-
+{-% DrIFT (Automatic class derivations for Haskell) v1.1 %-}
+%
% (c) The GRASP/AQUA Project, Glasgow University, 1992-1998
%
\begin{code}
module OccName (
- -- Modules
- Module, -- Abstract, instance of Outputable
- mkSrcModule, mkSrcModuleFS, mkSysModuleFS, mkImportModuleFS, mkBootModule, mkIfaceModuleFS,
- moduleString, moduleUserString, moduleIfaceFlavour,
- pprModule, pprModuleSep, pprModuleBoot,
-
- -- IfaceFlavour
- IfaceFlavour,
- hiFile, hiBootFile, bootFlavour,
-
-- The NameSpace type; abstact
- NameSpace, tcName, clsName, tcClsName, dataName, varName, tvName,
- nameSpaceString,
+ NameSpace, tcName, clsName, tcClsName, dataName, varName,
+ tvName, srcDataName, nameSpaceString,
-- The OccName type
OccName, -- Abstract, instance of Outputable
pprOccName,
- mkSrcOccFS, mkSysOcc, mkSysOccFS, mkSrcVarOcc, mkKindOccFS,
+ -- The OccEnv type
+ OccEnv, emptyOccEnv, unitOccEnv, extendOccEnv,
+ 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,
+
+ mkOccName, mkOccFS, mkSysOcc, mkSysOccFS, mkFCallOcc, mkKindOccFS,
+ mkVarOcc, mkVarOccEncoded,
mkSuperDictSelOcc, mkDFunOcc, mkForeignExportOcc,
- mkDictOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc,
- mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc,
+ mkDictOcc, mkIPOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc,
+ mkDerivedTyConOcc, mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc,
+ mkGenOcc1, mkGenOcc2, mkLocalOcc, mkDataTOcc, mkDataCOcc,
+ mkDataConWrapperOcc, mkDataConWorkerOcc,
- isTvOcc, isDataOcc, isDataSymOcc, isSymOcc,
+ isVarOcc, isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc,
+ reportIfUnused,
- occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour,
+ occNameFS, occNameString, occNameUserString, occNameSpace,
+ occNameFlavour, briefOccNameFlavour,
setOccNameSpace,
+ mkTupleOcc, isTupleOcc_maybe,
+
-- Tidying up
TidyOccEnv, emptyTidyOccEnv, tidyOccName, initTidyOccEnv,
-- Encoding
- EncodedString, EncodedFS, UserString, UserFS, encode, encodeFS, decode,
+ EncodedString, EncodedFS, UserString, UserFS, encode, encodeFS, decode, pprEncodedFS,
-- The basic form of names
isLexCon, isLexVar, isLexId, isLexSym,
#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 FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM )
+import Unique ( Unique, mkUnique, Uniquable(..) )
+import BasicTypes ( Boxity(..), Arity )
+import UniqFM
+import UniqSet
+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 :: EncodedFS -> SDoc
pprEncodedFS fs
= getPprStyle $ \ sty ->
- if userStyle sty then
- text (decode (_UNPK_ fs))
- else
- ptext fs
-\end{code}
-
-
-%************************************************************************
-%* *
-\subsection{Interface file flavour}
-%* *
-%************************************************************************
-
-The IfaceFlavour type is used mainly in an imported Name's Provenance
-to say whether the name comes from a regular .hi file, or whether it comes
-from a hand-written .hi-boot file. This is important, because it has to be
-propagated. Suppose
-
- C.hs imports B
- B.hs imports A
- A.hs imports C {-# SOURCE -#} ( f )
-
-Then in A.hi we may mention C.f, in an inlining. When compiling B we *must not*
-read C.f's details from C.hi, even if the latter happens to exist from an earlier
-compilation run. So we use the name "C!f" in A.hi, and when looking for an interface
-file with details of C!f we look in C.hi-boot. The "!" stuff is recorded in the
-IfaceFlavour in the Module of C.f in A.
-
-Not particularly beautiful, but it works.
