OccName.lhs 31.6 KB
Newer Older
1
%
Simon Marlow's avatar
Simon Marlow committed
2
% (c) The University of Glasgow 2006
3 4 5 6
% (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
%

\begin{code}
7 8
{-# LANGUAGE DeriveDataTypeable #-}

9 10 11 12 13
-- |
-- #name_types#
-- GHC uses several kinds of name internally:
--
-- * 'OccName.OccName' represents names as strings with just a little more information:
Thomas Schilling's avatar
Thomas Schilling committed
14
--   the \"namespace\" that the name came from, e.g. the namespace of value, type constructors or
15 16 17 18 19 20 21 22 23
--   data constructors
--
-- * 'RdrName.RdrName': see "RdrName#name_types"
--
-- * 'Name.Name': see "Name#name_types"
--
-- * 'Id.Id': see "Id#name_types"
--
-- * 'Var.Var': see "Var#name_types"
Ian Lynagh's avatar
Ian Lynagh committed
24

25
{-# OPTIONS_GHC -fno-warn-tabs #-}
Ian Lynagh's avatar
Ian Lynagh committed
26 27 28
-- The above warning supression flag is a temporary kludge.
-- While working on this module you are encouraged to remove it and
-- detab the module (please do the detabbing in a separate patch). See
29
--     http://ghc.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#TabsvsSpaces
Ian Lynagh's avatar
Ian Lynagh committed
30 31
-- for details

32
module OccName (
33 34
	-- * The 'NameSpace' type
	NameSpace, -- Abstract
35 36

        nameSpacesRelated,
37 38 39 40
	
	-- ** Construction
	-- $real_vs_source_data_constructors
	tcName, clsName, tcClsName, dataName, varName, 
41
	tvName, srcDataName,
42

43
	-- ** Pretty Printing
44 45
	pprNameSpace, pprNonVarNameSpace, pprNameSpaceBrief,

46
	-- * The 'OccName' type
47
	OccName, 	-- Abstract, instance of Outputable
48
	pprOccName, 
49

50 51 52
	-- ** Construction	
	mkOccName, mkOccNameFS, 
	mkVarOcc, mkVarOccFS,
53 54 55 56
	mkDataOcc, mkDataOccFS,
	mkTyVarOcc, mkTyVarOccFS,
	mkTcOcc, mkTcOccFS,
	mkClsOcc, mkClsOccFS,
57
        mkDFunOcc,
58
	setOccNameSpace,
dreixel's avatar
dreixel committed
59
        demoteOccName,
60
        HasOccName(..),
61

62
	-- ** Derived 'OccName's
63
        isDerivedOccName,
Gergő Érdi's avatar
Gergő Érdi committed
64
	mkDataConWrapperOcc, mkWorkerOcc, mkMatcherOcc, mkDefaultMethodOcc,
65
        mkGenDefMethodOcc, 
66
	mkDerivedTyConOcc, mkNewTyCoOcc, mkClassOpAuxOcc,
67
        mkCon2TagOcc, mkTag2ConOcc, mkMaxTagOcc,
68 69 70
	mkClassDataConOcc, mkDictOcc, mkIPOcc,
	mkSpecOcc, mkForeignExportOcc, mkRepEqOcc, mkGenOcc1, mkGenOcc2,
	mkGenD, mkGenR, mkGen1R, mkGenRCo, mkGenC, mkGenS,
71
        mkDataTOcc, mkDataCOcc, mkDataConWorkerOcc,
72
	mkSuperDictSelOcc, mkLocalOcc, mkMethodOcc, mkInstTyTcOcc,
73
	mkInstTyCoOcc, mkEqPredCoOcc,
74
        mkVectOcc, mkVectTyConOcc, mkVectDataConOcc, mkVectIsoOcc,
75 76
        mkPDataTyConOcc,  mkPDataDataConOcc,
	mkPDatasTyConOcc, mkPDatasDataConOcc,
77
        mkPReprTyConOcc, 
78
        mkPADFunOcc,
79 80 81 82 83

	-- ** Deconstruction
	occNameFS, occNameString, occNameSpace, 

	isVarOcc, isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc,
84
	parenSymOcc, startsWithUnderscore, 
85 86
	
	isTcClsNameSpace, isTvNameSpace, isDataConNameSpace, isVarNameSpace, isValNameSpace,
87

88
	-- * The 'OccEnv' type
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
89
	OccEnv, emptyOccEnv, unitOccEnv, extendOccEnv, mapOccEnv,
90
	lookupOccEnv, mkOccEnv, mkOccEnv_C, extendOccEnvList, elemOccEnv,
91
	occEnvElts, foldOccEnv, plusOccEnv, plusOccEnv_C, extendOccEnv_C,
92
        extendOccEnv_Acc, filterOccEnv, delListFromOccEnv, delFromOccEnv,
Simon Peyton Jones's avatar
Simon Peyton Jones committed
93
        alterOccEnv, 
94

95
	-- * The 'OccSet' type
96 97
	OccSet, emptyOccSet, unitOccSet, mkOccSet, extendOccSet, 
	extendOccSetList,
98 99
	unionOccSets, unionManyOccSets, minusOccSet, elemOccSet, occSetElts, 
	foldOccSet, isEmptyOccSet, intersectOccSet, intersectsOccSet,
100
                  
101
	-- * Tidying up
102 103
	TidyOccEnv, emptyTidyOccEnv, tidyOccName, initTidyOccEnv,

104
	-- * Lexical characteristics of Haskell names
105 106
	isLexCon, isLexVar, isLexId, isLexSym,
	isLexConId, isLexConSym, isLexVarId, isLexVarSym,
107 108 109 110
	startsVarSym, startsVarId, startsConSym, startsConId,

