll end up quantifying over (alpha ~ Maybe (E alpha)) Hence, 'solveWithIdentity' performs a small occurs check before  dimitris@microsoft.com committed Oct 06, 2010 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 actually solving. But this occurs check *must look through* flatten skolems. However, it may be the case that the flatten skolem in hand is equal to some other flatten skolem whith *does not* mention our unification variable. Here's a typical example: Original wanteds: g: F alpha ~ F beta w: alpha ~ F alpha After canonicalization: g: F beta ~ f1 g: F alpha ~ f1 w: alpha ~ f2 g: F alpha ~ f2 After some reactions: g: f1 ~ f2 g: F beta ~ f1 w: alpha ~ f2 g: F alpha ~ f2 At this point, we will try to spontaneously solve (alpha ~ f2) which remains as yet unsolved. We will look inside f2, which immediately mentions (F alpha), so it's not good to unify! However by looking at the equivalence class of the flatten skolems, we can see that it is fine to unify (alpha ~ f1) which solves our goals! A similar problem happens because of other spontaneous solving. Suppose we have the following wanteds, arriving in this exact order: (first) w: beta ~ alpha (second) w: alpha ~ fsk (third) g: F beta ~ fsk Then, we first spontaneously solve the first constraint, making (beta := alpha), and having (beta ~ alpha) as given. *Then* we encounter the second wanted (alpha ~ fsk). "fsk" does not obviously mention alpha, so naively we can also spontaneously solve (alpha := fsk). But that is wrong since fsk mentions beta, which has already secretly been unified to alpha! To avoid this problem, the same occurs check must unveil rewritings that can happen because of spontaneously having solved other constraints.  simonpj@microsoft.com committed Sep 13, 2010 663 664 665 666 667 668 669  Note [Avoid double unifications] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The spontaneous solver has to return a given which mentions the unified unification variable *on the left* of the equality. Here is what happens if not: Original wanted: (a ~ alpha), (alpha ~ Int) We spontaneously solve the first wanted, without changing the order!  simonpj@microsoft.com committed Oct 07, 2010 670  given : a ~ alpha [having unified alpha := a]  simonpj@microsoft.com committed Sep 13, 2010 671 672 673 Now the second wanted comes along, but he cannot rewrite the given, so we simply continue. At the end we spontaneously solve that guy, *reunifying* [alpha := Int]  simonpj@microsoft.com committed Oct 07, 2010 674 675 676 We avoid this problem by orienting the given so that the unification variable is on the left. [Note that alternatively we could attempt to enforce this at canonicalization]  simonpj@microsoft.com committed Sep 13, 2010 677   simonpj@microsoft.com committed Oct 07, 2010 678 679 680 See also Note [No touchables as FunEq RHS] in TcSMonad; avoiding double unifications is the main reason we disallow touchable unification variables as RHS of type family equations: F xis ~ alpha.  simonpj@microsoft.com committed Sep 13, 2010 681 682 683  \begin{code} ----------------  dimitris@microsoft.com committed Oct 04, 2010 684 685 686 solveWithIdentity :: InertSet -> CoVar -> CtFlavor -> TcTyVar -> Xi -> TcS (Maybe SWorkList)  simonpj@microsoft.com committed Sep 13, 2010 687 688 -- Solve with the identity coercion -- Precondition: kind(xi) is a sub-kind of kind(tv)  simonpj@microsoft.com committed Oct 07, 2010 689 690 691 -- Precondition: CtFlavor is Wanted or Derived -- See [New Wanted Superclass Work] to see why solveWithIdentity -- must work for Derived as well as Wanted  dimitris@microsoft.com committed Oct 04, 2010 692 solveWithIdentity inerts cv gw tv xi  simonpj@microsoft.com committed Oct 07, 2010 693  = do { tybnds <- getTcSTyBindsMap  simonpj@microsoft.com committed Oct 07, 2010 694 695 696 697 698 699  ; case occurCheck tybnds inerts tv xi of Nothing -> return Nothing Just (xi_unflat,coi) -> solve_with xi_unflat coi } where solve_with xi_unflat coi -- coi : xi_unflat ~ xi = do { traceTcS "Sneaky unification:"$  dimitris@microsoft.com committed Oct 04, 2010 700 701 702 703  vcat [text "Coercion variable: " <+> ppr gw, text "Coercion: " <+> pprEq (mkTyVarTy tv) xi, text "Left Kind is : " <+> ppr (typeKind (mkTyVarTy tv)), text "Right Kind is : " <+> ppr (typeKind xi)  simonpj@microsoft.com committed Oct 07, 2010 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722  ] ; setWantedTyBind tv xi_unflat -- Set tv := xi_unflat ; cv_given <- newGivOrDerCoVar (mkTyVarTy tv) xi_unflat xi_unflat ; let flav = mkGivenFlavor gw UnkSkol ; (cts, co) <- case coi of ACo co -> do { can_eqs <- canEq flav cv_given (mkTyVarTy tv) xi_unflat ; return (can_eqs, co) } IdCo co -> return $(singleCCan (CTyEqCan { cc_id = cv_given , cc_flavor = mkGivenFlavor gw UnkSkol , cc_tyvar = tv, cc_rhs = xi } -- xi, *not* xi_unflat because -- xi_unflat