-
-\begin{code}
-data IfaceFlavour = HiFile -- The thing comes from a standard interface file
- -- or from the source file itself
- | HiBootFile -- ... or from a handwritten "hi-boot" interface file
- deriving( Eq )
-
-hiFile = HiFile
-hiBootFile = HiBootFile
-
-instance Text IfaceFlavour where -- Just used in debug prints of lex tokens
- showsPrec n HiFile s = s
- showsPrec n HiBootFile s = "!" ++ s
-
-bootFlavour :: IfaceFlavour -> Bool
-bootFlavour HiBootFile = True
-bootFlavour HiFile = False
-\end{code}
-
-
-%************************************************************************
-%* *
-\subsection[Module]{The name of a module}
-%* *
-%************************************************************************
-
-\begin{code}
-data Module = Module
- EncodedFS
- IfaceFlavour
- -- Haskell module names can include the quote character ',
- -- so the module names have the z-encoding applied to them
-\end{code}
-
-\begin{code}
-instance Outputable Module where
- ppr = pprModule
-
--- Ignore the IfaceFlavour when comparing modules
-instance Eq Module where
- (Module m1 _) == (Module m2 _) = m1 == m2
-
-instance Ord Module where
- (Module m1 _) `compare` (Module m2 _) = m1 `compare` m2
-\end{code}
-
-
-\begin{code}
-pprModule :: Module -> SDoc
-pprModule (Module mod _) = pprEncodedFS mod
-
-pprModuleSep, pprModuleBoot :: Module -> SDoc
-pprModuleSep (Module mod HiFile) = dot
-pprModuleSep (Module mod HiBootFile) = char '!'
-
-pprModuleBoot (Module mod HiFile) = empty
-pprModuleBoot (Module mod HiBootFile) = char '!'
-\end{code}
-
-
-\begin{code}
-mkSrcModule :: UserString -> Module
-mkSrcModule s = Module (_PK_ (encode s)) HiFile
-
-mkSrcModuleFS :: UserFS -> Module
-mkSrcModuleFS s = Module (encodeFS s) HiFile
-
-mkImportModuleFS :: UserFS -> IfaceFlavour -> Module
-mkImportModuleFS s hif = Module (encodeFS s) hif
-
-mkSysModuleFS :: EncodedFS -> IfaceFlavour -> Module
-mkSysModuleFS s hif = Module s hif
-
-mkIfaceModuleFS :: EncodedFS -> Module
-mkIfaceModuleFS s = Module s HiFile
-
-mkBootModule :: Module -> Module
-mkBootModule (Module s _) = Module s HiBootFile
-
-moduleString :: Module -> EncodedString
-moduleString (Module mod _) = _UNPK_ mod
-
-moduleUserString :: Module -> UserString
-moduleUserString (Module mod _) = decode (_UNPK_ mod)
-
-moduleIfaceFlavour :: Module -> IfaceFlavour
-moduleIfaceFlavour (Module _ hif) = hif
+ if userStyle sty
+ -- ftext (decodeFS fs) would needlessly pack the string again
+ then text (decode (unpackFS fs))
+ else ftext fs
\end{code}
-
%************************************************************************
%* *
\subsection{Name space}
%************************************************************************
\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.
+ -- 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}
ppr = pprOccName
pprOccName :: OccName -> SDoc
-pprOccName (OccName sp occ) = pprEncodedFS occ
+pprOccName (OccName sp occ)
+ = getPprStyle $ \ sty ->
+ pprEncodedFS occ <> if debugStyle sty then
+ braces (text (briefNameSpaceFlavour sp))
+ else empty
\end{code}
%* *
\subsection{Construction}
%* *
-%************************************************************************
+%*****p*******************************************************************
*Sys* things do no encoding; the caller should ensure that the thing is
already encoded
\begin{code}
mkSysOcc :: NameSpace -> EncodedString -> OccName
-mkSysOcc occ_sp str = ASSERT( alreadyEncoded str )
- OccName occ_sp (_PK_ str)
+mkSysOcc occ_sp str = ASSERT2( alreadyEncoded str, text str )
+ OccName occ_sp (mkFastString str)
mkSysOccFS :: NameSpace -> EncodedFS -> OccName
mkSysOccFS occ_sp fs = ASSERT2( alreadyEncodedFS fs, ppr fs )
OccName occ_sp fs
--- Kind constructors get a speical function. Uniquely, they are not encoded,
+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!