        -- FsEnv
        FastStringEnv, emptyFsEnv, lookupFsEnv, extendFsEnv, mkFsEnv
111 112
    ) where

Simon Marlow's avatar
Simon Marlow committed
113 114
import Util
import Unique
115
import DynFlags
116 117
import UniqFM
import UniqSet
118
import FastString
119
import Outputable
120
import Binary
Simon Marlow's avatar
Simon Marlow committed
121
import Data.Char
122
import Data.Data
123
\end{code}
124

125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
%************************************************************************
%*									*
              FastStringEnv
%*									*
%************************************************************************

FastStringEnv can't be in FastString becuase the env depends on UniqFM

\begin{code}
type FastStringEnv a = UniqFM a         -- Keyed by FastString


emptyFsEnv  :: FastStringEnv a
lookupFsEnv :: FastStringEnv a -> FastString -> Maybe a
extendFsEnv :: FastStringEnv a -> FastString -> a -> FastStringEnv a
mkFsEnv     :: [(FastString,a)] -> FastStringEnv a

emptyFsEnv  = emptyUFM
lookupFsEnv = lookupUFM
extendFsEnv = addToUFM
mkFsEnv     = listToUFM
\end{code}

148 149 150 151 152 153 154
%************************************************************************
%*									*
\subsection{Name space}
%*									*
%************************************************************************

\begin{code}
155 156
data NameSpace = VarName	-- Variables, including "real" data constructors
	       | DataName	-- "Source" data constructors 
157 158
	       | TvName		-- Type variables
	       | TcClsName	-- Type constructors and classes; Haskell has them
159
				-- in the same name space for now.
160
	       deriving( Eq, Ord )
161
   {-! derive: Binary !-}
162

163 164 165
-- Note [Data Constructors]  
-- see also: Note [Data Constructor Naming] in DataCon.lhs
--
166 167 168 169 170 171 172 173 174 175 176 177 178
-- $real_vs_source_data_constructors
-- There are two forms of data constructor:
--
--	[Source data constructors] The data constructors mentioned in Haskell source code
--
--	[Real data constructors] The data constructors of the representation type, which may not be the same as the source type
--
-- For example:
--
-- > data T = T !(Int, Int)
--
-- The source datacon has type @(Int, Int) -> T@
-- The real   datacon has type @Int -> Int -> T@
179 180 181
--
-- GHC chooses a representation based on the strictness etc.

twanvl's avatar
twanvl committed
182 183 184
tcName, clsName, tcClsName :: NameSpace
dataName, srcDataName      :: NameSpace
tvName, varName            :: NameSpace
185

186 187 188 189 190 191
-- Though type constructors and classes are in the same name space now,
-- the NameSpace type is abstract, so we can easily separate them later
tcName    = TcClsName		-- Type constructors
clsName   = TcClsName		-- Classes
tcClsName = TcClsName		-- Not sure which!

192 193 194
dataName    = DataName
srcDataName = DataName	-- Haskell-source data constructors should be
			-- in the Data name space
195

196 197
tvName      = TvName
varName     = VarName
198

199 200 201 202 203 204 205 206 207 208 209
isDataConNameSpace :: NameSpace -> Bool
isDataConNameSpace DataName = True
isDataConNameSpace _        = False

isTcClsNameSpace :: NameSpace -> Bool
isTcClsNameSpace TcClsName = True
isTcClsNameSpace _         = False

isTvNameSpace :: NameSpace -> Bool
isTvNameSpace TvName = True
isTvNameSpace _      = False
210

211 212 213 214 215 216 217 218 219
isVarNameSpace :: NameSpace -> Bool	-- Variables or type variables, but not constructors
isVarNameSpace TvName  = True
isVarNameSpace VarName = True
isVarNameSpace _       = False

isValNameSpace :: NameSpace -> Bool
isValNameSpace DataName = True
isValNameSpace VarName  = True
isValNameSpace _        = False
220

221
pprNameSpace :: NameSpace -> SDoc
Ian Lynagh's avatar
Ian Lynagh committed
222 223 224 225
pprNameSpace DataName  = ptext (sLit "data constructor")
pprNameSpace VarName   = ptext (sLit "variable")
pprNameSpace TvName    = ptext (sLit "type variable")
pprNameSpace TcClsName = ptext (sLit "type constructor or class")
226 227 228 229

pprNonVarNameSpace :: NameSpace -> SDoc
pprNonVarNameSpace VarName = empty
pprNonVarNameSpace ns = pprNameSpace ns
230

twanvl's avatar
twanvl committed
231
pprNameSpaceBrief :: NameSpace -> SDoc
232 233
pprNameSpaceBrief DataName  = char 'd'
pprNameSpaceBrief VarName   = char 'v'
Ian Lynagh's avatar
Ian Lynagh committed
234 235
pprNameSpaceBrief TvName    = ptext (sLit "tv")
pprNameSpaceBrief TcClsName = ptext (sLit "tc")
dreixel's avatar
dreixel committed
236 237 238

-- demoteNameSpace lowers the NameSpace if possible.  We can not know
-- in advance, since a TvName can appear in an HsTyVar.
239
-- See Note [Demotion] in RnEnv
dreixel's avatar
dreixel committed
240 241 242 243 244
demoteNameSpace :: NameSpace -> Maybe NameSpace
demoteNameSpace VarName = Nothing
demoteNameSpace DataName = Nothing
demoteNameSpace TvName = Nothing
demoteNameSpace TcClsName = Just DataName
245 246 247 248 249 250 251 252 253 254
\end{code}


