Lib/io.py

"""New I/O library conforming to PEP 3116.

This is an early prototype; eventually some of this will be
reimplemented in C and the rest may be turned into a package.

Conformance of alternative implementations: all arguments are intended
to be positional-only except the arguments of the open() function.
Argument names except those of the open() function are not part of the
specification.  Instance variables and methods whose name starts with
a leading underscore are not part of the specification (except "magic"
names like __iter__).  Only the top-level names listed in the __all__
variable are part of the specification.

XXX need to default buffer size to 1 if isatty()
XXX need to support 1 meaning line-buffered
XXX don't use assert to validate input requirements
XXX whenever an argument is None, use the default value
XXX read/write ops should check readable/writable
"""

__author__ = ("Guido van Rossum <guido@python.org>, "
              "Mike Verdone <mike.verdone@gmail.com>, "
              "Mark Russell <mark.russell@zen.co.uk>")

__all__ = ["BlockingIOError", "open", "IOBase", "RawIOBase", "FileIO",
           "SocketIO", "BytesIO", "StringIO", "BufferedIOBase",
           "BufferedReader", "BufferedWriter", "BufferedRWPair",
           "BufferedRandom", "TextIOBase", "TextIOWrapper"]

import os
import sys
import codecs
import pickle
import _fileio
import warnings

# XXX Shouldn't we use st_blksize whenever we can?
DEFAULT_BUFFER_SIZE = 8 * 1024  # bytes


class BlockingIOError(IOError):

    """Exception raised when I/O would block on a non-blocking I/O stream."""

    def __init__(self, errno, strerror, characters_written=0):
        IOError.__init__(self, errno, strerror)
        self.characters_written = characters_written


def open(file, mode="r", buffering=None, *, encoding=None, newline=None):
    """Replacement for the built-in open function.

    Args:
      file: string giving the name of the file to be opened;
            or integer file descriptor of the file to be wrapped (*).
      mode: optional mode string; see below.
      buffering: optional int >= 0 giving the buffer size; values
                 can be: 0 = unbuffered, 1 = line buffered,
                 larger = fully buffered.
    Keywords (for text modes only; *must* be given as keyword arguments):
      encoding: optional string giving the text encoding.
      newline: optional newlines specifier; must be None, '\n' or '\r\n';
               specifies the line ending expected on input and written on
               output.  If None, use universal newlines on input and
               use os.linesep on output.

    (*) If a file descriptor is given, it is closed when the returned
    I/O object is closed.  If you don't want this to happen, use
    os.dup() to create a duplicate file descriptor.

    Mode strings characters:
      'r': open for reading (default)
      'w': open for writing, truncating the file first
      'a': open for writing, appending to the end if the file exists
      'b': binary mode
      't': text mode (default)
      '+': open a disk file for updating (implies reading and writing)
      'U': universal newline mode (for backwards compatibility)

    Constraints:
      - encoding must not be given when a binary mode is given
      - buffering must not be zero when a text mode is given

    Returns:
      Depending on the mode and buffering arguments, either a raw
      binary stream, a buffered binary stream, or a buffered text
      stream, open for reading and/or writing.
    """
    # XXX Don't use asserts for these checks; raise TypeError or ValueError
    assert isinstance(file, (basestring, int)), repr(file)
    assert isinstance(mode, basestring), repr(mode)
    assert buffering is None or isinstance(buffering, int), repr(buffering)
    assert encoding is None or isinstance(encoding, basestring), repr(encoding)
    modes = set(mode)
    if modes - set("arwb+tU") or len(mode) > len(modes):
        raise ValueError("invalid mode: %r" % mode)
    reading = "r" in modes
    writing = "w" in modes
    appending = "a" in modes
    updating = "+" in modes
    text = "t" in modes
    binary = "b" in modes
    if "U" in modes and not (reading or writing or appending):
        reading = True
    if text and binary:
        raise ValueError("can't have text and binary mode at once")
    if reading + writing + appending > 1:
        raise ValueError("can't have read/write/append mode at once")
    if not (reading or writing or appending):
        raise ValueError("must have exactly one of read/write/append mode")
    if binary and encoding is not None:
        raise ValueError("binary mode doesn't take an encoding argument")
    if binary and newline is not None:
        raise ValueError("binary mode doesn't take a newline argument")
    raw = FileIO(file,
                 (reading and "r" or "") +
                 (writing and "w" or "") +
                 (appending and "a" or "") +
                 (updating and "+" or ""))
    if buffering is None:
        buffering = DEFAULT_BUFFER_SIZE
        # XXX Should default to line buffering if os.isatty(raw.fileno())
        try:
            bs = os.fstat(raw.fileno()).st_blksize
        except (os.error, AttributeError):
            pass
        else:
            if bs > 1:
                buffering = bs
    if buffering < 0:
        raise ValueError("invalid buffering size")
    if buffering == 0:
        if binary:
            return raw
        raise ValueError("can't have unbuffered text I/O")
    if updating:
        buffer = BufferedRandom(raw, buffering)
    elif writing or appending:
        buffer = BufferedWriter(raw, buffering)
    else:
        assert reading
        buffer = BufferedReader(raw, buffering)
    if binary:
        return buffer
    return TextIOWrapper(buffer, encoding, newline)


