-
+%
% (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, nameSpaceString,
-- The OccName type
OccName, -- Abstract, instance of Outputable
pprOccName,
- mkSrcOccFS, mkSysOcc, mkSysOccFS, mkSrcVarOcc, mkKindOccFS,
+ mkOccFS, mkSysOcc, mkSysOccFS, mkFCallOcc, mkVarOcc, mkKindOccFS,
mkSuperDictSelOcc, mkDFunOcc, mkForeignExportOcc,
- mkDictOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc,
- mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc,
+ mkDictOcc, mkIPOcc, mkWorkerOcc, mkMethodOcc, mkDefaultMethodOcc,
+ mkDerivedTyConOcc, mkClassTyConOcc, mkClassDataConOcc, mkSpecOcc,
+ mkGenOcc1, mkGenOcc2, mkLocalOcc,
- isTvOcc, isDataOcc, isDataSymOcc, isSymOcc,
+ isTvOcc, isTcOcc, isDataOcc, isDataSymOcc, isSymOcc, isValOcc,
occNameFS, occNameString, occNameUserString, occNameSpace, occNameFlavour,
setOccNameSpace,
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 Unique ( Unique )
import FiniteMap ( FiniteMap, emptyFM, lookupFM, addToFM, elemFM )
import Outputable
import GlaExts
These type synonyms help documentation.
\begin{code}
-type UserFS = FAST_STRING -- As the user typed it
+type UserFS = FAST_STRING -- As the user typed it
type EncodedFS = FAST_STRING -- Encoded form
type UserString = String -- As the user typed it
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
+ -- ptext (decodeFS fs) would needlessly pack the string again
+ then text (decode (_UNPK_ fs))
+ else ptext fs
\end{code}
-
%************************************************************************
%* *
\subsection{Name space}
| DataName -- 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 )
-- Though type constructors and classes are in the same name space now,
\begin{code}
mkSysOcc :: NameSpace -> EncodedString -> OccName
-mkSysOcc occ_sp str = ASSERT( alreadyEncoded str )
+mkSysOcc occ_sp str = ASSERT2( alreadyEncoded str, text str )
OccName occ_sp (_PK_ 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 (_PK_ 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)
-mkSrcVarOcc :: UserFS -> OccName
-mkSrcVarOcc fs = mkSysOccFS varName (encodeFS fs)
+mkVarOcc :: UserFS -> OccName
+mkVarOcc fs = mkSysOccFS varName (encodeFS fs)
\end{code}
\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
+
-- Data constructor operator (starts with ':', or '[]')
-- Pretty inefficient!
isDataSymOcc (OccName DataName s) = isLexConSym (decodeFS s)
isDataSymOcc other = False
isDataOcc (OccName DataName _) = True
-isDataOcc oter = False
+isDataOcc other = False
-- Any operator (data constructor or variable)
-- Pretty inefficient!
(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"
-
+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"
+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)
\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
* 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 ' ' = 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
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
+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
str = _UNPK_ 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 '$' = "zd"
encode_ch '=' = "ze"
encode_ch '>' = "zg"
encode_ch '\\' = "zr"
encode_ch '/' = "zs"
encode_ch '*' = "zt"
-encode_ch c = ['z', 'x', intToDigit hi, intToDigit lo]
- where
- (hi,lo) = ord c `quotRem` 16
+encode_ch '_' = "zu"
+encode_ch '%' = "zv"
+encode_ch c = 'z' : shows (ord c) "U"
\end{code}
Decode is used for user printing.
decode_escape :: EncodedString -> UserString
-decode_escape ('Z' : rest) = 'Z' : decode rest
-decode_escape ('C' : rest) = ':' : decode rest
decode_escape ('L' : rest) = '(' : decode rest
decode_escape ('R' : rest) = ')' : decode rest
decode_escape ('M' : rest) = '[' : decode rest
decode_escape ('N' : rest) = ']' : decode rest
+decode_escape ('C' : rest) = ':' : decode rest
+decode_escape ('Z' : rest) = 'Z' : decode rest
decode_escape ('z' : rest) = 'z' : decode rest
decode_escape ('a' : rest) = '&' : decode rest
decode_escape ('b' : rest) = '|' : decode rest
+decode_escape ('c' : rest) = '^' : decode rest
decode_escape ('d' : rest) = '$' : decode rest
decode_escape ('e' : rest) = '=' : decode rest
decode_escape ('g' : 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
+decode_escape ('u' : rest) = '_' : decode rest
+decode_escape ('v' : rest) = '%' : decode rest
-- Tuples are coded as Z23T
+-- Characters not having a specific code are coded as z224U
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)
%************************************************************************
%* *
-n\subsection{Lexical categories}
+\subsection{Lexical categories}
%* *
%************************************************************************
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
+ | otherwise = startsConId (_HEAD_ cs)
isLexVarId cs -- Ordinary prefix identifiers
| _NULL_ cs = False -- e.g. "x", "_x"
- | otherwise = isLower c || isLowerISO c || c == '_'
- where
- c = _HEAD_ cs
+ | otherwise = startsVarId (_HEAD_ cs)
isLexConSym cs -- Infix type or data constructors
| _NULL_ cs = False -- e.g. ":-:", ":", "->"
- | otherwise = c == ':'
- || cs == SLIT("->")
- where
- c = _HEAD_ cs
+ | cs == SLIT("->") = True
+ | otherwise = startsConSym (_HEAD_ cs)
isLexVarSym cs -- Infix identifiers
| _NULL_ cs = False -- e.g. "+"
- | otherwise = isSymbolASCII c
- || isSymbolISO c
- where
- c = _HEAD_ cs
+ | otherwise = startsVarSym (_HEAD_ 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'#