+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*
-- we'll get kinds like (* -> (* -> *)). We can't use mkSysOcc because it
-- has an ASSERT that doesn't hold.
*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)
+
+mkOccName :: NameSpace -> String -> OccName
+mkOccName ns s = mkSysOcc ns (encode s)
-mkSrcVarOcc :: UserFS -> OccName
-mkSrcVarOcc fs = mkSysOccFS varName (encodeFS fs)
+mkVarOcc :: UserFS -> OccName
+mkVarOcc fs = mkSysOccFS varName (encodeFS fs)
+
+mkVarOccEncoded :: EncodedFS -> OccName
+mkVarOccEncoded fs = mkSysOccFS varName fs
\end{code}
%************************************************************************
%* *
+ Environments
+%* *
+%************************************************************************
+
+OccEnvs are used mainly for the envts in ModIfaces.
+
+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}
+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
+
+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
+
+
+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}
+
+
+%************************************************************************
+%* *
\subsection{Predicates and taking them apart}
%* *
%************************************************************************
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 -> SDoc
+occNameFlavour (OccName DataName _) = ptext SLIT("data constructor")
+occNameFlavour (OccName TvName _) = ptext SLIT("type variable")
+occNameFlavour (OccName TcClsName _) = ptext SLIT("type constructor or class")
+occNameFlavour (OccName VarName s) = empty
+
+-- briefOccNameFlavour is used in debug-printing of names
+briefOccNameFlavour :: OccName -> String
+briefOccNameFlavour (OccName sp _) = briefNameSpaceFlavour sp
+
+briefNameSpaceFlavour DataName = "d"
+briefNameSpaceFlavour VarName = "v"
+briefNameSpaceFlavour TvName = "tv"
+briefNameSpaceFlavour TcClsName = "tc"
\end{code}
\begin{code}
-isTvOcc, isDataSymOcc, isSymOcc :: OccName -> Bool
+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 (decodeFS s)
+isDataSymOcc (OccName VarName s) = isLexConSym (decodeFS s)
isDataSymOcc other = False
isDataOcc (OccName DataName _) = True
-isDataOcc oter = False
+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)
+isSymOcc other = False
+\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}
(local variables, so no name-clash worries)
$f... dict-fun identifiers (from inst decls)
- $m... default methods
+ $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
$sf.. specialised version of f
in encoded form these appear as Zdfxxx etc
:... keywords (export:, letrec: etc.)
+--- I THINK THIS IS WRONG!
This knowledge is encoded in the following functions.
-@mk_deriv@ generates an @OccName@ from the one-char prefix and a string.
+@mk_deriv@ generates an @OccName@ from the prefix and a string.
NB: The string must already be encoded!
\begin{code}
mk_deriv :: NameSpace
-> String -- Distinguishes one sort of derived name from another
-> EncodedString -- Must be already encoded!! We don't want to encode it a
- -- second time because encoding isn't itempotent
+ -- second time because encoding isn't idempotent
-> OccName
mk_deriv occ_sp sys_prefix str = mkSysOcc occ_sp (encode sys_prefix ++ str)
\end{code}
\begin{code}
-mkDictOcc, mkWorkerOcc, mkDefaultMethodOcc,
+mkDictOcc, mkIPOcc, mkWorkerOcc, mkDefaultMethodOcc,
mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc
:: OccName -> OccName
-- These derived variables have a prefix that no Haskell value could have
-mkWorkerOcc = mk_simple_deriv varName "$w"
-mkDefaultMethodOcc = mk_simple_deriv varName "$dm"
-mkClassTyConOcc = mk_simple_deriv tcName ":T" -- The : prefix makes sure it classifies
-mkClassDataConOcc = mk_simple_deriv dataName ":D" -- as a tycon/datacon
-mkDictOcc = mk_simple_deriv varName "$d"
-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 "$f"
+
+-- Generic derivable classes
+mkGenOcc1 = mk_simple_deriv varName "$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 "$c"
mk_simple_deriv sp px occ = mk_deriv sp px (occNameString occ)
+
+
+-- Data constructor workers are made by setting the name space
+-- of the data constructor OccName (which should be a DataName)
+-- to DataName
+mkDataConWorkerOcc datacon_occ = setOccNameSpace varName datacon_occ
\end{code}
\begin{code}
mkSuperDictSelOcc :: Int -- Index of superclass, eg 3
-> OccName -- Class, eg "Ord"
- -> OccName -- eg "p3Ord"
+ -> OccName -- eg "$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 ("$L" ++ show uniq) (occNameString occ)
+ -- The Unique might print with characters
+ -- that need encoding (e.g. 'z'!)
