2 % (c) The University of Glasgow 2006
3 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
8 -- * The NameSpace type; abstact
9 NameSpace, tcName, clsName, tcClsName, dataName, varName,
13 pprNameSpace, pprNonVarNameSpace, pprNameSpaceBrief,
16 OccName, -- Abstract, instance of Outputable
20 mkOccName, mkOccNameFS,
27 -- ** Derived OccNames
28 mkDataConWrapperOcc, mkWorkerOcc, mkDefaultMethodOcc,
29 mkDerivedTyConOcc, mkNewTyCoOcc,
30 mkClassTyConOcc, mkClassDataConOcc, mkDictOcc, mkIPOcc,
31 mkSpecOcc, mkForeignExportOcc, mkGenOcc1, mkGenOcc2,
32 mkDataTOcc, mkDataCOcc, mkDataConWorkerOcc,
33 mkSuperDictSelOcc, mkLocalOcc, mkMethodOcc, mkInstTyTcOcc,
37 occNameFS, occNameString, occNameSpace,
39 isVarOcc, isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc,
40 parenSymOcc, reportIfUnused, isTcClsName, isVarName,
45 OccEnv, emptyOccEnv, unitOccEnv, extendOccEnv, mapOccEnv,
46 lookupOccEnv, mkOccEnv, mkOccEnv_C, extendOccEnvList, elemOccEnv,
47 occEnvElts, foldOccEnv, plusOccEnv, plusOccEnv_C, extendOccEnv_C,
50 OccSet, emptyOccSet, unitOccSet, mkOccSet, extendOccSet,
52 unionOccSets, unionManyOccSets, minusOccSet, elemOccSet, occSetElts,
53 foldOccSet, isEmptyOccSet, intersectOccSet, intersectsOccSet,
56 TidyOccEnv, emptyTidyOccEnv, tidyOccName, initTidyOccEnv,
58 -- The basic form of names
59 isLexCon, isLexVar, isLexId, isLexSym,
60 isLexConId, isLexConSym, isLexVarId, isLexVarSym,
61 startsVarSym, startsVarId, startsConSym, startsConId
64 #include "HsVersions.h"
79 -- Unicode TODO: put isSymbol in libcompat
80 #if __GLASGOW_HASKELL__ > 604
81 import Data.Char ( isSymbol )
83 isSymbol = const False
88 %************************************************************************
90 \subsection{Name space}
92 %************************************************************************
95 data NameSpace = VarName -- Variables, including "source" data constructors
96 | DataName -- "Real" data constructors
97 | TvName -- Type variables
98 | TcClsName -- Type constructors and classes; Haskell has them
99 -- in the same name space for now.
101 {-! derive: Binary !-}
103 -- Note [Data Constructors]
104 -- see also: Note [Data Constructor Naming] in DataCon.lhs
106 -- "Source" data constructors are the data constructors mentioned
107 -- in Haskell source code
109 -- "Real" data constructors are the data constructors of the
110 -- representation type, which may not be the same as the source
114 -- data T = T !(Int,Int)
116 -- The source datacon has type (Int,Int) -> T
117 -- The real datacon has type Int -> Int -> T
118 -- GHC chooses a representation based on the strictness etc.
121 -- Though type constructors and classes are in the same name space now,
122 -- the NameSpace type is abstract, so we can easily separate them later
123 tcName = TcClsName -- Type constructors
124 clsName = TcClsName -- Classes
125 tcClsName = TcClsName -- Not sure which!