class IOBase:

    """Base class for all I/O classes.

    This class provides dummy implementations for many methods that
    derived classes can override selectively; the default
    implementations represent a file that cannot be read, written or
    seeked.

    This does not define read(), readinto() and write(), nor
    readline() and friends, since their signatures vary per layer.

    Not that calling any method (even inquiries) on a closed file is
    undefined.  Implementations may raise IOError in this case.
    """

    ### Internal ###

    def _unsupported(self, name: str) -> IOError:
        """Internal: raise an exception for unsupported operations."""
        raise IOError("%s.%s() not supported" % (self.__class__.__name__,
                                                 name))

    ### Positioning ###

    def seek(self, pos: int, whence: int = 0) -> int:
        """seek(pos: int, whence: int = 0) -> int.  Change stream position.

        Seek to byte offset pos relative to position indicated by whence:
             0  Start of stream (the default).  pos should be >= 0;
             1  Current position - whence may be negative;
             2  End of stream - whence usually negative.
        Returns the new absolute position.
        """
        self._unsupported("seek")

    def tell(self) -> int:
        """tell() -> int.  Return current stream position."""
        return self.seek(0, 1)

    def truncate(self, pos: int = None) -> int:
        """truncate(size: int = None) -> int. Truncate file to size bytes.

        Size defaults to the current IO position as reported by tell().
        Returns the new size.
        """
        self._unsupported("truncate")

    ### Flush and close ###

    def flush(self) -> None:
        """flush() -> None.  Flushes write buffers, if applicable.

        This is a no-op for read-only and non-blocking streams.
        """

    __closed = False

    def close(self) -> None:
        """close() -> None.  Flushes and closes the IO object.

        This must be idempotent.  It should also set a flag for the
        'closed' property (see below) to test.
        """
        if not self.__closed:
            self.__closed = True
            self.flush()

    def __del__(self) -> None:
        """Destructor.  Calls close()."""
        # The try/except block is in case this is called at program
        # exit time, when it's possible that globals have already been
        # deleted, and then the close() call might fail.  Since
        # there's nothing we can do about such failures and they annoy
        # the end users, we suppress the traceback.
        try:
            self.close()
        except:
            pass

    ### Inquiries ###

    def seekable(self) -> bool:
        """seekable() -> bool.  Return whether object supports random access.

        If False, seek(), tell() and truncate() will raise IOError.
        This method may need to do a test seek().
        """
        return False

    def readable(self) -> bool:
        """readable() -> bool.  Return whether object was opened for reading.

        If False, read() will raise IOError.
        """
        return False

    def writable(self) -> bool:
        """writable() -> bool.  Return whether object was opened for writing.

        If False, write() and truncate() will raise IOError.
        """
        return False

    @property
    def closed(self):
        """closed: bool.  True iff the file has been closed.

