2 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
5 \section[OccName]{@OccName@}
9 -- The NameSpace type; abstact
10 NameSpace, tcName, clsName, tcClsName, dataName, varName, ipName,
11 tvName, uvName, nameSpaceString,
14 OccName, -- Abstract, instance of Outputable
17 mkSrcOccFS, mkSysOcc, mkSysOccFS, mkSrcVarOcc, mkKindOccFS,
18 mkSuperDictSelOcc, mkDFunOcc, mkForeignExportOcc,
19 mkDictOcc, mkIPOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc,
20 mkDerivedTyConOcc, mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc,
22 isTvOcc, isUvOcc, isDataOcc, isDataSymOcc, isSymOcc, isIPOcc,
24 occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour,
28 TidyOccEnv, emptyTidyOccEnv, tidyOccName, initTidyOccEnv,
31 EncodedString, EncodedFS, UserString, UserFS, encode, encodeFS, decode, pprEncodedFS,
33 -- The basic form of names
34 isLexCon, isLexVar, isLexId, isLexSym,
35 isLexConId, isLexConSym, isLexVarId, isLexVarSym,
36 isLowerISO, isUpperISO
40 #include "HsVersions.h"
42 import Char ( isDigit, isAlpha, isUpper, isLower, ISALPHANUM, ord, chr, digitToInt, intToDigit )
43 import Util ( thenCmp )
44 import FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM )
49 We hold both module names and identifier names in a 'Z-encoded' form
50 that makes them acceptable both as a C identifier and as a Haskell
53 They can always be decoded again when printing error messages
54 or anything else for the user, but it does make sense for it
55 to be represented here in encoded form, so that when generating
56 code the encoding operation is not performed on each occurrence.
58 These type synonyms help documentation.
61 type UserFS = FAST_STRING -- As the user typed it
62 type EncodedFS = FAST_STRING -- Encoded form
64 type UserString = String -- As the user typed it
65 type EncodedString = String -- Encoded form
68 pprEncodedFS :: EncodedFS -> SDoc
70 = getPprStyle $ \ sty ->
72 text (decode (_UNPK_ fs))
77 %************************************************************************
79 \subsection{Name space}
81 %************************************************************************
84 data NameSpace = VarName -- Variables
85 | IPName -- Implicit Parameters
86 | DataName -- Data constructors
87 | TvName -- Type variables
88 | UvName -- Usage variables
89 | TcClsName -- Type constructors and classes; Haskell has them
90 -- in the same name space for now.
93 -- Though type constructors and classes are in the same name space now,
94 -- the NameSpace type is abstract, so we can easily separate them later
95 tcName = TcClsName -- Type constructors
96 clsName = TcClsName -- Classes
97 tcClsName = TcClsName -- Not sure which!
106 nameSpaceString :: NameSpace -> String
107 nameSpaceString DataName = "Data constructor"
108 nameSpaceString VarName = "Variable"
109 nameSpaceString IPName = "Implicit Param"
110 nameSpaceString TvName = "Type variable"
111 nameSpaceString UvName = "Usage variable"
112 nameSpaceString TcClsName = "Type constructor or class"
116 %************************************************************************
118 \subsection[Name-pieces-datatypes]{The @OccName@ datatypes}
120 %************************************************************************
123 data OccName = OccName
130 instance Eq OccName where
131 (OccName sp1 s1) == (OccName sp2 s2) = s1 == s2 && sp1 == sp2
133 instance Ord OccName where
134 compare (OccName sp1 s1) (OccName sp2 s2) = (s1 `compare` s2) `thenCmp`
139 %************************************************************************
141 \subsection{Printing}
143 %************************************************************************
146 instance Outputable OccName where
149 pprOccName :: OccName -> SDoc
150 pprOccName (OccName sp occ) = pprEncodedFS occ
154 %************************************************************************
156 \subsection{Construction}
158 %************************************************************************
160 *Sys* things do no encoding; the caller should ensure that the thing is
164 mkSysOcc :: NameSpace -> EncodedString -> OccName
165 mkSysOcc occ_sp str = ASSERT( alreadyEncoded str )
166 OccName occ_sp (_PK_ str)
168 mkSysOccFS :: NameSpace -> EncodedFS -> OccName
169 mkSysOccFS occ_sp fs = ASSERT2( alreadyEncodedFS fs, ppr fs )
172 -- Kind constructors get a speical function. Uniquely, they are not encoded,
173 -- so that they have names like '*'. This means that *even in interface files*
174 -- we'll get kinds like (* -> (* -> *)). We can't use mkSysOcc because it
175 -- has an ASSERT that doesn't hold.
