X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Fcompiler%2FbasicTypes%2FOccName.lhs;h=0faa8706963f8b60c81b19e70f6f04ebf4e8aed5;hb=b749b2c7fd7fb9cdd464c213672ded760f498dc9;hp=f91453575ace96c2ec3fa97e4f166c660b74d472;hpb=83eef621e4a4fbb6c1343304ec638cafd6c9dc09;p=ghc-hetmet.git diff --git a/ghc/compiler/basicTypes/OccName.lhs b/ghc/compiler/basicTypes/OccName.lhs index f914535..0faa870 100644 --- a/ghc/compiler/basicTypes/OccName.lhs +++ b/ghc/compiler/basicTypes/OccName.lhs @@ -1,3 +1,4 @@ +{-% DrIFT (Automatic class derivations for Haskell) v1.1 %-} % % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998 % @@ -7,22 +8,26 @@ \begin{code} module OccName ( -- The NameSpace type; abstact - NameSpace, tcName, clsName, tcClsName, dataName, varName, ipName, - tvName, nameSpaceString, + NameSpace, tcName, clsName, tcClsName, dataName, varName, + tvName, srcDataName, nameSpaceString, -- The OccName type OccName, -- Abstract, instance of Outputable pprOccName, - mkOccFS, mkSysOcc, mkSysOccFS, mkCCallOcc, mkVarOcc, mkKindOccFS, + mkOccFS, mkSysOcc, mkSysOccFS, mkFCallOcc, mkKindOccFS, + mkVarOcc, mkVarOccEncoded, mkSuperDictSelOcc, mkDFunOcc, mkForeignExportOcc, mkDictOcc, mkIPOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc, mkDerivedTyConOcc, mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc, - mkGenOcc1, mkGenOcc2, + mkGenOcc1, mkGenOcc2, mkLocalOcc, + mkDataConWrapperOcc, mkDataConWorkerOcc, - isSysOcc, isTvOcc, isDataOcc, isDataSymOcc, isSymOcc, isIPOcc, isValOcc, + isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc, + reportIfUnused, - occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour, + occNameFS, occNameString, occNameUserString, occNameSpace, + occNameFlavour, briefOccNameFlavour, setOccNameSpace, -- Tidying up @@ -40,11 +45,15 @@ module OccName ( #include "HsVersions.h" -import Char ( isDigit, isUpper, isLower, ISALPHANUM, ord, chr, digitToInt ) +import Char ( isDigit, isUpper, isLower, isAlphaNum, ord, chr, digitToInt ) import Util ( thenCmp ) +import Unique ( Unique ) import FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM ) +import FastString import Outputable -import GlaExts +import Binary + +import GLAEXTS \end{code} We hold both module names and identifier names in a 'Z-encoded' form @@ -59,8 +68,8 @@ code the encoding operation is not performed on each occurrence. These type synonyms help documentation. \begin{code} -type UserFS = FAST_STRING -- As the user typed it -type EncodedFS = FAST_STRING -- Encoded form +type UserFS = FastString -- As the user typed it +type EncodedFS = FastString -- Encoded form type UserString = String -- As the user typed it type EncodedString = String -- Encoded form @@ -70,9 +79,9 @@ pprEncodedFS :: EncodedFS -> SDoc pprEncodedFS fs = getPprStyle $ \ sty -> if userStyle sty - -- ptext (decodeFS fs) would needlessly pack the string again - then text (decode (_UNPK_ fs)) - else ptext fs + -- ftext (decodeFS fs) would needlessly pack the string again + then text (decode (unpackFS fs)) + else ftext fs \end{code} %************************************************************************ @@ -82,13 +91,31 @@ pprEncodedFS fs %************************************************************************ \begin{code} -data NameSpace = VarName -- Variables - | IPName -- Implicit Parameters - | DataName -- Data constructors +data NameSpace = VarName -- Variables, including "source" data constructors + | DataName -- "Real" data constructors | TvName -- Type variables | TcClsName -- Type constructors and classes; Haskell has them -- in the same name space for now. deriving( Eq, Ord ) + {-! derive: Binary !