\end{code}
\begin{code}
-mkDFunOcc :: OccName -- class, eg "Ord"
- -> OccName -- tycon (or something convenient from the instance type)
- -- eg "Maybe"
- -> Int -- Unique to distinguish dfuns which share the previous two
- -- eg 3
- -> OccName -- "dOrdMaybe3"
-
-mkDFunOcc cls_occ tycon_occ index
- = mk_deriv VarName "$f" (show_index ++ cls_str ++ tycon_str)
- where
- cls_str = occNameString cls_occ
- tycon_str = occNameString tycon_occ
- show_index | index == 0 = ""
- | otherwise = show index
+mkDFunOcc :: EncodedString -- Typically the class and type glommed together e.g. "OrdMaybe"
+ -> OccName -- "$fOrdMaybe"
+
+mkDFunOcc string = mk_deriv VarName "$f" string
\end{code}
We used to add a '$m' to indicate a method, but that gives rise to bad
\begin{code}
mkMethodOcc :: OccName -> OccName
mkMethodOcc occ@(OccName VarName fs) = occ
-mkMethodOcc occ = mk_simple_deriv varName "$m"
+mkMethodOcc occ = mk_simple_deriv varName "$m" occ
\end{code}
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 fs)
- | not (fs `elemFM` in_scope)
- = (addToFM in_scope fs 1, occ) -- First occurrence
-
- | otherwise -- Already occurs
- = go in_scope (_UNPK_ fs)
- 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, mkSysOccFS occ_sp pk_str)
- -- str is now unique
- where
- pk_str = _PK_ str
+ = case lookupOccEnv in_scope occ of
+ Nothing -> -- Not already used: make it used
+ (extendOccEnv in_scope occ 1, occ)
+
+ Just n -> -- Already used: make a new guess,
+ -- change the guess base, and try again
+ tidyOccName (extendOccEnv in_scope occ (n+1))
+ (mkSysOcc occ_sp (unpackFS fs ++ show n))
\end{code}
* Tuples (,,,) are coded as Z3T
-* Alphabetic characters (upper and lower), digits, and '_'
+* Alphabetic characters (upper and lower) and digits
all translate to themselves;
except 'Z', which translates to 'ZZ'
and 'z', which translates to 'zz'
We need both so that we can preserve the variable/tycon distinction
-* Most other printable characters translate to 'Zx' for some
+* 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
--------------------------
Trak Trak
- foo_wib foo_wib
- > Zg
- >1 Zg1
- foo# fooZh
- foo## fooZhZh
- foo##1 fooZhXh1
+ foo_wib foozuwib
+ > zg
+ >1 zg1
+ foo# foozh
+ 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
alreadyEncoded :: String -> Bool
alreadyEncoded s = all ok s
where
- ok '_' = True
- ok ch = ISALPHANUM ch
+ ok ' ' = True
+ -- This is a bit of a lie; if we really wanted spaces
+ -- 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
-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 cs
-maybe_tuple other = Nothing
-
-check_tuple n (',' : cs) = check_tuple (n+1) cs
-check_tuple n ")" = Just n
-check_tuple n other = Nothing
-