%************************************************************************
%*									*
\subsection[Name-pieces-datatypes]{The @OccName@ datatypes}
%*									*
%************************************************************************

\begin{code}
255
data OccName = OccName 
256
    { occNameSpace  :: !NameSpace
257
    , occNameFS     :: !FastString
258
    }
Ian Lynagh's avatar
Ian Lynagh committed
259
    deriving Typeable
260 261 262 263 264 265 266 267
\end{code}


\begin{code}
instance Eq OccName where
    (OccName sp1 s1) == (OccName sp2 s2) = s1 == s2 && sp1 == sp2

instance Ord OccName where
268 269 270
	-- Compares lexicographically, *not* by Unique of the string
    compare (OccName sp1 s1) (OccName sp2 s2) 
	= (s1  `compare` s2) `thenCmp` (sp1 `compare` sp2)
271 272 273 274 275 276

instance Data OccName where
  -- don't traverse?
  toConstr _   = abstractConstr "OccName"
  gunfold _ _  = error "gunfold"
  dataTypeOf _ = mkNoRepType "OccName"
277 278 279

instance HasOccName OccName where
  occName = id
280 281 282 283 284 285 286 287 288 289 290
\end{code}


%************************************************************************
%*									*
\subsection{Printing}
%*									*
%************************************************************************
 
\begin{code}
instance Outputable OccName where
291
    ppr = pprOccName
292

293 294 295 296 297
instance OutputableBndr OccName where
    pprBndr _ = ppr
    pprInfixOcc n = pprInfixVar (isSymOcc n) (ppr n)
    pprPrefixOcc n = pprPrefixVar (isSymOcc n) (ppr n)

298
pprOccName :: OccName -> SDoc
299 300
pprOccName (OccName sp occ) 
  = getPprStyle $ \ sty ->
301
    if codeStyle sty 
Ian Lynagh's avatar
Ian Lynagh committed
302
    then ztext (zEncodeFS occ)
303 304 305 306 307
    else pp_occ <> pp_debug sty
  where
    pp_debug sty | debugStyle sty = braces (pprNameSpaceBrief sp)
	         | otherwise      = empty

308
    pp_occ = sdocWithDynFlags $ \dflags ->
ian@well-typed.com's avatar
ian@well-typed.com committed
309
             if gopt Opt_SuppressUniques dflags
310 311
             then text (strip_th_unique (unpackFS occ))
             else ftext occ
312 313 314 315 316

	-- See Note [Suppressing uniques in OccNames]
    strip_th_unique ('[' : c : _) | isAlphaNum c = []
    strip_th_unique (c : cs) = c : strip_th_unique cs
    strip_th_unique []       = []
317 318
\end{code}

319 320 321 322 323
Note [Suppressing uniques in OccNames]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This is a hack to de-wobblify the OccNames that contain uniques from
Template Haskell that have been turned into a string in the OccName.
See Note [Unique OccNames from Template Haskell] in Convert.hs
324 325 326 327 328

%************************************************************************
%*									*
\subsection{Construction}
%*									*
329
%************************************************************************
330 331

\begin{code}
332
mkOccName :: NameSpace -> String -> OccName
333
mkOccName occ_sp str = OccName occ_sp (mkFastString str)
334

335 336
mkOccNameFS :: NameSpace -> FastString -> OccName
mkOccNameFS occ_sp fs = OccName occ_sp fs
337

338 339
mkVarOcc :: String -> OccName
mkVarOcc s = mkOccName varName s
340

341 342
mkVarOccFS :: FastString -> OccName
mkVarOccFS fs = mkOccNameFS varName fs
343

344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
mkDataOcc :: String -> OccName
mkDataOcc = mkOccName dataName

mkDataOccFS :: FastString -> OccName
mkDataOccFS = mkOccNameFS dataName

mkTyVarOcc :: String -> OccName
mkTyVarOcc = mkOccName tvName

mkTyVarOccFS :: FastString -> OccName
mkTyVarOccFS fs = mkOccNameFS tvName fs

mkTcOcc :: String -> OccName
mkTcOcc = mkOccName tcName

mkTcOccFS :: FastString -> OccName
mkTcOccFS = mkOccNameFS tcName

mkClsOcc :: String -> OccName
mkClsOcc = mkOccName clsName

mkClsOccFS :: FastString -> OccName
mkClsOccFS = mkOccNameFS clsName
dreixel's avatar
dreixel committed
367 368 369 370 371 372 373

-- demoteOccName lowers the Namespace of OccName.
-- see Note [Demotion]
demoteOccName :: OccName -> Maybe OccName
demoteOccName (OccName space name) = do
  space' <- demoteNameSpace space
  return $ OccName space' name
374

375 376 377 378
-- Name spaces are related if there is a chance to mean the one when one writes
-- the other, i.e. variables <-> data construtors and type variables <-> type constructors
nameSpacesRelated :: NameSpace -> NameSpace -> Bool
nameSpacesRelated ns1 ns2 = ns1 == ns2 || otherNameSpace ns1 == ns2
379 380 381 382 383 384 385 386

otherNameSpace :: NameSpace -> NameSpace
otherNameSpace VarName = DataName
otherNameSpace DataName = VarName
otherNameSpace TvName = TcClsName
otherNameSpace TcClsName = TvName


387

388 389 390 391
{- | Other names in the compiler add aditional information to an OccName.
This class provides a consistent way to access the underlying OccName. -}
class HasOccName name where
  occName :: name -> OccName
392
\end{code}
393

394

395 396 397 398 399 400 401 402
%************************************************************************
%*									*
		Environments
%*									*
%************************************************************************

OccEnvs are used mainly for the envts in ModIfaces.

