TcInteract.lhs 85.5 KB
Newer Older
1 2 3
\begin{code}
module TcInteract ( 
     solveInteract, AtomicInert, 
4
     InertSet, emptyInert, updInertSet, extractUnsolved, solveOne 
5 6 7 8
  ) where  

#include "HsVersions.h"

9

10 11 12 13
import BasicTypes 
import TcCanonical
import VarSet
import Type
14
import TypeRep 
15 16

import Id 
17
import VarEnv
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 
38
import Bag
39 40 41
import qualified Data.Map as Map 
import Maybes 

42 43 44 45 46
import Control.Monad( zipWithM, unless )
import FastString ( sLit ) 
import DynFlags
\end{code}

47
Note [InertSet invariants]
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).

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] 
109 110 111
\begin{code}

-- See Note [InertSet invariants]
112
data InertSet 
113 114 115 116
  = IS { inert_eqs  :: Bag.Bag CanonicalCt   -- Equalities only (CTyEqCan,CFunEqCan)
       , 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 
117
       , inert_fsks :: Map.Map TcTyVar [(TcTyVar,Coercion)] }
118
       -- See Note [InertSet FlattenSkolemEqClass] 
119

120 121 122
type FDImprovement  = (PredType,PredType) 
type FDImprovements = [(PredType,PredType)] 

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))
126 127 128 129 130
                , vcat (map (\(v,rest) -> ppr v <+> text "|->" <+> hsep (map (ppr.fst) rest)) 
                       (Map.toList $ inert_fsks is)
                       )
                ]
                       
131
emptyInert :: InertSet
132
emptyInert = IS { inert_eqs = Bag.emptyBag
133
                , inert_cts = Bag.emptyBag, inert_fsks = Map.empty, inert_fds = [] }
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 })  
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 
145
        fsks' = Map.insertWith (++) tv2 [(tv1, mkCoVarCoercion cv)] fsks
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 
150 151
  | isEqCCan item 
  = let eqs' = eqs `Bag.snocBag` item 
152
    in IS { inert_eqs = eqs', inert_cts = cts, inert_fsks = fsks, inert_fds = fdis } 
153
  | otherwise 
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 
160

161
foldlInertSetM :: (Monad m) => (a -> AtomicInert -> m a) -> a -> InertSet -> m a 
162 163 164 165
-- Prioritize over the equalities see Note [Prioritizing Equalities]
foldlInertSetM k z (IS { inert_eqs = eqs, inert_cts = cts }) 
  = do { z' <- Bag.foldlBagM k z eqs
       ; Bag.foldlBagM k z' cts } 
166 167

extractUnsolved :: InertSet -> (InertSet, CanonicalCts)
168
extractUnsolved is@(IS {inert_eqs = eqs, inert_cts = cts, inert_fds = fdis }) 
169
  = let is_init  = is { inert_eqs = emptyCCan 
170
                      , inert_cts = solved_cts, inert_fsks = Map.empty, inert_fds = fdis }
171 172 173 174 175
        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
176

177 178

getFskEqClass :: InertSet -> TcTyVar -> [(TcTyVar,Coercion)] 
179
-- Precondition: tv is a FlatSkol. See Note [InertSet FlattenSkolemEqClass] 
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
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 

195 196 197 198 199 200 201 202 203 204 205 206 207 208
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 

209

210 211
isWantedCt :: CanonicalCt -> Bool 
isWantedCt ct = isWanted (cc_flavor ct)
212 213 214 215 216 217 218 219 220 221 222

{- 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?
-}

223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
\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
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

286 287 288 289 290 291
-- 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 
292 293

unionWorkLists :: WorkList -> WorkList -> WorkList 
294 295 296
unionWorkLists wl1 wl2 
  = WL { wl_eqs   = andCCan (wl_eqs wl1) (wl_eqs wl2)
       , wl_other = andCCan (wl_other wl1) (wl_other wl2) }
297 298

foldlWorkListM :: (Monad m) => (a -> WorkItem -> m a) -> a -> WorkList -> m a 
299 300 301 302
-- This fold prioritizes equalities 
foldlWorkListM f r wl 
  = do { r' <- Bag.foldlBagM f r (wl_eqs wl) 
       ; Bag.foldlBagM f r' (wl_other wl) }
303 304

isEmptyWorkList :: WorkList -> Bool 
305
isEmptyWorkList wl = isEmptyCCan (wl_eqs wl) && isEmptyCCan (wl_other wl) 
306 307

emptyWorkList :: WorkList
308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327
emptyWorkList = WL { wl_eqs = emptyCCan, wl_other = emptyCCan } 

mkEqWorkList :: CanonicalCts -> WorkList
-- Precondition: *ALL* equalities
mkEqWorkList eqs = WL { wl_eqs = eqs, wl_other = emptyCCan } 

mkNonEqWorkList :: CanonicalCts -> WorkList 
-- Precondition: *NO* equalities 
mkNonEqWorkList cts = WL { wl_eqs = emptyCCan, wl_other = cts } 

workListFromCCans :: CanonicalCts -> WorkList 
-- Generic, no precondition 
workListFromCCans cts = WL eqs others 
  where (eqs, others) = Bag.partitionBag isEqCCan cts