128 srcDataName = DataName -- Haskell-source data constructors should be
129 -- in the Data name space
134 isTcClsName :: NameSpace -> Bool
135 isTcClsName TcClsName = True
136 isTcClsName _ = False
138 isVarName :: NameSpace -> Bool -- Variables or type variables, but not constructors
139 isVarName TvName = True
140 isVarName VarName = True
141 isVarName other = False
143 pprNameSpace :: NameSpace -> SDoc
144 pprNameSpace DataName = ptext SLIT("data constructor")
145 pprNameSpace VarName = ptext SLIT("variable")
146 pprNameSpace TvName = ptext SLIT("type variable")
147 pprNameSpace TcClsName = ptext SLIT("type constructor or class")
149 pprNonVarNameSpace :: NameSpace -> SDoc
150 pprNonVarNameSpace VarName = empty
151 pprNonVarNameSpace ns = pprNameSpace ns
153 pprNameSpaceBrief DataName = char 'd'
154 pprNameSpaceBrief VarName = char 'v'
155 pprNameSpaceBrief TvName = ptext SLIT("tv")
156 pprNameSpaceBrief TcClsName = ptext SLIT("tc")
160 %************************************************************************
162 \subsection[Name-pieces-datatypes]{The @OccName@ datatypes}
164 %************************************************************************
167 data OccName = OccName
168 { occNameSpace :: !NameSpace
169 , occNameFS :: !FastString
175 instance Eq OccName where
176 (OccName sp1 s1) == (OccName sp2 s2) = s1 == s2 && sp1 == sp2
178 instance Ord OccName where
179 compare (OccName sp1 s1) (OccName sp2 s2) = (s1 `compare` s2) `thenCmp`
184 %************************************************************************
186 \subsection{Printing}
188 %************************************************************************
191 instance Outputable OccName where
194 pprOccName :: OccName -> SDoc
195 pprOccName (OccName sp occ)
196 = getPprStyle $ \ sty ->
198 then ftext (zEncodeFS occ)
199 else ftext occ <> if debugStyle sty
200 then braces (pprNameSpaceBrief sp)
205 %************************************************************************
207 \subsection{Construction}
209 %************************************************************************
212 mkOccName :: NameSpace -> String -> OccName
213 mkOccName occ_sp str = OccName occ_sp (mkFastString str)
215 mkOccNameFS :: NameSpace -> FastString -> OccName
216 mkOccNameFS occ_sp fs = OccName occ_sp fs
218 mkVarOcc :: String -> OccName
219 mkVarOcc s = mkOccName varName s
221 mkVarOccFS :: FastString -> OccName
222 mkVarOccFS fs = mkOccNameFS varName fs
224 mkTyVarOcc :: FastString -> OccName
225 mkTyVarOcc fs = mkOccNameFS tvName fs
229 %************************************************************************
233 %************************************************************************
235 OccEnvs are used mainly for the envts in ModIfaces.
237 They are efficient, because FastStrings have unique Int# keys. We assume
238 this key is less than 2^24, so we can make a Unique using
239 mkUnique ns key :: Unique
240 where 'ns' is a Char reprsenting the name space. This in turn makes it
241 easy to build an OccEnv.
244 instance Uniquable OccName where
245 getUnique (OccName ns fs)
246 = mkUnique char (I# (uniqueOfFS fs))
247 where -- See notes above about this getUnique function
254 type OccEnv a = UniqFM a
256 emptyOccEnv :: OccEnv a
257 unitOccEnv :: OccName -> a -> OccEnv a
258 extendOccEnv :: OccEnv a -> OccName -> a -> OccEnv a
259 extendOccEnvList :: OccEnv a -> [(OccName, a)] -> OccEnv a
260 lookupOccEnv :: OccEnv a -> OccName -> Maybe a
261 mkOccEnv :: [(OccName,a)] -> OccEnv a
262 mkOccEnv_C :: (a -> a -> a) -> [(OccName,a)] -> OccEnv a
263 elemOccEnv :: OccName -> OccEnv a -> Bool
264 foldOccEnv :: (a -> b -> b) -> b -> OccEnv a -> b
265 occEnvElts :: OccEnv a -> [a]
266 extendOccEnv_C :: (a->a->a) -> OccEnv a -> OccName -> a -> OccEnv a
267 plusOccEnv :: OccEnv a -> OccEnv a -> OccEnv a
268 plusOccEnv_C :: (a->a->a) -> OccEnv a -> OccEnv a -> OccEnv a
269 mapOccEnv :: (a->b) -> OccEnv a -> OccEnv b
271 emptyOccEnv = emptyUFM
273 extendOccEnv = addToUFM
274 extendOccEnvList = addListToUFM