403 404
Note [The Unique of an OccName]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
405
They are efficient, because FastStrings have unique Int# keys.  We assume
406 407 408 409
this key is less than 2^24, and indeed FastStrings are allocated keys 
sequentially starting at 0.

So we can make a Unique using
410
	mkUnique ns key  :: Unique
411
where 'ns' is a Char representing the name space.  This in turn makes it
412 413 414 415
easy to build an OccEnv.

\begin{code}
instance Uniquable OccName where
416 417 418 419 420
      -- See Note [The Unique of an OccName]
  getUnique (OccName VarName   fs) = mkVarOccUnique  fs
  getUnique (OccName DataName  fs) = mkDataOccUnique fs
  getUnique (OccName TvName    fs) = mkTvOccUnique   fs
  getUnique (OccName TcClsName fs) = mkTcOccUnique   fs
421

422
newtype OccEnv a = A (UniqFM a)
423 424 425 426 427 428 429

emptyOccEnv :: OccEnv a
unitOccEnv  :: OccName -> a -> OccEnv a
extendOccEnv :: OccEnv a -> OccName -> a -> OccEnv a
extendOccEnvList :: OccEnv a -> [(OccName, a)] -> OccEnv a
lookupOccEnv :: OccEnv a -> OccName -> Maybe a
mkOccEnv     :: [(OccName,a)] -> OccEnv a
430
mkOccEnv_C   :: (a -> a -> a) -> [(OccName,a)] -> OccEnv a
431 432 433 434
elemOccEnv   :: OccName -> OccEnv a -> Bool
foldOccEnv   :: (a -> b -> b) -> b -> OccEnv a -> b
occEnvElts   :: OccEnv a -> [a]
extendOccEnv_C :: (a->a->a) -> OccEnv a -> OccName -> a -> OccEnv a
435
extendOccEnv_Acc :: (a->b->b) -> (a->b) -> OccEnv b -> OccName -> a -> OccEnv b
436
plusOccEnv     :: OccEnv a -> OccEnv a -> OccEnv a
437
plusOccEnv_C   :: (a->a->a) -> OccEnv a -> OccEnv a -> OccEnv a
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
438
mapOccEnv      :: (a->b) -> OccEnv a -> OccEnv b
439 440 441
delFromOccEnv 	   :: OccEnv a -> OccName -> OccEnv a
delListFromOccEnv :: OccEnv a -> [OccName] -> OccEnv a
filterOccEnv	   :: (elt -> Bool) -> OccEnv elt -> OccEnv elt
Simon Peyton Jones's avatar
Simon Peyton Jones committed
442
alterOccEnv	   :: (Maybe elt -> Maybe elt) -> OccEnv elt -> OccName -> OccEnv elt
443 444 445 446 447 448 449 450 451 452 453 454 455

emptyOccEnv  	 = A emptyUFM
unitOccEnv x y = A $ unitUFM x y 
extendOccEnv (A x) y z = A $ addToUFM x y z
extendOccEnvList (A x) l = A $ addListToUFM x l
lookupOccEnv (A x) y = lookupUFM x y
mkOccEnv     l    = A $ listToUFM l
elemOccEnv x (A y) 	 = elemUFM x y
foldOccEnv a b (A c)	 = foldUFM a b c 
occEnvElts (A x)	 = eltsUFM x
plusOccEnv (A x) (A y)	 = A $ plusUFM x y 
plusOccEnv_C f (A x) (A y)	 = A $ plusUFM_C f x y 
extendOccEnv_C f (A x) y z   = A $ addToUFM_C f x y z
456
extendOccEnv_Acc f g (A x) y z   = A $ addToUFM_Acc f g x y z
457 458 459 460 461
mapOccEnv f (A x)	 = A $ mapUFM f x
mkOccEnv_C comb l = A $ addListToUFM_C comb emptyUFM l
delFromOccEnv (A x) y    = A $ delFromUFM x y
delListFromOccEnv (A x) y  = A $ delListFromUFM x y
filterOccEnv x (A y)       = A $ filterUFM x y
Simon Peyton Jones's avatar
Simon Peyton Jones committed
462
alterOccEnv fn (A y) k     = A $ alterUFM fn y k
463 464 465

instance Outputable a => Outputable (OccEnv a) where
    ppr (A x) = ppr x
466

467
type OccSet = UniqSet OccName
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500

emptyOccSet	  :: OccSet
unitOccSet	  :: OccName -> OccSet
mkOccSet          :: [OccName] -> OccSet
extendOccSet      :: OccSet -> OccName -> OccSet
extendOccSetList  :: OccSet -> [OccName] -> OccSet
unionOccSets	  :: OccSet -> OccSet -> OccSet
unionManyOccSets  :: [OccSet] -> OccSet
minusOccSet 	  :: OccSet -> OccSet -> OccSet
elemOccSet	  :: OccName -> OccSet -> Bool
occSetElts	  :: OccSet -> [OccName]
foldOccSet	  :: (OccName -> b -> b) -> b -> OccSet -> b
isEmptyOccSet	  :: OccSet -> Bool
intersectOccSet   :: OccSet -> OccSet -> OccSet
intersectsOccSet  :: OccSet -> OccSet -> Bool

emptyOccSet	  = emptyUniqSet
unitOccSet	  = unitUniqSet
mkOccSet          = mkUniqSet
extendOccSet	  = addOneToUniqSet
extendOccSetList  = addListToUniqSet
unionOccSets      = unionUniqSets
unionManyOccSets  = unionManyUniqSets
minusOccSet	  = minusUniqSet
elemOccSet        = elementOfUniqSet
occSetElts        = uniqSetToList
foldOccSet	  = foldUniqSet
isEmptyOccSet     = isEmptyUniqSet
intersectOccSet   = intersectUniqSets
intersectsOccSet s1 s2 = not (isEmptyOccSet (s1 `intersectOccSet` s2))
\end{code}


