2 -- -----------------------------------------------------------------------------
4 -- (c) The University of Glasgow, 1997-2003
8 -- -----------------------------------------------------------------------------
31 #define COMPILING_FAST_STRING
32 #include "HsVersions.h"
34 import Data.Char ( ord, chr, isDigit, digitToInt, isHexDigit )
35 import Numeric ( showHex )
37 import GHC.Ptr ( Ptr(..) )
40 -- -----------------------------------------------------------------------------
43 latin1DecodeChar ptr = do
45 return (unsafeChr (fromIntegral w), ptr `plusPtr` 1)
47 latin1EncodeChar c ptr = do
48 poke ptr (fromIntegral (ord c))
49 return (ptr `plusPtr` 1)
51 -- -----------------------------------------------------------------------------
54 -- We can't write the decoder as efficiently as we'd like without
55 -- resorting to unboxed extensions, unfortunately. I tried to write
56 -- an IO version of this function, but GHC can't eliminate boxed
57 -- results from an IO-returning function.
59 -- We assume we can ignore overflow when parsing a multibyte character here.
60 -- To make this safe, we add extra sentinel bytes to unparsed UTF-8 sequences
61 -- before decoding them (see StringBuffer.hs).
63 {-# INLINE utf8DecodeChar# #-}
64 utf8DecodeChar# :: Addr# -> (# Char#, Addr# #)
66 let ch0 = word2Int# (indexWord8OffAddr# a# 0#) in
68 _ | ch0 <=# 0x7F# -> (# chr# ch0, a# `plusAddr#` 1# #)
70 | ch0 >=# 0xC0# && ch0 <=# 0xDF# ->
71 let ch1 = word2Int# (indexWord8OffAddr# a# 1#) in
72 if ch1 <# 0x80# || ch1 >=# 0xC0# then fail 1# else
73 (# chr# (((ch0 -# 0xC0#) `uncheckedIShiftL#` 6#) +#
77 | ch0 >=# 0xE0# && ch0 <=# 0xEF# ->
78 let ch1 = word2Int# (indexWord8OffAddr# a# 1#) in
79 if ch1 <# 0x80# || ch1 >=# 0xC0# then fail 1# else
80 let ch2 = word2Int# (indexWord8OffAddr# a# 2#) in
81 if ch2 <# 0x80# || ch2 >=# 0xC0# then fail 2# else
82 (# chr# (((ch0 -# 0xE0#) `uncheckedIShiftL#` 12#) +#
83 ((ch1 -# 0x80#) `uncheckedIShiftL#` 6#) +#
87 | ch0 >=# 0xF0# && ch0 <=# 0xF8# ->
88 let ch1 = word2Int# (indexWord8OffAddr# a# 1#) in
89 if ch1 <# 0x80# || ch1 >=# 0xC0# then fail 1# else
90 let ch2 = word2Int# (indexWord8OffAddr# a# 2#) in
91 if ch2 <# 0x80# || ch2 >=# 0xC0# then fail 2# else
92 let ch3 = word2Int# (indexWord8OffAddr# a# 3#) in
93 if ch3 <# 0x80# || ch3 >=# 0xC0# then fail 3# else
94 (# chr# (((ch0 -# 0xF0#) `uncheckedIShiftL#` 18#) +#
95 ((ch1 -# 0x80#) `uncheckedIShiftL#` 12#) +#
96 ((ch2 -# 0x80#) `uncheckedIShiftL#` 6#) +#
100 | otherwise -> fail 1#
102 -- all invalid sequences end up here:
103 fail n = (# '\0'#, a# `plusAddr#` n #)
104 -- '\xFFFD' would be the usual replacement character, but
105 -- that's a valid symbol in Haskell, so will result in a
106 -- confusing parse error later on. Instead we use '\0' which
107 -- will signal a lexer error immediately.
109 utf8DecodeChar :: Ptr Word8 -> (Char, Ptr Word8)
110 utf8DecodeChar (Ptr a#) = ( C# c#, Ptr b# )
111 where (# c#, b# #) = utf8DecodeChar# a#
113 -- UTF-8 is cleverly designed so that we can always figure out where
114 -- the start of the current character is, given any position in a
115 -- stream. This function finds the start of the previous character,
116 -- assuming there *is* a previous character.