176 mkKindOccFS :: NameSpace -> EncodedFS -> OccName
177 mkKindOccFS occ_sp fs = OccName occ_sp fs
180 *Source-code* things are encoded.
183 mkSrcOccFS :: NameSpace -> UserFS -> OccName
184 mkSrcOccFS occ_sp fs = mkSysOccFS occ_sp (encodeFS fs)
186 mkSrcVarOcc :: UserFS -> OccName
187 mkSrcVarOcc fs = mkSysOccFS varName (encodeFS fs)
192 %************************************************************************
194 \subsection{Predicates and taking them apart}
196 %************************************************************************
199 occNameFS :: OccName -> EncodedFS
200 occNameFS (OccName _ s) = s
202 occNameString :: OccName -> EncodedString
203 occNameString (OccName _ s) = _UNPK_ s
205 occNameUserString :: OccName -> UserString
206 occNameUserString occ = decode (occNameString occ)
208 occNameSpace :: OccName -> NameSpace
209 occNameSpace (OccName sp _) = sp
211 setOccNameSpace :: OccName -> NameSpace -> OccName
212 setOccNameSpace (OccName _ occ) sp = OccName sp occ
214 -- occNameFlavour is used only to generate good error messages
215 occNameFlavour :: OccName -> String
216 occNameFlavour (OccName sp _) = nameSpaceString sp
220 isTvOcc, isDataSymOcc, isSymOcc, isUvOcc :: OccName -> Bool
222 isTvOcc (OccName TvName _) = True
223 isTvOcc other = False
225 isUvOcc (OccName UvName _) = True
226 isUvOcc other = False
228 -- Data constructor operator (starts with ':', or '[]')
229 -- Pretty inefficient!
230 isDataSymOcc (OccName DataName s) = isLexConSym (decodeFS s)
231 isDataSymOcc other = False
233 isDataOcc (OccName DataName _) = True
234 isDataOcc oter = False
236 -- Any operator (data constructor or variable)
237 -- Pretty inefficient!
238 isSymOcc (OccName DataName s) = isLexConSym (decodeFS s)
239 isSymOcc (OccName VarName s) = isLexSym (decodeFS s)
241 isIPOcc (OccName IPName _) = True
246 %************************************************************************
248 \subsection{Making system names}
250 %************************************************************************
252 Here's our convention for splitting up the interface file name space:
254 d... dictionary identifiers
255 (local variables, so no name-clash worries)
257 $f... dict-fun identifiers (from inst decls)
258 $dm... default methods
259 $p... superclass selectors
261 $T... compiler-generated tycons for dictionaries
262 $D... ...ditto data cons
263 $sf.. specialised version of f
265 in encoded form these appear as Zdfxxx etc
267 :... keywords (export:, letrec: etc.)
269 This knowledge is encoded in the following functions.
272 @mk_deriv@ generates an @OccName@ from the one-char prefix and a string.
