X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Fcompiler%2FbasicTypes%2FOccName.lhs;h=a3661a9ab0d093eba7e43ee1574dcc9a8a8c4f33;hb=c5b03909e7c630a874f6f1abf76d28baf4b19d55;hp=4ecd069360b90634324d1d0b36fd434e400528fb;hpb=3160f854580e6d8df412c8cd34d93bae27175d67;p=ghc-hetmet.git diff --git a/ghc/compiler/basicTypes/OccName.lhs b/ghc/compiler/basicTypes/OccName.lhs index 4ecd069..a3661a9 100644 --- a/ghc/compiler/basicTypes/OccName.lhs +++ b/ghc/compiler/basicTypes/OccName.lhs @@ -1,4 +1,5 @@ - +{-% DrIFT (Automatic class derivations for Haskell) v1.1 %-} +% % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998 % @@ -6,86 +7,154 @@ \begin{code} module OccName ( - -- Modules - Module, -- Abstract, instance of Outputable - mkModule, mkModuleFS, moduleString, moduleCString, pprModule, + -- * The NameSpace type; abstact + NameSpace, tcName, clsName, tcClsName, dataName, varName, + tvName, srcDataName, + + -- ** Printing + pprNameSpace, pprNonVarNameSpace, pprNameSpaceBrief, - -- The OccName type + -- * The OccName type OccName, -- Abstract, instance of Outputable - varOcc, tcOcc, tvOcc, -- Occ constructors - srcVarOcc, srcTCOcc, srcTvOcc, -- For Occs arising from source code + pprOccName, + + -- ** Construction + mkOccName, mkOccNameFS, + mkVarOcc, mkVarOccFS, + mkTyVarOcc, + mkDFunOcc, + mkTupleOcc, + setOccNameSpace, + + -- ** Derived OccNames + mkDataConWrapperOcc, mkWorkerOcc, mkDefaultMethodOcc, mkDerivedTyConOcc, + mkClassTyConOcc, mkClassDataConOcc, mkDictOcc, mkIPOcc, + mkSpecOcc, mkForeignExportOcc, mkGenOcc1, mkGenOcc2, + mkDataTOcc, mkDataCOcc, mkDataConWorkerOcc, + mkSuperDictSelOcc, mkLocalOcc, mkMethodOcc, + + -- ** Deconstruction + occNameFS, occNameString, occNameSpace, + + isVarOcc, isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc, + parenSymOcc, reportIfUnused, isTcClsName, isVarName, + + isTupleOcc_maybe, + + -- The OccEnv type + OccEnv, emptyOccEnv, unitOccEnv, extendOccEnv, mapOccEnv, + lookupOccEnv, mkOccEnv, extendOccEnvList, elemOccEnv, + occEnvElts, foldOccEnv, plusOccEnv, plusOccEnv_C, extendOccEnv_C, + + -- The OccSet type + OccSet, emptyOccSet, unitOccSet, mkOccSet, extendOccSet, + extendOccSetList, + unionOccSets, unionManyOccSets, minusOccSet, elemOccSet, occSetElts, + foldOccSet, isEmptyOccSet, intersectOccSet, intersectsOccSet, - mkSuperDictSelOcc, mkDFunOcc, - mkDictOcc, mkWorkerOcc, mkDefaultMethodOcc, - mkClassTyConOcc, mkClassDataConOcc, - - isTvOcc, isTCOcc, isVarOcc, isConSymOcc, isConOcc, isSymOcc, - isWildCardOcc, isAnonOcc, - pprOccName, occNameString, occNameFlavour, + -- Tidying up + TidyOccEnv, emptyTidyOccEnv, tidyOccName, initTidyOccEnv, -- The basic form of names isLexCon, isLexVar, isLexId, isLexSym, isLexConId, isLexConSym, isLexVarId, isLexVarSym, - isLowerISO, isUpperISO, - - -- Tidying up - TidyOccEnv, emptyTidyOccEnv, tidyOccName, initTidyOccEnv, - - -- Junk - identToC - + startsVarSym, startsVarId, startsConSym, startsConId ) where #include "HsVersions.h" -import Char ( isAlpha, isUpper, isLower, isAlphanum, ord ) -import Util ( thenCmp ) -import FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM ) +import Util ( thenCmp ) +import Unique ( Unique, mkUnique, Uniquable(..) ) +import BasicTypes ( Boxity(..), Arity ) +import StaticFlags ( opt_PprStyle_Debug ) +import UniqFM +import UniqSet +import FastString import Outputable -import GlaExts -\end{code} +import Binary + +import GLAEXTS +import Data.Char ( isUpper, isLower, ord ) + +-- Unicode TODO: put isSymbol in libcompat +#if __GLASGOW_HASKELL__ > 604 +import Data.Char ( isSymbol ) +#else +isSymbol = const False +#endif + +\end{code} %************************************************************************ %* * -\subsection[Module]{The name of a module} +\subsection{Name space} %* * %************************************************************************ \begin{code} -data Module = Module FAST_STRING -- User and interface files - FAST_STRING -- Print this in C files - - -- The C version has quote chars