{-

Copyright (c) 2002, members of the Haskell Internationalisation Working
Group All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

* Redistributions of source code must retain the above copyright notice,
   this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright notice,
   this list of conditions and the following disclaimer in the
   documentation and/or other materials provided with the distribution.
* Neither the name of the Haskell Internationalisation Working Group nor
   the names of its contributors may be used to endorse or promote products
   derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.

This module provides lazy stream encoding/decoding facilities for UTF-8,
the Unicode Transformation Format with 8-bit words.

2002-09-02  Sven Moritz Hallberg <pesco@gmx.de>

-}

{-

2007-04-30 Henning Thielemann:
Slight changes to make decode lazy.
The calls of 'reverse' in the original version have broken laziness
and thus had memory leaks.

-}

module Data.String.UTF8
  ( encode
  , decode
  , decodeEmbedErrors
  , encodeOne
  , decodeOne
  , Error
         -- Haddock does not want to document signatures with private types
         -- these functions should be moved to a utility module
  ) where

import Data.Char (ord, chr)
import Data.Word (Word8, Word16, Word32)
import Data.Bits (Bits, shiftL, shiftR, (.&.), (.|.))

import Data.List (unfoldr)

-- - UTF-8 in General -

-- Adapted from the Unicode standard, version 3.2,
-- Table 3.1 "UTF-8 Bit Distribution" (excluded are UTF-16 encodings):

--   Scalar                    1st Byte  2nd Byte  3rd Byte  4th Byte
--           000000000xxxxxxx  0xxxxxxx
--           00000yyyyyxxxxxx  110yyyyy  10xxxxxx
--           zzzzyyyyyyxxxxxx  1110zzzz  10yyyyyy  10xxxxxx
--   000uuuzzzzzzyyyyyyxxxxxx  11110uuu  10zzzzzz  10yyyyyy  10xxxxxx

-- Also from the Unicode standard, version 3.2,
-- Table 3.1B "Legal UTF-8 Byte Sequences":

--   Code Points         1st Byte  2nd Byte  3rd Byte  4th Byte
--     U+0000..U+007F    00..7F
--     U+0080..U+07FF    C2..DF    80..BF
--     U+0800..U+0FFF    E0        A0..BF    80..BF
--     U+1000..U+CFFF    E1..EC    80..BF    80..BF
--     U+D000..U+D7FF    ED        80..9F    80..BF
--     U+D800..U+DFFF    ill-formed
--     U+E000..U+FFFF    EE..EF    80..BF    80..BF
--    U+10000..U+3FFFF   F0        90..BF    80..BF    80..BF
--    U+40000..U+FFFFF   F1..F3    80..BF    80..BF    80..BF
--   U+100000..U+10FFFF  F4        80..8F    80..BF    80..BF



-- - Encoding Functions -

-- Must the encoder ensure that no illegal byte sequences are output or
-- can we trust the Haskell system to supply only legal values?
-- For now I include error case for the surrogate values U+D800..U+DFFF and
-- out-of-range scalars.

-- The function is pretty much a transscript of table 3.1B with error checks.
-- It dispatches the actual encoding to functions specific to the number of
-- required bytes.

encodeOne :: Char -> [Word8]
encodeOne c
    -- The report guarantees in (6.1.2) that this won't happen:
    --   | n < 0       = error "encodeUTF8: ord returned a negative value"
    | n < 0x0080  = encodeOne_onebyte n8
    | n < 0x0800  = encodeOne_twobyte n16
    | n < 0xD800  = encodeOne_threebyte n16
    | n < 0xE000  = error "encodeUTF8: ord returned a surrogate value"
    | n < 0x10000       = encodeOne_threebyte n16
    -- Haskell 98 only talks about 16 bit characters, but ghc handles 20.1.
    | n < 0x10FFFF      = encodeOne_fourbyte n32
    | otherwise  = error "encodeUTF8: ord returned a value above 0x10FFFF"
    where
    n = ord c            :: Int
    n8 = fromIntegral n  :: Word8
    n16 = fromIntegral n :: Word16
    n32 = fromIntegral n :: Word32


-- With the above, a stream decoder is trivial:

encode :: [Char] -> [Word8]
encode = concatMap encodeOne


-- Now follow the individual encoders for certain numbers of bytes...
--           _
--          / |  __  ___  __ __
--         / ^| //  /__/ // //
--        /.==| \\ //_  // //
-- It's  //  || // \_/_//_//_  and it's here to stay!

encodeOne_onebyte :: Word8 -> [Word8]
encodeOne_onebyte cp = [cp]


-- 00000yyyyyxxxxxx -> 110yyyyy 10xxxxxx

encodeOne_twobyte :: Word16 -> [Word8]
encodeOne_twobyte cp = [(0xC0.|.ys), (0x80.|.xs)]
    where
    xs, ys :: Word8
    ys = fromIntegral (shiftR cp 6)
    xs = (fromIntegral cp) .&. 0x3F