-- Convenience 
singleEqWL    :: CanonicalCt -> WorkList 
singleNonEqWL :: CanonicalCt -> WorkList 
singleEqWL    = mkEqWorkList . singleCCan 
singleNonEqWL = mkNonEqWorkList . singleCCan 
328

329

330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
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])

357 358 359
instance Outputable WorkList where 
  ppr (WL eqcts othercts) = vcat [ppr eqcts, ppr othercts] 

360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
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
379
                  ContinueWith item -> sr_inerts itr_out `updInertSet` item
380 381 382 383 384 385 386 387 388 389 390 391 392
       ; 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)
393
           ; let itr' = itr { sr_new_work = accum_work `unionWorkLists` sr_new_work itr }
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
           ; 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].
425
solveInteract :: InertSet -> CanonicalCts -> TcS InertSet
426 427
solveInteract inert ws 
  = do { dyn_flags <- getDynFlags
428 429
       ; let worklist = workListFromCCans ws 
       ; solveInteractWithDepth (ctxtStkDepth dyn_flags,0,[]) inert worklist
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
       }
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
              ]
       ; foldlWorkListM (solveOneWithDepth ctxt) inert ws }

------------------
-- 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)]
thePipeline = [ ("interact with inerts", interactWithInertsStage)
              , ("spontaneous solve",    spontaneousSolveStage)
              , ("top-level reactions",  topReactionsStage) ]
\end{code}

*********************************************************************************
*                                                                               * 
                       The spontaneous-solve Stage
*                                                                               *
*********************************************************************************

\begin{code}
spontaneousSolveStage :: SimplifierStage 
spontaneousSolveStage workItem inerts 
490
  = do { mSolve <- trySpontaneousSolve workItem inerts
491 492
       ; case mSolve of 
           Nothing -> -- no spontaneous solution for him, keep going
493 494
               return $ SR { sr_new_work   = emptyWorkList
                           , sr_inerts     = inerts
495 496
                           , sr_stop       = ContinueWith workItem }

497
           Just workList' -> -- He has been solved; workList' are all givens
498 499
               return $ SR { sr_new_work = workList'
                           , sr_inerts   = inerts 
500
                           , sr_stop     = Stop }
501
       }
502 503 504 505 506 507 508 509 510

{-- This is all old code, but does not quite work now. The problem is that due to 
    Note [Loopy Spontaneous Solving] we may have unflattened a type, to be able to 
    perform a sneaky unification. This unflattening means that we may have to recanonicalize
    a given (solved) equality, this is why the result of trySpontaneousSolve is now a list
    of constraints (instead of an atomic solved constraint). We would have to react all of 
    them once again with the worklist but that is very tiresome. Instead we throw them back
    in the worklist. 

511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
               | isWantedCt workItem 
                           -- Original was wanted we have now made him given so 
                           -- we have to ineract him with the inerts again because 
                           -- of the change in his status. This may produce some work. 
                   -> do { traceTcS "recursive interact with inerts {" $ vcat
                               [ text "work = " <+> ppr workItem'
                               , text "inerts = " <+> ppr inerts ]
                         ; itr_again <- interactWithInertsStage workItem' inerts 
                         ; case sr_stop itr_again of 
                            Stop -> pprPanic "BUG: Impossible to happen" $ 
                                    vcat [ text "Original workitem:" <+> ppr workItem
                                         , text "Spontaneously solved:" <+> ppr workItem'
                                         , text "Solved was consumed, when reacting with inerts:"
                                         , nest 2 (ppr inerts) ]
                            ContinueWith workItem'' -- Now *this* guy is inert wrt to inerts
                                ->  do { traceTcS "end recursive interact }" $ ppr workItem''
                                       ; return $ SR { sr_new_work = sr_new_work itr_again
                                                     , sr_inerts   = sr_inerts itr_again 
                                                                     `extendInertSet` workItem'' 
                                                     , sr_stop     = Stop } }
                         }
               | otherwise
                   -> return $ SR { sr_new_work   = emptyWorkList 
                                  , sr_inerts     = inerts `extendInertSet` workItem' 
                                  , sr_stop       = Stop } }
536
--} 
537 538 539 540 541 542