Z-encoded - -instance Outputable Module where - ppr = pprModule - -instance Eq Module where - (Module m1 _) == (Module m2 _) = m1 == m2 - -instance Ord Module where - (Module m1 _) `compare` (Module m2 _) = m1 `compare` m2 - -pprModule :: Module -> SDoc -pprModule (Module real code) - = getPprStyle $ \ sty -> - if codeStyle sty then - ptext code - else - ptext real - -mkModule :: String -> Module -mkModule s = Module (_PK_ s) (identToC s) - -mkModuleFS :: FAST_STRING -> Module -mkModuleFS s = Module s (identFsToC s) - -moduleString :: Module -> String -moduleString (Module mod _) = _UNPK_ mod +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 -moduleCString :: Module -> String -moduleCString (Module _ code) = _UNPK_ code +-- 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 +tcName = TcClsName -- Type constructors +clsName = TcClsName -- Classes +tcClsName = TcClsName -- Not sure which! + +dataName = DataName +srcDataName = DataName -- Haskell-source data constructors should be + -- in the Data name space + +tvName = TvName +varName = VarName + +isTcClsName :: NameSpace -> Bool +isTcClsName TcClsName = True +isTcClsName _ = False + +isVarName :: NameSpace -> Bool -- Variables or type variables, but not constructors +isVarName TvName = True +isVarName VarName = True +isVarName other = False + +pprNameSpace :: NameSpace -> SDoc +pprNameSpace DataName = ptext SLIT("data constructor") +pprNameSpace VarName = ptext SLIT("variable") +pprNameSpace TvName = ptext SLIT("type variable") +pprNameSpace TcClsName = ptext SLIT("type constructor or class") + +pprNonVarNameSpace :: NameSpace -> SDoc +pprNonVarNameSpace VarName = empty +pprNonVarNameSpace ns = pprNameSpace ns + +pprNameSpaceBrief DataName = char 'd' +pprNameSpaceBrief VarName = char 'v' +pprNameSpaceBrief TvName = ptext SLIT("tv") +pprNameSpaceBrief TcClsName = ptext SLIT("tc") \end{code} @@ -96,44 +165,20 @@ moduleCString (Module _ code) = _UNPK_ code %************************************************************************ \begin{code} -data OccName = OccName - OccSpace - FAST_STRING -- The 'real name' - FAST_STRING -- Print this in interface files - FAST_STRING -- Print this in C/asm code - --- The OccSpace/real-name pair define the OccName --- The iface and c/asm versions are simply derived from the --- other two. They are cached here simply to avoid recomputing --- them repeatedly when printing - --- The latter two are irrelevant in RdrNames; on the other hand, --- the OccSpace field is irrelevant after RdrNames. --- So the OccName type might be refined a bit. --- It is now abstract so that's easier than before - - --- Why three print-names? --- Real Iface C --- --------------------- --- foo foo foo --- --- + + Zp Operators OK in interface files; --- 'Z' is the escape char for C names --- --- x# x# xZh Trailing # lexed ok by GHC -fglasgow-exts --- --- _foo _ufoo _ufoo Leading '_' is the escape char in interface files --- --- _vfoo _vfoo _vfoo Worker for foo --- --- _wp _wp _wp Worker for + +data OccName = OccName + { occNameSpace :: !NameSpace + , occNameFS :: !FastString + } +\end{code} + +\begin{code} +instance Eq OccName where + (OccName sp1 s1) == (OccName sp2 s2) = s1 == s2 && sp1 == sp2 -data OccSpace = VarOcc -- Variables and data constructors - | TvOcc -- Type variables - | TCOcc -- Type constructors and classes - deriving( Eq, Ord ) +instance Ord OccName where + compare (OccName sp1 s1) (OccName sp2 s2) = (s1 `compare` s2) `thenCmp` + (sp1 `compare` sp2) \end{code} @@ -145,17 +190,16 @@ data OccSpace = VarOcc -- Variables and data constructors \begin{code} instance Outputable OccName where - ppr = pprOccName + ppr = pprOccName pprOccName :: OccName -> SDoc -pprOccName (OccName space real iface code) +pprOccName (OccName sp occ) = getPprStyle $ \ sty -> - if codeStyle sty then - ptext code - else if ifaceStyle sty then - ptext iface - else - ptext real + if codeStyle sty + then ftext (zEncodeFS occ) + else ftext occ <> if debugStyle sty + then braces (pprNameSpaceBrief