-} + +-- Note [Data Constructors] +-- see also: Note [Data Constructor Naming] in DataCon.lhs +-- +-- "Source" data constructors are the data constructors mentioned +-- in Haskell source code +-- +-- "Real" data constructors are the data constructors of the +-- representation type, which may not be the same as the source +-- type + +-- Example: +-- data T = T !(Int,Int) +-- +-- The source datacon has type (Int,Int) -> T +-- The real datacon has type Int -> Int -> T +-- GHC chooses a representation based on the strictness etc. + -- Though type constructors and classes are in the same name space now, -- the NameSpace type is abstract, so we can easily separate them later @@ -96,16 +123,16 @@ tcName = TcClsName -- Type constructors clsName = TcClsName -- Classes tcClsName = TcClsName -- Not sure which! -dataName = DataName -tvName = TvName -varName = VarName -ipName = IPName +dataName = DataName +srcDataName = DataName -- Haskell-source data constructors should be + -- in the Data name space +tvName = TvName +varName = VarName nameSpaceString :: NameSpace -> String nameSpaceString DataName = "Data constructor" nameSpaceString VarName = "Variable" -nameSpaceString IPName = "Implicit Param" nameSpaceString TvName = "Type variable" nameSpaceString TcClsName = "Type constructor or class" \end{code} @@ -121,6 +148,7 @@ nameSpaceString TcClsName = "Type constructor or class" data OccName = OccName NameSpace EncodedFS + {-! derive : Binary !-} \end{code} @@ -161,18 +189,18 @@ already encoded \begin{code} mkSysOcc :: NameSpace -> EncodedString -> OccName mkSysOcc occ_sp str = ASSERT2( alreadyEncoded str, text str ) - OccName occ_sp (_PK_ str) + OccName occ_sp (mkFastString str) 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 (mkFastString str) -- Kind constructors get a special function. Uniquely, they are not encoded, -- so that they have names like '*'. This means that *even in interface files* @@ -190,6 +218,9 @@ mkOccFS occ_sp fs = mkSysOccFS occ_sp (encodeFS fs) mkVarOcc :: UserFS -> OccName mkVarOcc fs = mkSysOccFS varName (encodeFS fs) + +mkVarOccEncoded :: EncodedFS -> OccName +mkVarOccEncoded fs = mkSysOccFS varName fs \end{code} @@ -205,7 +236,7 @@ occNameFS :: OccName -> EncodedFS occNameFS (OccName _ s) = s occNameString :: OccName -> EncodedString -occNameString (OccName _ s) = _UNPK_ s +occNameString (OccName _ s) = unpackFS s occNameUserString :: OccName -> UserString occNameUserString occ = decode (occNameString occ) @@ -213,20 +244,33 @@ occNameUserString occ = decode (occNameString occ) occNameSpace :: OccName -> NameSpace occNameSpace (OccName sp _) = sp -setOccNameSpace :: OccName -> NameSpace -> OccName -setOccNameSpace (OccName _ occ) sp = OccName sp occ +setOccNameSpace :: NameSpace -> OccName -> OccName +setOccNameSpace sp (OccName _ occ) = OccName sp occ -- occNameFlavour is used only to generate good error messages occNameFlavour :: OccName -> String -occNameFlavour (OccName sp _) = nameSpaceString sp +occNameFlavour (OccName DataName _) = "Data constructor" +occNameFlavour (OccName TvName _) = "Type variable" +occNameFlavour (OccName TcClsName _) = "Type constructor or class" +occNameFlavour (OccName VarName s) = "Variable" + +-- briefOccNameFlavour is used in debug-printing of names +briefOccNameFlavour :: OccName -> String +briefOccNameFlavour (OccName DataName _) = "d" +briefOccNameFlavour (OccName VarName _) = "v" +briefOccNameFlavour (OccName TvName _) = "tv" +briefOccNameFlavour (OccName TcClsName _) = "tc" \end{code} \begin{code} -isTvOcc, isDataSymOcc, isSymOcc :: OccName -> Bool +isTvOcc, isDataSymOcc, isSymOcc, isTcOcc :: OccName -> Bool isTvOcc (OccName TvName _) = True isTvOcc other = False +isTcOcc (OccName TcClsName _) = True +isTcOcc other = False + isValOcc (OccName VarName _) = True isValOcc (OccName DataName _) = True isValOcc other = False @@ -234,21 +278,31 @@ isValOcc other = False -- Data constructor operator (starts with ':', or '[]') -- Pretty inefficient! isDataSymOcc (OccName DataName s) = isLexConSym (decodeFS s) +isDataSymOcc (OccName VarName s) = isLexConSym (decodeFS s) isDataSymOcc other = False isDataOcc (OccName DataName _) = True +isDataOcc (OccName VarName s) = isLexCon (decodeFS s) isDataOcc other = False -- Any operator (data constructor or variable) -- Pretty inefficient! isSymOcc (OccName DataName s) = isLexConSym (decodeFS s) isSymOcc (OccName VarName s) = isLexSym (decodeFS s) +\end{code} + -isIPOcc (OccName IPName _) = True -isIPOcc _ = False +\begin{code} +reportIfUnused :: OccName -> Bool + -- Haskell 98 encourages compilers to suppress warnings about + -- unused names in a pattern if they start with "_". +reportIfUnused occ = case occNameUserString occ of + ('_' : _) -> False + zz_other -> True \end{code} + %************************************************************************ %* * \subsection{Making system names} @@ -264,25 +318,26 @@ 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 $sf.. specialised version of f in encoded form these appear as Zdfxxx etc :... keywords (export:, letrec: etc.) +--- I THINK THIS IS WRONG! This knowledge is encoded in the following functions. -@mk_deriv@ generates an @OccName@ from the one-char prefix and a string. +@mk_deriv@ generates an @OccName@ from the prefix and a string. NB: The string must already be encoded! \begin{code} mk_deriv :: NameSpace -> String -- Distinguishes one sort of derived name from another -> EncodedString -- Must be already encoded!! We don't want to encode it a - -- second time because encoding isn't itempotent + -- second time because encoding isn't idempotent -> OccName mk_deriv occ_sp sys_prefix str = mkSysOcc occ_sp (encode sys_prefix ++ str) @@ -294,33 +349,42 @@ mkDictOcc, mkIPOcc, mkWorkerOcc, mkDefaultMethodOcc, :: OccName -> OccName -- These derived variables have a prefix that no Haskell value could have -mkWorkerOcc = mk_simple_deriv varName "$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 "$f" +mkDataConWrapperOcc = mk_simple_deriv varName "$W" +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" -- We go straight to the "real" data con + -- for datacons from classes +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 +-- Data constructor workers are made by setting the name space +-- of the data constructor OccName (which should be a DataName) +-- to DataName +mkDataConWorkerOcc datacon_occ = setOccNameSpace varName datacon_occ \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} @@ -375,7 +439,7 @@ because that isn't a single lexeme. So we encode it to 'lle' and *then* tack on the '1', if necessary. \begin{code} -type TidyOccEnv = FiniteMap FAST_STRING Int -- The in-scope OccNames +type TidyOccEnv = FiniteMap FastString Int -- The in-scope OccNames emptyTidyOccEnv = emptyFM initTidyOccEnv :: [OccName] -> TidyOccEnv -- Initialise with names to avoid! @@ -388,7 +452,7 @@ tidyOccName in_scope occ@(OccName occ_sp fs) = (addToFM in_scope fs 1, occ) -- First occurrence | otherwise -- Already occurs - = go in_scope (_UNPK_ fs) + = go in_scope (unpackFS fs) where go in_scope str = case lookupFM in_scope pk_str of @@ -399,7 +463,7 @@ tidyOccName in_scope occ@(OccName occ_sp fs) Nothing -> (addToFM in_scope pk_str 1, mkSysOccFS occ_sp pk_str) -- str is now unique where - pk_str = _PK_ str + pk_str = mkFastString str \end{code} @@ -439,10 +503,12 @@ 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 @@ -456,38 +522,45 @@ 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) +alreadyEncodedFS :: FastString -> Bool +alreadyEncodedFS fs = alreadyEncoded (unpackFS 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 - | otherwise = _PK_ (encode str) + | otherwise = mkFastString (encode str) where - str = _UNPK_ fast_str + str = unpackFS 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 @@ -526,8 +599,8 @@ encode_ch c = 'z' : shows (ord c) "U" Decode is used for user printing. \begin{code} -decodeFS :: FAST_STRING -> FAST_STRING -decodeFS fs = _PK_ (decode (_UNPK_ fs)) +decodeFS :: FastString -> FastString +decodeFS fs = mkFastString (decode (unpackFS fs)) decode :: EncodedString -> UserString decode [] = [] @@ -570,17 +643,21 @@ 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} %* * %************************************************************************ @@ -588,8 +665,8 @@ These functions test strings to see if they fit the lexical categories defined in the Haskell report. \begin{code} -isLexCon, isLexVar, isLexId, isLexSym :: FAST_STRING -> Bool -isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FAST_STRING -> Bool +isLexCon, isLexVar, isLexId, isLexSym :: FastString -> Bool +isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FastString -> Bool isLexCon cs = isLexConId cs || isLexConSym cs isLexVar cs = isLexVarId cs || isLexVarSym cs @@ -600,22 +677,22 @@ isLexSym cs = isLexConSym cs || isLexVarSym cs ------------- isLexConId cs -- Prefix type or data constructors - | _NULL_ cs = False -- e.g. "Foo", "[]", "(,)" - | cs == SLIT("[]") = True - | otherwise = startsConId (_HEAD_ cs) + | nullFastString cs = False -- e.g. "Foo", "[]", "(,)" + | cs == FSLIT("[]") = True + | otherwise = startsConId (headFS cs) isLexVarId cs -- Ordinary prefix identifiers - | _NULL_ cs = False -- e.g. "x", "_x" - | otherwise = startsVarId (_HEAD_ cs) + | nullFastString cs = False -- e.g. "x", "_x" + | otherwise = startsVarId (headFS cs) isLexConSym cs -- Infix type or data constructors - | _NULL_ cs = False -- e.g. ":-:", ":", "->" - | cs == SLIT("->") = True - | otherwise = startsConSym (_HEAD_ cs) + | nullFastString cs = False -- e.g. ":-:", ":", "->" + | cs == FSLIT("->") = True + | otherwise = startsConSym (headFS cs) isLexVarSym cs -- Infix identifiers - | _NULL_ cs = False -- e.g. "+" - | otherwise = startsVarSym (_HEAD_ cs) + | nullFastString cs = False -- e.g. "+" + | otherwise = startsVarSym (headFS cs) ------------- startsVarSym, startsVarId, startsConSym, startsConId :: Char -> Bool @@ -632,3 +709,34 @@ isUpperISO (C# c#) = c# `geChar#` '\xc0'# && c# `leChar#` '\xde'# && c# `neCh isLowerISO (C# c#) = c# `geChar#` '\xdf'# && c# `leChar#` '\xff'# && c# `neChar#` '\xf7'# --0xdf <= oc && oc <= 0xff && oc /= 0xf7 where oc = ord c \end{code} +\begin{code} +{-* Generated by DrIFT-v1.0 : Look, but Don't Touch. *-} +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) + +-- Imported from other files :- + +\end{code}