-- @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] 
543
-- NB: just passing the inerts through for the skolem equivalence classes
544 545
trySpontaneousSolve :: WorkItem -> InertSet -> TcS (Maybe SWorkList)
trySpontaneousSolve (CTyEqCan { cc_id = cv, cc_flavor = gw, cc_tyvar = tv1, cc_rhs = xi }) inerts 
546 547
  | isGiven gw
  = return Nothing
548 549 550 551
  | Just tv2 <- tcGetTyVar_maybe xi
  = do { tch1 <- isTouchableMetaTyVar tv1
       ; tch2 <- isTouchableMetaTyVar tv2
       ; case (tch1, tch2) of
552 553
           (True,  True)  -> trySpontaneousEqTwoWay inerts cv gw tv1 tv2
           (True,  False) -> trySpontaneousEqOneWay inerts cv gw tv1 xi
554
           (False, True)  -> trySpontaneousEqOneWay inerts cv gw tv2 (mkTyVarTy tv1)
555 556 557
	   _ -> return Nothing }
  | otherwise
  = do { tch1 <- isTouchableMetaTyVar tv1
558
       ; if tch1 then trySpontaneousEqOneWay inerts cv gw tv1 xi
559 560 561 562 563
                 else return Nothing }

  -- No need for 
  --      trySpontaneousSolve (CFunEqCan ...) = ...
  -- See Note [No touchables as FunEq RHS] in TcSMonad
564
trySpontaneousSolve _ _ = return Nothing 
565 566

----------------
567 568
trySpontaneousEqOneWay :: InertSet -> CoVar -> CtFlavor -> TcTyVar -> Xi
                       -> TcS (Maybe SWorkList)
569
-- tv is a MetaTyVar, not untouchable
570
trySpontaneousEqOneWay inerts cv gw tv xi	
571 572
  | not (isSigTyVar tv) || isTyVarTy xi, 
    typeKind xi `isSubKind` tyVarKind tv 
573
  = solveWithIdentity inerts cv gw tv xi
574 575 576 577
  | otherwise
  = return Nothing

----------------
578 579
trySpontaneousEqTwoWay :: InertSet -> CoVar -> CtFlavor -> TcTyVar -> TcTyVar
                       -> TcS (Maybe SWorkList)
580 581
-- Both tyvars are *touchable* MetaTyvars
-- By the CTyEqCan invariant, k2 `isSubKind` k1
582
trySpontaneousEqTwoWay inerts cv gw tv1 tv2
583
  | k1 `isSubKind` k2
584
  , nicer_to_update_tv2 = solveWithIdentity inerts cv gw tv2 (mkTyVarTy tv1)
585 586 587
  | k2 `isSubKind` k1 
  = solveWithIdentity inerts cv gw tv1 (mkTyVarTy tv2) 
  | otherwise = return Nothing 
588 589 590 591 592 593
  where
    k1 = tyVarKind tv1
    k2 = tyVarKind tv2
    nicer_to_update_tv2 = isSigTyVar tv1 || isSystemName (Var.varName tv2)
\end{code}

594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621

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:
        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@. 

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' 

622 623 624 625 626 627 628 629 630 631 632 633 634
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
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670
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. 

671 672 673 674 675 676 677

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! 
678
      given : a ~ alpha      [having unified alpha := a] 
679 680 681
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] 

682 683 684
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]
685

686 687 688
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.
689 690 691

\begin{code}
----------------
692 693 694
solveWithIdentity :: InertSet 
                  -> CoVar -> CtFlavor -> TcTyVar -> Xi 
                  -> TcS (Maybe SWorkList)
695 696
-- Solve with the identity coercion 
-- Precondition: kind(xi) is a sub-kind of kind(tv)
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
697 698 699
-- Precondition: CtFlavor is Wanted or Derived
-- See [New Wanted Superclass Work] to see why solveWithIdentity 
--     must work for Derived as well as Wanted
700
solveWithIdentity inerts cv gw tv xi 
701
  = do { tybnds <- getTcSTyBindsMap
702 703 704 705 706 707
       ; 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:" $ 
708 709 710 711
                       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)
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730
                  ]
           ; 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] 
731
           ; return $ Just (mkEqWorkList cts)  }
732

733
occurCheck :: VarEnv (TcTyVar, TcType) -> InertSet
734 735 736 737 738 739 740 741 742 743
           -> 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!

