Buffers for scanning string input stored in external arrays.
\begin{code}
-{-# OPTIONS -fno-prune-tydecls #-}
+
+{-# OPTIONS -fvia-C #-}
+
module StringBuffer
(
StringBuffer,
- -- creation
- hGetStringBuffer, -- :: FilePath -> IO StringBuffer
+ -- creation/destruction
+ hGetStringBuffer, -- :: FilePath -> IO StringBuffer
+ stringToStringBuffer, -- :: String -> IO StringBuffer
+ freeStringBuffer, -- :: StringBuffer -> IO ()
-- Lookup
currentChar, -- :: StringBuffer -> Char
lookAhead, -- :: StringBuffer -> Int -> Char
lookAhead#, -- :: StringBuffer -> Int# -> Char#
+ -- offsets
+ currentIndex#, -- :: StringBuffer -> Int#
+ lexemeIndex, -- :: StringBuffer -> Int#
+
-- moving the end point of the current lexeme.
setCurrentPos#, -- :: StringBuffer -> Int# -> StringBuffer
incLexeme, -- :: StringBuffer -> StringBuffer
stepOnBy#, -- :: StringBuffer -> Int# -> StringBuffer
stepOnTo#, -- :: StringBuffer -> Int# -> StringBuffer
stepOnUntil, -- :: (Char -> Bool) -> StringBuffer -> StringBuffer
+ stepOnUntilChar#, -- :: StringBuffer -> Char# -> StringBuffer
stepOverLexeme, -- :: StringBuffer -> StringBuffer
scanNumLit, -- :: Int -> StringBuffer -> (Int, StringBuffer)
+ squeezeLexeme, -- :: StringBuffer -> Int# -> StringBuffer
+ mergeLexemes, -- :: StringBuffer -> StringBuffer -> StringBuffer
expandWhile, -- :: (Char -> Bool) -> StringBuffer -> StringBuffer
expandWhile#, -- :: (Char# -> Bool) -> StringBuffer -> StringBuffer
expandUntilMatch, -- :: StrinBuffer -> String -> StringBuffer
-- matching
prefixMatch, -- :: StringBuffer -> String -> Bool
untilEndOfString#, -- :: StringBuffer -> Int#
- untilEndOfChar#, -- :: StringBuffer -> Int#
- untilChar#, -- :: StringBuffer -> Char# -> Int#
-- conversion
lexemeToString, -- :: StringBuffer -> String
#include "HsVersions.h"
import GlaExts
-import Addr ( Addr(..) )
-import Foreign
-import ST
-
-#if __GLASGOW_HASKELL__ >= 303
-import IO ( slurpFile )
+#if __GLASGOW_HASKELL__ < 411
+import PrelAddr ( Addr(..) )
#else
-import IO ( openFile, hFileSize, hClose, IOMode(..) )
+import Addr ( Addr(..) )
+#endif
+import Foreign
+import Char ( chr )
+import Panic ( panic )
+
+import IO ( openFile )
+import IOExts ( slurpFile )
+import PrelIOBase
+import PrelHandle
+import Addr
+#if __GLASGOW_HASKELL__ >= 411
+import Ptr ( Ptr(..) )
#endif
-#if __GLASGOW_HASKELL__ < 301
-import IOBase ( IOError(..), IOErrorType(..) )
-import IOHandle ( readHandle, writeHandle, filePtr )
-import PackBase ( unpackCStringBA )
-#else
-# if __GLASGOW_HASKELL__ <= 302
-import PrelIOBase ( IOError(..), IOErrorType(..) )
-import PrelHandle ( readHandle, writeHandle, filePtr )
-# endif
import PrelPack ( unpackCStringBA )
+
+#if __GLASGOW_HASKELL__ >= 501
+import PrelIO ( hGetcBuffered )
+import PrelCError ( throwErrnoIfMinus1RetryMayBlock )
+import PrelConc ( threadWaitRead )
#endif
+import Exception ( bracket )
import PrimPacked
import FastString
-import Char (isDigit)
+import Char ( isDigit )
\end{code}
\begin{code}
\end{code}
\begin{code}
-instance Text StringBuffer where
+instance Show StringBuffer where
showsPrec _ s = showString ""
\end{code}
\begin{code}
-hGetStringBuffer :: FilePath -> IO StringBuffer
-hGetStringBuffer fname =
-#if __GLASGOW_HASKELL__ >= 303
- slurpFile fname >>= \ (a , read) ->
- let (A# a#) = a
- (I# read#) = read
- in
- _casm_ `` ((char *)%0)[(int)%1]=(char)0; '' a (I# (read# -# 1#)) >>= \ () ->
- return (StringBuffer a# read# 0# 0#)
+hGetStringBuffer :: Bool -> FilePath -> IO StringBuffer
+hGetStringBuffer expand_tabs fname = do
+ (a, read) <- if expand_tabs
+ then slurpFileExpandTabs fname
+#if __GLASGOW_HASKELL__ < 411
+ else slurpFile fname
#else
- openFile fname ReadMode >>= \ hndl ->
- hFileSize hndl >>= \ len@(J# _ _ d#) ->
- let len_i = fromInteger len in
- -- Allocate an array for system call to store its bytes into.