501 502
%************************************************************************
%*									*
503
\subsection{Predicates and taking them apart}
504 505 506
%*									*
%************************************************************************

507 508
\begin{code}
occNameString :: OccName -> String
509
occNameString (OccName _ s) = unpackFS s
510

511 512
setOccNameSpace :: NameSpace -> OccName -> OccName
setOccNameSpace sp (OccName _ occ) = OccName sp occ
513

514
isVarOcc, isTvOcc, isTcOcc, isDataOcc :: OccName -> Bool
515 516

isVarOcc (OccName VarName _) = True
twanvl's avatar
twanvl committed
517
isVarOcc _                   = False
518

519
isTvOcc (OccName TvName _) = True
twanvl's avatar
twanvl committed
520
isTvOcc _                  = False
521

522
isTcOcc (OccName TcClsName _) = True
twanvl's avatar
twanvl committed
523
isTcOcc _                     = False
524

525 526 527
-- | /Value/ 'OccNames's are those that are either in 
-- the variable or data constructor namespaces
isValOcc :: OccName -> Bool
528 529
isValOcc (OccName VarName  _) = True
isValOcc (OccName DataName _) = True
twanvl's avatar
twanvl committed
530
isValOcc _                    = False
531

532
isDataOcc (OccName DataName _) = True
twanvl's avatar
twanvl committed
533
isDataOcc _                    = False
534

535 536 537 538 539
-- | Test if the 'OccName' is a data constructor that starts with
-- a symbol (e.g. @:@, or @[]@)
isDataSymOcc :: OccName -> Bool
isDataSymOcc (OccName DataName s) = isLexConSym s
isDataSymOcc _                    = False
540
-- Pretty inefficient!
541 542 543 544

-- | Test if the 'OccName' is that for any operator (whether 
-- it is a data constructor or variable or whatever)
isSymOcc :: OccName -> Bool
545
isSymOcc (OccName DataName s)  = isLexConSym s
546
isSymOcc (OccName TcClsName s) = isLexSym s
547
isSymOcc (OccName VarName s)   = isLexSym s
548
isSymOcc (OccName TvName s)    = isLexSym s
549
-- Pretty inefficient!
550 551

parenSymOcc :: OccName -> SDoc -> SDoc
552
-- ^ Wrap parens around an operator
553 554
parenSymOcc occ doc | isSymOcc occ = parens doc
		    | otherwise    = doc
555
\end{code}
556 557


558
\begin{code}
559 560 561 562 563 564
startsWithUnderscore :: OccName -> Bool
-- ^ Haskell 98 encourages compilers to suppress warnings about unsed
-- names in a pattern if they start with @_@: this implements that test
startsWithUnderscore occ = case occNameString occ of
			     ('_' : _) -> True
			     _other    -> False
565 566 567
\end{code}


568 569 570 571 572
%************************************************************************
%*									*
\subsection{Making system names}
%*									*
%************************************************************************
573

574
Here's our convention for splitting up the interface file name space:
575

576 577
   d...		dictionary identifiers
   		(local variables, so no name-clash worries)
578

579 580 581
All of these other OccNames contain a mixture of alphabetic
and symbolic characters, and hence cannot possibly clash with
a user-written type or function name
582

583 584 585 586 587 588 589 590 591 592 593
   $f...	Dict-fun identifiers (from inst decls)
   $dmop	Default method for 'op'
   $pnC		n'th superclass selector for class C
   $wf		Worker for functtoin 'f'
   $sf..	Specialised version of f
   T:C		Tycon for dictionary for class C
   D:C		Data constructor for dictionary for class C
   NTCo:T       Coercion connecting newtype T with its representation type
   TFCo:R       Coercion connecting a data family to its respresentation type R

In encoded form these appear as Zdfxxx etc
594

595
	:...		keywords (export:, letrec: etc.)
apt's avatar
apt committed
596
--- I THINK THIS IS WRONG!
597

598 599
This knowledge is encoded in the following functions.

apt's avatar
apt committed
600
@mk_deriv@ generates an @OccName@ from the prefix and a string.
601
NB: The string must already be encoded!
602 603

\begin{code}
604 605
mk_deriv :: NameSpace 
	 -> String		-- Distinguishes one sort of derived name from another
606
	 -> String
607
	 -> OccName
608

609
mk_deriv occ_sp sys_prefix str = mkOccName occ_sp (sys_prefix ++ str)
610 611 612 613 614 615 616

isDerivedOccName :: OccName -> Bool
isDerivedOccName occ = 
   case occNameString occ of
     '$':c:_ | isAlphaNum c -> True
     ':':c:_ | isAlphaNum c -> True
     _other                 -> False
617
\end{code}
618