744 745
occurCheck ty_binds inerts the_tv the_ty
  = ok emptyVarSet the_ty 
746
  where 
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
747 748
    -- If (fsk `elem` bad) then tv occurs in any rendering
    -- of the type under the expansion of fsk
749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
    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 
770 771 772 773 774 775
    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)
776 777 778 779 780

    -- Check if there exists a ty bind already, as a result of sneaky unification. 
    -- Fall through
    ok _bad _ty = Nothing 

781
    -----------
782 783 784 785 786 787 788 789 790 791 792 793
    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 

794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812
    -----------
    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)
813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837
\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
838 839

        , ir_improvement  :: Maybe FDImprovement -- In case improvement kicked in
840 841 842 843 844 845 846
        }

-- What to do with the inert reactant.
data InertAction = KeepInert | DropInert
  deriving Eq

mkIRContinue :: Monad m => WorkItem -> InertAction -> WorkList -> m InteractResult
847
mkIRContinue wi keep newWork = return $ IR (ContinueWith wi) keep newWork Nothing 
848 849

mkIRStop :: Monad m => InertAction -> WorkList -> m InteractResult
850 851 852 853 854
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) 

855 856

dischargeWorkItem :: Monad m => m InteractResult
857
dischargeWorkItem = mkIRStop KeepInert emptyWorkList
858 859

noInteraction :: Monad m => WorkItem -> m InteractResult
860
noInteraction workItem = mkIRContinue workItem KeepInert emptyWorkList
861

dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
862
data WhichComesFromInert = LeftComesFromInert | RightComesFromInert 
863 864 865 866 867 868 869 870 871 872 873 874 875

---------------------------------------------------
-- Interact a single WorkItem with an InertSet as far as possible, i.e. until we get a Stop 
-- result from an individual interaction (i.e. when the WorkItem is consumed), or until we've 
-- interacted the WorkItem with the entire InertSet.
--
-- Postcondition: the new InertSet in the resulting StageResult is subset 
-- of the input InertSet.

interactWithInertsStage :: SimplifierStage
interactWithInertsStage workItem inert
  = foldlInertSetM interactNext initITR inert
  where 
876
    initITR = SR { sr_inerts   = emptyInert { inert_fds = inert_fds inert } -- Pick up the improvements!
877
                 , sr_new_work = emptyWorkList
878 879
                 , sr_stop     = ContinueWith workItem }

880

881 882 883
    interactNext :: StageResult -> AtomicInert -> TcS StageResult 
    interactNext it inert  
      | ContinueWith workItem <- sr_stop it
884 885 886 887 888 889 890 891 892 893 894 895
        = 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
896 897 898 899 900 901
                           , sr_new_work = sr_new_work it `unionWorkLists` ir_new_work ir
                           , sr_stop     = ir_stop ir } }
      | otherwise = return $ itrAddInert inert it
    
                             
    itrAddInert :: AtomicInert -> StageResult -> StageResult
902
    itrAddInert inert itr = itr { sr_inerts = (sr_inerts itr) `updInertSet` inert }
903 904

-- Do a single interaction of two constraints.
905 906
interactWithInert :: FDImprovements -> AtomicInert -> WorkItem -> TcS InteractResult
interactWithInert fdimprs inert workitem 
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923
  =  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 
              (Wanted loc, Derived _) -> isGoodRecEv work_ev  (WantedEvVar inert_ev loc)
              (Derived _, Wanted loc) -> isGoodRecEv inert_ev (WantedEvVar work_ev loc)
              _                       -> return True 

        ; if is_allowed && rec_ev_ok then 
924
              doInteractWithInert fdimprs inert workitem 
925 926 927 928 929 930 931 932 933 934 935
          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 

--------------------------------------------
936
doInteractWithInert :: FDImprovements -> CanonicalCt -> CanonicalCt -> TcS InteractResult
937 938
-- Identical class constraints.

939
doInteractWithInert fdimprs
940 941 942 943 944 945 946
           (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]
947 948 949 950
    do { let pty1 = ClassP cls1 tys1 
             pty2 = ClassP cls2 tys2 
             work_item_pred_loc = (pty2, ppr d2)
             inert_pred_loc     = (pty1, ppr d1)
951 952
	     loc                = combineCtLoc fl1 fl2
             eqn_pred_locs = improveFromAnother work_item_pred_loc inert_pred_loc         
953

954
       ; wevvars <- mkWantedFunDepEqns loc eqn_pred_locs 
955 956
       ; fd_cts <- canWanteds wevvars 
       ; let fd_work = mkEqWorkList fd_cts 
957
              	 -- See Note [Generating extra equalities]
958 959 960 961
       ; traceTcS "Checking if improvements existed." (ppr fdimprs) 
--       ; if isEmptyCCan fd_cts || not (isWanted fl2) || haveBeenImproved fdimprs pty1 pty2 then
       ; if isEmptyCCan fd_cts || haveBeenImproved fdimprs pty1 pty2 then
             -- Must keep going
962
             mkIRContinue workItem KeepInert fd_work 
963 964 965 966
         else do { traceTcS "Recording improvement and throwing item back in worklist." (ppr (pty1,pty2))
                 ; mkIRStop_RecordImprovement KeepInert 
                      (fd_work `unionWorkLists` singleNonEqWL workItem) (pty1,pty2)
                 }
967 968 969 970 971
         -- See Note [FunDep Reactions] 
       }