-- zzzzyyyyyyxxxxxx -> 1110zzzz 10yyyyyy 10xxxxxx

encodeOne_threebyte :: Word16 -> [Word8]
encodeOne_threebyte cp = [(0xE0.|.zs), (0x80.|.ys), (0x80.|.xs)]
    where
    xs, ys, zs :: Word8
    xs = (fromIntegral cp) .&. 0x3F
    ys = (fromIntegral (shiftR cp 6)) .&. 0x3F
    zs = fromIntegral (shiftR cp 12)


-- 000uuuzzzzzzyyyyyyxxxxxx -> 11110uuu 10zzzzzz 10yyyyyy 10xxxxxx

encodeOne_fourbyte :: Word32 -> [Word8]
encodeOne_fourbyte cp = [0xF0.|.us, 0x80.|.zs, 0x80.|.ys, 0x80.|.xs]
    where
    xs, ys, zs, us :: Word8
    xs = (fromIntegral cp) .&. 0x3F
    ys = (fromIntegral (shiftR cp 6)) .&. 0x3F
    zs = (fromIntegral (shiftR cp 12)) .&. 0x3F
    us = fromIntegral (shiftR cp 18)



-- - Decoding -

-- The decoding is a bit more involved. The byte sequence could contain all
-- sorts of corruptions. The user must be able to either notice or ignore these
-- errors.

-- I will first look at the decoding of a single character. The process
-- consumes a certain number of bytes from the input. It returns the
-- remaining input and either an error and the index of its occurance in the
-- byte sequence or the decoded character.

data Error

-- The first byte in a sequence starts with either zero, two, three, or four
-- ones and one zero to indicate the length of the sequence. If it doesn't,
-- it is invalid. It is dropped and the next byte interpreted as the start
-- of a new sequence.

    = InvalidFirstByte

-- All bytes in the sequence except the first match the bit pattern 10xxxxxx.
-- If one doesn't, it is invalid. The sequence up to that point is dropped
-- and the "invalid" byte interpreted as the start of a new sequence. The error
-- includes the length of the partial sequence and the number of expected bytes.

    | InvalidLaterByte Int      -- the byte at relative index n was invalid

-- If a sequence ends prematurely, it has been truncated. It dropped and
-- decoding stops. The error reports the actual and expected lengths of the
-- sequence.

    | Truncated Int Int         -- only n of m expected bytes were present

-- Some sequences would represent code points which would be encoded as a
-- shorter sequence by a conformant encoder. Such non-shortest sequences are
-- considered erroneous and dropped. The error reports the actual and
-- expected number of bytes used.

    | NonShortest Int Int       -- n instead of m bytes were used

-- Unicode code points are in the range of [0..0x10FFFF]. Any values outside
-- of those bounds are simply invalid.

    | ValueOutOfBounds

-- There is no such thing as "surrogate pairs" any more in UTF-8. The
-- corresponding code points now form illegal byte sequences.

    | Surrogate
      deriving (Show, Eq)


-- Second, third, and fourth bytes share the common requirement to start
-- with the bit sequence 10. So, here's the function to check that property.

first_bits_not_10 :: Word8 -> Bool
first_bits_not_10 b
    | (b.&.0xC0) /= 0x80  = True
    | otherwise           = False


-- Erm, OK, the single-character decoding function's return type is a bit
-- longish. It is a tripel:

--  - The first component contains the decoded character or an error
--    if the byte sequence was erroneous.
--  - The second component contains the number of bytes that were consumed
--    from the input.
--  - The third component contains the remaining bytes of input.

decodeOne :: [Word8] -> (Either Error Char, Int, [Word8])
decodeOne bs@(b1:rest)
    | b1 < 0x80   = decodeOne_onebyte bs
    | b1 < 0xC0   = (Left InvalidFirstByte, 1, rest)
    | b1 < 0xE0   = decodeOne_twobyte bs
    | b1 < 0xF0   = decodeOne_threebyte bs
    | b1 < 0xF5   = decodeOne_fourbyte bs
    | otherwise   = (Left ValueOutOfBounds, 1, rest)
decodeOne [] = error "UTF8.decodeOne: No input"


-- 0xxxxxxx -> 000000000xxxxxxx

decodeOne_onebyte :: [Word8] -> (Either Error Char, Int, [Word8])
decodeOne_onebyte (b:bs) = (Right (cpToChar b), 1, bs)
decodeOne_onebyte[] = error "UTF8.decodeOne_onebyte: No input (can't happen)"

cpToChar :: Integral a => a -> Char
cpToChar = chr . fromIntegral


-- 110yyyyy 10xxxxxx -> 00000yyyyyxxxxxx

decodeOne_twobyte :: [Word8] -> (Either Error Char, Int, [Word8])
decodeOne_twobyte (_:[])
    = (Left (Truncated 1 2), 1, [])
decodeOne_twobyte (b1:b2:bs)
    | b1 < 0xC2            = (Left (NonShortest 2 1), 2, bs)
    | first_bits_not_10 b2 = (Left (InvalidLaterByte 1), 1, (b2:bs))
    | otherwise            = (Right (cpToChar result), 2, bs)
    where
    xs, ys, result :: Word32
    xs = fromIntegral (b2.&.0x3F)
    ys = fromIntegral (b1.&.0x1F)
    result = shiftL ys 6 .|. xs
decodeOne_twobyte[] = error "UTF8.decodeOne_twobyte: No input (can't happen)"