may require flattening! ), co) ; case gw of Wanted {} -> setWantedCoBind cv co Derived {} -> setDerivedCoBind cv co _ -> pprPanic "Can't spontaneously solve *given*" empty -- See Note [Avoid double unifications]  723  ; return$ Just (workListFromCCans cts) }  simonpj@microsoft.com committed Oct 07, 2010 724   simonpj@microsoft.com committed Oct 07, 2010 725 occurCheck :: VarEnv (TcTyVar, TcType) -> InertSet  simonpj@microsoft.com committed Oct 07, 2010 726 727 728 729 730 731 732 733 734 735  -> TcTyVar -> TcType -> Maybe (TcType,CoercionI) -- Traverse @ty@ to make sure that @tv@ does not appear under some flatten skolem. -- If it appears under some flatten skolem look in that flatten skolem equivalence class -- (see Note [InertSet FlattenSkolemEqClass], [Loopy Spontaneous Solving]) to see if you -- can find a different flatten skolem to use, that is, one that does not mention @tv@. -- -- Postcondition: Just (ty', coi) = occurCheck binds inerts tv ty -- coi :: ty' ~ ty -- NB: The returned type ty' may not be flat!  simonpj@microsoft.com committed Oct 07, 2010 736 737 occurCheck ty_binds inerts the_tv the_ty = ok emptyVarSet the_ty  simonpj@microsoft.com committed Oct 07, 2010 738  where  simonpj@microsoft.com committed Oct 07, 2010 739 740  -- If (fsk elem bad) then tv occurs in any rendering -- of the type under the expansion of fsk  simonpj@microsoft.com committed Oct 07, 2010 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761  ok bad this_ty@(TyConApp tc tys) | Just tys_cois <- allMaybes (map (ok bad) tys) , (tys',cois') <- unzip tys_cois = Just (TyConApp tc tys', mkTyConAppCoI tc cois') | isSynTyCon tc, Just ty_expanded <- tcView this_ty = ok bad ty_expanded -- See Note [Type synonyms and the occur check] in TcUnify ok bad (PredTy sty) | Just (sty',coi) <- ok_pred bad sty = Just (PredTy sty', coi) ok bad (FunTy arg res) | Just (arg', coiarg) <- ok bad arg, Just (res', coires) <- ok bad res = Just (FunTy arg' res', mkFunTyCoI coiarg coires) ok bad (AppTy fun arg) | Just (fun', coifun) <- ok bad fun, Just (arg', coiarg) <- ok bad arg = Just (AppTy fun' arg', mkAppTyCoI coifun coiarg) ok bad (ForAllTy tv1 ty1) -- WARNING: What if it is a (t1 ~ t2) => t3? It's not handled properly at the moment. | Just (ty1', coi) <- ok bad ty1 = Just (ForAllTy tv1 ty1', mkForAllTyCoI tv1 coi) -- Variable cases  simonpj@microsoft.com committed Oct 07, 2010 762 763 764 765 766 767  ok bad this_ty@(TyVarTy tv) | tv == the_tv = Nothing -- Occurs check error | not (isTcTyVar tv) = Just (this_ty, IdCo this_ty) -- Bound var | FlatSkol zty <- tcTyVarDetails tv = ok_fsk bad tv zty | Just (_,ty) <- lookupVarEnv ty_binds tv = ok bad ty | otherwise = Just (this_ty, IdCo this_ty)  simonpj@microsoft.com committed Oct 07, 2010 768 769 770 771 772  -- Check if there exists a ty bind already, as a result of sneaky unification. -- Fall through ok _bad _ty = Nothing  simonpj@microsoft.com committed Oct 07, 2010 773  -----------  simonpj@microsoft.com committed Oct 07, 2010 774 775 776 777 778 779 780 781 782 783 784 785  ok_pred bad (ClassP cn tys) | Just tys_cois <- allMaybes $map (ok bad) tys = let (tys', cois') = unzip tys_cois in Just (ClassP cn tys', mkClassPPredCoI cn cois') ok_pred bad (IParam nm ty) | Just (ty',co') <- ok bad ty = Just (IParam nm ty', mkIParamPredCoI nm co') ok_pred bad (EqPred ty1 ty2) | Just (ty1',coi1) <- ok bad ty1, Just (ty2',coi2) <- ok bad ty2 = Just (EqPred ty1' ty2', mkEqPredCoI coi1 coi2) ok_pred _ _ = Nothing  simonpj@microsoft.com committed Oct 07, 2010 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804  ----------- ok_fsk bad fsk zty | fsk elemVarSet bad -- We are already trying to find a rendering of fsk, -- and to do that it seems we need a rendering, so fail = Nothing | otherwise = firstJusts (ok new_bad zty : map (go_under_fsk new_bad) fsk_equivs) where fsk_equivs = getFskEqClass inerts fsk new_bad = bad extendVarSetList (fsk : map fst fsk_equivs) ----------- go_under_fsk bad_tvs (fsk,co) | FlatSkol zty <- tcTyVarDetails fsk = case ok bad_tvs zty of Nothing -> Nothing Just (ty,coi') -> Just (ty, mkTransCoI coi' (ACo co)) | otherwise = pprPanic "go_down_equiv" (ppr fsk)  simonpj@microsoft.com committed Sep 13, 2010 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 \end{code} ********************************************************************************* * * The interact-with-inert Stage * * ********************************************************************************* \begin{code} -- Interaction result of WorkItem <~> AtomicInert data InteractResult = IR { ir_stop :: StopOrContinue -- Stop -- => Reagent (work item) consumed. -- ContinueWith new_reagent -- => Reagent transformed but keep gathering interactions. -- The transformed item remains inert with respect -- to any previously encountered inerts. , ir_inert_action :: InertAction -- Whether the inert item