273 NB: The string must already be encoded!
276 mk_deriv :: NameSpace
277 -> String -- Distinguishes one sort of derived name from another
278 -> EncodedString -- Must be already encoded!! We don't want to encode it a
279 -- second time because encoding isn't itempotent
282 mk_deriv occ_sp sys_prefix str = mkSysOcc occ_sp (encode sys_prefix ++ str)
286 mkDictOcc, mkIPOcc, mkWorkerOcc, mkDefaultMethodOcc,
287 mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc
288 :: OccName -> OccName
290 -- These derived variables have a prefix that no Haskell value could have
291 mkWorkerOcc = mk_simple_deriv varName "$w"
292 mkDefaultMethodOcc = mk_simple_deriv varName "$dm"
293 mkDerivedTyConOcc = mk_simple_deriv tcName ":" -- The : prefix makes sure it classifies
294 mkClassTyConOcc = mk_simple_deriv tcName ":T" -- as a tycon/datacon
295 mkClassDataConOcc = mk_simple_deriv dataName ":D" --
296 mkDictOcc = mk_simple_deriv varName "$d"
297 mkIPOcc = mk_simple_deriv varName "$i"
298 mkSpecOcc = mk_simple_deriv varName "$s"
299 mkForeignExportOcc = mk_simple_deriv varName "$f"
301 mk_simple_deriv sp px occ = mk_deriv sp px (occNameString occ)
305 mkSuperDictSelOcc :: Int -- Index of superclass, eg 3
306 -> OccName -- Class, eg "Ord"
307 -> OccName -- eg "p3Ord"
308 mkSuperDictSelOcc index cls_occ
309 = mk_deriv varName "$p" (show index ++ occNameString cls_occ)
314 mkDFunOcc :: EncodedString -- Typically the class and type glommed together e.g. "OrdMaybe"
315 -> Int -- Unique to distinguish dfuns which share the previous two
317 -- The requirement is that the (string,index) pair be unique in this module
319 -> OccName -- "$fOrdMaybe3"
321 mkDFunOcc string index
322 = mk_deriv VarName "$f" (show_index ++ string)
324 show_index | index == 0 = ""
325 | otherwise = show index
328 We used to add a '$m' to indicate a method, but that gives rise to bad
329 error messages from the type checker when we print the function name or pattern
330 of an instance-decl binding. Why? Because the binding is zapped
331 to use the method name in place of the selector name.
332 (See TcClassDcl.tcMethodBind)
334 The way it is now, -ddump-xx output may look confusing, but
335 you can always say -dppr-debug to get the uniques.
337 However, we *do* have to zap the first character to be lower case,
338 because overloaded constructors (blarg) generate methods too.
339 And convert to VarName space
341 e.g. a call to constructor MkFoo where
342 data (Ord a) => Foo a = MkFoo a
344 If this is necessary, we do it by prefixing '$m'. These
345 guys never show up in error messages. What a hack.
348 mkMethodOcc :: OccName -> OccName
349 mkMethodOcc occ@(OccName VarName fs) = occ
350 mkMethodOcc occ = mk_simple_deriv varName "$m" occ
354 %************************************************************************
356 \subsection{Tidying them up}
358 %************************************************************************
360 Before we print chunks of code we like to rename it so that
361 we don't have to print lots of silly uniques in it. But we mustn't
362 accidentally introduce name clashes! So the idea is that we leave the
363 OccName alone unless it accidentally clashes with one that is already
364 in scope; if so, we tack on '1' at the end and try again, then '2', and
365 so on till we find a unique one.
367 There's a wrinkle for operators. Consider '>>='. We can't use '>>=1'
368 because that isn't a single lexeme. So we encode it to 'lle' and *then*
369 tack on the '1', if necessary.
372 type TidyOccEnv = FiniteMap FAST_STRING Int -- The in-scope OccNames
373 emptyTidyOccEnv = emptyFM
375 initTidyOccEnv :: [OccName] -> TidyOccEnv -- Initialise with names to avoid!