- -- ToDo: make it robust
--- trace (show ((len_i::Int)+1)) $
- _casm_ `` %r=(char *)malloc(sizeof(char)*(int)%0); '' (len_i::Int) >>= \ arr@(A# a#) ->
- if addr2Int# a# ==# 0# then
- fail (userError ("hGetStringBuffer: Could not allocate "++show len_i ++ " bytes"))
- else
- readHandle hndl >>= \ hndl_ ->
- writeHandle hndl hndl_ >>
- let ptr = filePtr hndl_ in
-#if __GLASGOW_HASKELL__ <= 302
- _ccall_ fread arr (1::Int) len_i (ptr::ForeignObj) >>= \ (I# read#) ->
-#else
- _ccall_ fread arr (1::Int) len_i (ptr::Addr) >>= \ (I# read#) ->
+ else do
+ (Ptr a#, read) <- slurpFile fname
+ return (A# a#, read)
#endif
- hClose hndl >>
- if read# ==# 0# then -- EOF or some other error
- fail (userError ("hGetStringBuffer: failed to slurp in interface file "++fname))
- else
- -- Add a sentinel NUL
- _casm_ `` ((char *)%0)[(int)%1]=(char)0; '' arr (I# (read# -# 1#)) >>= \ () ->
- return (StringBuffer a# read# 0# 0#)
-#endif
+ let (A# a#) = a; (I# read#) = read
+
+ -- add sentinel '\NUL'
+ _casm_ `` ((char *)%0)[(int)%1]=(char)0; '' (A# a#) (I# (read# -# 1#))
+ return (StringBuffer a# read# 0# 0#)
unsafeWriteBuffer :: StringBuffer -> Int# -> Char# -> StringBuffer
unsafeWriteBuffer s@(StringBuffer a _ _ _) i# ch# =
)
\end{code}
+-----------------------------------------------------------------------------
+-- Turn a String into a StringBuffer
+
+\begin{code}
+stringToStringBuffer :: String -> IO StringBuffer
+freeStringBuffer :: StringBuffer -> IO ()
+
+#if __GLASGOW_HASKELL__ >= 411
+stringToStringBuffer str =
+ do let sz@(I# sz#) = length str
+ (Ptr a#) <- mallocBytes (sz+1)
+ fill_in str (A# a#)
+ writeCharOffAddr (A# a#) sz '\0' -- sentinel
+ return (StringBuffer a# sz# 0# 0#)
+ where
+ fill_in [] _ = return ()
+ fill_in (c:cs) a = do
+ writeCharOffAddr a 0 c
+ fill_in cs (a `plusAddr` 1)
+
+freeStringBuffer (StringBuffer a# _ _ _) = Foreign.free (Ptr a#)
+#else
+stringToStringBuffer = panic "stringToStringBuffer: not implemented"
+freeStringBuffer sb = return ()
+#endif
+
+\end{code}
+
+-----------------------------------------------------------------------------
+This very disturbing bit of code is used for expanding the tabs in a
+file before we start parsing it. Expanding the tabs early makes the
+lexer a lot simpler: we only have to record the beginning of the line
+in order to be able to calculate the column offset of the current
+token.
+
+We guess the size of the buffer required as 20% extra for
+expanded tabs, and enlarge it if necessary.
