X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Fcompiler%2FbasicTypes%2FOccName.lhs;h=faf7aa8988c94c6dbe4d0c10cf49823a73569fc5;hb=70b66a28aa5ef80bababb1ca239b2c69e1362147;hp=e2da548cb2b593caf4e5d812b02591b5a8e1adf9;hpb=1f3a9da2d3b6dcaf0634b3d4f5dbe54f069eebb8;p=ghc-hetmet.git diff --git a/ghc/compiler/basicTypes/OccName.lhs b/ghc/compiler/basicTypes/OccName.lhs index e2da548..faf7aa8 100644 --- a/ghc/compiler/basicTypes/OccName.lhs +++ b/ghc/compiler/basicTypes/OccName.lhs @@ -7,19 +7,20 @@ \begin{code} module OccName ( -- The NameSpace type; abstact - NameSpace, tcName, clsName, tcClsName, dataName, varName, ipName, - tvName, uvName, nameSpaceString, + NameSpace, tcName, clsName, tcClsName, dataName, varName, + tvName, nameSpaceString, -- The OccName type OccName, -- Abstract, instance of Outputable pprOccName, - mkSrcOccFS, mkSysOcc, mkSysOccFS, mkCCallOcc, mkSrcVarOcc, mkKindOccFS, + mkOccFS, mkSysOcc, mkSysOccFS, mkFCallOcc, mkVarOcc, mkKindOccFS, mkSuperDictSelOcc, mkDFunOcc, mkForeignExportOcc, mkDictOcc, mkIPOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc, mkDerivedTyConOcc, mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc, + mkGenOcc1, mkGenOcc2, mkLocalOcc, - isSysOcc, isTvOcc, isUvOcc, isDataOcc, isDataSymOcc, isSymOcc, isIPOcc, isValOcc, + isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc, occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour, setOccNameSpace, @@ -39,8 +40,9 @@ module OccName ( #include "HsVersions.h" -import Char ( isDigit, isAlpha, isUpper, isLower, ISALPHANUM, ord, chr, digitToInt, intToDigit ) +import Char ( isDigit, isUpper, isLower, isAlphaNum, ord, chr, digitToInt ) import Util ( thenCmp ) +import Unique ( Unique ) import FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM ) import Outputable import GlaExts @@ -82,10 +84,8 @@ pprEncodedFS fs \begin{code} data NameSpace = VarName -- Variables - | IPName -- Implicit Parameters | DataName -- Data constructors | TvName -- Type variables - | UvName -- Usage variables | TcClsName -- Type constructors and classes; Haskell has them -- in the same name space for now. deriving( Eq, Ord ) @@ -98,17 +98,13 @@ tcClsName = TcClsName -- Not sure which! dataName = DataName tvName = TvName -uvName = UvName varName = VarName -ipName = IPName nameSpaceString :: NameSpace -> String nameSpaceString DataName = "Data constructor" nameSpaceString VarName = "Variable" -nameSpaceString IPName = "Implicit Param" nameSpaceString TvName = "Type variable" -nameSpaceString UvName = "Usage variable" nameSpaceString TcClsName = "Type constructor or class" \end{code} @@ -169,14 +165,14 @@ mkSysOccFS :: NameSpace -> EncodedFS -> OccName mkSysOccFS occ_sp fs = ASSERT2( alreadyEncodedFS fs, ppr fs ) OccName occ_sp fs -mkCCallOcc :: EncodedString -> OccName +mkFCallOcc :: EncodedString -> OccName -- This version of mkSysOcc doesn't check that the string is already encoded, -- because it will be something like "{__ccall f dyn Int# -> Int#}" -- This encodes a lot into something that then parses like an Id. -- But then alreadyEncoded complains about the braces! -mkCCallOcc str = OccName varName (_PK_ str) +mkFCallOcc