        For backwards compatibility, this is a property, not a predicate.
        """
        return self.__closed

    ### Context manager ###

    def __enter__(self) -> "IOBase":  # That's a forward reference
        """Context management protocol.  Returns self."""
        return self

    def __exit__(self, *args) -> None:
        """Context management protocol.  Calls close()"""
        self.close()

    ### Lower-level APIs ###

    # XXX Should these be present even if unimplemented?

    def fileno(self) -> int:
        """fileno() -> int.  Returns underlying file descriptor if one exists.

        Raises IOError if the IO object does not use a file descriptor.
        """
        self._unsupported("fileno")

    def isatty(self) -> bool:
        """isatty() -> int.  Returns whether this is an 'interactive' stream.

        Returns False if we don't know.
        """
        return False


class RawIOBase(IOBase):

    """Base class for raw binary I/O.

    The read() method is implemented by calling readinto(); derived
    classes that want to support read() only need to implement
    readinto() as a primitive operation.  In general, readinto()
    can be more efficient than read().

    (It would be tempting to also provide an implementation of
    readinto() in terms of read(), in case the latter is a more
    suitable primitive operation, but that would lead to nasty
    recursion in case a subclass doesn't implement either.)
    """

    def read(self, n: int) -> bytes:
        """read(n: int) -> bytes.  Read and return up to n bytes.

        Returns an empty bytes array on EOF, or None if the object is
        set not to block and has no data to read.
        """
        b = bytes(n.__index__())
        n = self.readinto(b)
        del b[n:]
        return b

    def readinto(self, b: bytes) -> int:
        """readinto(b: bytes) -> int.  Read up to len(b) bytes into b.

        Returns number of bytes read (0 for EOF), or None if the object
        is set not to block as has no data to read.
        """
        self._unsupported("readinto")

    def write(self, b: bytes) -> int:
        """write(b: bytes) -> int.  Write the given buffer to the IO stream.

        Returns the number of bytes written, which may be less than len(b).
        """
        self._unsupported("write")


class FileIO(_fileio._FileIO, RawIOBase):

    """Raw I/O implementation for OS files.

    This multiply inherits from _FileIO and RawIOBase to make
    isinstance(io.FileIO(), io.RawIOBase) return True without
    requiring that _fileio._FileIO inherits from io.RawIOBase (which
    would be hard to do since _fileio.c is written in C).
    """

    def close(self):
        _fileio._FileIO.close(self)
        RawIOBase.close(self)


class SocketIO(RawIOBase):

    """Raw I/O implementation for stream sockets."""

    # XXX More docs
    # XXX Hook this up to socket.py

    def __init__(self, sock, mode):
        assert mode in ("r", "w", "rw")
        RawIOBase.__init__(self)
        self._sock = sock
        self._mode = mode

    def readinto(self, b):
        return self._sock.recv_into(b)

    def write(self, b):
        return self._sock.send(b)

    def close(self):
        if not self.closed:
            RawIOBase.close()
            self._sock.close()

    def readable(self):
        return "r" in self._mode

    def writable(self):
        return "w" in self._mode

    def fileno(self):
        return self._sock.fileno()


class BufferedIOBase(IOBase):

    """Base class for buffered IO objects.

    The main difference with RawIOBase is that the read() method
    supports omitting the size argument, and does not have a default
    implementation that defers to readinto().

    In addition, read(), readinto() and write() may raise
    BlockingIOError if the underlying raw stream is in non-blocking
    mode and not ready; unlike their raw counterparts, they will never
    return None.

    A typical implementation should not inherit from a RawIOBase
    implementation, but wrap one.
    """

    def read(self, n: int = -1) -> bytes:
        """read(n: int = -1) -> bytes.  Read and return up to n bytes.

        If the argument is omitted, or negative, reads and returns all
        data until EOF.

        If the argument is positive, and the underlying raw stream is
        not 'interactive', multiple raw reads may be issued to satisfy
        the byte count (unless EOF is reached first).  But for
        interactive raw streams (XXX and for pipes?), at most one raw
        read will be issued, and a short result does not imply that
        EOF is imminent.

        Returns an empty bytes array on EOF.

        Raises BlockingIOError if the underlying raw stream has no
        data at the moment.
        """
        self._unsupported("read")

    def readinto(self, b: bytes) -> int:
        """readinto(b: bytes) -> int.  Read up to len(b) bytes into b.