+
+\begin{code}
+getErrType :: IO Int
+getErrType = _ccall_ getErrType__
+
+slurpFileExpandTabs :: FilePath -> IO (Addr,Int)
+slurpFileExpandTabs fname = do
+ bracket (openFile fname ReadMode) (hClose)
+ (\ handle ->
+ do sz <- hFileSize handle
+ if sz > toInteger (maxBound::Int)
+ then IOERROR (userError "slurpFile: file too big")
+ else do
+ let sz_i = fromInteger sz
+ sz_i' = (sz_i * 12) `div` 10 -- add 20% for tabs
+ chunk <- allocMem sz_i'
+ trySlurp handle sz_i' chunk
+ )
+
+trySlurp :: Handle -> Int -> Addr -> IO (Addr, Int)
+trySlurp handle sz_i chunk =
+#if __GLASGOW_HASKELL__ < 501
+ wantReadableHandle "hGetChar" handle $ \ handle_ ->
+ let fo = haFO__ handle_ in
+#else
+ wantReadableHandle "hGetChar" handle $
+ \ handle_@Handle__{ haFD=fd, haBuffer=ref, haBufferMode=mode } ->
+#endif
+ let
+ (I# chunk_sz) = sz_i
+
+ tAB_SIZE = 8#
+
+ slurpFile :: Int# -> Int# -> Addr -> Int# -> Int# -> IO (Addr, Int)
+ slurpFile c off chunk chunk_sz max_off = slurp c off
+ where
+
+ slurp :: Int# -> Int# -> IO (Addr, Int)
+ slurp c off | off >=# max_off = do
+ let new_sz = chunk_sz *# 2#
+ chunk' <- reAllocMem chunk (I# new_sz)
+ slurpFile c off chunk' new_sz (new_sz -# (tAB_SIZE +# 1#))
+ slurp c off = do
+#if __GLASGOW_HASKELL__ < 501
+ intc <- mayBlock fo (_ccall_ fileGetc fo)
+ if intc == ((-1)::Int)
+ then do errtype <- getErrType
+ if errtype == (19{-ERR_EOF-} :: Int)
+ then return (chunk, I# off)
+ else constructErrorAndFail "slurpFile"
+ else case chr intc of
+#else
+ buf <- readIORef ref
+ ch <- (if not (bufferEmpty buf)
+ then hGetcBuffered fd ref buf
+ else do new_buf <- fillReadBuffer fd True buf
+ hGetcBuffered fd ref new_buf)
+ `catch` \e -> if isEOFError e
+ then return '\xFFFF'
+ else ioError e
+ case ch of
+ '\xFFFF' -> return (chunk, I# off)
+#endif
+ '\t' -> tabIt c off
+ ch -> do writeCharOffAddr chunk (I# off) ch
+ let c' | ch == '\n' = 0#
+ | otherwise = c +# 1#
+ slurp c' (off +# 1#)
+
+ tabIt :: Int# -> Int# -> IO (Addr, Int)
+ -- can't run out of buffer in here, because we reserved an
+ -- extra tAB_SIZE bytes at the end earlier.
+ tabIt c off = do
+ writeCharOffAddr chunk (I# off) ' '
+ let c' = c +# 1#
+ off' = off +# 1#
+ if c' `remInt#` tAB_SIZE ==# 0#
+ then slurp c' off'
+ else tabIt c' off'
+ in do
+
+ -- allow space for a full tab at the end of the buffer
+ -- (that's what the max_off thing is for),
+ -- and add 1 to allow room for the final sentinel \NUL at
+ -- the end of the file.
+ (chunk', rc) <- slurpFile 0# 0# chunk chunk_sz (chunk_sz -# (tAB_SIZE +# 1#))
+#if __GLASGOW_HASKELL__ < 404
+ writeHandle handle handle_
+#endif
+ return (chunk', rc+1 {-room for sentinel-})
+
+
+reAllocMem :: Addr -> Int -> IO Addr
+reAllocMem ptr sz = do
+ chunk <- _ccall_ realloc ptr sz
+ if chunk == nullAddr
+ then fail "reAllocMem"
+ else return chunk
+
+allocMem :: Int -> IO Addr
+allocMem sz = do
+ chunk <- _ccall_ malloc sz
+ if chunk == nullAddr
+#if __GLASGOW_HASKELL__ < 501
+ then constructErrorAndFail "allocMem"
+#else
+ then ioException (IOError Nothing ResourceExhausted "malloc"
+ "out of memory" Nothing)
+#endif
+ else return chunk
+\end{code}
+
Lookup
\begin{code}
-- relative lookup, i.e, currentChar = lookAhead 0
lookAhead# (StringBuffer fo# _ _ c#) i# = indexCharOffAddr# fo# (c# +# i#)
+currentIndex# :: StringBuffer -> Int#
+currentIndex# (StringBuffer fo# _ _ c#) = c#
+
+lexemeIndex :: StringBuffer -> Int#
+lexemeIndex (StringBuffer fo# _ c# _) = c#
\end{code}
moving the start point of the current lexeme.
