X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Fcompiler%2FbasicTypes%2FOccName.lhs;h=c10e1c4bba127c3d66e031ee4d4642bd0944c3e2;hb=1dfaee318171836b32f6b33a14231c69adfdef2f;hp=5eb623b12570a654222bf59cbcd564d7cbaef6a7;hpb=4b17269854ccf10df8b3ca1711410a5ca439ea8a;p=ghc-hetmet.git diff --git a/ghc/compiler/basicTypes/OccName.lhs b/ghc/compiler/basicTypes/OccName.lhs index 5eb623b..c10e1c4 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, - isSysOcc, isTvOcc, isUvOcc, isDataOcc, isDataSymOcc, isSymOcc, isIPOcc, isValOcc, + isTvOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc, occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour, setOccNameSpace, @@ -39,7 +40,7 @@ module OccName ( #include "HsVersions.h" -import Char ( isDigit, isAlpha, isUpper, isLower, ISALPHANUM, ord, chr, digitToInt ) +import Char ( isDigit, isUpper, isLower, ISALPHANUM, ord, chr, digitToInt ) import Util ( thenCmp ) import FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM ) import Outputable @@ -82,10 +83,8 @@ 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 +97,13 @@ 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 +164,14 @@ 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 +182,11 @@ 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} @@ -224,14 +219,11 @@ occNameFlavour (OccName sp _) = nameSpaceString sp \end{code} \begin{code} -isTvOcc, isDataSymOcc, isSymOcc, isUvOcc :: OccName -> Bool +isTvOcc, isDataSymOcc, isSymOcc :: OccName -> Bool isTvOcc (OccName TvName _) = True isTvOcc other = False -isUvOcc (OccName UvName _) = True -isUvOcc other = False - isValOcc (OccName VarName _) = True isValOcc (OccName DataName _) = True isValOcc other = False @@ -248,9 +240,6 @@ 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} @@ -269,18 +258,19 @@ 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 $sf.. specialised version of f in encoded form these appear as Zdfxxx etc :... keywords (export:, letrec: etc.) +--- I THINK THIS IS WRONG! This knowledge is encoded in the following functions. -@mk_deriv@ generates an @OccName@ from the one-char prefix and a string. +@mk_deriv@ generates an @OccName@ from the prefix and a string. NB: The string must already be encoded! \begin{code} @@ -308,15 +298,9 @@ mkDictOcc = mk_simple_deriv varName "$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 - ('$' : _ ) -> True - other -> False -- We don't care about the ':' ones - -- isSysOcc is only called for Ids anyway \end{code} \begin{code} @@ -330,17 +314,9 @@ mkSuperDictSelOcc index cls_occ \begin{code} mkDFunOcc :: EncodedString -- Typically the class and type glommed together e.g. "OrdMaybe" - -> Int -- Unique to distinguish dfuns which share the previous two - -- eg 3 - -- The requirement is that the (string,index) pair be unique in this module - - -> OccName -- "$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 @@ -451,10 +427,12 @@ 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 @@ -475,20 +453,25 @@ 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 +482,9 @@ 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 @@ -582,7 +567,10 @@ 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)) @@ -592,7 +580,7 @@ decode_escape (c : rest) = pprTrace "decode_escape" (char c) (decode rest) %************************************************************************ %* * -n\subsection{Lexical categories} +\subsection{Lexical categories} %* * %************************************************************************