should remain in the InertSet. , ir_new_work :: WorkList -- new work items to add to the WorkList  830 831  , ir_improvement :: Maybe FDImprovement -- In case improvement kicked in  simonpj@microsoft.com committed Sep 13, 2010 832 833 834 835 836 837 838  } -- What to do with the inert reactant. data InertAction = KeepInert | DropInert deriving Eq mkIRContinue :: Monad m => WorkItem -> InertAction -> WorkList -> m InteractResult  839 mkIRContinue wi keep newWork = return$ IR (ContinueWith wi) keep newWork Nothing  simonpj@microsoft.com committed Sep 13, 2010 840 841  mkIRStop :: Monad m => InertAction -> WorkList -> m InteractResult  842 843 844 845 846 mkIRStop keep newWork = return $IR Stop keep newWork Nothing mkIRStop_RecordImprovement :: Monad m => InertAction -> WorkList -> FDImprovement -> m InteractResult mkIRStop_RecordImprovement keep newWork fdimpr = return$ IR Stop keep newWork (Just fdimpr)  simonpj@microsoft.com committed Sep 13, 2010 847 848  dischargeWorkItem :: Monad m => m InteractResult  849 dischargeWorkItem = mkIRStop KeepInert emptyWorkList  simonpj@microsoft.com committed Sep 13, 2010 850 851  noInteraction :: Monad m => WorkItem -> m InteractResult  852 noInteraction workItem = mkIRContinue workItem KeepInert emptyWorkList  simonpj@microsoft.com committed Sep 13, 2010 853   dimitris@microsoft.com committed Sep 23, 2010 854 data WhichComesFromInert = LeftComesFromInert | RightComesFromInert  simonpj@microsoft.com committed Sep 13, 2010 855   856   simonpj@microsoft.com committed Sep 13, 2010 857 ---------------------------------------------------  858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 -- Interact a single WorkItem with the equalities of an inert set as far as possible, i.e. until we -- get a Stop result from an individual reaction (i.e. when the WorkItem is consumed), or until we've -- interact the WorkItem with the entire equalities of the InertSet interactWithInertEqsStage :: SimplifierStage interactWithInertEqsStage workItem inert = foldISEqCtsM interactNext initITR inert where initITR = SR { sr_inerts = IS { inert_eqs = emptyCCan -- We will fold over the equalities , inert_fsks = Map.empty -- which will generate those two again , inert_cts = inert_cts inert , inert_fds = inert_fds inert } , sr_new_work = emptyWorkList , sr_stop = ContinueWith workItem }  simonpj@microsoft.com committed Sep 13, 2010 873   874 875 876 877 878 --------------------------------------------------- -- Interact a single WorkItem with *non-equality* constraints in the inert set. -- Precondition: equality interactions must have already happened, hence we have -- to pick up some information from the incoming inert, before folding over the -- "Other" constraints it contains!  simonpj@microsoft.com committed Sep 13, 2010 879 880 interactWithInertsStage :: SimplifierStage interactWithInertsStage workItem inert  881  = foldISOtherCtsM interactNext initITR inert  simonpj@microsoft.com committed Sep 13, 2010 882  where  883 884 885 886 887  initITR = SR { -- Pick up: (1) equations, (2) FD improvements, (3) FlatSkol equiv. classes sr_inerts = IS { inert_eqs = inert_eqs inert , inert_cts = emptyCCan , inert_fds = inert_fds inert , inert_fsks = inert_fsks inert }  888  , sr_new_work = emptyWorkList  simonpj@microsoft.com committed Sep 13, 2010 889 890  , sr_stop = ContinueWith workItem }  891 892 893 interactNext :: StageResult -> AtomicInert -> TcS StageResult interactNext it inert | ContinueWith workItem <- sr_stop it  simonpj@microsoft.com committed Oct 19, 2010 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909  = do { let inerts = sr_inerts it fdimprs_old = getFDImprovements inerts ; ir <- interactWithInert fdimprs_old inert workItem -- New inerts depend on whether we KeepInert or not and must -- be updated with FD improvement information from the interaction result (ir) ; let inerts_new = updInertSetFDImprs upd_inert (ir_improvement ir) upd_inert = if ir_inert_action ir == KeepInert then inerts updInertSet inert else inerts ; return $SR { sr_inerts = inerts_new , sr_new_work = sr_new_work it unionWorkLists ir_new_work ir , sr_stop = ir_stop ir } } | otherwise = return$ it { sr_inerts = (sr_inerts it) updInertSet inert }  simonpj@microsoft.com committed Sep 13, 2010 910 911  -- Do a single interaction of two constraints.  