117 utf8PrevChar :: Ptr Word8 -> IO (Ptr Word8)
118 utf8PrevChar p = utf8CharStart (p `plusPtr` (-1))
120 utf8CharStart :: Ptr Word8 -> IO (Ptr Word8)
121 utf8CharStart p = go p
122 where go p = do w <- peek p
123 if (w .&. 0xC0) == 0x80
124 then go (p `plusPtr` (-1))
127 utf8DecodeString :: Ptr Word8 -> Int -> IO [Char]
128 STRICT2(utf8DecodeString)
129 utf8DecodeString (Ptr a#) (I# len#)
132 end# = addr2Int# (a# `plusAddr#` len#)
135 | addr2Int# p# >=# end# = return []
137 case utf8DecodeChar# p# of
142 countUTF8Chars :: Ptr Word8 -> Int -> IO Int
143 countUTF8Chars ptr bytes = go ptr 0
145 end = ptr `plusPtr` bytes
149 | ptr >= end = return n
151 case utf8DecodeChar# (unPtr ptr) of
152 (# c, a #) -> go (Ptr a) (n+1)
156 utf8EncodeChar c ptr =
159 _ | x > 0 && x <= 0x007f -> do
160 poke ptr (fromIntegral x)
161 return (ptr `plusPtr` 1)
162 -- NB. '\0' is encoded as '\xC0\x80', not '\0'. This is so that we
163 -- can have 0-terminated UTF-8 strings (see GHC.Base.unpackCStringUtf8).
165 poke ptr (fromIntegral (0xC0 .|. ((x `shiftR` 6) .&. 0x1F)))
166 pokeElemOff ptr 1 (fromIntegral (0x80 .|. (x .&. 0x3F)))
167 return (ptr `plusPtr` 2)
169 poke ptr (fromIntegral (0xE0 .|. (x `shiftR` 12) .&. 0x0F))
170 pokeElemOff ptr 1 (fromIntegral (0x80 .|. (x `shiftR` 6) .&. 0x3F))
171 pokeElemOff ptr 2 (fromIntegral (0x80 .|. (x .&. 0x3F)))
172 return (ptr `plusPtr` 3)
174 poke ptr (fromIntegral (0xF0 .|. (x `shiftR` 18)))
175 pokeElemOff ptr 1 (fromIntegral (0x80 .|. ((x `shiftR` 12) .&. 0x3F)))
176 pokeElemOff ptr 2 (fromIntegral (0x80 .|. ((x `shiftR` 6) .&. 0x3F)))
177 pokeElemOff ptr 3 (fromIntegral (0x80 .|. (x .&. 0x3F)))
178 return (ptr `plusPtr` 4)
180 utf8EncodeString :: Ptr Word8 -> String -> IO ()
181 utf8EncodeString ptr str = go ptr str
183 go ptr [] = return ()
185 ptr' <- utf8EncodeChar c ptr
188 utf8EncodedLength :: String -> Int
189 utf8EncodedLength str = go 0 str
193 | ord c > 0 && ord c <= 0x007f = go (n+1) cs
194 | ord c <= 0x07ff = go (n+2) cs
195 | ord c <= 0xffff = go (n+3) cs
196 | otherwise = go (n+4) cs
198 -- -----------------------------------------------------------------------------
202 This is the main name-encoding and decoding function. It encodes any
203 string into a string that is acceptable as a C name. This is the name
204 by which things are known right through the compiler.
206 The basic encoding scheme is this.
208 * Tuples (,,,) are coded as Z3T
210 * Alphabetic characters (upper and lower) and digits
211 all translate to themselves;
212 except 'Z', which translates to 'ZZ'
213 and 'z', which translates to 'zz'
214 We need both so that we can preserve the variable/tycon distinction
216 * Most other printable characters translate to 'zx' or 'Zx' for some
217 alphabetic character x
219 * The others translate as 'znnnU' where 'nnn' is the decimal number
223 --------------------------
235 (# #) Z1H unboxed 1-tuple (note the space)
236 (#,,,,#) Z5H unboxed 5-tuple
237 (NB: There is no Z1T nor Z0H.)
240 type UserString = String -- As the user typed it
241 type EncodedString = String -- Encoded form
244 zEncodeString :: UserString -> EncodedString
245 zEncodeString cs = case maybe_tuple cs of
246 Just n -> n -- Tuples go to Z2T etc
250 go (c:cs) = encode_ch c ++ go cs
252 unencodedChar :: Char -> Bool -- True for chars that don't need encoding
253 unencodedChar 'Z' = False
254 unencodedChar 'z' = False
255 unencodedChar c = c >= 'a' && c <= 'z'
256 || c >= 'A' && c <= 'Z'
257 || c >= '0' && c <= '9'
259 encode_ch :: Char -> EncodedString
260 encode_ch c | unencodedChar c = [c] -- Common case first
263 encode_ch '(' = "ZL" -- Needed for things like (,), and (->)
264 encode_ch ')' = "ZR" -- For symmetry with (
284 encode_ch '\'' = "zq"
285 encode_ch '\\' = "zr"
290 encode_ch c = 'z' : if isDigit (head hex_str) then hex_str
292 where hex_str = showHex (ord c) "U"
293 -- ToDo: we could improve the encoding here in various ways.