-- Class constraint and given equality: use the equality to rewrite
-- the class constraint. 
972 973
doInteractWithInert _fdimprs
                    (CTyEqCan { cc_id = cv, cc_flavor = ifl, cc_tyvar = tv, cc_rhs = xi }) 
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990
                    (CDictCan { cc_id = dv, cc_flavor = wfl, cc_class = cl, cc_tyargs = xis }) 
  | ifl `canRewrite` wfl 
  , tv `elemVarSet` tyVarsOfTypes xis
    -- substitute for tv in xis.  Note that the resulting class
    -- constraint is still canonical, since substituting xi-types in
    -- xi-types generates xi-types.  However, it may no longer be
    -- inert with respect to the inert set items we've already seen.
    -- For example, consider the inert set
    --
    --   D Int (g)
    --   a ~g Int
    --
    -- and the work item D a (w). D a does not interact with D Int.
    -- Next, it does interact with a ~g Int, getting rewritten to D
    -- Int (w).  But now we must go back through the rest of the inert
    -- set again, to find that it can now be discharged by the given D
    -- Int instance.
991 992 993 994 995

    -- DV: Update to the comment above:  
    --     This situation can no longer happen, see Note [Prioritizing equalities]
    --     so we are good to simply keep going with the rewritten dictionary to rewrite 
    --     it as much as possible before reaching any other dictionary constraints! 
996
  = do { rewritten_dict <- rewriteDict (cv,tv,xi) (dv,wfl,cl,xis)
997 998
       ; mkIRContinue rewritten_dict KeepInert emptyWorkList } 
--        ; mkIRStop KeepInert $ singleNonEqWL rewritten_dict }
999
    
1000 1001
doInteractWithInert _fdimprs 
                    (CDictCan { cc_id = dv, cc_flavor = ifl, cc_class = cl, cc_tyargs = xis }) 
1002 1003 1004 1005
           workItem@(CTyEqCan { cc_id = cv, cc_flavor = wfl, cc_tyvar = tv, cc_rhs = xi })
  | wfl `canRewrite` ifl
  , tv `elemVarSet` tyVarsOfTypes xis
  = do { rewritten_dict <- rewriteDict (cv,tv,xi) (dv,ifl,cl,xis) 
1006
       ; mkIRContinue workItem DropInert (singleNonEqWL rewritten_dict) }
1007 1008 1009

-- Class constraint and given equality: use the equality to rewrite
-- the class constraint.
1010 1011
doInteractWithInert _fdimprs 
                    (CTyEqCan { cc_id = cv, cc_flavor = ifl, cc_tyvar = tv, cc_rhs = xi }) 
1012 1013 1014 1015
                    (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) 
1016 1017
       ; mkIRContinue rewritten_ip KeepInert emptyWorkList } 
--        ; mkIRStop KeepInert (singleNonEqWL rewritten_ip) }
1018

1019 1020
doInteractWithInert _fdimprs 
                    (CIPCan { cc_id = ipid, cc_flavor = ifl, cc_ip_nm = nm, cc_ip_ty = ty }) 
1021 1022 1023 1024
           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) 
1025
       ; mkIRContinue workItem DropInert (singleNonEqWL rewritten_ip) }
1026 1027 1028 1029 1030 1031

-- 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.
1032 1033
doInteractWithInert _fdimprs 
                    (CIPCan { cc_id = id1, cc_flavor = ifl, cc_ip_nm = nm1, cc_ip_ty = ty1 }) 
1034
           workItem@(CIPCan { cc_flavor = wfl, cc_ip_nm = nm2, cc_ip_ty = ty2 })
1035 1036 1037 1038
  | 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
1039
    mkIRContinue workItem DropInert emptyWorkList
1040

1041 1042 1043
  | nm1 == nm2 && ty1 `tcEqType` ty2 
  = solveOneFromTheOther (id1,ifl) workItem 

1044
  | nm1 == nm2
1045 1046 1047
  =  	-- See Note [When improvement happens]
    do { co_var <- newWantedCoVar ty1 ty2 
       ; let flav = Wanted (combineCtLoc ifl wfl) 
1048 1049
       ; cans <- mkCanonical flav co_var 
       ; mkIRContinue workItem KeepInert (mkEqWorkList cans) }
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060


-- 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.
1061 1062
doInteractWithInert _fdimprs
                    (CTyEqCan { cc_id = cv1, cc_flavor = ifl, cc_tyvar = tv, cc_rhs = xi1 }) 
1063 1064 1065 1066 1067
                    (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) 
1068
       ; mkIRStop KeepInert (singleEqWL rewritten_funeq) } -- DV: must stop here, because we may no longer be inert after the rewritting.
1069 1070 1071