912 913 interactWithInert :: FDImprovements -> AtomicInert -> WorkItem -> TcS InteractResult interactWithInert fdimprs inert workitem  simonpj@microsoft.com committed Sep 13, 2010 914 915 916 917 918 919 920 921 922 923 924 925  = do { ctxt <- getTcSContext ; let is_allowed = allowedInteraction (simplEqsOnly ctxt) inert workitem inert_ev = cc_id inert work_ev = cc_id workitem -- Never interact a wanted and a derived where the derived's evidence -- mentions the wanted evidence in an unguarded way. -- See Note [Superclasses and recursive dictionaries] -- and Note [New Wanted Superclass Work] -- We don't have to do this for givens, as we fully know the evidence for them. ; rec_ev_ok <- case (cc_flavor inert, cc_flavor workitem) of  926 927 928  (Wanted loc, Derived {}) -> isGoodRecEv work_ev (WantedEvVar inert_ev loc) (Derived {}, Wanted loc) -> isGoodRecEv inert_ev (WantedEvVar work_ev loc) _ -> return True  simonpj@microsoft.com committed Sep 13, 2010 929 930  ; if is_allowed && rec_ev_ok then  931  doInteractWithInert fdimprs inert workitem  simonpj@microsoft.com committed Sep 13, 2010 932 933 934 935 936 937 938 939 940 941 942  else noInteraction workitem } allowedInteraction :: Bool -> AtomicInert -> WorkItem -> Bool -- Allowed interactions allowedInteraction eqs_only (CDictCan {}) (CDictCan {}) = not eqs_only allowedInteraction eqs_only (CIPCan {}) (CIPCan {}) = not eqs_only allowedInteraction _ _ _ = True --------------------------------------------  943 doInteractWithInert :: FDImprovements -> CanonicalCt -> CanonicalCt -> TcS InteractResult  simonpj@microsoft.com committed Sep 13, 2010 944 945 -- Identical class constraints.  946 doInteractWithInert fdimprs  simonpj@microsoft.com committed Sep 13, 2010 947 948 949 950 951 952 953  (CDictCan { cc_id = d1, cc_flavor = fl1, cc_class = cls1, cc_tyargs = tys1 }) workItem@(CDictCan { cc_id = d2, cc_flavor = fl2, cc_class = cls2, cc_tyargs = tys2 }) | cls1 == cls2 && (and $zipWith tcEqType tys1 tys2) = solveOneFromTheOther (d1,fl1) workItem | cls1 == cls2 && (not (isGiven fl1 && isGiven fl2)) = -- See Note [When improvement happens]  954 955 956 957  do { let pty1 = ClassP cls1 tys1 pty2 = ClassP cls2 tys2 work_item_pred_loc = (pty2, ppr d2) inert_pred_loc = (pty1, ppr d1)  simonpj@microsoft.com committed Sep 13, 2010 958 959  loc = combineCtLoc fl1 fl2 eqn_pred_locs = improveFromAnother work_item_pred_loc inert_pred_loc  960   simonpj@microsoft.com committed Sep 13, 2010 961  ; wevvars <- mkWantedFunDepEqns loc eqn_pred_locs  962  ; fd_cts <- canWanteds wevvars  963  ; let fd_work = workListFromCCans fd_cts  simonpj@microsoft.com committed Sep 13, 2010 964  -- See Note [Generating extra equalities]  965 966 967  ; traceTcS "Checking if improvements existed." (ppr fdimprs) ; if isEmptyCCan fd_cts || haveBeenImproved fdimprs pty1 pty2 then -- Must keep going  968  mkIRContinue workItem KeepInert fd_work  969 970  else do { traceTcS "Recording improvement and throwing item back in worklist." (ppr (pty1,pty2)) ; mkIRStop_RecordImprovement KeepInert  971  (fd_work unionWorkLists workListFromCCan workItem) (pty1,pty2)  972  }  simonpj@microsoft.com committed Sep 13, 2010 973 974 975 976 977  -- See Note [FunDep Reactions] } -- Class constraint and given equality: use the equality to rewrite -- the class constraint.  978 979 doInteractWithInert _fdimprs (CTyEqCan { cc_id = cv, cc_flavor = ifl, cc_tyvar = tv, cc_rhs = xi })  simonpj@microsoft.com committed Sep 13, 2010 980 981 982  (CDictCan { cc_id = dv, cc_flavor = wfl, cc_class = cl, cc_tyargs = xis }) | ifl canRewrite wfl , tv elemVarSet tyVarsOfTypes xis  983 984 985 986 987 988  = if isDerivedSC wfl then mkIRStop KeepInert$ emptyWorkList -- See Note [Adding Derived Superclasses] else do { rewritten_dict <- rewriteDict (cv,tv,xi) (dv,wfl,cl,xis) -- Continue with rewritten Dictionary because we can only be in the -- interactWithEqsStage, so the dictionary is inert. ; mkIRContinue rewritten_dict KeepInert emptyWorkList }  simonpj@microsoft.com committed Sep 13, 2010 989   990 991 doInteractWithInert _fdimprs (CDictCan { cc_id = dv, cc_flavor = ifl, cc_class = cl, cc_tyargs = xis })  simonpj@microsoft.com committed Sep 13, 2010 992 993 994  workItem@(CTyEqCan { cc_id = cv, cc_flavor = wfl, cc_tyvar = tv, cc_rhs = xi }) | wfl canRewrite ifl , tv elemVarSet tyVarsOfTypes xis  995 996 997 998 999  = if isDerivedSC ifl then mkIRContinue workItem DropInert emptyWorkList -- No need to do any rewriting, -- see Note [Adding Derived Superclasses] else do { rewritten_dict <- rewriteDict (cv,tv,xi) (dv,ifl,cl,xis) ; mkIRContinue workItem DropInert (workListFromCCan rewritten_dict) }  simonpj@microsoft.com committed Sep 13, 2010 1000 1001 1002  -- Class constraint and given equality: use the equality to rewrite -- the class constraint.  