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 '_' = True
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
encode_ch 'z' = "zz"
encode_ch '&' = "za"
encode_ch '|' = "zb"
+encode_ch '^' = "zc"
encode_ch '$' = "zd"
encode_ch '=' = "ze"
encode_ch '>' = "zg"
encode_ch '\\' = "zr"
encode_ch '/' = "zs"
encode_ch '*' = "zt"
-encode_ch c = ['z', 'x', intToDigit hi, intToDigit lo]
- where
- (hi,lo) = ord c `quotRem` 16
+encode_ch '_' = "zu"
+encode_ch '%' = "zv"
+encode_ch c = 'z' : shows (ord c) "U"
\end{code}
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 [] = []
-decode ('Z' : rest) = decode_escape rest
-decode ('z' : rest) = decode_escape rest
+decode ('Z' : d : rest) | isDigit d = decode_tuple d rest
+ | otherwise = decode_upper d : decode rest
+decode ('z' : d : rest) | isDigit d = decode_num_esc d rest
+ | otherwise = decode_lower d : decode rest
decode (c : rest) = c : decode rest
-decode_escape :: EncodedString -> UserString
-
-decode_escape ('Z' : rest) = 'Z' : decode rest
-decode_escape ('C' : rest) = ':' : decode rest
-decode_escape ('L' : rest) = '(' : decode rest
-decode_escape ('R' : rest) = ')' : decode rest
-decode_escape ('M' : rest) = '[' : decode rest
-decode_escape ('N' : rest) = ']' : decode rest
-
-decode_escape ('z' : rest) = 'z' : decode rest
-decode_escape ('a' : rest) = '&' : decode rest
-decode_escape ('b' : rest) = '|' : decode rest
-decode_escape ('d' : rest) = '$' : decode rest
-decode_escape ('e' : rest) = '=' : decode rest
-decode_escape ('g' : rest) = '>' : decode rest
-decode_escape ('h' : rest) = '#' : decode rest
-decode_escape ('i' : rest) = '.' : decode rest
-decode_escape ('l' : rest) = '<' : decode rest
-decode_escape ('m' : rest) = '-' : decode rest
-decode_escape ('n' : rest) = '!' : decode rest
-decode_escape ('p' : rest) = '+' : decode rest
-decode_escape ('q' : rest) = '\'' : decode rest
-decode_escape ('r' : rest) = '\\' : decode rest
-decode_escape ('s' : rest) = '/' : decode rest
-decode_escape ('t' : rest) = '*' : decode rest
-decode_escape ('x' : d1 : d2 : rest) = chr (digitToInt d1 * 16 + digitToInt d2) : decode rest
-
--- Tuples are coded as Z23T
-decode_escape (c : rest)
- | isDigit c = go (digitToInt c) rest
+decode_upper, decode_lower :: Char -> Char
+
+decode_upper 'L' = '('
+decode_upper 'R' = ')'
+decode_upper 'M' = '['
+decode_upper 'N' = ']'
+decode_upper 'C' = ':'
+decode_upper 'Z' = 'Z'
+decode_upper ch = pprTrace "decode_upper" (char ch) ch
+
+decode_lower 'z' = 'z'
+decode_lower 'a' = '&'
+decode_lower 'b' = '|'
+decode_lower 'c' = '^'
+decode_lower 'd' = '$'
+decode_lower 'e' = '='
+decode_lower 'g' = '>'
+decode_lower 'h' = '#'
+decode_lower 'i' = '.'
+decode_lower 'l' = '<'
+decode_lower 'm' = '-'
+decode_lower 'n' = '!'