275 lookupOccEnv = lookupUFM
281 plusOccEnv_C = plusUFM_C
282 extendOccEnv_C = addToUFM_C
285 mkOccEnv_C comb l = addListToUFM_C comb emptyOccEnv l
287 type OccSet = UniqFM OccName
289 emptyOccSet :: OccSet
290 unitOccSet :: OccName -> OccSet
291 mkOccSet :: [OccName] -> OccSet
292 extendOccSet :: OccSet -> OccName -> OccSet
293 extendOccSetList :: OccSet -> [OccName] -> OccSet
294 unionOccSets :: OccSet -> OccSet -> OccSet
295 unionManyOccSets :: [OccSet] -> OccSet
296 minusOccSet :: OccSet -> OccSet -> OccSet
297 elemOccSet :: OccName -> OccSet -> Bool
298 occSetElts :: OccSet -> [OccName]
299 foldOccSet :: (OccName -> b -> b) -> b -> OccSet -> b
300 isEmptyOccSet :: OccSet -> Bool
301 intersectOccSet :: OccSet -> OccSet -> OccSet
302 intersectsOccSet :: OccSet -> OccSet -> Bool
304 emptyOccSet = emptyUniqSet
305 unitOccSet = unitUniqSet
307 extendOccSet = addOneToUniqSet
308 extendOccSetList = addListToUniqSet
309 unionOccSets = unionUniqSets
310 unionManyOccSets = unionManyUniqSets
311 minusOccSet = minusUniqSet
312 elemOccSet = elementOfUniqSet
313 occSetElts = uniqSetToList
314 foldOccSet = foldUniqSet
315 isEmptyOccSet = isEmptyUniqSet
316 intersectOccSet = intersectUniqSets
317 intersectsOccSet s1 s2 = not (isEmptyOccSet (s1 `intersectOccSet` s2))
321 %************************************************************************
323 \subsection{Predicates and taking them apart}
325 %************************************************************************
328 occNameString :: OccName -> String
329 occNameString (OccName _ s) = unpackFS s
331 setOccNameSpace :: NameSpace -> OccName -> OccName
332 setOccNameSpace sp (OccName _ occ) = OccName sp occ
334 isVarOcc, isTvOcc, isDataSymOcc, isSymOcc, isTcOcc :: OccName -> Bool
336 isVarOcc (OccName VarName _) = True
337 isVarOcc other = False
339 isTvOcc (OccName TvName _) = True
340 isTvOcc other = False
342 isTcOcc (OccName TcClsName _) = True
343 isTcOcc other = False
345 isValOcc (OccName VarName _) = True
346 isValOcc (OccName DataName _) = True
347 isValOcc other = False
349 -- Data constructor operator (starts with ':', or '[]')
350 -- Pretty inefficient!
351 isDataSymOcc (OccName DataName s) = isLexConSym s
352 isDataSymOcc (OccName VarName s)
353 | isLexConSym s = pprPanic "isDataSymOcc: check me" (ppr s)
354 -- Jan06: I don't think this should happen
355 isDataSymOcc other = False
357 isDataOcc (OccName DataName _) = True
358 isDataOcc (OccName VarName s)
359 | isLexCon s = pprPanic "isDataOcc: check me" (ppr s)
360 -- Jan06: I don't think this should happen
361 isDataOcc other = False
363 -- Any operator (data constructor or variable)
364 -- Pretty inefficient!
365 isSymOcc (OccName DataName s) = isLexConSym s
366 isSymOcc (OccName TcClsName s) = isLexConSym s
367 isSymOcc (OccName VarName s) = isLexSym s
368 isSymOcc other = False
370 parenSymOcc :: OccName -> SDoc -> SDoc
371 -- Wrap parens around an operator
372 parenSymOcc occ doc | isSymOcc occ = parens doc
378 reportIfUnused :: OccName -> Bool
379 -- Haskell 98 encourages compilers to suppress warnings about
380 -- unused names in a pattern if they start with "_".
381 reportIfUnused occ = case occNameString occ of
387 %************************************************************************
389 \subsection{Making system names}
391 %************************************************************************
393 Here's our convention for splitting up the interface file name space:
395 d... dictionary identifiers
396 (local variables, so no name-clash worries)
398 $f... dict-fun identifiers (from inst decls)
399 $dm... default methods
400 $p... superclass selectors
402 :T... compiler-generated tycons for dictionaries
403 :D... ...ditto data cons
404 $sf.. specialised version of f
406 in encoded form these appear as Zdfxxx etc
408 :... keywords (export:, letrec: etc.)
409 --- I THINK THIS IS WRONG!
411 This knowledge is encoded in the following functions.
414 @mk_deriv@ generates an @OccName@ from the prefix and a string.