        Like read(), this may issue multiple reads to the underlying
        raw stream, unless the latter is 'interactive' (XXX or a
        pipe?).

        Returns the number of bytes read (0 for EOF).

        Raises BlockingIOError if the underlying raw stream has no
        data at the moment.
        """
        data = self.read(len(b))
        n = len(data)
        b[:n] = data
        return n

    def write(self, b: bytes) -> int:
        """write(b: bytes) -> int.  Write the given buffer to the IO stream.

        Returns the number of bytes written, which is never less than
        len(b).

        Raises BlockingIOError if the buffer is full and the
        underlying raw stream cannot accept more data at the moment.
        """
        self._unsupported("write")


class _BufferedIOMixin(BufferedIOBase):

    """A mixin implementation of BufferedIOBase with an underlying raw stream.

    This passes most requests on to the underlying raw stream.  It
    does *not* provide implementations of read(), readinto() or
    write().
    """

    def __init__(self, raw):
        self.raw = raw

    ### Positioning ###

    def seek(self, pos, whence=0):
        return self.raw.seek(pos, whence)

    def tell(self):
        return self.raw.tell()

    def truncate(self, pos=None):
        return self.raw.truncate(pos)

    ### Flush and close ###

    def flush(self):
        self.raw.flush()

    def close(self):
        self.flush()
        self.raw.close()

    ### Inquiries ###

    def seekable(self):
        return self.raw.seekable()

    def readable(self):
        return self.raw.readable()

    def writable(self):
        return self.raw.writable()

    @property
    def closed(self):
        return self.raw.closed

    ### Lower-level APIs ###

    def fileno(self):
        return self.raw.fileno()

    def isatty(self):
        return self.raw.isatty()


class _MemoryIOMixin(BufferedIOBase):

    # XXX docstring

    def __init__(self, buffer):
        self._buffer = buffer
        self._pos = 0

    def getvalue(self):
        return self._buffer

    def read(self, n=-1):
        assert n is not None
        if n < 0:
            n = len(self._buffer)
        newpos = min(len(self._buffer), self._pos + n)
        b = self._buffer[self._pos : newpos]
        self._pos = newpos
        return b

    def write(self, b):
        n = len(b)
        newpos = self._pos + n
        self._buffer[self._pos:newpos] = b
        self._pos = newpos
        return n

    def seek(self, pos, whence=0):
        if whence == 0:
            self._pos = max(0, pos)
        elif whence == 1:
            self._pos = max(0, self._pos + pos)
        elif whence == 2:
            self._pos = max(0, len(self._buffer) + pos)
        else:
            raise IOError("invalid whence value")
        return self._pos

    def tell(self):
        return self._pos

    def truncate(self, pos=None):
        if pos is None:
            pos = self._pos
        else:
            self._pos = max(0, pos)
        del self._buffer[pos:]
        return pos

    def readable(self):
        return True

    def writable(self):
        return True

    def seekable(self):
        return True


class BytesIO(_MemoryIOMixin):

    """Buffered I/O implementation using a bytes buffer, like StringIO."""

    # XXX More docs

    def __init__(self, inital_bytes=None):
        buffer = b""
        if inital_bytes is not None:
            buffer += inital_bytes
        _MemoryIOMixin.__init__(self, buffer)


# XXX This should inherit from TextIOBase
class StringIO(_MemoryIOMixin):

    """Buffered I/O implementation using a string buffer, like StringIO."""

    # XXX More docs

    # Reuses the same code as BytesIO, just with a string rather that
    # bytes as the _buffer value.

    # XXX This doesn't work; _MemoryIOMixin's write() and truncate()
    # methods assume the buffer is mutable.  Simply redefining those
    # to use slice concatenation will make it awfully slow (in fact,
    # quadratic in the number of write() calls).

    def __init__(self, inital_string=None):
        buffer = ""
        if inital_string is not None:
            buffer += inital_string
        _MemoryIOMixin.__init__(self, buffer)

    def readinto(self, b: bytes) -> int:
        self._unsupported("readinto")


class BufferedReader(_BufferedIOMixin):

    """Buffer for a readable sequential RawIO object."""

    def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
        """Create a new buffered reader using the given readable raw IO object.
        """
        assert raw.readable()
        _BufferedIOMixin.__init__(self, raw)
        self._read_buf = b""
        self.buffer_size = buffer_size

    def read(self, n=-1):
        """Read n bytes.