+decode_lower 'p' = '+'
+decode_lower 'q' = '\''
+decode_lower 'r' = '\\'
+decode_lower 's' = '/'
+decode_lower 't' = '*'
+decode_lower 'u' = '_'
+decode_lower 'v' = '%'
+decode_lower ch = pprTrace "decode_lower" (char ch) ch
+
+-- Characters not having a specific code are coded as z224U
+decode_num_esc d rest
+ = go (digitToInt d) rest
where
go n (c : rest) | isDigit c = go (10*n + digitToInt c) rest
- go n ('T' : rest) = '(' : replicate n ',' ++ ')' : decode rest
- go n other = pprPanic "decode_escape" (ppr n <+> text (c:rest))
+ go n ('U' : rest) = chr n : decode rest
+ go n other = pprPanic "decode_num_esc" (ppr n <+> text other)
-decode_escape (c : rest) = pprTrace "decode_escape" (char c) (decode rest)
+decode_tuple :: Char -> EncodedString -> UserString
+decode_tuple d rest
+ = go (digitToInt d) rest
+ where
+ -- NB. recurse back to decode after decoding the tuple, because
+ -- the tuple might be embedded in a longer name.
+ go n (c : rest) | isDigit c = go (10*n + digitToInt c) 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 other = pprPanic "decode_tuple" (ppr n <+> text other)
\end{code}
%************************************************************************
%* *
-n\subsection{Lexical categories}
+ Stuff for dealing with tuples
+%* *
+%************************************************************************
+
+Tuples are encoded as
+ Z3T or Z3H
+for 3-tuples or unboxed 3-tuples respectively. No other encoding starts
+ Z<digit>
+
+* "(# #)" is the tycon for an unboxed 1-tuple (not 0-tuple)
+ There are no unboxed 0-tuples.
+
+* "()" is the tycon for a boxed 0-tuple.
+ There are no boxed 1-tuples.
+
+
+\begin{code}
+maybe_tuple :: UserString -> Maybe EncodedString
+
+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
+
+count_commas :: Int -> String -> (Int, String)
+count_commas n (',' : cs) = count_commas (n+1) cs
+count_commas n cs = (n,cs)
+\end{code}
+
+\begin{code}
+mkTupleOcc :: NameSpace -> Boxity -> Arity -> OccName
+mkTupleOcc ns bx ar
+ = OccName ns (mkFastString ('Z' : (show ar ++ bx_char)))
+ where
+ bx_char = case bx of
+ Boxed -> "T"
+ Unboxed -> "H"
+
+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
+ ('Z':d:rest) | isDigit d -> Just (decode_tup (digitToInt d) rest)
+ other -> Nothing
+ where
+ decode_tup n "H" = (ns, Unboxed, n)
+ decode_tup n "T" = (ns, Boxed, n)
+ decode_tup n (d:rest) = decode_tup (n*10 + digitToInt d) rest
+\end{code}
+
+%************************************************************************
+%* *
+\subsection{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
-------------
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
+ | 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 = isLower c || isLowerISO c || c == '_'
- where
- c = _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. ":-:", ":", "->"
- | otherwise = c == ':'
- || cs == SLIT("->")
- where
- c = _HEAD_ cs
+ | nullFastString cs = False -- e.g. ":-:", ":", "->"
+ | cs == FSLIT("->") = True
+ | otherwise = startsConSym (headFS cs)
isLexVarSym cs -- Infix identifiers
- | _NULL_ cs = False -- e.g. "+"
- | otherwise = isSymbolASCII c
- || isSymbolISO c
- where
- c = _HEAD_ cs
+ | nullFastString cs = False -- e.g. "+"
+ | otherwise = startsVarSym (headFS cs)
-------------
+startsVarSym, startsVarId, startsConSym, startsConId :: Char -> Bool
+startsVarSym c = isSymbolASCII c || isSymbolISO c -- Infix Ids
+startsConSym c = c == ':' -- Infix data constructors
+startsVarId c = isLower c || isLowerISO c || c == '_' -- Ordinary Ids
+startsConId c = isUpper c || isUpperISO c || c == '(' -- Ordinary type constructors and data constructors
+
+
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'#
isLowerISO (C# c#) = c# `geChar#` '\xdf'# && c# `leChar#` '\xff'# && c# `neChar#` '\xf7'#
--0xdf <= oc && oc <= 0xff && oc /= 0xf7 where oc = ord c
\end{code}
+
+%************************************************************************
+%* *
+ Binary instance
+ Here rather than BinIface because OccName is abstract
+%* *
+%************************************************************************
+
+\begin{code}
+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}