+{-% DrIFT (Automatic class derivations for Haskell) v1.1 %-}
%
% (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
%
module OccName (
-- The NameSpace type; abstact
NameSpace, tcName, clsName, tcClsName, dataName, varName,
- tvName, nameSpaceString,
+ tvName, srcDataName, nameSpaceString,
-- The OccName type
OccName, -- Abstract, instance of Outputable
pprOccName,
- mkOccFS, mkSysOcc, mkSysOccFS, mkFCallOcc, mkVarOcc, mkKindOccFS,
+ mkOccFS, mkSysOcc, mkSysOccFS, mkFCallOcc, mkKindOccFS,
+ mkVarOcc, mkVarOccEncoded,
mkSuperDictSelOcc, mkDFunOcc, mkForeignExportOcc,
mkDictOcc, mkIPOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc,
mkDerivedTyConOcc, mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc,
- mkGenOcc1, mkGenOcc2, mkLocalOcc,
+ mkGenOcc1, mkGenOcc2, mkLocalOcc,
+ mkDataConWrapperOcc, mkDataConWorkerOcc,
isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc,
+ reportIfUnused,
- occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour,
+ occNameFS, occNameString, occNameUserString, occNameSpace,
+ occNameFlavour, briefOccNameFlavour,
setOccNameSpace,
-- Tidying up
#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
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
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}
%************************************************************************
%************************************************************************
\begin{code}
-data NameSpace = VarName -- Variables
- | 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
clsName = TcClsName -- Classes
tcClsName = TcClsName -- Not sure which!
-dataName = DataName
-tvName = TvName
-varName = VarName
+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"
data OccName = OccName
NameSpace
EncodedFS
+ {-! derive : Binary !-}
\end{code}
\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 )
-- 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!
-mkFCallOcc 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*
mkVarOcc :: UserFS -> OccName
mkVarOcc fs = mkSysOccFS varName (encodeFS fs)
+
+mkVarOccEncoded :: EncodedFS -> OccName
+mkVarOccEncoded fs = mkSysOccFS varName fs
\end{code}
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)
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 _) = "Real 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}
-- 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)
\end{code}
+\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}
:: OccName -> OccName
-- 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 "$dm"
mkDerivedTyConOcc = mk_simple_deriv tcName ":" -- The : prefix makes sure it classifies
mkClassTyConOcc = mk_simple_deriv tcName ":T" -- as a tycon/datacon
-mkClassDataConOcc = mk_simple_deriv dataName ":D" --
+mkClassDataConOcc = mk_simple_deriv dataName ":D" -- We go straight to the "real" data con
+ -- for datacons from classes
mkDictOcc = mk_simple_deriv varName "$d"
mkIPOcc = mk_simple_deriv varName "$i"
mkSpecOcc = mk_simple_deriv varName "$s"
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)
+
+
+-- 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}
-> OccName -- Local name (e.g. "sat")
-> OccName -- Nice unique version ("$L23sat")
mkLocalOcc uniq occ
- = mk_deriv varName "$L" (show uniq ++ occNameString occ)
+ = mk_deriv varName ("$L" ++ show uniq) (occNameString occ)
+ -- The Unique might print with characters
+ -- that need encoding (e.g. 'z'!)
\end{code}
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!
= (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
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}
-- 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
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
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 [] = []
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}
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
-------------
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
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}