-- Inert: function equality, work item: equality

1072 1073
doInteractWithInert _fdimprs
                    (CFunEqCan {cc_id = cv1, cc_flavor = ifl, cc_fun = tc
1074 1075 1076 1077 1078
                              , 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) 
1079
       ; mkIRContinue workItem DropInert (singleEqWL rewritten_funeq) } 
1080

1081 1082
doInteractWithInert _fdimprs
                    (CFunEqCan { cc_id = cv1, cc_flavor = fl1, cc_fun = tc1
1083 1084 1085
                               , cc_tyargs = args1, cc_rhs = xi1 }) 
           workItem@(CFunEqCan { cc_id = cv2, cc_flavor = fl2, cc_fun = tc2
                               , cc_tyargs = args2, cc_rhs = xi2 })
1086
  | fl1 `canSolve` fl2 && lhss_match
dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
1087
  = do { cans <- rewriteEqLHS LeftComesFromInert  (mkCoVarCoercion cv1,xi1) (cv2,fl2,xi2) 
1088
       ; mkIRStop KeepInert (mkEqWorkList cans) } 
1089
  | fl2 `canSolve` fl1 && lhss_match
dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
1090
  = do { cans <- rewriteEqLHS RightComesFromInert (mkCoVarCoercion cv2,xi2) (cv1,fl1,xi1) 
1091
       ; mkIRContinue workItem DropInert (mkEqWorkList cans) }
1092 1093 1094
  where
    lhss_match = tc1 == tc2 && and (zipWith tcEqType args1 args2) 

1095
doInteractWithInert _fdimprs 
1096
           inert@(CTyEqCan { cc_id = cv1, cc_flavor = fl1, cc_tyvar = tv1, cc_rhs = xi1 }) 
1097 1098
           workItem@(CTyEqCan { cc_id = cv2, cc_flavor = fl2, cc_tyvar = tv2, cc_rhs = xi2 })
-- Check for matching LHS 
1099
  | fl1 `canSolve` fl2 && tv1 == tv2 
dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
1100
  = do { cans <- rewriteEqLHS LeftComesFromInert (mkCoVarCoercion cv1,xi1) (cv2,fl2,xi2) 
1101
       ; mkIRStop KeepInert (mkEqWorkList cans) } 
1102

1103
  | fl2 `canSolve` fl1 && tv1 == tv2 
dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
1104
  = do { cans <- rewriteEqLHS RightComesFromInert (mkCoVarCoercion cv2,xi2) (cv1,fl1,xi1) 
1105
       ; mkIRContinue workItem DropInert (mkEqWorkList cans) } 
1106 1107 1108 1109

-- Check for rewriting RHS 
  | fl1 `canRewrite` fl2 && tv1 `elemVarSet` tyVarsOfType xi2 
  = do { rewritten_eq <- rewriteEqRHS (cv1,tv1,xi1) (cv2,fl2,tv2,xi2) 
1110
       ; mkIRStop KeepInert (mkEqWorkList rewritten_eq) }
1111 1112
  | fl2 `canRewrite` fl1 && tv2 `elemVarSet` tyVarsOfType xi1
  = do { rewritten_eq <- rewriteEqRHS (cv2,tv2,xi2) (cv1,fl1,tv1,xi1) 
1113
       ; mkIRContinue workItem DropInert (mkEqWorkList rewritten_eq) } 
1114 1115 1116 1117 1118 1119

-- 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] 
1120 1121 1122 1123 1124

  | 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' 
1125
  = mkIRContinue workItem DropInert (singleEqWL inert)   
1126 1127


1128
-- Fall-through case for all other situations
1129
doInteractWithInert _fdimprs _ workItem = noInteraction workItem
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 1200 1201 1202 1203 1204 1205 1206 1207
-- 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 
1208
                                       , cc_rhs   = xi2'' })
1209 1210 1211 1212 1213
       }
  where 
    xi2' = substTyWith [tv1] [xi1] xi2 
    co2' = substTyWith [tv1] [mkCoVarCoercion cv1] xi2  -- xi2 ~ xi2[xi1/tv1]

dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
1214 1215

rewriteEqLHS :: WhichComesFromInert -> (Coercion,Xi) -> (CoVar,CtFlavor,Xi) -> TcS CanonicalCts
1216 1217 1218
-- Used to ineratct two equalities of the following form: 
-- First Equality:   co1: (XXX ~ xi1)  
-- Second Equality:  cv2: (XXX ~ xi2) 
1219
-- Where the cv1 `canSolve` cv2 equality 
dimitris@microsoft.com's avatar
dimitris@microsoft.com committed
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242
-- 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
1243 1244 1245
       ; mkCanonical gw cv2'
       }
                                           
1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
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 

1261
  | ifl `canSolve` wfl
1262 1263 1264 1265 1266
  = do { unless (isGiven wfl) $ setEvBind wid (EvId iid) 
           -- Overwrite the binding, if one exists
	   -- For Givens, which are lambda-bound, nothing to overwrite,
       ; dischargeWorkItem }

