2 % (c) The University of Glasgow 2006
3 % (c) The University of Glasgow, 1997-2006
6 Buffers for scanning string input stored in external arrays.
12 -- non-abstract for vs\/HaskellService
14 -- * Creation\/destruction
16 hGetStringBufferBlock,
26 -- * Moving and comparison
39 #include "HsVersions.h"
42 import FastString hiding ( buf )
47 import System.IO ( hGetBuf, hFileSize,IOMode(ReadMode), hClose
52 #if !defined(__GLASGOW_HASKELL__) || __GLASGOW_HASKELL__ >= 601
53 import System.IO ( openBinaryFile )
55 import IOExts ( openFileEx, IOModeEx(..) )
58 #if defined(__GLASGOW_HASKELL__) && __GLASGOW_HASKELL__ < 601
59 openBinaryFile fp mode = openFileEx fp (BinaryMode mode)
62 -- -----------------------------------------------------------------------------
63 -- The StringBuffer type
65 -- |A StringBuffer is an internal pointer to a sized chunk of bytes.
66 -- The bytes are intended to be *immutable*. There are pure
67 -- operations to read the contents of a StringBuffer.
69 -- A StringBuffer may have a finalizer, depending on how it was
74 buf :: {-# UNPACK #-} !(ForeignPtr Word8),
75 len :: {-# UNPACK #-} !Int, -- length
76 cur :: {-# UNPACK #-} !Int -- current pos
78 -- The buffer is assumed to be UTF-8 encoded, and furthermore
79 -- we add three '\0' bytes to the end as sentinels so that the
80 -- decoder doesn't have to check for overflow at every single byte
81 -- of a multibyte sequence.
83 instance Show StringBuffer where
84 showsPrec _ s = showString "<stringbuffer("
85 . shows (len s) . showString "," . shows (cur s)
88 -- -----------------------------------------------------------------------------
89 -- Creation / Destruction
91 hGetStringBuffer :: FilePath -> IO StringBuffer
92 hGetStringBuffer fname = do
93 h <- openBinaryFile fname ReadMode
95 let size = fromIntegral size_i
96 buf <- mallocForeignPtrArray (size+3)
97 withForeignPtr buf $ \ptr -> do
98 r <- if size == 0 then return 0 else hGetBuf h ptr size
101 then ioError (userError "short read of file")
102 else newUTF8StringBuffer buf ptr size
104 hGetStringBufferBlock :: Handle -> Int -> IO StringBuffer
105 hGetStringBufferBlock handle wanted
106 = do size_i <- hFileSize handle
107 offset_i <- hTell handle
108 let size = min wanted (fromIntegral $ size_i-offset_i)
109 buf <- mallocForeignPtrArray (size+3)
110 withForeignPtr buf $ \ptr ->
111 do r <- if size == 0 then return 0 else hGetBuf handle ptr size
113 then ioError (userError $ "short read of file: "++show(r,size,size_i,handle))
114 else newUTF8StringBuffer buf ptr size
116 newUTF8StringBuffer :: ForeignPtr Word8 -> Ptr Word8 -> Int -> IO StringBuffer
117 newUTF8StringBuffer buf ptr size = do
118 pokeArray (ptr `plusPtr` size :: Ptr Word8) [0,0,0]
119 -- sentinels for UTF-8 decoding
121 sb0 = StringBuffer buf size 0
122 (first_char, sb1) = nextChar sb0
123 -- skip the byte-order mark if there is one (see #1744)
124 -- This is better than treating #FEFF as whitespace,
125 -- because that would mess up layout. We don't have a concept
126 -- of zero-width whitespace in Haskell: all whitespace codepoints
127 -- have a width of one column.