-- 1110zzzz 10yyyyyy 10xxxxxx -> zzzzyyyyyyxxxxxx

decodeOne_threebyte :: [Word8] -> (Either Error Char, Int, [Word8])
decodeOne_threebyte (_:[])   = threebyte_truncated 1
decodeOne_threebyte (_:_:[]) = threebyte_truncated 2
decodeOne_threebyte bs@(b1:b2:b3:rest)
    | first_bits_not_10 b2
        = (Left (InvalidLaterByte 1), 1, drop 1 bs)
    | first_bits_not_10 b3
        = (Left (InvalidLaterByte 2), 2, drop 2 bs)
    | result < 0x0080
        = (Left (NonShortest 3 1), 3, rest)
    | result < 0x0800
        = (Left (NonShortest 3 2), 3, rest)
    | result >= 0xD800 && result < 0xE000
        = (Left Surrogate, 3, rest)
    | otherwise
        = (Right (cpToChar result), 3, rest)
    where
    xs, ys, zs, result :: Word32
    xs = fromIntegral (b3.&.0x3F)
    ys = fromIntegral (b2.&.0x3F)
    zs = fromIntegral (b1.&.0x0F)
    result = shiftL zs 12 .|. shiftL ys 6 .|. xs
decodeOne_threebyte[]
 = error "UTF8.decodeOne_threebyte: No input (can't happen)"

threebyte_truncated :: Int -> (Either Error Char, Int, [Word8])
threebyte_truncated n = (Left (Truncated n 3), n, [])


-- 11110uuu 10zzzzzz 10yyyyyy 10xxxxxx -> 000uuuzzzzzzyyyyyyxxxxxx

decodeOne_fourbyte :: [Word8] -> (Either Error Char, Int, [Word8])
decodeOne_fourbyte (_:[])     = fourbyte_truncated 1
decodeOne_fourbyte (_:_:[])   = fourbyte_truncated 2
decodeOne_fourbyte (_:_:_:[]) = fourbyte_truncated 3
decodeOne_fourbyte bs@(b1:b2:b3:b4:rest)
    | first_bits_not_10 b2
        = (Left (InvalidLaterByte 1), 1, drop 1 bs)
    | first_bits_not_10 b3
        = (Left (InvalidLaterByte 2), 2, drop 2 bs)
    | first_bits_not_10 b4
        = (Left (InvalidLaterByte 3), 3, drop 3 bs)
    | result < 0x0080
        = (Left (NonShortest 4 1), 4, rest)
    | result < 0x0800
        = (Left (NonShortest 4 2), 4, rest)
    | result < 0x10000
        = (Left (NonShortest 4 3), 4, rest)
    | result > 0x10FFFF
        = (Left ValueOutOfBounds, 4, rest)
    | otherwise
        = (Right (cpToChar result), 4, rest)
    where
    xs, ys, zs, us, result :: Word32
    xs = fromIntegral (b4 .&. 0x3F)
    ys = fromIntegral (b3 .&. 0x3F)
    zs = fromIntegral (b2 .&. 0x3F)
    us = fromIntegral (b1 .&. 0x07)
    result = xs .|. shiftL ys 6 .|. shiftL zs 12 .|. shiftL us 18
decodeOne_fourbyte[]
 = error "UTF8.decodeOne_fourbyte: No input (can't happen)"

fourbyte_truncated :: Int -> (Either Error Char, Int, [Word8])
fourbyte_truncated n = (Left (Truncated n 4), n, [])


-- The decoder examines all input, recording decoded characters as well as
-- error-index pairs along the way.

decode :: [Word8] -> ([Char], [(Error,Int)])
decode = swap . partitionEither . decodeEmbedErrors

decodeEmbedErrors :: [Word8] -> [Either (Error,Int) Char]
decodeEmbedErrors =
   unfoldr (\(pos,xs) ->
       toMaybe
          (not $ null xs)
          (let (c,n,rest) = decodeOne xs
           in  (either (\err -> Left (err,pos)) Right c,
                (pos+n,rest)))) .
   (,) 0

swap :: (a,b) -> (b,a)
swap (x,y) = (y,x)
{-# INLINE swap #-}

partitionEither :: [Either a b] -> ([a], [b])
partitionEither =
   foldr (\x ~(ls,rs) -> either (\l -> (l:ls,rs)) (\r -> (ls,r:rs)) x) ([],[])
{-# INLINE partitionEither #-}

toMaybe :: Bool -> a -> Maybe a
toMaybe False _ = Nothing
toMaybe True  x = Just x
{-# INLINE toMaybe #-}

-- ------------------------------------------------------------