1003 1004 doInteractWithInert _fdimprs (CTyEqCan { cc_id = cv, cc_flavor = ifl, cc_tyvar = tv, cc_rhs = xi })  simonpj@microsoft.com committed Sep 13, 2010 1005 1006 1007 1008  (CIPCan { cc_id = ipid, cc_flavor = wfl, cc_ip_nm = nm, cc_ip_ty = ty }) | ifl canRewrite wfl , tv elemVarSet tyVarsOfType ty = do { rewritten_ip <- rewriteIP (cv,tv,xi) (ipid,wfl,nm,ty)  1009  ; mkIRContinue rewritten_ip KeepInert emptyWorkList }  simonpj@microsoft.com committed Sep 13, 2010 1010   1011 1012 doInteractWithInert _fdimprs (CIPCan { cc_id = ipid, cc_flavor = ifl, cc_ip_nm = nm, cc_ip_ty = ty })  simonpj@microsoft.com committed Sep 13, 2010 1013 1014 1015 1016  workItem@(CTyEqCan { cc_id = cv, cc_flavor = wfl, cc_tyvar = tv, cc_rhs = xi }) | wfl canRewrite ifl , tv elemVarSet tyVarsOfType ty = do { rewritten_ip <- rewriteIP (cv,tv,xi) (ipid,ifl,nm,ty)  1017  ; mkIRContinue workItem DropInert (workListFromCCan rewritten_ip) }  simonpj@microsoft.com committed Sep 13, 2010 1018 1019 1020 1021 1022 1023  -- Two implicit parameter constraints. If the names are the same, -- but their types are not, we generate a wanted type equality -- that equates the type (this is "improvement"). -- However, we don't actually need the coercion evidence, -- so we just generate a fresh coercion variable that isn't used anywhere.  1024 1025 doInteractWithInert _fdimprs (CIPCan { cc_id = id1, cc_flavor = ifl, cc_ip_nm = nm1, cc_ip_ty = ty1 })  simonpj@microsoft.com committed Sep 13, 2010 1026  workItem@(CIPCan { cc_flavor = wfl, cc_ip_nm = nm2, cc_ip_ty = ty2 })  simonpj@microsoft.com committed Sep 17, 2010 1027 1028 1029 1030  | nm1 == nm2 && isGiven wfl && isGiven ifl = -- See Note [Overriding implicit parameters] -- Dump the inert item, override totally with the new one -- Do not require type equality  1031  mkIRContinue workItem DropInert emptyWorkList  simonpj@microsoft.com committed Sep 17, 2010 1032   simonpj@microsoft.com committed Sep 13, 2010 1033 1034 1035  | nm1 == nm2 && ty1 tcEqType ty2 = solveOneFromTheOther (id1,ifl) workItem  simonpj@microsoft.com committed Sep 17, 2010 1036  | nm1 == nm2  simonpj@microsoft.com committed Sep 13, 2010 1037 1038 1039  = -- See Note [When improvement happens] do { co_var <- newWantedCoVar ty1 ty2 ; let flav = Wanted (combineCtLoc ifl wfl)  1040  ; cans <- mkCanonical flav co_var  1041  ; mkIRContinue workItem KeepInert (workListFromCCans cans) }  simonpj@microsoft.com committed Sep 13, 2010 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052  -- Inert: equality, work item: function equality -- Never rewrite a given with a wanted equality, and a type function -- equality can never rewrite an equality. Note also that if we have -- F x1 ~ x2 and a ~ x3, and a occurs in x2, we don't rewrite it. We -- can wait until F x1 ~ x2 matches another F x1 ~ x4, and only then -- we will expose'' x2 and x4 to rewriting. -- Otherwise, we can try rewriting the type function equality with the equality.  1053 1054 doInteractWithInert _fdimprs (CTyEqCan { cc_id = cv1, cc_flavor = ifl, cc_tyvar = tv, cc_rhs = xi1 })  simonpj@microsoft.com committed Sep 13, 2010 1055 1056 1057 1058 1059  (CFunEqCan { cc_id = cv2, cc_flavor = wfl, cc_fun = tc , cc_tyargs = args, cc_rhs = xi2 }) | ifl canRewrite wfl , tv elemVarSet tyVarsOfTypes args = do { rewritten_funeq <- rewriteFunEq (cv1,tv,xi1) (cv2,wfl,tc,args,xi2)  1060 1061  ; mkIRStop KeepInert (workListFromCCan rewritten_funeq) } -- must Stop here, because we may no longer be inert after the rewritting.  simonpj@microsoft.com committed Sep 13, 2010 1062 1063  -- Inert: function equality, work item: equality  1064 1065 doInteractWithInert _fdimprs (CFunEqCan {cc_id = cv1, cc_flavor = ifl, cc_fun = tc  simonpj@microsoft.com committed Sep 13, 2010 1066 1067 1068 1069 1070  , cc_tyargs = args, cc_rhs = xi1 }) workItem@(CTyEqCan { cc_id = cv2, cc_flavor = wfl, cc_tyvar = tv, cc_rhs = xi2 }) | wfl canRewrite ifl , tv elemVarSet tyVarsOfTypes args = do { rewritten_funeq <- rewriteFunEq (cv2,tv,xi2) (cv1,ifl,tc,args,xi1)  1071  ; mkIRContinue workItem DropInert (workListFromCCan rewritten_funeq) }  simonpj@microsoft.com committed Sep 13, 2010 1072   1073 1074 doInteractWithInert _fdimprs (CFunEqCan { cc_id = cv1, cc_flavor = fl1, cc_fun = tc1  simonpj@microsoft.com committed Sep 13, 2010 1075 1076 1077  , cc_tyargs = args1, cc_rhs = xi1 }) workItem@(CFunEqCan { cc_id = cv2, cc_flavor = fl2, cc_fun = tc2 , cc_tyargs = args2, cc_rhs = xi2 })  1078  | fl1 canSolve fl2 && lhss_match  dimitris@microsoft.com committed Sep 23, 2010 1079  = do { cans <- rewriteEqLHS LeftComesFromInert (mkCoVarCoercion cv1,xi1) (cv2,fl2,xi2)  1080  ; mkIRStop KeepInert (workListFromCCans cans) }  1081  | fl2 canSolve fl1 && lhss_match  dimitris@microsoft.com committed Sep 23, 2010 1082  = do { cans <- rewriteEqLHS RightComesFromInert (mkCoVarCoercion cv2,xi2) (cv1,fl1,xi1)  1083  ; mkIRContinue workItem DropInert (workListFromCCans cans) }  simonpj@microsoft.com committed Sep 13, 2010 1084 1085 1086  where lhss_match = tc1 == tc2 && and (zipWith tcEqType args1 args2)  1087 doInteractWithInert _fdimprs  dimitris@microsoft.com committed Oct 04, 2010 1088  inert@(CTyEqCan { cc_id = cv1, cc_flavor = fl1, cc_tyvar = tv1, cc_rhs = xi1 })  simonpj@microsoft.com committed Sep 13, 2010 1089 1090  workItem@(CTyEqCan { cc_id = cv2, cc_flavor = fl2, cc_tyvar = tv2, cc_rhs = xi2 }) -- Check for matching LHS  1091  | fl1 canSolve fl2 && tv1 == tv2  dimitris@microsoft.com committed Sep 23, 2010 1092  = do { cans <- rewriteEqLHS LeftComesFromInert (mkCoVarCoercion cv1,xi1) (cv2,fl2,xi2)  1093  ; mkIRStop KeepInert (workListFromCCans cans) }  simonpj@microsoft.com committed Sep 13, 2010 1094   1095  | fl2 canSolve fl1 && tv1 == tv2  dimitris@microsoft.com committed Sep 23, 2010 1096  = do { cans <- rewriteEqLHS RightComesFromInert (mkCoVarCoercion cv2,xi2) (cv1,fl1,xi1)  1097  ; mkIRContinue workItem DropInert (workListFromCCans cans) }  simonpj@microsoft.com committed Sep 13, 2010 1098 1099 1100 1101  -- Check for rewriting RHS | fl1 canRewrite fl2 && tv1 elemVarSet tyVarsOfType xi2 = do { rewritten_eq <- rewriteEqRHS (cv1,tv1,xi1) (cv2,fl2,tv2,xi2)  1102  ; mkIRStop KeepInert (workListFromCCans rewritten_eq) }  simonpj@microsoft.com committed Sep 13, 2010 1103 1104  | fl2 canRewrite fl1 && tv2 elemVarSet tyVarsOfType xi1 = do { rewritten_eq <- rewriteEqRHS (cv2,tv2,xi2) (cv1,fl1,tv1,xi1)  1105  ; mkIRContinue workItem DropInert (workListFromCCans rewritten_eq) }  dimitris@microsoft.com committed Oct 06, 2010 1106 1107 1108 1109 1110 1111  -- Finally, if workitem is a Flatten Equivalence Class constraint and the -- inert is a wanted constraint, even when the workitem cannot rewrite the -- inert, drop the inert out because you may have to reconsider solving the -- inert *using* the equivalence class you created. See note [Loopy Spontaneous Solving] -- and [InertSet FlattenSkolemEqClass]  dimitris@microsoft.com committed Oct 04, 2010 1112 1113 1114 1115 1116  | not $isGiven fl1, -- The inert is wanted or derived isMetaTyVar tv1, -- and has a unification variable lhs FlatSkol {} <- tcTyVarDetails tv2, -- And workitem is a flatten skolem equality Just tv2' <- tcGetTyVar_maybe xi2, FlatSkol {} <- tcTyVarDetails tv2'  1117  = mkIRContinue workItem DropInert (workListFromCCan inert)  simonpj@microsoft.com committed Sep 13, 2010 1118 1119   dimitris@microsoft.com committed Oct 06, 2010 1120 -- Fall-through case for all other situations  1121 doInteractWithInert _fdimprs _ workItem = noInteraction workItem  simonpj@microsoft.com committed Sep 13, 2010 1122   simonpj@microsoft.com committed Oct 08, 2010 1123 -------------------------  simonpj@microsoft.com committed Sep 13, 2010 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 -- Equational Rewriting rewriteDict :: (CoVar, TcTyVar, Xi) -> (DictId, CtFlavor, Class, [Xi]) -> TcS CanonicalCt rewriteDict (cv,tv,xi) (dv,gw,cl,xis) = do { let cos = substTysWith [tv] [mkCoVarCoercion cv] xis -- xis[tv] ~ xis[xi] args = substTysWith [tv] [xi] xis con = classTyCon cl dict_co = mkTyConCoercion con cos ; dv' <- newDictVar cl args ; case gw of Wanted {} -> setDictBind dv (EvCast dv' (mkSymCoercion dict_co)) _given_or_derived -> setDictBind dv' (EvCast dv dict_co) ; return (CDictCan { cc_id = dv' , cc_flavor = gw , cc_class = cl , cc_tyargs = args }) } rewriteIP :: (CoVar,TcTyVar,Xi) -> (EvVar,CtFlavor, IPName Name, TcType) -> TcS CanonicalCt rewriteIP (cv,tv,xi) (ipid,gw,nm,ty) = do { let ip_co = substTyWith [tv] [mkCoVarCoercion cv] ty -- ty[tv] ~ t[xi] ty' = substTyWith [tv] [xi] ty ; ipid' <- newIPVar nm ty' ; case gw of Wanted {} -> setIPBind ipid (EvCast ipid' (mkSymCoercion ip_co)) _given_or_derived -> setIPBind ipid' (EvCast ipid ip_co) ; return (CIPCan { cc_id = ipid' , cc_flavor = gw , cc_ip_nm = nm , cc_ip_ty = ty' }) } rewriteFunEq :: (CoVar,TcTyVar,Xi) -> (CoVar,CtFlavor,TyCon, [Xi], Xi) -> TcS CanonicalCt rewriteFunEq (cv1,tv,xi1) (cv2,gw, tc,args,xi2) = do { let arg_cos = substTysWith [tv] [mkCoVarCoercion cv1] args args' = substTysWith [tv] [xi1] args fun_co = mkTyConCoercion tc arg_cos ; cv2' <- case gw of Wanted {} -> do { cv2' <- newWantedCoVar (mkTyConApp tc args') xi2 ; setWantedCoBind cv2$ mkTransCoercion fun_co (mkCoVarCoercion cv2') ; return cv2' } _giv_or_der -> newGivOrDerCoVar (mkTyConApp tc args') xi2 $mkTransCoercion (mkSymCoercion fun_co) (mkCoVarCoercion cv2) ; return (CFunEqCan { cc_id = cv2' , cc_flavor = gw , cc_tyargs = args' , cc_fun = tc , cc_rhs = xi2 }) } rewriteEqRHS :: (CoVar,TcTyVar,Xi) -> (CoVar,CtFlavor,TcTyVar,Xi) -> TcS CanonicalCts -- Use the first equality to rewrite the second, flavors already checked. -- E.g. c1 : tv1 ~ xi1 c2 : tv2 ~ xi2 -- rewrites c2 to give -- c2' : tv2 ~ xi2[xi1/tv1] -- We must do an occurs check to sure the new constraint is canonical -- So we might return an empty bag rewriteEqRHS (cv1,tv1,xi1) (cv2,gw,tv2,xi2) | Just tv2' <- tcGetTyVar_maybe xi2' , tv2 == tv2' -- In this case xi2[xi1/tv1] = tv2, so we have tv2~tv2 = do { when (isWanted gw) (setWantedCoBind cv2 (mkSymCoercion co2')) ; return emptyCCan } | otherwise = do { cv2' <- case gw of Wanted {} -> do { cv2' <- newWantedCoVar (mkTyVarTy tv2) xi2' ; setWantedCoBind cv2$ mkCoVarCoercion cv2' mkTransCoercion mkSymCoercion co2' ; return cv2' } _giv_or_der -> newGivOrDerCoVar (mkTyVarTy tv2) xi2' $mkCoVarCoercion cv2 mkTransCoercion