376 initTidyOccEnv = foldl (\env (OccName _ fs) -> addToFM env fs 1) emptyTidyOccEnv
378 tidyOccName :: TidyOccEnv -> OccName -> (TidyOccEnv, OccName)
380 tidyOccName in_scope occ@(OccName occ_sp fs)
381 | not (fs `elemFM` in_scope)
382 = (addToFM in_scope fs 1, occ) -- First occurrence
384 | otherwise -- Already occurs
385 = go in_scope (_UNPK_ fs)
388 go in_scope str = case lookupFM in_scope pk_str of
389 Just n -> go (addToFM in_scope pk_str (n+1)) (str ++ show n)
390 -- Need to go round again, just in case "t3" (say)
391 -- clashes with a "t3" that's already in scope
393 Nothing -> (addToFM in_scope pk_str 1, mkSysOccFS occ_sp pk_str)
400 %************************************************************************
402 \subsection{The 'Z' encoding}
404 %************************************************************************
406 This is the main name-encoding and decoding function. It encodes any
407 string into a string that is acceptable as a C name. This is the name
408 by which things are known right through the compiler.
410 The basic encoding scheme is this.
412 * Tuples (,,,) are coded as Z3T
414 * Alphabetic characters (upper and lower), digits, and '_'
415 all translate to themselves;
416 except 'Z', which translates to 'ZZ'
417 and 'z', which translates to 'zz'
418 We need both so that we can preserve the variable/tycon distinction
420 * Most other printable characters translate to 'Zx' for some
421 alphabetic character x
423 * The others translate as 'Zxdd' where 'dd' is exactly two hexadecimal
424 digits for the ord of the character
427 --------------------------
442 -- alreadyEncoded is used in ASSERTs to check for encoded
443 -- strings. It isn't fail-safe, of course, because, say 'zh' might
444 -- be encoded or not.
445 alreadyEncoded :: String -> Bool
446 alreadyEncoded s = all ok s
449 ok ch = ISALPHANUM ch
451 alreadyEncodedFS :: FAST_STRING -> Bool
452 alreadyEncodedFS fs = alreadyEncoded (_UNPK_ fs)
454 encode :: UserString -> EncodedString
455 encode cs = case maybe_tuple cs of
456 Just n -> 'Z' : show n ++ "T" -- Tuples go to Z2T etc
460 go (c:cs) = encode_ch c ++ go cs
462 -- ToDo: Unboxed tuples too, perhaps?
463 maybe_tuple ('(' : cs) = check_tuple (0::Int) cs
464 maybe_tuple other = Nothing
466 check_tuple :: Int -> String -> Maybe Int
467 check_tuple n (',' : cs) = check_tuple (n+1) cs
468 check_tuple n ")" = Just n
469 check_tuple n other = Nothing
471 encodeFS :: UserFS -> EncodedFS
472 encodeFS fast_str | all unencodedChar str = fast_str
473 | otherwise = _PK_ (encode str)
475 str = _UNPK_ fast_str
477 unencodedChar :: Char -> Bool -- True for chars that don't need encoding
478 unencodedChar '_' = True
479 unencodedChar 'Z' = False
480 unencodedChar 'z' = False
481 unencodedChar c = ISALPHANUM c
483 encode_ch :: Char -> EncodedString
484 encode_ch c | unencodedChar c = [c] -- Common case first
487 encode_ch '(' = "ZL" -- Needed for things like (,), and (->)
488 encode_ch ')' = "ZR" -- For symmetry with (
507 encode_ch '\'' = "zq"
508 encode_ch '\\' = "zr"
513 encode_ch c = ['z', 'x', intToDigit hi, intToDigit lo]