415 NB: The string must already be encoded!
418 mk_deriv :: NameSpace
419 -> String -- Distinguishes one sort of derived name from another
423 mk_deriv occ_sp sys_prefix str = mkOccName occ_sp (sys_prefix ++ str)
427 mkDataConWrapperOcc, mkWorkerOcc, mkDefaultMethodOcc, mkDerivedTyConOcc,
428 mkClassTyConOcc, mkClassDataConOcc, mkDictOcc, mkIPOcc,
429 mkSpecOcc, mkForeignExportOcc, mkGenOcc1, mkGenOcc2,
430 mkDataTOcc, mkDataCOcc, mkDataConWorkerOcc, mkNewTyCoOcc
431 :: OccName -> OccName
433 -- These derived variables have a prefix that no Haskell value could have
434 mkDataConWrapperOcc = mk_simple_deriv varName "$W"
435 mkWorkerOcc = mk_simple_deriv varName "$w"
436 mkDefaultMethodOcc = mk_simple_deriv varName "$dm"
437 mkDerivedTyConOcc = mk_simple_deriv tcName ":" -- The : prefix makes sure it classifies
438 mkClassTyConOcc = mk_simple_deriv tcName ":T" -- as a tycon/datacon
439 mkClassDataConOcc = mk_simple_deriv dataName ":D" -- We go straight to the "real" data con
440 -- for datacons from classes
441 mkDictOcc = mk_simple_deriv varName "$d"
442 mkIPOcc = mk_simple_deriv varName "$i"
443 mkSpecOcc = mk_simple_deriv varName "$s"
444 mkForeignExportOcc = mk_simple_deriv varName "$f"
445 mkNewTyCoOcc = mk_simple_deriv tcName ":Co"
446 mkInstTyCoOcc = mk_simple_deriv tcName ":Co" -- derived from rep ty
448 -- Generic derivable classes
449 mkGenOcc1 = mk_simple_deriv varName "$gfrom"
450 mkGenOcc2 = mk_simple_deriv varName "$gto"
452 -- data T = MkT ... deriving( Data ) needs defintions for
453 -- $tT :: Data.Generics.Basics.DataType
454 -- $cMkT :: Data.Generics.Basics.Constr
455 mkDataTOcc = mk_simple_deriv varName "$t"
456 mkDataCOcc = mk_simple_deriv varName "$c"
458 mk_simple_deriv sp px occ = mk_deriv sp px (occNameString occ)
460 -- Data constructor workers are made by setting the name space
461 -- of the data constructor OccName (which should be a DataName)
463 mkDataConWorkerOcc datacon_occ = setOccNameSpace varName datacon_occ
467 mkSuperDictSelOcc :: Int -- Index of superclass, eg 3
468 -> OccName -- Class, eg "Ord"
469 -> OccName -- eg "$p3Ord"
470 mkSuperDictSelOcc index cls_occ
471 = mk_deriv varName "$p" (show index ++ occNameString cls_occ)
473 mkLocalOcc :: Unique -- Unique
474 -> OccName -- Local name (e.g. "sat")
475 -> OccName -- Nice unique version ("$L23sat")
477 = mk_deriv varName ("$L" ++ show uniq) (occNameString occ)
478 -- The Unique might print with characters
479 -- that need encoding (e.g. 'z'!)
482 Derive a name for the representation type constructor of a data/newtype
486 mkInstTyTcOcc :: Int -- Index
487 -> OccName -- Family name (e.g. "Map")
488 -> OccName -- Nice unique version (":R23Map")
489 mkInstTyTcOcc index occ
490 = mk_deriv tcName (":R" ++ show index) (occNameString occ)
494 mkDFunOcc :: String -- Typically the class and type glommed together e.g. "OrdMaybe"
495 -- Only used in debug mode, for extra clarity
496 -> Bool -- True <=> hs-boot instance dfun
497 -> Int -- Unique index
498 -> OccName -- "$f3OrdMaybe"
500 -- In hs-boot files we make dict funs like $fx7ClsTy, which get bound to the real
501 -- thing when we compile the mother module. Reason: we don't know exactly
502 -- what the mother module will call it.
504 mkDFunOcc info_str is_boot index
505 = mk_deriv VarName prefix string
507 prefix | is_boot = "$fx"
509 string | opt_PprStyle_Debug = show index ++ info_str
510 | otherwise = show index
513 We used to add a '$m' to indicate a method, but that gives rise to bad
514 error messages from the type checker when we print the function name or pattern
515 of an instance-decl binding. Why? Because the binding is zapped
516 to use the method name in place of the selector name.
517 (See TcClassDcl.tcMethodBind)
519 The way it is now, -ddump-xx output may look confusing, but
520 you can always say -dppr-debug to get the uniques.
522 However, we *do* have to zap the first character to be lower case,
523 because overloaded constructors (blarg) generate methods too.
524 And convert to VarName space
526 e.g. a call to constructor MkFoo where
527 data (Ord a) => Foo a = MkFoo a
529 If this is necessary, we do it by prefixing '$m'. These
530 guys never show up in error messages. What a hack.
533 mkMethodOcc :: OccName -> OccName
534 mkMethodOcc occ@(OccName VarName fs) = occ
535 mkMethodOcc occ = mk_simple_deriv varName "$m" occ
539 %************************************************************************
541 \subsection{Tidying them up}
543 %************************************************************************
545 Before we print chunks of code we like to rename it so that
546 we don't have to print lots of silly uniques in it. But we mustn't
547 accidentally introduce name clashes! So the idea is that we leave the
548 OccName alone unless it accidentally clashes with one that is already
549 in scope; if so, we tack on '1' at the end and try again, then '2', and
550 so on till we find a unique one.