co2' ; xi2'' <- canOccursCheck gw tv2 xi2' -- we know xi2' is *not* tv2 ; return (singleCCan$ CTyEqCan { cc_id = cv2' , cc_flavor = gw , cc_tyvar = tv2  1200  , cc_rhs = xi2'' })  simonpj@microsoft.com committed Sep 13, 2010 1201 1202 1203 1204 1205  } where xi2' = substTyWith [tv1] [xi1] xi2 co2' = substTyWith [tv1] [mkCoVarCoercion cv1] xi2 -- xi2 ~ xi2[xi1/tv1]  dimitris@microsoft.com committed Sep 23, 2010 1206 1207  rewriteEqLHS :: WhichComesFromInert -> (Coercion,Xi) -> (CoVar,CtFlavor,Xi) -> TcS CanonicalCts  1208 -- Used to ineract two equalities of the following form:  simonpj@microsoft.com committed Sep 13, 2010 1209 1210 -- First Equality: co1: (XXX ~ xi1) -- Second Equality: cv2: (XXX ~ xi2)  1211 -- Where the cv1 canSolve cv2 equality  dimitris@microsoft.com committed Sep 23, 2010 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 -- We have an option of creating new work (xi1 ~ xi2) OR (xi2 ~ xi1). This -- depends on whether the left or the right equality comes from the inert set. -- We must: -- prefer to create (xi2 ~ xi1) if the first comes from the inert -- prefer to create (xi1 ~ xi2) if the second comes from the inert rewriteEqLHS which (co1,xi1) (cv2,gw,xi2) = do { cv2' <- case (isWanted gw, which) of (True,LeftComesFromInert) -> do { cv2' <- newWantedCoVar xi2 xi1 ; setWantedCoBind cv2 $co1 mkTransCoercion mkSymCoercion (mkCoVarCoercion cv2') ; return cv2' } (True,RightComesFromInert) -> do { cv2' <- newWantedCoVar xi1 xi2 ; setWantedCoBind cv2$ co1 mkTransCoercion mkCoVarCoercion cv2' ; return cv2' } (False,LeftComesFromInert) -> newGivOrDerCoVar xi2 xi1 $mkSymCoercion (mkCoVarCoercion cv2) mkTransCoercion co1 (False,RightComesFromInert) -> newGivOrDerCoVar xi1 xi2$ mkSymCoercion co1 mkTransCoercion mkCoVarCoercion cv2  1235 1236 1237  ; mkCanonical gw cv2' }  simonpj@microsoft.com committed Sep 13, 2010 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 solveOneFromTheOther :: (EvVar, CtFlavor) -> CanonicalCt -> TcS InteractResult -- First argument inert, second argument workitem. They both represent -- wanted/given/derived evidence for the *same* predicate so we try here to -- discharge one directly from the other. -- -- Precondition: value evidence only (implicit parameters, classes) -- not coercion solveOneFromTheOther (iid,ifl) workItem -- Both derived needs a special case. You might think that we do not need -- two evidence terms for the same claim. But, since the evidence is partial, -- either evidence may do in some cases; see TcSMonad.isGoodRecEv. -- See also Example 3 in Note [Superclasses and recursive dictionaries] | isDerived ifl && isDerived wfl = noInteraction workItem  1253  | ifl canSolve wfl  simonpj@microsoft.com committed Sep 17, 2010 1254 1255 1256 1257 1258  = do { unless (isGiven wfl) $setEvBind wid (EvId iid) -- Overwrite the binding, if one exists -- For Givens, which are lambda-bound, nothing to overwrite, ; dischargeWorkItem }  1259  | otherwise -- wfl canSolve ifl  simonpj@microsoft.com committed Sep 13, 2010 1260  = do { unless (isGiven ifl)$ setEvBind iid (EvId wid)  1261  ; mkIRContinue workItem DropInert emptyWorkList }  simonpj@microsoft.com committed Sep 13, 2010 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684  where wfl = cc_flavor workItem wid = cc_id workItem \end{code} Note [Superclasses and recursive dictionaries] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Overlaps with Note [SUPERCLASS-LOOP 1] Note [SUPERCLASS-LOOP 2] Note [Recursive instances and superclases] ToDo: check overlap and delete redundant stuff Right before adding a given into the inert set, we must produce some more work, that will bring the superclasses of the given into scope. The superclass constraints go into our worklist. When we simplify a wanted constraint, if we first see a matching instance, we may produce new wanted work. To (1) avoid doing this work twice in the future and (2) to handle recursive dictionaries we may cache'' this item as solved (in effect, given) into our inert set and with that add its superclass constraints (as given) in our worklist. But now we have added partially solved constraints to the worklist which may interact with other wanteds. Consider the example: Example 1: class Eq b => Foo a b --- 0-th selector instance Eq a => Foo [a] a --- fooDFun and wanted (Foo [t] t). We are first going to see that the instance matches and create an inert set that includes the solved (Foo [t] t) and its superclasses. d1 :_g Foo [t] t d1 := EvDFunApp fooDFun d3 d2 :_g Eq t d2 := EvSuperClass d1 0 Our work list is going to contain a new *wanted* goal d3 :_w Eq t It is wrong to react the wanted (Eq t) with the given (Eq t) because that would construct loopy evidence. Hence the check isGoodRecEv in doInteractWithInert. OK, so we have ruled out bad behaviour, but how do we ge recursive dictionaries, at all? Consider Example 