515 (hi,lo) = ord c `quotRem` 16
518 Decode is used for user printing.
521 decodeFS :: FAST_STRING -> FAST_STRING
522 decodeFS fs = _PK_ (decode (_UNPK_ fs))
524 decode :: EncodedString -> UserString
526 decode ('Z' : rest) = decode_escape rest
527 decode ('z' : rest) = decode_escape rest
528 decode (c : rest) = c : decode rest
530 decode_escape :: EncodedString -> UserString
532 decode_escape ('Z' : rest) = 'Z' : decode rest
533 decode_escape ('C' : rest) = ':' : decode rest
534 decode_escape ('L' : rest) = '(' : decode rest
535 decode_escape ('R' : rest) = ')' : decode rest
536 decode_escape ('M' : rest) = '[' : decode rest
537 decode_escape ('N' : rest) = ']' : decode rest
539 decode_escape ('z' : rest) = 'z' : decode rest
540 decode_escape ('a' : rest) = '&' : decode rest
541 decode_escape ('b' : rest) = '|' : decode rest
542 decode_escape ('d' : rest) = '$' : decode rest
543 decode_escape ('e' : rest) = '=' : decode rest
544 decode_escape ('g' : rest) = '>' : decode rest
545 decode_escape ('h' : rest) = '#' : decode rest
546 decode_escape ('i' : rest) = '.' : decode rest
547 decode_escape ('l' : rest) = '<' : decode rest
548 decode_escape ('m' : rest) = '-' : decode rest
549 decode_escape ('n' : rest) = '!' : decode rest
550 decode_escape ('p' : rest) = '+' : decode rest
551 decode_escape ('q' : rest) = '\'' : decode rest
552 decode_escape ('r' : rest) = '\\' : decode rest
553 decode_escape ('s' : rest) = '/' : decode rest
554 decode_escape ('t' : rest) = '*' : decode rest
555 decode_escape ('u' : rest) = '^' : decode rest
556 decode_escape ('v' : rest) = '%' : decode rest
557 decode_escape ('x' : d1 : d2 : rest) = chr (digitToInt d1 * 16 + digitToInt d2) : decode rest
559 -- Tuples are coded as Z23T
560 decode_escape (c : rest)
561 | isDigit c = go (digitToInt c) rest
563 go n (c : rest) | isDigit c = go (10*n + digitToInt c) rest
564 go n ('T' : rest) = '(' : replicate n ',' ++ ')' : decode rest
565 go n other = pprPanic "decode_escape" (ppr n <+> text (c:rest))
567 decode_escape (c : rest) = pprTrace "decode_escape" (char c) (decode rest)
571 %************************************************************************
573 n\subsection{Lexical categories}
575 %************************************************************************
577 These functions test strings to see if they fit the lexical categories
578 defined in the Haskell report.
581 isLexCon, isLexVar, isLexId, isLexSym :: FAST_STRING -> Bool
582 isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FAST_STRING -> Bool
584 isLexCon cs = isLexConId cs || isLexConSym cs
585 isLexVar cs = isLexVarId cs || isLexVarSym cs
587 isLexId cs = isLexConId cs || isLexVarId cs
588 isLexSym cs = isLexConSym cs || isLexVarSym cs
592 isLexConId cs -- Prefix type or data constructors
593 | _NULL_ cs = False -- e.g. "Foo", "[]", "(,)"
594 | cs == SLIT("[]") = True
595 | c == '(' = True -- (), (,), (,,), ...
596 | otherwise = isUpper c || isUpperISO c
600 isLexVarId cs -- Ordinary prefix identifiers
601 | _NULL_ cs = False -- e.g. "x", "_x"
602 | otherwise = isLower c || isLowerISO c || c == '_'
606 isLexConSym cs -- Infix type or data constructors
607 | _NULL_ cs = False -- e.g. ":-:", ":", "->"
608 | otherwise = c == ':'
613 isLexVarSym cs -- Infix identifiers
614 | _NULL_ cs = False -- e.g. "+"
615 | otherwise = isSymbolASCII c
621 isSymbolASCII c = c `elem` "!#$%&*+./<=>?@\\^|~-"
622 isSymbolISO c = ord c `elem` (0xd7 : 0xf7 : [0xa1 .. 0xbf])
623 isUpperISO (C# c#) = c# `geChar#` '\xc0'# && c# `leChar#` '\xde'# && c# `neChar#` '\xd7'#
624 --0xc0 <= oc && oc <= 0xde && oc /= 0xd7 where oc = ord c
625 isLowerISO (C# c#) = c# `geChar#` '\xdf'# && c# `leChar#` '\xff'# && c# `neChar#` '\xf7'#
626 --0xdf <= oc && oc <= 0xff && oc /= 0xf7 where oc = ord c