str = OccName varName (_PK_ str) --- Kind constructors get a speical function. Uniquely, they are not encoded, +-- Kind constructors get a special function. Uniquely, they are not encoded, -- so that they have names like '*'. This means that *even in interface files* -- we'll get kinds like (* -> (* -> *)). We can't use mkSysOcc because it -- has an ASSERT that doesn't hold. @@ -187,11 +183,11 @@ mkKindOccFS occ_sp fs = OccName occ_sp fs *Source-code* things are encoded. \begin{code} -mkSrcOccFS :: NameSpace -> UserFS -> OccName -mkSrcOccFS occ_sp fs = mkSysOccFS occ_sp (encodeFS fs) +mkOccFS :: NameSpace -> UserFS -> OccName +mkOccFS occ_sp fs = mkSysOccFS occ_sp (encodeFS fs) -mkSrcVarOcc :: UserFS -> OccName -mkSrcVarOcc fs = mkSysOccFS varName (encodeFS fs) +mkVarOcc :: UserFS -> OccName +mkVarOcc fs = mkSysOccFS varName (encodeFS fs) \end{code} @@ -224,13 +220,13 @@ occNameFlavour (OccName sp _) = nameSpaceString sp \end{code} \begin{code} -isTvOcc, isDataSymOcc, isSymOcc, isUvOcc :: OccName -> Bool +isTvOcc, isDataSymOcc, isSymOcc, isTcOcc :: OccName -> Bool isTvOcc (OccName TvName _) = True isTvOcc other = False -isUvOcc (OccName UvName _) = True -isUvOcc other = False +isTcOcc (OccName TcClsName _) = True +isTcOcc other = False isValOcc (OccName VarName _) = True isValOcc (OccName DataName _) = True @@ -248,9 +244,6 @@ isDataOcc other = False -- Pretty inefficient! isSymOcc (OccName DataName s) = isLexConSym (decodeFS s) isSymOcc (OccName VarName s) = isLexSym (decodeFS s) - -isIPOcc (OccName IPName _) = True -isIPOcc _ = False \end{code} @@ -269,25 +262,26 @@ Here's our convention for splitting up the interface file name space: $dm... default methods $p... superclass selectors $w... workers - $T... compiler-generated tycons for dictionaries - $D... ...ditto data cons + :T... compiler-generated tycons for dictionaries + :D... ...ditto data cons $sf.. specialised version of f in encoded form these appear as Zdfxxx etc :... keywords (export:, letrec: etc.) +--- I THINK THIS IS WRONG! This knowledge is encoded in the following functions. -@mk_deriv@ generates an @OccName@ from the one-char prefix and a string. +@mk_deriv@ generates an @OccName@ from the prefix and a string. NB: The string must already be encoded! \begin{code} mk_deriv :: NameSpace -> String -- Distinguishes one sort of derived name from another -> EncodedString -- Must be already encoded!! We don't want to encode it a - -- second time because encoding isn't itempotent + -- second time because encoding isn't idempotent -> OccName mk_deriv occ_sp sys_prefix str = mkSysOcc occ_sp (encode sys_prefix ++ str) @@ -299,48 +293,42 @@ mkDictOcc, mkIPOcc, mkWorkerOcc, mkDefaultMethodOcc, :: OccName -> OccName -- These derived variables have a prefix that no Haskell value could have -mkWorkerOcc = mk_simple_deriv varName "$w" -mkDefaultMethodOcc = mk_simple_deriv varName "$dm" -mkDerivedTyConOcc = mk_simple_deriv tcName ":" -- The : prefix makes sure it classifies -mkClassTyConOcc = mk_simple_deriv tcName ":T" -- as a tycon/datacon -mkClassDataConOcc = mk_simple_deriv dataName ":D" -- -mkDictOcc = mk_simple_deriv varName "$d" -mkIPOcc = mk_simple_deriv varName "$i" -mkSpecOcc = mk_simple_deriv varName "$s" -mkForeignExportOcc = mk_simple_deriv varName "$f" - +mkWorkerOcc = mk_simple_deriv varName "$w" +mkDefaultMethodOcc = mk_simple_deriv varName "$dm" +mkDerivedTyConOcc = mk_simple_deriv tcName ":" -- The : prefix makes sure it classifies +mkClassTyConOcc = mk_simple_deriv tcName ":T" -- as a tycon/datacon +mkClassDataConOcc = mk_simple_deriv dataName ":D" -- +mkDictOcc = mk_simple_deriv varName "$d" +mkIPOcc = mk_simple_deriv varName "$i" +mkSpecOcc = mk_simple_deriv varName "$s" +mkForeignExportOcc = mk_simple_deriv varName "$f" +mkGenOcc1 = mk_simple_deriv varName "$gfrom" -- Generics +mkGenOcc2 = mk_simple_deriv varName "$gto" -- Generics mk_simple_deriv sp px occ = mk_deriv sp px (occNameString occ) - - -isSysOcc :: OccName -> Bool -- True for all these '$' things -isSysOcc occ = case occNameUserString occ of - ('$' : _ ) -> True - other -> False -- We don't care about the ':' ones - -- isSysOcc is only called for Ids anyway \end{code} \begin{code} mkSuperDictSelOcc :: Int -- Index of superclass, eg 3 -> OccName -- Class, eg "Ord" - -> OccName -- eg "p3Ord" + -> OccName -- eg "$p3Ord" mkSuperDictSelOcc index cls_occ = mk_deriv varName "$p" (show index ++ occNameString cls_occ) + +mkLocalOcc :: Unique -- Unique + -> OccName -- Local name (e.g. "sat") + -> OccName -- Nice unique version ("$L23sat") +mkLocalOcc uniq occ + = mk_deriv varName ("$L" ++ show uniq) (occNameString occ) + -- The Unique might print with characters + -- that need encoding (e.g. 'z'!) \end{code} \begin{code} mkDFunOcc :: EncodedString -- Typically the class and type glommed together e.g. "OrdMaybe" - -> Int -- Unique to distinguish dfuns which share the previous two - -- eg 3 - -- The requirement is that the (string,index) pair be unique in this module - - -> OccName -- "$fOrdMaybe3" + -> OccName -- "$fOrdMaybe" -mkDFunOcc string index - = mk_deriv VarName "$f" (show_index ++ string) - where - show_index | index == 0 = "" - | otherwise = show index +mkDFunOcc string = mk_deriv VarName "$f" string \end{code} We used to add a '$m' to indicate a method, but that gives rise to bad @@ -438,8 +426,8 @@ The basic encoding scheme is this. * Most other printable characters translate to 'zx' or 'Zx' for some alphabetic character x -* The others translate as 'zxdd' where 'dd' is exactly two hexadecimal - digits for the ord of the character +* The others translate as 'znnnU' where 'nnn' is the decimal number + of the character Before After -------------------------- @@ -451,10 +439,12 @@ The basic encoding scheme is this. foo## foozhzh foo##1 foozhzh1 fooZ fooZZ - :+ Zczp - () Z0T - (,,,,) Z4T - + :+ ZCzp + () Z0T 0-tuple + (,,,,) Z5T 5-tuple + (# #) Z1H unboxed 1-tuple (note the space) + (#,,,,#) Z5H unboxed 5-tuple + (NB: There is no Z1T nor Z0H.) \begin{code} -- alreadyEncoded is used in ASSERTs to check for encoded @@ -468,27 +458,32 @@ alreadyEncoded s = all ok s -- in names we'd have to encode them. But we do put -- spaces in ccall "occurrences", and we don't want to -- reject them here - ok ch = ISALPHANUM ch + ok ch = isAlphaNum ch alreadyEncodedFS :: FAST_STRING -> Bool alreadyEncodedFS fs = alreadyEncoded (_UNPK_ fs) encode :: UserString -> EncodedString encode cs = case maybe_tuple cs of - Just n -> 'Z' : show n ++ "T" -- Tuples go to Z2T etc + Just n -> n -- Tuples go to Z2T etc Nothing -> go cs where go [] = [] go (c:cs) = encode_ch c ++ go cs --- ToDo: Unboxed tuples too, perhaps? -maybe_tuple ('(' : cs) = check_tuple (0::Int) cs -maybe_tuple other = Nothing +maybe_tuple "(# #)" = Just("Z1H") +maybe_tuple ('(' : '#' : cs) = case count_commas (0::Int) cs of + (n, '#' : ')' : cs) -> Just ('Z' : shows (n+1) "H") + other -> Nothing +maybe_tuple "()" = Just("Z0T") +maybe_tuple ('(' : cs) = case count_commas (0::Int) cs of + (n, ')' : cs) -> Just ('Z' : shows (n+1) "T") + other -> Nothing +maybe_tuple other = Nothing -check_tuple :: Int -> String -> Maybe Int -check_tuple n (',' : cs) = check_tuple (n+1) cs -check_tuple n ")" = Just n -check_tuple n other = Nothing +count_commas :: Int -> String -> (Int, String) +count_commas n (',' : cs) = count_commas (n+1) cs +count_commas n cs = (n,cs) encodeFS :: UserFS -> EncodedFS encodeFS fast_str | all unencodedChar str = fast_str @@ -499,7 +494,9 @@ encodeFS fast_str | all unencodedChar str = fast_str unencodedChar :: Char -> Bool -- True for chars that don't need encoding unencodedChar 'Z' = False unencodedChar 'z' = False -unencodedChar c = ISALPHANUM c +unencodedChar c = c >= 'a' && c <= 'z' + || c >= 'A' && c <= 'Z' + || c >= '0' && c <= '9' encode_ch :: Char -> EncodedString encode_ch c | unencodedChar c = [c] -- Common case first @@ -532,9 +529,7 @@ encode_ch '/' = "zs" encode_ch '*' = "zt" encode_ch '_' = "zu" encode_ch '%' = "zv" -encode_ch c = ['z', 'x', intToDigit hi, intToDigit lo] - where - (hi,lo) = ord c `quotRem` 16 +encode_ch c = 'z' : shows (ord c) "U" \end{code} Decode is used for user printing. @@ -577,23 +572,28 @@ decode_escape ('s' : rest) = '/' : decode rest decode_escape ('t' : rest) = '*' : decode rest decode_escape ('u' : rest) = '_' : decode rest decode_escape ('v' : rest) = '%' : decode rest -decode_escape ('x' : d1 : d2 : rest) = chr (digitToInt d1 * 16 + digitToInt d2) : decode rest -- Tuples are coded as Z23T +-- Characters not having a specific code are coded as z224U decode_escape (c : rest) | isDigit c = go (digitToInt c) rest where go n (c : rest) | isDigit c = go (10*n + digitToInt c) rest - go n ('T' : rest) = '(' : replicate n ',' ++ ')' : decode rest + go 0 ('T' : rest) = "()" ++ (decode rest) + go n ('T' : rest) = '(' : replicate (n-1) ',' ++ ')' : decode rest + go 1 ('H' : rest) = "(# #)" ++ (decode rest) + go n ('H' : rest) = '(' : '#' : replicate (n-1) ',' ++ '#' : ')' : decode rest + go n ('U' : rest) = chr n : decode rest go n other = pprPanic "decode_escape" (ppr n <+> text (c:rest)) decode_escape (c : rest) = pprTrace "decode_escape" (char c) (decode rest) +decode_escape [] = pprTrace "decode_escape" (text "empty") "" \end{code} %************************************************************************ %* * -n\subsection{Lexical categories} +\subsection{Lexical categories} %* * %************************************************************************