619
\begin{code}
Gergő Érdi's avatar
Gergő Érdi committed
620
mkDataConWrapperOcc, mkWorkerOcc, mkMatcherOcc, mkDefaultMethodOcc,
621
        mkGenDefMethodOcc, mkDerivedTyConOcc, mkClassDataConOcc, mkDictOcc,
622
	mkIPOcc, mkSpecOcc, mkForeignExportOcc, mkRepEqOcc, mkGenOcc1, mkGenOcc2,
jpm@cs.ox.ac.uk's avatar
jpm@cs.ox.ac.uk committed
623
 	mkGenD, mkGenR, mkGen1R, mkGenRCo,
batterseapower's avatar
batterseapower committed
624 625 626
	mkDataTOcc, mkDataCOcc, mkDataConWorkerOcc, mkNewTyCoOcc,
	mkInstTyCoOcc, mkEqPredCoOcc, mkClassOpAuxOcc,
        mkCon2TagOcc, mkTag2ConOcc, mkMaxTagOcc
627 628
   :: OccName -> OccName

629
-- These derived variables have a prefix that no Haskell value could have
630
mkDataConWrapperOcc = mk_simple_deriv varName  "$W"
631
mkWorkerOcc         = mk_simple_deriv varName  "$w"
Gergő Érdi's avatar
Gergő Érdi committed
632
mkMatcherOcc        = mk_simple_deriv varName  "$m"
633
mkDefaultMethodOcc  = mk_simple_deriv varName  "$dm"
634
mkGenDefMethodOcc   = mk_simple_deriv varName  "$gdm"
635
mkClassOpAuxOcc     = mk_simple_deriv varName  "$c"
batterseapower's avatar
batterseapower committed
636
mkDerivedTyConOcc   = mk_simple_deriv tcName   ":"	-- The : prefix makes sure it classifies as a tycon/datacon
637
mkClassDataConOcc   = mk_simple_deriv dataName "D:"	-- We go straight to the "real" data con
638
							-- for datacons from classes
639 640 641 642
mkDictOcc	    = mk_simple_deriv varName  "$d"
mkIPOcc		    = mk_simple_deriv varName  "$i"
mkSpecOcc	    = mk_simple_deriv varName  "$s"
mkForeignExportOcc  = mk_simple_deriv varName  "$f"
643
mkRepEqOcc          = mk_simple_deriv tvName   "$r"      -- In RULES involving Coercible
644 645 646
mkNewTyCoOcc        = mk_simple_deriv tcName   "NTCo:"	-- Coercion for newtypes
mkInstTyCoOcc       = mk_simple_deriv tcName   "TFCo:"   -- Coercion for type functions
mkEqPredCoOcc	    = mk_simple_deriv tcName   "$co"
647

648 649 650 651 652
-- used in derived instances
mkCon2TagOcc        = mk_simple_deriv varName  "$con2tag_"
mkTag2ConOcc        = mk_simple_deriv varName  "$tag2con_"
mkMaxTagOcc         = mk_simple_deriv varName  "$maxtag_"

653
-- Generic derivable classes (old)
654 655 656
mkGenOcc1           = mk_simple_deriv varName  "$gfrom"
mkGenOcc2           = mk_simple_deriv varName  "$gto" 

657 658
-- Generic deriving mechanism (new)
mkGenD         = mk_simple_deriv tcName "D1"
659 660

mkGenC :: OccName -> Int -> OccName
661
mkGenC occ m   = mk_deriv tcName ("C1_" ++ show m) (occNameString occ)
662 663

mkGenS :: OccName -> Int -> Int -> OccName
664 665 666
mkGenS occ m n = mk_deriv tcName ("S1_" ++ show m ++ "_" ++ show n)
                   (occNameString occ)

667
mkGenR   = mk_simple_deriv tcName "Rep_"
jpm@cs.ox.ac.uk's avatar
jpm@cs.ox.ac.uk committed
668
mkGen1R  = mk_simple_deriv tcName "Rep1_"
669
mkGenRCo = mk_simple_deriv tcName "CoRep_"
670

Gabor Greif's avatar
Gabor Greif committed
671
-- data T = MkT ... deriving( Data ) needs definitions for 
672 673 674 675 676
--	$tT   :: Data.Generics.Basics.DataType
--	$cMkT :: Data.Generics.Basics.Constr
mkDataTOcc = mk_simple_deriv varName  "$t"
mkDataCOcc = mk_simple_deriv varName  "$c"

677
-- Vectorisation
678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
mkVectOcc, mkVectTyConOcc, mkVectDataConOcc, mkVectIsoOcc,
 mkPADFunOcc,      mkPReprTyConOcc,
 mkPDataTyConOcc,  mkPDataDataConOcc,
 mkPDatasTyConOcc, mkPDatasDataConOcc
  :: Maybe String -> OccName -> OccName
mkVectOcc          = mk_simple_deriv_with varName  "$v"
mkVectTyConOcc     = mk_simple_deriv_with tcName   "V:"
mkVectDataConOcc   = mk_simple_deriv_with dataName "VD:"
mkVectIsoOcc       = mk_simple_deriv_with varName  "$vi"
mkPADFunOcc        = mk_simple_deriv_with varName  "$pa"
mkPReprTyConOcc    = mk_simple_deriv_with tcName   "VR:"
mkPDataTyConOcc    = mk_simple_deriv_with tcName   "VP:"
mkPDatasTyConOcc   = mk_simple_deriv_with tcName   "VPs:"
mkPDataDataConOcc  = mk_simple_deriv_with dataName "VPD:"
mkPDatasDataConOcc = mk_simple_deriv_with dataName "VPDs:"
693

twanvl's avatar
twanvl committed
694
mk_simple_deriv :: NameSpace -> String -> OccName -> OccName
695
mk_simple_deriv sp px occ = mk_deriv sp px (occNameString occ)
696