128 return (if first_char == '\xfeff' then sb1 else sb0)
130 appendStringBuffers :: StringBuffer -> StringBuffer -> IO StringBuffer
131 appendStringBuffers sb1 sb2
132 = do newBuf <- mallocForeignPtrArray (size+3)
133 withForeignPtr newBuf $ \ptr ->
134 withForeignPtr (buf sb1) $ \sb1Ptr ->
135 withForeignPtr (buf sb2) $ \sb2Ptr ->
136 do copyArray (sb1Ptr `advancePtr` cur sb1) ptr (calcLen sb1)
137 copyArray (sb2Ptr `advancePtr` cur sb2) (ptr `advancePtr` cur sb1) (calcLen sb2)
138 pokeArray (ptr `advancePtr` size) [0,0,0]
139 return (StringBuffer newBuf size 0)
140 where calcLen sb = len sb - cur sb
141 size = calcLen sb1 + calcLen sb2
143 stringToStringBuffer :: String -> IO StringBuffer
144 stringToStringBuffer str = do
145 let size = utf8EncodedLength str
146 buf <- mallocForeignPtrArray (size+3)
147 withForeignPtr buf $ \ptr -> do
148 utf8EncodeString ptr str
149 pokeArray (ptr `plusPtr` size :: Ptr Word8) [0,0,0]
150 -- sentinels for UTF-8 decoding
151 return (StringBuffer buf size 0)
153 -- -----------------------------------------------------------------------------
156 -- Getting our fingers dirty a little here, but this is performance-critical
157 {-# INLINE nextChar #-}
158 nextChar :: StringBuffer -> (Char,StringBuffer)
159 nextChar (StringBuffer buf len (I# cur#)) =
161 withForeignPtr buf $ \(Ptr a#) -> do
162 case utf8DecodeChar# (a# `plusAddr#` cur#) of
164 let cur' = I# (b# `minusAddr#` a#) in
165 return (C# c#, StringBuffer buf len cur')
167 currentChar :: StringBuffer -> Char
168 currentChar = fst . nextChar
170 prevChar :: StringBuffer -> Char -> Char
171 prevChar (StringBuffer _ _ 0) deflt = deflt
172 prevChar (StringBuffer buf _ cur) _ =
174 withForeignPtr buf $ \p -> do
175 p' <- utf8PrevChar (p `plusPtr` cur)
176 return (fst (utf8DecodeChar p'))
178 -- -----------------------------------------------------------------------------
181 stepOn :: StringBuffer -> StringBuffer
182 stepOn s = snd (nextChar s)
184 offsetBytes :: Int -> StringBuffer -> StringBuffer
185 offsetBytes i s = s { cur = cur s + i }
187 byteDiff :: StringBuffer -> StringBuffer -> Int
188 byteDiff s1 s2 = cur s2 - cur s1
190 atEnd :: StringBuffer -> Bool
191 atEnd (StringBuffer _ l c) = l == c
193 -- -----------------------------------------------------------------------------
196 lexemeToString :: StringBuffer -> Int {-bytes-} -> String
197 lexemeToString _ 0 = ""
198 lexemeToString (StringBuffer buf _ cur) bytes =
200 withForeignPtr buf $ \ptr ->
201 utf8DecodeString (ptr `plusPtr` cur) bytes
203 lexemeToFastString :: StringBuffer -> Int {-bytes-} -> FastString
204 lexemeToFastString _ 0 = nilFS
205 lexemeToFastString (StringBuffer buf _ cur) len =
207 withForeignPtr buf $ \ptr ->
208 return $! mkFastStringBytes (ptr `plusPtr` cur) len
210 -- -----------------------------------------------------------------------------
211 -- Parsing integer strings in various bases
213 byteOff :: StringBuffer -> Int -> Char
214 byteOff (StringBuffer buf _ cur) i =
215 inlinePerformIO $ withForeignPtr buf $ \ptr -> do
216 -- return $! cBox (indexWord8OffFastPtrAsFastChar
217 -- (pUnbox ptr) (iUnbox (cur+i)))
219 -- w <- peek (ptr `plusPtr` (cur+i))
220 -- return (unsafeChr (fromIntegral (w::Word8)))
222 -- | XXX assumes ASCII digits only (by using byteOff)
223 parseUnsignedInteger :: StringBuffer -> Int -> Integer -> (Char->Int) -> Integer
224 parseUnsignedInteger (StringBuffer buf _ cur) len radix char_to_int
225 = inlinePerformIO $ withForeignPtr buf $ \ptr -> return $! let
226 --LOL, in implementations where the indexing needs slow unsafePerformIO,
227 --this is less (not more) efficient than using the IO monad explicitly
230 byteOff i = cBox (indexWord8OffFastPtrAsFastChar ptr' (iUnbox (cur + i)))
231 go i x | i == len = x
232 | otherwise = case byteOff i of
233 char -> go (i + 1) (x * radix + toInteger (char_to_int char))