stepOnTo# :: StringBuffer -> Int# -> StringBuffer
stepOnTo# (StringBuffer fo l _ _) s# = StringBuffer fo l s# s#
+squeezeLexeme :: StringBuffer -> Int# -> StringBuffer
+squeezeLexeme (StringBuffer fo l s# c#) i# = StringBuffer fo l (s# +# i#) c#
+
+mergeLexemes :: StringBuffer -> StringBuffer -> StringBuffer
+mergeLexemes (StringBuffer fo l s# _) (StringBuffer _ _ _ c#)
+ = StringBuffer fo l s# c#
+
stepOnUntil :: (Char -> Bool) -> StringBuffer -> StringBuffer
stepOnUntil pred (StringBuffer fo l# s# c#) =
| ch# `eqChar#` '\NUL'# && c# >=# l# -> StringBuffer fo l# s# c# -- EOB, return immediately.
| otherwise -> StringBuffer fo l# s# c#
-scanNumLit :: Int -> StringBuffer -> (Int,StringBuffer)
-scanNumLit (I# acc#) (StringBuffer fo l# s# c#) =
- loop acc# c#
+scanNumLit :: Integer -> StringBuffer -> (Integer,StringBuffer)
+scanNumLit acc (StringBuffer fo l# s# c#) =
+ loop acc c#
where
- loop acc# c# =
+ loop acc c# =
case indexCharOffAddr# fo c# of
- ch# | isDigit (C# ch#) -> loop (acc# *# 10# +# (ord# ch# -# ord# '0'#)) (c# +# 1#)
- | ch# `eqChar#` '\NUL'# && c# >=# l# -> (I# acc#, StringBuffer fo l# s# c#) -- EOB, return immediately.
- | otherwise -> (I# acc#,StringBuffer fo l# s# c#)
+ ch# | isDigit (C# ch#) -> loop (acc*10 + (toInteger (I# (ord# ch# -# ord# '0'#)))) (c# +# 1#)
+ | ch# `eqChar#` '\NUL'# && c# >=# l# -> (acc, StringBuffer fo l# s# c#) -- EOB, return immediately.
+ | otherwise -> (acc,StringBuffer fo l# s# c#)
-expandUntilMatch :: StringBuffer -> String -> StringBuffer
+expandUntilMatch :: StringBuffer -> String -> Maybe StringBuffer
expandUntilMatch (StringBuffer fo l# s# c#) str =
loop c# str
where
- loop c# [] = StringBuffer fo l# s# c#
- loop c# ((C# x#):xs)
- | indexCharOffAddr# fo c# `eqChar#` x#
- = loop (c# +# 1#) xs
- | otherwise
- = loop (c# +# 1#) str
+ loop c# [] = Just (StringBuffer fo l# s# c#)
+ loop c# ((C# x#):xs) =
+ case indexCharOffAddr# fo c# of
+ ch# | ch# `eqChar#` '\NUL'# && c# >=# l# -> Nothing
+ | ch# `eqChar#` x# -> loop (c# +# 1#) xs
+ | otherwise -> loop (c# +# 1#) str
\end{code}
_ -> loop (c# +# 1#)
-untilEndOfChar# :: StringBuffer -> StringBuffer
-untilEndOfChar# (StringBuffer fo l# s# c#) =
- loop c#
- where
- getch# i# = indexCharOffAddr# fo i#
-
- loop c# =
- case getch# c# of
- '\''# ->
- case getch# (c# -# 1#) of
- '\\'# ->
- case getch# (c# -# 2#) of
- '\\'# -> -- end of char
- StringBuffer fo l# s# c#
- _ -> loop (c# +# 1#) -- false alarm
- _ -> StringBuffer fo l# s# c#
- '\NUL'# ->
- if c# >=# l# then -- hit sentinel, this doesn't look too good..
- StringBuffer fo l# l# l#
- else
- loop (c# +# 1#)
- _ -> loop (c# +# 1#)
-
-untilChar# :: StringBuffer -> Char# -> StringBuffer
-untilChar# (StringBuffer fo l# s# c#) x# =
+stepOnUntilChar# :: StringBuffer -> Char# -> StringBuffer
+stepOnUntilChar# (StringBuffer fo l# s# c#) x# =
loop c#
where
loop c#
- | indexCharOffAddr# fo c# `eqChar#` x#
- = StringBuffer fo l# s# c#
+ | c# >=# l# || indexCharOffAddr# fo c# `eqChar#` x#
+ = StringBuffer fo l# c# c#
| otherwise
= loop (c# +# 1#)
else
unpackCStringBA (copySubStr (A# fo) (I# start_pos#) (I# (current# -# start_pos#)))
-lexemeToByteArray :: StringBuffer -> _ByteArray Int
+lexemeToByteArray :: StringBuffer -> ByteArray Int
lexemeToByteArray (StringBuffer fo _ start_pos# current#) =
if start_pos# ==# current# then
error "lexemeToByteArray"