2: data D r = ZeroD | SuccD (r (D r)); instance (Eq (r (D r))) => Eq (D r) where ZeroD == ZeroD = True (SuccD a) == (SuccD b) = a == b _ == _ = False; equalDC :: D [] -> D [] -> Bool; equalDC = (==); We need to prove (Eq (D [])). Here's how we go: d1 :_w Eq (D []) by instance decl, holds if d2 :_w Eq [D []] where d1 = dfEqD d2 *BUT* we have an inert set which gives us (no superclasses): d1 :_g Eq (D []) By the instance declaration of Eq we can show the 'd2' goal if d3 :_w Eq (D []) where d2 = dfEqList d3 d1 = dfEqD d2 Now, however this wanted can interact with our inert d1 to set: d3 := d1 and solve the goal. Why was this interaction OK? Because, if we chase the evidence of d1 ~~> dfEqD d2 ~~-> dfEqList d3, so by setting d3 := d1 we are really setting d3 := dfEqD2 (dfEqList d3) which is FINE because the use of d3 is protected by the instance function applications. So, our strategy is to try to put solved wanted dictionaries into the inert set along with their superclasses (when this is meaningful, i.e. when new wanted goals are generated) but solve a wanted dictionary from a given only in the case where the evidence variable of the wanted is mentioned in the evidence of the given (recursively through the evidence binds) in a protected way: more instance function applications than superclass selectors. Here are some more examples from GHC's previous type checker Example 3: This code arises in the context of "Scrap Your Boilerplate with Class" class Sat a class Data ctx a instance Sat (ctx Char) => Data ctx Char -- dfunData1 instance (Sat (ctx [a]), Data ctx a) => Data ctx [a] -- dfunData2 class Data Maybe a => Foo a instance Foo t => Sat (Maybe t) -- dfunSat instance Data Maybe a => Foo a -- dfunFoo1 instance Foo a => Foo [a] -- dfunFoo2 instance Foo [Char] -- dfunFoo3 Consider generating the superclasses of the instance declaration instance Foo a => Foo [a] So our problem is this d0 :_g Foo t d1 :_w Data Maybe [t] We may add the given in the inert set, along with its superclasses [assuming we don't fail because there is a matching instance, see tryTopReact, given case ] Inert: d0 :_g Foo t WorkList d01 :_g Data Maybe t -- d2 := EvDictSuperClass d0 0 d1 :_w Data Maybe [t] Then d2 can readily enter the inert, and we also do solving of the wanted Inert: d0 :_g Foo t d1 :_s Data Maybe [t] d1 := dfunData2 d2 d3 WorkList d2 :_w Sat (Maybe [t]) d3 :_w Data Maybe t d01 :_g Data Maybe t Now, we may simplify d2 more: Inert: d0 :_g Foo t d1 :_s Data Maybe [t] d1 := dfunData2 d2 d3 d1 :_g Data Maybe [t] d2 :_g Sat (Maybe [t]) d2 := dfunSat d4 WorkList: d3 :_w Data Maybe t d4 :_w Foo [t] d01 :_g Data Maybe t Now, we can just solve d3. Inert d0 :_g Foo t d1 :_s Data Maybe [t] d1 := dfunData2 d2 d3 d2 :_g Sat (Maybe [t]) d2 := dfunSat d4 WorkList d4 :_w Foo [t] d01 :_g Data Maybe t And now we can simplify d4 again, but since it has superclasses we *add* them to the worklist: Inert d0 :_g Foo t d1 :_s Data Maybe [t] d1 := dfunData2 d2 d3 d2 :_g Sat (Maybe [t]) d2 := dfunSat d4 d4 :_g Foo [t] d4 := dfunFoo2 d5 WorkList: d5 :_w Foo t d6 :_g Data Maybe [t] d6 := EvDictSuperClass d4 0 d01 :_g Data Maybe t Now, d5 can be solved! (and its superclass enter scope) Inert d0 :_g Foo t d1 :_s Data Maybe [t] d1 := dfunData2 d2 d3 d2 :_g Sat (Maybe [t]) d2 := dfunSat d4 d4 :_g Foo [t] d4 := dfunFoo2 d5 d5 :_g Foo t d5 := dfunFoo1 d7 WorkList: d7 :_w Data Maybe t d6 :_g Data Maybe [t] d8 :_g Data Maybe t d8 := EvDictSuperClass d5 0 d01 :_g Data Maybe t Now, two problems: [1] Suppose we pick d8 and we react him with d01. Which of the two givens should we keep? Well, we *MUST NOT* drop d01 because d8 contains recursive evidence that must not be used (look at case interactInert where both inert and workitem are givens). So we have several options: - Drop the workitem always (this will drop d8) This feels very unsafe -- what if the work item was the "good" one that should be used later to solve another wanted? - Don't drop anyone: the inert set may contain multiple givens! [This is currently implemented] The "don't drop anyone" seems the most safe thing to do, so now we come to problem 2: [2] We have added both d6 and d01 in the inert set, and we are interacting our wanted d7. Now the [isRecDictEv] function in the ineration solver [case inert-given workitem-wanted] will prevent us from interacting d7 := d8 precisely because chasing the evidence of d8 leads us to an unguarded use of d7. So, no interaction happens there. Then we meet d01 and there is no recursion problem there [isRectDictEv] gives us the OK to interact and we do solve d7 := d01! Note [SUPERCLASS-LOOP 1] ~~~~~~~~~~~~~~~~~~~~~~~~ We have to be very, very careful when generating superclasses, lest we accidentally build a loop. Here's an example: class S a class S a => C a where { opc :: a -> a } class S b => D b where { opd :: b -> b } instance C Int where