697 698 699 700
mk_simple_deriv_with :: NameSpace -> String -> Maybe String -> OccName -> OccName
mk_simple_deriv_with sp px Nothing     occ = mk_deriv sp px                  (occNameString occ)
mk_simple_deriv_with sp px (Just with) occ = mk_deriv sp (px ++ with ++ "_") (occNameString occ)

701 702
-- Data constructor workers are made by setting the name space
-- of the data constructor OccName (which should be a DataName)
703
-- to VarName
704
mkDataConWorkerOcc datacon_occ = setOccNameSpace varName datacon_occ 
705 706 707
\end{code}

\begin{code}
708 709 710
mkSuperDictSelOcc :: Int 	-- ^ Index of superclass, e.g. 3
		  -> OccName 	-- ^ Class, e.g. @Ord@
		  -> OccName	-- ^ Derived 'Occname', e.g. @$p3Ord@
batterseapower's avatar
batterseapower committed
711 712
mkSuperDictSelOcc index cls_tc_occ
  = mk_deriv varName "$p" (show index ++ occNameString cls_tc_occ)
713

714 715 716
mkLocalOcc :: Unique 		-- ^ Unique to combine with the 'OccName'
	   -> OccName		-- ^ Local name, e.g. @sat@
	   -> OccName		-- ^ Nice unique version, e.g. @$L23sat@
717
mkLocalOcc uniq occ
718 719 720
   = mk_deriv varName ("$L" ++ show uniq) (occNameString occ)
	-- The Unique might print with characters 
	-- that need encoding (e.g. 'z'!)
721 722
\end{code}

723
\begin{code}
Ian Lynagh's avatar
Ian Lynagh committed
724 725
-- | Derive a name for the representation type constructor of a
-- @data@\/@newtype@ instance.
726 727 728 729 730
mkInstTyTcOcc :: String 		-- ^ Family name, e.g. @Map@
              -> OccSet                 -- ^ avoid these Occs
	      -> OccName		-- ^ @R:Map@
mkInstTyTcOcc str set =
  chooseUniqueOcc tcName ('R' : ':' : str) set
731
\end{code}
732 733

\begin{code}
734
mkDFunOcc :: String		-- ^ Typically the class and type glommed together e.g. @OrdMaybe@.
735
				-- Only used in debug mode, for extra clarity
736
	  -> Bool		-- ^ Is this a hs-boot instance DFun?
737
          -> OccSet             -- ^ avoid these Occs
738
	  -> OccName		-- ^ E.g. @$f3OrdMaybe@
739 740 741 742 743

-- In hs-boot files we make dict funs like $fx7ClsTy, which get bound to the real
-- thing when we compile the mother module. Reason: we don't know exactly
-- what the  mother module will call it.

744 745
mkDFunOcc info_str is_boot set
  = chooseUniqueOcc VarName (prefix ++ info_str) set
746 747 748
  where
    prefix | is_boot   = "$fx"
	   | otherwise = "$f"
749 750 751 752 753 754 755 756 757 758 759 760
\end{code}

Sometimes we need to pick an OccName that has not already been used,
given a set of in-use OccNames.

\begin{code}
chooseUniqueOcc :: NameSpace -> String -> OccSet -> OccName
chooseUniqueOcc ns str set = loop (mkOccName ns str) (0::Int)
  where
  loop occ n
   | occ `elemOccSet` set = loop (mkOccName ns (str ++ show n)) (n+1)
   | otherwise            = occ
761 762
\end{code}

763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783
We used to add a '$m' to indicate a method, but that gives rise to bad
error messages from the type checker when we print the function name or pattern
of an instance-decl binding.  Why? Because the binding is zapped
to use the method name in place of the selector name.
(See TcClassDcl.tcMethodBind)

The way it is now, -ddump-xx output may look confusing, but
you can always say -dppr-debug to get the uniques.

However, we *do* have to zap the first character to be lower case,
because overloaded constructors (blarg) generate methods too.
And convert to VarName space

e.g. a call to constructor MkFoo where
	data (Ord a) => Foo a = MkFoo a

If this is necessary, we do it by prefixing '$m'.  These 
guys never show up in error messages.  What a hack.

\begin{code}
mkMethodOcc :: OccName -> OccName
twanvl's avatar
twanvl committed
784 785
mkMethodOcc occ@(OccName VarName _) = occ
mkMethodOcc occ                     = mk_simple_deriv varName "$m" occ
786 787
\end{code}

788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805

%************************************************************************
%*									*
\subsection{Tidying them up}
%*									*
%************************************************************************

Before we print chunks of code we like to rename it so that
we don't have to print lots of silly uniques in it.  But we mustn't
accidentally introduce name clashes!  So the idea is that we leave the
OccName alone unless it accidentally clashes with one that is already
in scope; if so, we tack on '1' at the end and try again, then '2', and
so on till we find a unique one.

There's a wrinkle for operators.  Consider '>>='.  We can't use '>>=1' 
because that isn't a single lexeme.  So we encode it to 'lle' and *then*
tack on the '1', if necessary.