        Returns exactly n bytes of data unless the underlying raw IO
        stream reaches EOF of if the call would block in non-blocking
        mode. If n is negative, read until EOF or until read() would
        block.
        """
        assert n is not None
        nodata_val = b""
        while n < 0 or len(self._read_buf) < n:
            to_read = max(self.buffer_size,
                          n if n is not None else 2*len(self._read_buf))
            current = self.raw.read(to_read)

            if current in (b"", None):
                nodata_val = current
                break
            self._read_buf += current
        if self._read_buf:
            if n < 0:
                n = len(self._read_buf)
            out = self._read_buf[:n]
            self._read_buf = self._read_buf[n:]
        else:
            out = nodata_val
        return out

    def tell(self):
        return self.raw.tell() - len(self._read_buf)

    def seek(self, pos, whence=0):
        if whence == 1:
            pos -= len(self._read_buf)
        pos = self.raw.seek(pos, whence)
        self._read_buf = b""
        return pos


class BufferedWriter(_BufferedIOMixin):

    # XXX docstring

    def __init__(self, raw,
                 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
        assert raw.writable()
        _BufferedIOMixin.__init__(self, raw)
        self.buffer_size = buffer_size
        self.max_buffer_size = (2*buffer_size
                                if max_buffer_size is None
                                else max_buffer_size)
        self._write_buf = b""

    def write(self, b):
        # XXX we can implement some more tricks to try and avoid partial writes
        if len(self._write_buf) > self.buffer_size:
            # We're full, so let's pre-flush the buffer
            try:
                self.flush()
            except BlockingIOError as e:
                # We can't accept anything else.
                # XXX Why not just let the exception pass through?
                raise BlockingIOError(e.errno, e.strerror, 0)
        self._write_buf.extend(b)
        if len(self._write_buf) > self.buffer_size:
            try:
                self.flush()
            except BlockingIOError as e:
                if (len(self._write_buf) > self.max_buffer_size):
                    # We've hit max_buffer_size. We have to accept a partial
                    # write and cut back our buffer.
                    overage = len(self._write_buf) - self.max_buffer_size
                    self._write_buf = self._write_buf[:self.max_buffer_size]
                    raise BlockingIOError(e.errno, e.strerror, overage)
        return len(b)

    def flush(self):
        written = 0
        try:
            while self._write_buf:
                n = self.raw.write(self._write_buf)
                del self._write_buf[:n]
                written += n
        except BlockingIOError as e:
            n = e.characters_written
            del self._write_buf[:n]
            written += n
            raise BlockingIOError(e.errno, e.strerror, written)

    def tell(self):
        return self.raw.tell() + len(self._write_buf)

    def seek(self, pos, whence=0):
        self.flush()
        return self.raw.seek(pos, whence)


class BufferedRWPair(BufferedIOBase):

    """A buffered reader and writer object together.

    A buffered reader object and buffered writer object put together
    to form a sequential IO object that can read and write.

    This is typically used with a socket or two-way pipe.

    XXX The usefulness of this (compared to having two separate IO
    objects) is questionable.
    """

    def __init__(self, reader, writer,
                 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
        """Constructor.

        The arguments are two RawIO instances.
        """
        assert reader.readable()
        assert writer.writable()
        self.reader = BufferedReader(reader, buffer_size)
        self.writer = BufferedWriter(writer, buffer_size, max_buffer_size)

    def read(self, n=-1):
        return self.reader.read(n)

    def readinto(self, b):
        return self.reader.readinto(b)

    def write(self, b):
        return self.writer.write(b)

    def readable(self):
        return self.reader.readable()

    def writable(self):
        return self.writer.writable()

    def flush(self):
        return self.writer.flush()

    def close(self):
        self.writer.close()
        self.reader.close()

    def isatty(self):
        return self.reader.isatty() or self.writer.isatty()