sp) + else empty \end{code} @@ -164,277 +208,317 @@ pprOccName (OccName space real iface code) \subsection{Construction} %* * %************************************************************************ - -*Source-code* things beginning with '_' are zapped to begin with '_u' \begin{code} -mkSrcOcc :: OccSpace -> FAST_STRING -> OccName -mkSrcOcc occ_sp real - = case _UNPK_ real of - - '_' : rest -> OccName occ_sp real (_PK_ zapped_str) (identToC zapped_str) - where - zapped_str = '_' : 'u' : rest +mkOccName :: NameSpace -> String -> OccName +mkOccName occ_sp str = OccName occ_sp (mkFastString str) - other -> OccName occ_sp real real (identFsToC real) +mkOccNameFS :: NameSpace -> FastString -> OccName +mkOccNameFS occ_sp fs = OccName occ_sp fs -srcVarOcc, srcTCOcc, srcTvOcc :: FAST_STRING -> OccName -srcVarOcc = mkSrcOcc VarOcc -srcTCOcc = mkSrcOcc TCOcc -srcTvOcc = mkSrcOcc TvOcc -\end{code} +mkVarOcc :: String -> OccName +mkVarOcc s = mkOccName varName s -However, things that don't come from Haskell source code aren't -treated specially. +mkVarOccFS :: FastString -> OccName +mkVarOccFS fs = mkOccNameFS varName fs -\begin{code} -mkOcc :: OccSpace -> String -> OccName -mkOcc occ_sp str = OccName occ_sp fs fs (identToC str) - where - fs = _PK_ str - -mkFsOcc :: OccSpace -> FAST_STRING -> OccName -mkFsOcc occ_sp real = OccName occ_sp real real (identFsToC real) - -varOcc, tcOcc, tvOcc :: FAST_STRING -> OccName -varOcc = mkFsOcc VarOcc -tcOcc = mkFsOcc TCOcc -tvOcc = mkFsOcc TvOcc +mkTyVarOcc :: FastString -> OccName +mkTyVarOcc fs = mkOccNameFS tvName fs \end{code} %************************************************************************ %* * -\subsection{Making system names} + Environments %* * %************************************************************************ -Here's our convention for splitting up the interface file name space: - - _d... dictionary identifiers +OccEnvs are used mainly for the envts in ModIfaces. - _f... dict-fun identifiers (from inst decls) - _g... ditto, when the tycon has symbols - - _t... externally visible (non-user visible) names - - _m... default methods - _n... default methods (encoded symbols, eg. <= becomes _nle) - - _p... superclass selectors - - _v... workers - _w... workers (encoded symbols) - - _x... local variables - - _u... user-defined names that previously began with '_' - - _T... compiler-generated tycons for dictionaries - _D.. ...ditto data cons - - __.... keywords (__export, __letrec etc.) - -This knowledge is encoded in the following functions. - - - - -@mkDerivedOcc@ generates an @OccName@ from an existing @OccName@; - eg: workers, derived methods - -We pass a character to use as the prefix. So, for example, - "f" gets derived to "_vf", if the prefix char is 'v' +They are efficient, because FastStrings have unique Int# keys. We assume +this key is less than 2^24, so we can make a Unique using + mkUnique ns key :: Unique +where 'ns' is a Char reprsenting the name space. This in turn makes it +easy to build an OccEnv. \begin{code} -mk_deriv :: OccSpace -> Char -> String -> OccName -mk_deriv occ_sp sys_ch str = mkOcc occ_sp ('_' : sys_ch : str) +instance Uniquable OccName where + getUnique (OccName ns fs) + = mkUnique char (I# (uniqueOfFS fs)) + where -- See notes above about this getUnique function + char = case ns of + VarName -> 'i' + DataName -> 'd' + TvName -> 'v' + TcClsName -> 't' + +type OccEnv a = UniqFM a + +emptyOccEnv :: OccEnv a +unitOccEnv :: OccName -> a -> OccEnv a +extendOccEnv :: OccEnv a -> OccName -> a -> OccEnv a +extendOccEnvList :: OccEnv a -> [(OccName, a)] -> OccEnv a +lookupOccEnv :: OccEnv a -> OccName -> Maybe a +mkOccEnv :: [(OccName,a)] -> OccEnv a +elemOccEnv :: OccName -> OccEnv a -> Bool +foldOccEnv :: (a -> b -> b) -> b -> OccEnv a -> b +occEnvElts :: OccEnv a -> [a] +extendOccEnv_C :: (a->a->a) -> OccEnv a -> OccName -> a -> OccEnv a +plusOccEnv :: OccEnv a -> OccEnv a -> OccEnv a +plusOccEnv_C :: (a->a->a) -> OccEnv a -> OccEnv a -> OccEnv a +mapOccEnv :: (a->b) -> OccEnv a -> OccEnv b + +emptyOccEnv = emptyUFM +unitOccEnv = unitUFM +extendOccEnv = addToUFM +extendOccEnvList = addListToUFM +lookupOccEnv = lookupUFM +mkOccEnv = listToUFM +elemOccEnv = elemUFM +foldOccEnv = foldUFM +occEnvElts = eltsUFM +plusOccEnv = plusUFM +plusOccEnv_C = plusUFM_C +extendOccEnv_C = addToUFM_C +mapOccEnv = mapUFM + +type OccSet = UniqFM OccName + +emptyOccSet :: OccSet +unitOccSet :: OccName -> OccSet +mkOccSet :: [OccName] -> OccSet +extendOccSet :: OccSet -> OccName -> OccSet +extendOccSetList :: OccSet -> [OccName] -> OccSet +unionOccSets :: OccSet -> OccSet -> OccSet +unionManyOccSets :: [OccSet] -> OccSet +minusOccSet :: OccSet -> OccSet -> OccSet +elemOccSet :: OccName -> OccSet -> Bool +occSetElts :: OccSet -> [OccName] +foldOccSet :: (OccName -> b -> b) -> b -> OccSet -> b +isEmptyOccSet :: OccSet -> Bool +intersectOccSet :: OccSet -> OccSet -> OccSet +intersectsOccSet :: OccSet -> OccSet -> Bool + +emptyOccSet = emptyUniqSet +unitOccSet = unitUniqSet +mkOccSet = mkUniqSet +extendOccSet = addOneToUniqSet +extendOccSetList = addListToUniqSet +unionOccSets = unionUniqSets +unionManyOccSets = unionManyUniqSets +minusOccSet = minusUniqSet +elemOccSet = elementOfUniqSet +occSetElts = uniqSetToList +foldOccSet = foldUniqSet +isEmptyOccSet = isEmptyUniqSet +intersectOccSet = intersectUniqSets +intersectsOccSet s1 s2 = not (isEmptyOccSet (s1 `intersectOccSet` s2)) \end{code} -Things are a bit more complicated if the thing is an operator; then -we must encode it into a normal identifier first. We do this in -a simple way, and use a different character prefix (one after the one -suggested). For example - "<" gets derived to "_wl", if the prefix char is 'v' - -\begin{code} -mk_enc_deriv :: OccSpace - -> Char -- The system-name-space character (see list above) - -> OccName -- The OccName from which we are deriving - -> OccName - -mk_enc_deriv occ_sp sys_ch occ - | needs_encoding real_str = mk_deriv occ_sp sys_op_ch (encode_operator real_str) - | otherwise = mk_deriv occ_sp sys_ch real_str - where - real_str = occNameString occ - sys_op_ch = succ sys_ch - - -mkDictOcc, mkWorkerOcc, mkDefaultMethodOcc, - mkClassTyConOcc, mkClassDataConOcc - :: OccName -> OccName -mkWorkerOcc = mk_enc_deriv VarOcc 'v' -- v,w -mkDefaultMethodOcc = mk_enc_deriv VarOcc 'm' -- m,n -mkClassTyConOcc = mk_enc_deriv TCOcc 'T' -- not U -mkClassDataConOcc = mk_enc_deriv VarOcc 'D' -- not E -mkDictOcc = mk_enc_deriv VarOcc 'd' -- not e -\end{code} +%************************************************************************ +%* * +\subsection{Predicates and taking them apart} +%* * +%************************************************************************ \begin{code} -mkSuperDictSelOcc :: Int -- Index of superclass, eg 3 - -> OccName -- Class, eg "Ord" - -> OccName -- eg "p3Ord" -mkSuperDictSelOcc index cls_occ - = mk_deriv VarOcc 'p' (show index ++ occNameString cls_occ) +occNameString :: OccName -> String +occNameString (OccName _ s) = unpackFS s + +setOccNameSpace :: NameSpace -> OccName -> OccName +setOccNameSpace sp (OccName _ occ) = OccName sp occ + +isVarOcc, isTvOcc, isDataSymOcc, isSymOcc, isTcOcc :: OccName -> Bool + +isVarOcc (OccName VarName _) = True +isVarOcc other = False + +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 + +-- Data constructor operator (starts with ':', or '[]') +-- Pretty inefficient! +isDataSymOcc (OccName DataName s) = isLexConSym s +isDataSymOcc (OccName VarName s) + | isLexConSym s = pprPanic "isDataSymOcc: check me" (ppr s) + -- Jan06: I don't think this should happen +isDataSymOcc other = False + +isDataOcc (OccName DataName _) = True +isDataOcc (OccName VarName s) + | isLexCon s = pprPanic "isDataOcc: check me" (ppr s) + -- Jan06: I don't think this should happen +isDataOcc other = False + +-- Any operator (data constructor or variable) +-- Pretty inefficient! +isSymOcc (OccName DataName s) = isLexConSym s +isSymOcc (OccName TcClsName s) = isLexConSym s +isSymOcc (OccName