1267
  | otherwise  -- wfl `canSolve` ifl 
1268
  = do { unless (isGiven ifl) $ setEvBind iid (EvId wid)
1269
       ; mkIRContinue workItem DropInert emptyWorkList }
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 1685 1686 1687 1688 1689 1690 1691 1692

  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
     opc = opd
  
  instance D Int where
     opd = opc

From (instance C Int) we get the constraint set {ds1:S Int, dd:D Int}
Simplifying, we may well get:
	$dfCInt = :C ds1 (opd dd)
	dd  = $dfDInt
	ds1 = $p1 dd
Notice that we spot that we can extract ds1 from dd.  

Alas!  Alack! We can do the same for (instance D Int):

	$dfDInt = :D ds2 (opc dc)
	dc  = $dfCInt
	ds2 = $p1 dc

And now we've defined the superclass in terms of itself.
Two more nasty cases are in
	tcrun021
	tcrun033

Solution: 
  - Satisfy the superclass context *all by itself* 
    (tcSimplifySuperClasses)
  - And do so completely; i.e. no left-over constraints
    to mix with the constraints arising from method declarations


Note [SUPERCLASS-LOOP 2]
~~~~~~~~~~~~~~~~~~~~~~~~
We need to be careful when adding "the constaint we are trying to prove".
Suppose we are *given* d1:Ord a, and want to deduce (d2:C [a]) where

	class Ord a => C a where
	instance Ord [a] => C [a] where ...

Then we'll use the instance decl to deduce C [a] from Ord [a], and then add the
superclasses of C [a] to avails.  But we must not overwrite the binding
for Ord [a] (which is obtained from Ord a) with a superclass selection or we'll just
build a loop! 

Here's another variant, immortalised in tcrun020
	class Monad m => C1 m
	class C1 m => C2 m x
	instance C2 Maybe Bool
For the instance decl we need to build (C1 Maybe), and it's no good if
we run around and add (C2 Maybe Bool) and its superclasses to the avails 
before we search for C1 Maybe.

Here's another example 
 	class Eq b => Foo a b
	instance Eq a => Foo [a] a
If we are reducing
	(Foo [t] t)

we'll first deduce that it holds (via the instance decl).  We must not
then overwrite the Eq t constraint with a superclass selection!

At first I had a gross hack, whereby I simply did not add superclass constraints
in addWanted, though I did for addGiven and addIrred.  This was sub-optimal,
becuase it lost legitimate superclass sharing, and it still didn't do the job:
I found a very obscure program (now tcrun021) in which improvement meant the
simplifier got two bites a the cherry... so something seemed to be an Stop
first time, but reducible next time.

Now we implement the Right Solution, which is to check for loops directly 
when adding superclasses.  It's a bit like the occurs check in unification.

Note [Recursive instances and superclases]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Consider this code, which arises in the context of "Scrap Your 
Boilerplate with Class".  

    class Sat a
    class Data ctx a
    instance  Sat (ctx Char)             => Data ctx Char
    instance (Sat (ctx [a]), Data ctx a) => Data ctx [a]

    class Data Maybe a => Foo a

    instance Foo t => Sat (Maybe t)

    instance Data Maybe a => Foo a
    instance Foo a        => Foo [a]
    instance                 Foo [Char]

In the instance for Foo [a], when generating evidence for the superclasses
(ie in tcSimplifySuperClasses) we need a superclass (Data Maybe [a]).
Using the instance for Data, we therefore need
        (Sat (Maybe [a], Data Maybe a)
But we are given (Foo a), and hence its superclass (Data Maybe a).
So that leaves (Sat (Maybe [a])).  Using the instance for Sat means
we need (Foo [a]).  And that is the very dictionary we are bulding
an instance for!  So we must put that in the "givens".  So in this
case we have
	Given:  Foo a, Foo [a]
	Wanted: Data Maybe [a]

BUT we must *not not not* put the *superclasses* of (Foo [a]) in
the givens, which is what 'addGiven' would normally do. Why? Because
(Data Maybe [a]) is the superclass, so we'd "satisfy" the wanted 
by selecting a superclass from Foo [a], which simply makes a loop.

On the other hand we *must* put the superclasses of (Foo a) in
the givens, as you can see from the derivation described above.

Conclusion: in the very special case of tcSimplifySuperClasses
we have one 'given' (namely the "this" dictionary) whose superclasses
must not be added to 'givens' by addGiven.  

There is a complication though.  Suppose there are equalities
      instance (Eq a, a~b) => Num (a,b)
Then we normalise the 'givens' wrt the equalities, so the original
given "this" dictionary is cast to one of a different type.  So it's a
bit trickier than before to identify the "special" dictionary whose
superclasses must not be added. See test
   indexed-types/should_run/EqInInstance

We need a persistent property of the dictionary to record this
special-ness.  Current I'm using the InstLocOrigin (a bit of a hack,
but cool), which is maintained by dictionary normalisation.
Specifically, the InstLocOrigin is
	     NoScOrigin
then the no-superclass thing kicks in.  WATCH OUT if you fiddle
with InstLocOrigin!

Note [MATCHING-SYNONYMS]
~~~~~~~~~~~~~~~~~~~~~~~~
When trying to match a dictionary (D tau) to a top-level instance, or a 
type family equation (F taus_1 ~ tau_2) to a top-level family instance, 
we do *not* need to expand type synonyms because the matcher will do that for us.