294 -- eg. strings of unicode characters come out as 'z1234Uz5678U', we
295 -- could remove the 'U' in the middle (the 'z' works as a separator).
297 zDecodeString :: EncodedString -> UserString
298 zDecodeString [] = []
299 zDecodeString ('Z' : d : rest)
300 | isDigit d = decode_tuple d rest
301 | otherwise = decode_upper d : zDecodeString rest
302 zDecodeString ('z' : d : rest)
303 | isDigit d = decode_num_esc d rest
304 | otherwise = decode_lower d : zDecodeString rest
305 zDecodeString (c : rest) = c : zDecodeString rest
307 decode_upper, decode_lower :: Char -> Char
309 decode_upper 'L' = '('
310 decode_upper 'R' = ')'
311 decode_upper 'M' = '['
312 decode_upper 'N' = ']'
313 decode_upper 'C' = ':'
314 decode_upper 'Z' = 'Z'
315 decode_upper ch = {-pprTrace "decode_upper" (char ch)-} ch
317 decode_lower 'z' = 'z'
318 decode_lower 'a' = '&'
319 decode_lower 'b' = '|'
320 decode_lower 'c' = '^'
321 decode_lower 'd' = '$'
322 decode_lower 'e' = '='
323 decode_lower 'g' = '>'
324 decode_lower 'h' = '#'
325 decode_lower 'i' = '.'
326 decode_lower 'l' = '<'
327 decode_lower 'm' = '-'
328 decode_lower 'n' = '!'
329 decode_lower 'p' = '+'
330 decode_lower 'q' = '\''
331 decode_lower 'r' = '\\'
332 decode_lower 's' = '/'
333 decode_lower 't' = '*'
334 decode_lower 'u' = '_'
335 decode_lower 'v' = '%'
336 decode_lower ch = {-pprTrace "decode_lower" (char ch)-} ch
338 -- Characters not having a specific code are coded as z224U (in hex)
339 decode_num_esc d rest
340 = go (digitToInt d) rest
342 go n (c : rest) | isHexDigit c = go (16*n + digitToInt c) rest
343 go n ('U' : rest) = chr n : zDecodeString rest
344 go n other = error ("decode_num_esc: " ++ show n ++ ' ':other)
346 decode_tuple :: Char -> EncodedString -> UserString
348 = go (digitToInt d) rest
350 -- NB. recurse back to zDecodeString after decoding the tuple, because
351 -- the tuple might be embedded in a longer name.
352 go n (c : rest) | isDigit c = go (10*n + digitToInt c) rest
353 go 0 ('T':rest) = "()" ++ zDecodeString rest
354 go n ('T':rest) = '(' : replicate (n-1) ',' ++ ")" ++ zDecodeString rest
355 go 1 ('H':rest) = "(# #)" ++ zDecodeString rest
356 go n ('H':rest) = '(' : '#' : replicate (n-1) ',' ++ "#)" ++ zDecodeString rest
357 go n other = error ("decode_tuple: " ++ show n ++ ' ':other)
360 Tuples are encoded as
362 for 3-tuples or unboxed 3-tuples respectively. No other encoding starts
365 * "(# #)" is the tycon for an unboxed 1-tuple (not 0-tuple)
366 There are no unboxed 0-tuples.
368 * "()" is the tycon for a boxed 0-tuple.
369 There are no boxed 1-tuples.
372 maybe_tuple :: UserString -> Maybe EncodedString
374 maybe_tuple "(# #)" = Just("Z1H")
375 maybe_tuple ('(' : '#' : cs) = case count_commas (0::Int) cs of
376 (n, '#' : ')' : cs) -> Just ('Z' : shows (n+1) "H")
378 maybe_tuple "()" = Just("Z0T")
379 maybe_tuple ('(' : cs) = case count_commas (0::Int) cs of
380 (n, ')' : cs) -> Just ('Z' : shows (n+1) "T")
382 maybe_tuple other = Nothing
384 count_commas :: Int -> String -> (Int, String)
385 count_commas n (',' : cs) = count_commas (n+1) cs
386 count_commas n cs = (n,cs)