VarName s) = isLexSym s +isSymOcc other = False + +parenSymOcc :: OccName -> SDoc -> SDoc +-- Wrap parens around an operator +parenSymOcc occ doc | isSymOcc occ = parens doc + | otherwise = doc \end{code} \begin{code} -mkDFunOcc :: OccName -- class, eg "Ord" - -> OccName -- tycon (or something convenient from the instance type) - -- eg "Maybe" - -> Int -- Unique to distinguish dfuns which share the previous two - -- eg 3 - -> OccName -- "dOrdMaybe3" - -mkDFunOcc cls_occ tycon_occ index - | needs_encoding tycon_str -- Drat! Have to encode the tycon - = mk_deriv VarOcc 'g' (show_index ++ cls_str ++ encode_operator tycon_str) - | otherwise -- Normal case - = mk_deriv VarOcc 'f' (show_index ++ cls_str ++ tycon_str) - where - cls_str = occNameString cls_occ - tycon_str = occNameString tycon_occ - -- NB: if a non-operator the tycon has a trailing # we don't encode. - show_index | index == 0 = "" - | otherwise = show index +reportIfUnused :: OccName -> Bool + -- Haskell 98 encourages compilers to suppress warnings about + -- unused names in a pattern if they start with "_". +reportIfUnused occ = case occNameString occ of + ('_' : _) -> False + _other -> True \end{code} %************************************************************************ %* * -\subsection{Lexical categories} +\subsection{Making system names} %* * %************************************************************************ -These functions test strings to see if they fit the lexical categories -defined in the Haskell report. +Here's our convention for splitting up the interface file name space: -\begin{code} -isLexCon, isLexVar, isLexId, isLexSym :: FAST_STRING -> Bool -isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FAST_STRING -> Bool + d... dictionary identifiers + (local variables, so no name-clash worries) -isLexCon cs = isLexConId cs || isLexConSym cs -isLexVar cs = isLexVarId cs || isLexVarSym cs + $f... dict-fun identifiers (from inst decls) + $dm... default methods + $p... superclass selectors + $w... workers + :T... compiler-generated tycons for dictionaries + :D... ...ditto data cons + $sf.. specialised version of f -isLexId cs = isLexConId cs || isLexVarId cs -isLexSym cs = isLexConSym cs || isLexVarSym cs + in encoded form these appear as Zdfxxx etc -------------- + :... keywords (export:, letrec: etc.) +--- I THINK THIS IS WRONG! -isLexConId cs -- Prefix type or data constructors - | _NULL_ cs = False -- e.g. "Foo", "[]", "(,)" - | cs == SLIT("[]") = True - | c == '(' = True -- (), (,), (,,), ... - | otherwise = isUpper c || isUpperISO c - where - c = _HEAD_ cs +This knowledge is encoded in the following functions. -isLexVarId cs -- Ordinary prefix identifiers - | _NULL_ cs = False -- e.g. "x", "_x" - | otherwise = isLower c || isLowerISO c || c == '_' - where - c = _HEAD_ cs -isLexConSym cs -- Infix type or data constructors - | _NULL_ cs = False -- e.g. ":-:", ":", "->" - | otherwise = c == ':' - || cs == SLIT("->") - where - c = _HEAD_ cs +@mk_deriv@ generates an @OccName@ from the prefix and a string. +NB: The string must already be encoded! -isLexVarSym cs -- Infix identifiers - | _NULL_ cs = False -- e.g. "+" - | otherwise = isSymbolASCII c - || isSymbolISO c - where - c = _HEAD_ cs +\begin{code} +mk_deriv :: NameSpace + -> String -- Distinguishes one sort of derived name from another + -> String + -> OccName -------------- -isSymbolASCII c = c `elem` "!#$%&*+./<=>?@\\^|~-" -isSymbolISO c = ord c `elem` (0xd7 : 0xf7 : [0xa1 .. 