Note [RHS-FAMILY-SYNONYMS] 
~~~~~~~~~~~~~~~~~~~~~~~~~~
The RHS of a family instance is represented as yet another constructor which is 
like a type synonym for the real RHS the programmer declared. Eg: 
    type instance F (a,a) = [a] 
Becomes: 
    :R32 a = [a]      -- internal type synonym introduced
    F (a,a) ~ :R32 a  -- instance 

When we react a family instance with a type family equation in the work list 
we keep the synonym-using RHS without expansion. 


*********************************************************************************
*                                                                               * 
                       The top-reaction Stage
*                                                                               *
*********************************************************************************

\begin{code}
-- If a work item has any form of interaction with top-level we get this 
data TopInteractResult 
  = NoTopInt 	           -- No top-level interaction
  | SomeTopInt 
      { tir_new_work  :: WorkList	-- Sub-goals or new work (could be given, 
                                        --                        for superclasses)
      , tir_new_inert :: StopOrContinue -- The input work item, ready to become *inert* now: 
      }                       		-- NB: in ``given'' (solved) form if the 
                              		-- original was wanted or given and instance match
                              		-- was found, but may also be in wanted form if we 
                                        -- only reacted with functional dependencies 
					-- arising from top-level instances.

topReactionsStage :: SimplifierStage 
topReactionsStage workItem inerts 
  = do { tir <- tryTopReact workItem 
       ; case tir of 
           NoTopInt -> 
               return $ SR { sr_inerts   = inerts 
                           , sr_new_work = emptyWorkList 
                           , sr_stop     = ContinueWith workItem } 
           SomeTopInt tir_new_work tir_new_inert -> 
               return $ SR { sr_inerts   = inerts 
                           , sr_new_work = tir_new_work
                           , sr_stop     = tir_new_inert
                           }
       }

tryTopReact :: WorkItem -> TcS TopInteractResult 
tryTopReact workitem 
  = do {  -- A flag controls the amount of interaction allowed
          -- See Note [Simplifying RULE lhs constraints]
         ctxt <- getTcSContext
       ; if allowedTopReaction (simplEqsOnly ctxt) workitem 
         then do { traceTcS "tryTopReact / calling doTopReact" (ppr workitem)
                 ; doTopReact workitem }
         else return NoTopInt 
       } 

allowedTopReaction :: Bool -> WorkItem -> Bool 
allowedTopReaction eqs_only (CDictCan {}) = not eqs_only
allowedTopReaction _        _             = True 


doTopReact :: WorkItem -> TcS TopInteractResult 
-- The work item does not react with the inert set, 
-- so try interaction with top-level instances
doTopReact workItem@(CDictCan { cc_id = dv, cc_flavor = Wanted loc
                              , cc_class = cls, cc_tyargs = xis }) 
  = do { -- See Note [MATCHING-SYNONYMS]
       ; lkp_inst_res <- matchClassInst cls xis loc
       ; case lkp_inst_res of 
           NoInstance -> do { traceTcS "doTopReact/ no class instance for" (ppr dv) 
                            ; funDepReact }
           GenInst wtvs ev_term ->  -- Solved 
	   	   -- No need to do fundeps stuff here; the instance 
		   -- matches already so we won't get any more info
		   -- from functional dependencies
               do { traceTcS "doTopReact/ found class instance for" (ppr dv) 
                  ; setDictBind dv ev_term 
                  ; workList <- canWanteds wtvs
                  ; if null wtvs
                    -- Solved in one step and no new wanted work produced. 
                    -- i.e we directly matched a top-level instance
		    -- No point in caching this in 'inert', nor in adding superclasses
1693
                    then return $ SomeTopInt { tir_new_work  = emptyWorkList 
1694 1695 1696 1697 1698 1699
                                             , tir_new_inert = Stop }

                    -- Solved and new wanted work produced, you may cache the 
		    -- (tentatively solved) dictionary as Derived and its superclasses
                    else do { let solved = makeSolved workItem
                            ; sc_work <- newSCWorkFromFlavored dv (Derived loc) cls xis