552 There's a wrinkle for operators. Consider '>>='. We can't use '>>=1'
553 because that isn't a single lexeme. So we encode it to 'lle' and *then*
554 tack on the '1', if necessary.
557 type TidyOccEnv = OccEnv Int -- The in-scope OccNames
558 -- Range gives a plausible starting point for new guesses
560 emptyTidyOccEnv = emptyOccEnv
562 initTidyOccEnv :: [OccName] -> TidyOccEnv -- Initialise with names to avoid!
563 initTidyOccEnv = foldl (\env occ -> extendOccEnv env occ 1) emptyTidyOccEnv
565 tidyOccName :: TidyOccEnv -> OccName -> (TidyOccEnv, OccName)
567 tidyOccName in_scope occ@(OccName occ_sp fs)
568 = case lookupOccEnv in_scope occ of
569 Nothing -> -- Not already used: make it used
570 (extendOccEnv in_scope occ 1, occ)
572 Just n -> -- Already used: make a new guess,
573 -- change the guess base, and try again
574 tidyOccName (extendOccEnv in_scope occ (n+1))
575 (mkOccName occ_sp (unpackFS fs ++ show n))
578 %************************************************************************
580 Stuff for dealing with tuples
582 %************************************************************************
585 mkTupleOcc :: NameSpace -> Boxity -> Arity -> OccName
586 mkTupleOcc ns bx ar = OccName ns (mkFastString str)
588 -- no need to cache these, the caching is done in the caller
589 -- (TysWiredIn.mk_tuple)
591 Boxed -> '(' : commas ++ ")"
592 Unboxed -> '(' : '#' : commas ++ "#)"
594 commas = take (ar-1) (repeat ',')
596 isTupleOcc_maybe :: OccName -> Maybe (NameSpace, Boxity, Arity)
597 -- Tuples are special, because there are so many of them!
598 isTupleOcc_maybe (OccName ns fs)
599 = case unpackFS fs of
600 '(':'#':',':rest -> Just (ns, Unboxed, 2 + count_commas rest)
601 '(':',':rest -> Just (ns, Boxed, 2 + count_commas rest)
604 count_commas (',':rest) = 1 + count_commas rest
608 %************************************************************************
610 \subsection{Lexical categories}
612 %************************************************************************
614 These functions test strings to see if they fit the lexical categories
615 defined in the Haskell report.
618 isLexCon, isLexVar, isLexId, isLexSym :: FastString -> Bool
619 isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FastString -> Bool
621 isLexCon cs = isLexConId cs || isLexConSym cs
622 isLexVar cs = isLexVarId cs || isLexVarSym cs
624 isLexId cs = isLexConId cs || isLexVarId cs
625 isLexSym cs = isLexConSym cs || isLexVarSym cs
629 isLexConId cs -- Prefix type or data constructors
630 | nullFS cs = False -- e.g. "Foo", "[]", "(,)"
631 | cs == FSLIT("[]") = True
632 | otherwise = startsConId (headFS cs)
634 isLexVarId cs -- Ordinary prefix identifiers
635 | nullFS cs = False -- e.g. "x", "_x"
636 | otherwise = startsVarId (headFS cs)
638 isLexConSym cs -- Infix type or data constructors
639 | nullFS cs = False -- e.g. ":-:", ":", "->"
640 | cs == FSLIT("->") = True
641 | otherwise = startsConSym (headFS cs)
643 isLexVarSym cs -- Infix identifiers
644 | nullFS cs = False -- e.g. "+"
645 | otherwise = startsVarSym (headFS cs)
648 startsVarSym, startsVarId, startsConSym, startsConId :: Char -> Bool
649 startsVarSym c = isSymbolASCII c || (ord c > 0x7f && isSymbol c) -- Infix Ids
650 startsConSym c = c == ':' -- Infix data constructors
651 startsVarId c = isLower c || c == '_' -- Ordinary Ids
652 startsConId c = isUpper c || c == '(' -- Ordinary type constructors and data constructors
654 isSymbolASCII c = c `elem` "!#$%&*+./<=>?@\\^|~-"
657 %************************************************************************
660 Here rather than BinIface because OccName is abstract
662 %************************************************************************
665 instance Binary NameSpace where
668 put_ bh DataName = do
672 put_ bh TcClsName = do
677 0 -> do return VarName
678 1 -> do return DataName
679 2 -> do return TvName
680 _ -> do return TcClsName
682 instance Binary OccName where
683 put_ bh (OccName aa ab) = do
689 return (OccName aa ab)