0xbf]) -isUpperISO (C# c#) = c# `geChar#` '\xc0'# && c# `leChar#` '\xde'# && c# `neChar#` '\xd7'# - --0xc0 <= oc && oc <= 0xde && oc /= 0xd7 where oc = ord c -isLowerISO (C# c#) = c# `geChar#` '\xdf'# && c# `leChar#` '\xff'# && c# `neChar#` '\xf7'# - --0xdf <= oc && oc <= 0xff && oc /= 0xf7 where oc = ord c +mk_deriv occ_sp sys_prefix str = mkOccName occ_sp (sys_prefix ++ str) \end{code} -%************************************************************************ -%* * -\subsection{Predicates and taking them apart} -%* * -%************************************************************************ - -\begin{code} -occNameString :: OccName -> String -occNameString (OccName _ s _ _) = _UNPK_ s - --- occNameFlavour is used only to generate good error messages, so it doesn't matter --- that the VarOcc case isn't mega-efficient. We could have different Occ constructors for --- data constructors and values, but that makes everything else a bit more complicated. -occNameFlavour :: OccName -> String -occNameFlavour (OccName VarOcc s _ _) | isLexConId s = "Data constructor" - | otherwise = "Value" -occNameFlavour (OccName TvOcc _ _ _) = "Type variable" -occNameFlavour (OccName TCOcc s _ _) = "Type constructor or class" - -isVarOcc, isTCOcc, isTvOcc, - isConSymOcc, isSymOcc, isWildCardOcc :: OccName -> Bool - -isVarOcc (OccName VarOcc _ _ _) = True -isVarOcc other = False +\begin{code} +mkDataConWrapperOcc, mkWorkerOcc, mkDefaultMethodOcc, mkDerivedTyConOcc, + mkClassTyConOcc, mkClassDataConOcc, mkDictOcc, mkIPOcc, + mkSpecOcc, mkForeignExportOcc, mkGenOcc1, mkGenOcc2, + mkDataTOcc, mkDataCOcc, mkDataConWorkerOcc + :: OccName -> OccName -isTvOcc (OccName TvOcc _ _ _) = True -isTvOcc other = False +-- These derived variables have a prefix that no Haskell value could have +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 "$f" + +-- Generic derivable classes +mkGenOcc1 = mk_simple_deriv varName "$gfrom" +mkGenOcc2 = mk_simple_deriv varName "$gto" + +-- data T = MkT ... deriving( Data ) needs defintions for +-- $tT :: Data.Generics.Basics.DataType +-- $cMkT :: Data.Generics.Basics.Constr +mkDataTOcc = mk_simple_deriv varName "$t" +mkDataCOcc = mk_simple_deriv varName "$c" + +mk_simple_deriv sp px occ = mk_deriv sp px (occNameString occ) + +-- Data constructor workers are made by setting the name space +-- of the data constructor OccName (which should be a DataName) +-- to VarName +mkDataConWorkerOcc datacon_occ = setOccNameSpace varName datacon_occ +\end{code} -isTCOcc (OccName TCOcc _ _ _) = True -isTCOcc other = False +\begin{code} +mkSuperDictSelOcc :: Int -- Index of superclass, eg 3 + -> OccName -- Class, eg "Ord" + -> OccName -- eg "$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 ("$L" ++ show uniq) (occNameString occ) + -- The Unique might print with characters + -- that need encoding (e.g. 'z'!) +\end{code} -isConSymOcc (OccName _ s _ _) = isLexConSym s -isSymOcc (OccName _ s _ _) = isLexSym s +\begin{code} +mkDFunOcc :: String -- Typically the class and type glommed together e.g. "OrdMaybe" + -- Only used in debug mode, for extra clarity + -> Bool -- True <=> hs-boot instance dfun + -> Int -- Unique index + -> OccName -- "$f3OrdMaybe" + +-- In hs-boot files we make dict funs like $fx7ClsTy, which get bound to the real +-- thing when we compile the mother module. Reason: we don't know exactly +-- what the mother module will call it. + +mkDFunOcc info_str is_boot index + = mk_deriv VarName prefix string + where + prefix | is_boot = "$fx" + | otherwise = "$f" + string | opt_PprStyle_Debug = show index ++ info_str + | otherwise = show index +\end{code} -isConOcc (OccName _ s _ _) = isLexCon s +We used to add a '$m' to indicate a method, but that gives rise to bad +error messages from the type checker when we print the function name or pattern +of an instance-decl binding. Why? Because the binding is zapped +to use the method name in place of the selector name. +(See TcClassDcl.tcMethodBind) -isWildCardOcc (OccName _ s _ _) = (_HEAD_ s) == '_' && _LENGTH_ s == 1 +The way it is now, -ddump-xx output may look confusing, but +you can always say -dppr-debug to get the uniques. -isAnonOcc (OccName _ s _ _) = (_HEAD_ s) == '_' -\end{code} +However, we *do* have to zap the first character to be lower case, +because overloaded constructors (blarg) generate methods too. +And convert to VarName space +e.g. a call to constructor MkFoo where + data (Ord a) => Foo a = MkFoo a -%************************************************************************ -%* * -\subsection{Comparison} -%* * -%************************************************************************ - -Comparison is done by space and 'real' name +If this is necessary, we do it by prefixing '$m'. These +guys never show up in error messages. What a hack. \begin{code} -instance Eq OccName where - a == b = case (a `compare` b) of { EQ -> True; _ -> False } - a /= b = case (a `compare` b) of { EQ -> False; _ -> True } - -instance Ord OccName where - a <= b = case (a `compare` b) of { LT -> True; EQ -> True; GT -> False } - a < b = case (a `compare` b) of { LT -> True; EQ -> False; GT -> False } - a >= b = case (a `compare` b) of { LT -> False; EQ -> True; GT -> True } - a > b = case (a `compare` b) of { LT -> False; EQ -> False; GT -> True } - - compare (OccName sp1 r1 _ _) (OccName sp2 r2 _ _) - = (sp1 `compare` sp2) `thenCmp` (r1 `compare` r2) +mkMethodOcc :: OccName -> OccName +mkMethodOcc occ@(OccName VarName fs) = occ +mkMethodOcc occ = mk_simple_deriv varName "$m" occ \end{code} @@ -456,148 +540,137 @@ 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 -emptyTidyOccEnv = emptyFM +type TidyOccEnv = OccEnv Int -- The in-scope OccNames + -- Range gives a plausible starting point for new guesses + +emptyTidyOccEnv = emptyOccEnv initTidyOccEnv :: [OccName] -> TidyOccEnv -- Initialise with names to avoid! -initTidyOccEnv = foldl (\env (OccName _ fs _ _) -> addToFM env fs 1) emptyTidyOccEnv +initTidyOccEnv = foldl (\env occ -> extendOccEnv env occ 1) emptyTidyOccEnv tidyOccName :: TidyOccEnv -> OccName -> (TidyOccEnv, OccName) -tidyOccName in_scope occ@(OccName occ_sp real _ _) - | not (real `elemFM` in_scope) && - not (isLexCon real) -- Hack alert! Specialised versions of overloaded - -- constructors end up as ordinary Ids, but we don't - -- want them as ConIds in interface files. - - = (addToFM in_scope real 1, occ) -- First occurrence +tidyOccName in_scope occ@(OccName occ_sp fs) + = case lookupOccEnv in_scope occ of + Nothing -> -- Not already used: make it used + (extendOccEnv in_scope occ 1, occ) - | otherwise -- Already occurs - = -- First encode, to deal with - -- a) operators, and - -- b) trailing # signs - -- so that we can then append '1', '2', etc - go in_scope (encode_operator (_UNPK_ real)) - where - - go in_scope str = case lookupFM in_scope pk_str of - Just n -> go (addToFM in_scope pk_str (n+1)) (str ++ show n) - -- Need to go round again, just in case "t3" (say) - -- clashes with a "t3" that's already in scope - - Nothing -> (addToFM in_scope pk_str 1, mkFsOcc occ_sp pk_str) - -- str is now unique - where - pk_str = _PK_ str + Just n -> -- Already used: make a new guess, + -- change the guess base, and try again + tidyOccName (extendOccEnv in_scope occ (n+1)) + (mkOccName occ_sp (unpackFS fs ++ show n)) \end{code} - %************************************************************************ %* * -\subsection{Encoding for operators in derived names} + Stuff for dealing with tuples %* * %************************************************************************ -See comments with mk_enc_deriv - \begin{code} -needs_encoding :: String -> Bool -- Needs encoding when embedded in a derived name - -- Just look at the first character -needs_encoding (c:cs) = not (isAlpha c || c == '_') - -encode_operator :: String -> String -encode_operator nm = foldr tran "" nm - where - tran c cs = case trChar c of - '\0' -> '_' : show (ord c) ++ cs -- No translation - tr_c -> tr_c : cs - - trChar '&' = 'a' - trChar '|' = 'b' - trChar ':' = 'c' - trChar '/' = 'd' - trChar '=' = 'e' - trChar '>' = 'g' - trChar '#' = 'h' - trChar '@' = 'i' - trChar '<' = 'l' - trChar '-' = 'm' - trChar '!' = 'n' - trChar '+' = 'p' - trChar '\'' = 'q' - trChar '$' = 'r' - trChar '?' = 's' - trChar '*' = 't' - trChar '_' = 'u' - trChar '.' = 'v' - trChar '\\' = 'w' - trChar '%' = 'x' - trChar '~' = 'y' - trChar '^' = 'z' - trChar _ = '\0' -- No translation +mkTupleOcc :: NameSpace -> Boxity -> Arity -> OccName +mkTupleOcc ns bx ar = OccName ns (mkFastString str) + where + -- no need to cache these, the caching is done in the caller + -- (TysWiredIn.mk_tuple) + str = case bx of + Boxed -> '(' : commas ++ ")" + Unboxed -> '(' : '#' : commas ++ "#)" + + commas = take (ar-1) (repeat ',') + +isTupleOcc_maybe :: OccName -> Maybe (NameSpace, Boxity, Arity) +-- Tuples are special, because there are so many of them! +isTupleOcc_maybe (OccName ns fs) + = case unpackFS fs of + '(':'#':',':rest -> Just (ns, Unboxed, 2 + count_commas rest) + '(':',':rest -> Just (ns, Boxed, 2 + count_commas rest) + _other -> Nothing + where + count_commas (',':rest) = 1 + count_commas rest + count_commas _ = 0 \end{code} - %************************************************************************ %* * -\subsection{The 'Z' encoding} +\subsection{Lexical categories} %* * %************************************************************************ -We provide two interfaces for efficiency. +These functions test strings to see if they fit the lexical categories +defined in the Haskell report. \begin{code} -identToC :: String -> FAST_STRING -identToC str - | all isAlphanum str && not std = _PK_ str - | std = _PK_ ("Zs" ++ encode str) - | otherwise = _PK_ (encode str) - where - std = has_std_prefix str +isLexCon, isLexVar, isLexId, isLexSym :: FastString -> Bool +isLexConId, isLexConSym, isLexVarId, isLexVarSym :: FastString -> Bool -identFsToC :: FAST_STRING -> FAST_STRING -identFsToC fast_str - | all isAlphanum str && not std = fast_str - | std = _PK_ ("Zs" ++ encode str) - | otherwise = _PK_ (encode str) - where - std = has_std_prefix str - str = _UNPK_ fast_str - --- avoid "stdin", "stdout", and "stderr"... -has_std_prefix ('s':'t':'d':_) = True -has_std_prefix _ = False - -encode :: String -> String -encode [] = [] -encode (c:cs) = encode_ch c ++ encode cs - -encode_ch :: Char -> String -encode_ch c | isAlphanum c = [c] - -- Common case first -encode_ch 'Z' = "ZZ" -encode_ch '&' = "Za" -encode_ch '|' = "Zb" -encode_ch ':' = "Zc" -encode_ch '/' = "Zd" -encode_ch '=' = "Ze" -encode_ch '>' = "Zg" -encode_ch '#' = "Zh" -encode_ch '<' = "Zl" -encode_ch '-' = "Zm" -encode_ch '!' = "Zn" -encode_ch '.' = "Zs" -encode_ch '\'' = "Zq" -encode_ch '*' = "Zt" -encode_ch '+' = "Zp" -encode_ch '_' = "_" -encode_ch c = 'Z':show (ord c) +isLexCon cs = isLexConId cs || isLexConSym cs +isLexVar cs = isLexVarId cs || isLexVarSym cs + +isLexId cs = isLexConId cs || isLexVarId cs +isLexSym cs = isLexConSym cs || isLexVarSym cs + +------------- + +isLexConId cs -- Prefix type or data constructors + | nullFS cs = False -- e.g. "Foo", "[]", "(,)" + | cs == FSLIT("[]") = True + | otherwise = startsConId (headFS cs) + +isLexVarId cs -- Ordinary prefix identifiers + | nullFS cs = False -- e.g. "x", "_x" + | otherwise = startsVarId (headFS cs) + +isLexConSym cs -- Infix type or data constructors + | nullFS cs = False -- e.g. ":-:", ":", "->" + | cs == FSLIT("->") = True + | otherwise = startsConSym (headFS cs) + +isLexVarSym cs -- Infix identifiers + | nullFS cs = False -- e.g. "+" + | otherwise = startsVarSym (headFS cs) + +------------- +startsVarSym, startsVarId, startsConSym, startsConId :: Char -> Bool +startsVarSym c = isSymbolASCII c || (ord c > 0x7f && isSymbol c) -- Infix Ids +startsConSym c = c == ':' -- Infix data constructors +startsVarId c = isLower c || c == '_' -- Ordinary Ids +startsConId c = isUpper c || c == '(' -- Ordinary type constructors and data constructors + +isSymbolASCII c = c `elem` "!#$%&*+./<=>?@\\^|~-" \end{code} -For \tr{modnameToC}, we really only have to worry about \tr{'}s -(quote chars) in the name. Rare. +%************************************************************************ +%* * + Binary instance + Here rather than BinIface because OccName is abstract +%* * +%************************************************************************ \begin{code} -modnameToC :: FAST_STRING -> FAST_STRING -modnameToC fast_str = identFsToC fast_str +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) \end{code}