806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
Note [TidyOccEnv]
~~~~~~~~~~~~~~~~~
type TidyOccEnv = UniqFM Int

* Domain = The OccName's FastString. These FastStrings are "taken";
           make sure that we don't re-use

* Int, n = A plausible starting point for new guesses
           There is no guarantee that "FSn" is available; 
           you must look that up in the TidyOccEnv.  But
           it's a good place to start looking.

* When looking for a renaming for "foo2" we strip off the "2" and start
  with "foo".  Otherwise if we tidy twice we get silly names like foo23.

821
\begin{code}
822 823
type TidyOccEnv = UniqFM Int	-- The in-scope OccNames
  -- See Note [TidyOccEnv]
824

twanvl's avatar
twanvl committed
825
emptyTidyOccEnv :: TidyOccEnv
826
emptyTidyOccEnv = emptyUFM
827 828

initTidyOccEnv :: [OccName] -> TidyOccEnv	-- Initialise with names to avoid!
829 830 831
initTidyOccEnv = foldl add emptyUFM
  where
    add env (OccName _ fs) = addToUFM env fs 1
832 833

tidyOccName :: TidyOccEnv -> OccName -> (TidyOccEnv, OccName)
834 835 836 837
tidyOccName env occ@(OccName occ_sp fs)
  = case lookupUFM env fs of
	Just n  -> find n
	Nothing -> (addToUFM env fs 1, occ)
838
  where
839 840 841 842 843 844 845 846 847 848 849 850 851 852 853
    base :: String  -- Drop trailing digits (see Note [TidyOccEnv])
    base = reverse (dropWhile isDigit (reverse (unpackFS fs)))
 
    find n 
      = case lookupUFM env new_fs of
          Just n' -> find (n1 `max` n')
                     -- The max ensures that n increases, avoiding loops
          Nothing -> (addToUFM (addToUFM env fs n1) new_fs n1,
                      OccName occ_sp new_fs)
                     -- We update only the beginning and end of the
                     -- chain that find explores; it's a little harder to
                     -- update the middle and there's no real need.
       where
         n1 = n+1
         new_fs = mkFastString (base ++ show n)
854 855
\end{code}

856
%************************************************************************
857
%*									*
apt's avatar
apt committed
858
\subsection{Lexical categories}
859 860 861
%*									*
%************************************************************************

862 863
These functions test strings to see if they fit the lexical categories
defined in the Haskell report.
864

865 866 867 868 869 870 871 872 873
Note [Classification of generated names]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Some names generated for internal use can show up in debugging output,
e.g.  when using -ddump-simpl. These generated names start with a $
but should still be pretty-printed using prefix notation. We make sure
this is the case in isLexVarSym by only classifying a name as a symbol
if all its characters are symbols, not just its first one.

874
\begin{code}
875 876
isLexCon,   isLexVar,    isLexId,    isLexSym    :: FastString -> Bool
isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FastString -> Bool
877

878 879
isLexCon cs = isLexConId  cs || isLexConSym cs
isLexVar cs = isLexVarId  cs || isLexVarSym cs
880

881 882
isLexId  cs = isLexConId  cs || isLexVarId  cs
isLexSym cs = isLexConSym cs || isLexVarSym cs
883

884
-------------
885

886
isLexConId cs				-- Prefix type or data constructors
Ian Lynagh's avatar
Ian Lynagh committed
887 888 889
  | nullFS cs	       = False		-- 	e.g. "Foo", "[]", "(,)" 
  | cs == (fsLit "[]") = True
  | otherwise	       = startsConId (headFS cs)
890

891
isLexVarId cs				-- Ordinary prefix identifiers
892
  | nullFS cs	      = False		-- 	e.g. "x", "_x"
893
  | otherwise         = startsVarId (headFS cs)
894

895
isLexConSym cs				-- Infix type or data constructors
Ian Lynagh's avatar
Ian Lynagh committed
896 897 898
  | nullFS cs	       = False		--	e.g. ":-:", ":", "->"
  | cs == (fsLit "->") = True
  | otherwise	       = startsConSym (headFS cs)
899

900
isLexVarSym fs				-- Infix identifiers e.g. "+"
eir@cis.upenn.edu's avatar
eir@cis.upenn.edu committed
901 902
  | fs == (fsLit "~R#") = True
  | otherwise
903 904 905
  = case (if nullFS fs then [] else unpackFS fs) of
      [] -> False
      (c:cs) -> startsVarSym c && all isVarSymChar cs
eir@cis.upenn.edu's avatar
eir@cis.upenn.edu committed
906
        -- See Note [Classification of generated names]
907 908

-------------
909
startsVarSym, startsVarId, startsConSym, startsConId :: Char -> Bool
910 911
startsVarSym c = isSymbolASCII c || (ord c > 0x7f && isSymbol c)  -- Infix Ids
startsConSym c = c == ':'		-- Infix data constructors
912 913
startsVarId c  = isLower c || c == '_'	-- Ordinary Ids
startsConId c  = isUpper c || c == '('	-- Ordinary type constructors and data constructors
914

twanvl's avatar
twanvl committed
915
isSymbolASCII :: Char -> Bool
916
isSymbolASCII c = c `elem` "!#$%&*+./<=>?@\\^|~-"
917 918 919

isVarSymChar :: Char -> Bool
isVarSymChar c = c == ':' || startsVarSym c
920
\end{code}
921 922 923 924 925 926 927 928

%************************************************************************
%*									*
		Binary instance
    Here rather than BinIface because OccName is abstract
%*									*
%************************************************************************

929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
\begin{code}
instance Binary NameSpace where
    put_ bh VarName = do
	    putByte bh 0
    put_ bh DataName = do
	    putByte bh 1
    put_ bh TvName = do
	    putByte bh 2
    put_ bh TcClsName = do
	    putByte bh 3
    get bh = do
	    h <- getByte bh
	    case h of
	      0 -> do return VarName
	      1 -> do return DataName
	      2 -> do return TvName
	      _ -> do return TcClsName

instance Binary OccName where
    put_ bh (OccName aa ab) = do
	    put_ bh aa
	    put_ bh ab
    get bh = do
	  aa <- get bh
	  ab <- get bh
	  return (OccName aa ab)
\end{code}