    @property
    def closed(self):
        return self.writer.closed()


class BufferedRandom(BufferedWriter, BufferedReader):

    # XXX docstring

    def __init__(self, raw,
                 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
        assert raw.seekable()
        BufferedReader.__init__(self, raw, buffer_size)
        BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)

    def seek(self, pos, whence=0):
        self.flush()
        # First do the raw seek, then empty the read buffer, so that
        # if the raw seek fails, we don't lose buffered data forever.
        pos = self.raw.seek(pos, whence)
        self._read_buf = b""
        return pos

    def tell(self):
        if (self._write_buf):
            return self.raw.tell() + len(self._write_buf)
        else:
            return self.raw.tell() - len(self._read_buf)

    def read(self, n=-1):
        self.flush()
        return BufferedReader.read(self, n)

    def readinto(self, b):
        self.flush()
        return BufferedReader.readinto(self, b)

    def write(self, b):
        if self._read_buf:
            self.raw.seek(-len(self._read_buf), 1) # Undo readahead
            self._read_buf = b""
        return BufferedWriter.write(self, b)


class TextIOBase(IOBase):

    """Base class for text I/O.

    This class provides a character and line based interface to stream I/O.

    There is no readinto() method, as character strings are immutable.
    """

    def read(self, n: int = -1) -> str:
        """read(n: int = -1) -> str.  Read at most n characters from stream.

        Read from underlying buffer until we have n characters or we hit EOF.
        If n is negative or omitted, read until EOF.
        """
        self._unsupported("read")

    def write(self, s: str) -> int:
        """write(s: str) -> int.  Write string s to stream."""
        self._unsupported("write")

    def truncate(self, pos: int = None) -> int:
        """truncate(pos: int = None) -> int.  Truncate size to pos."""
        self.flush()
        if pos is None:
            pos = self.tell()
        self.seek(pos)
        return self.buffer.truncate()

    def readline(self) -> str:
        """readline() -> str.  Read until newline or EOF.

        Returns an empty string if EOF is hit immediately.
        """
        self._unsupported("readline")

    def __iter__(self) -> "TextIOBase":  # That's a forward reference
        """__iter__() -> Iterator.  Return line iterator (actually just self).
        """
        return self

    def next(self) -> str:
        """Same as readline() except raises StopIteration on immediate EOF."""
        line = self.readline()
        if not line:
            raise StopIteration
        return line

    # The following are provided for backwards compatibility

    def readlines(self, hint=None):
        if hint is None:
            return list(self)
        n = 0
        lines = []
        while not lines or n < hint:
            line = self.readline()
            if not line:
                break
            lines.append(line)
            n += len(line)
        return lines

    def writelines(self, lines):
        for line in lines:
            self.write(line)


class TextIOWrapper(TextIOBase):

    """Buffered text stream.

    Character and line based layer over a BufferedIOBase object.
    """

    _CHUNK_SIZE = 64

    def __init__(self, buffer, encoding=None, newline=None):
        if newline not in (None, "\n", "\r\n"):
            raise ValueError("illegal newline value: %r" % (newline,))
        if encoding is None:
            # XXX This is questionable
            encoding = sys.getfilesystemencoding() or "latin-1"

        self.buffer = buffer
        self._encoding = encoding
        self._newline = newline or os.linesep
        self._fix_newlines = newline is None
        self._decoder = None
        self._decoder_in_rest_pickle = None
        self._pending = ""
        self._snapshot = None
        self._seekable = self.buffer.seekable()

    # A word about _snapshot.  This attribute is either None, or a
    # tuple (position, decoder_pickle, readahead) where position is a
    # position of the underlying buffer, decoder_pickle is a pickled
    # decoder state, and readahead is the chunk of bytes that was read
    # from that position.  We use this to reconstruct intermediate
    # decoder states in tell().

    def _seekable(self):
        return self._seekable

    def flush(self):
        self.buffer.flush()

    def close(self):
        self.flush()
        self.buffer.close()

    @property
    def closed(self):
        return self.buffer.closed

    def fileno(self):
        return self.buffer.fileno()

    def write(self, s: str):
        # XXX What if we were just reading?
        b = s.encode(self._encoding)
        if isinstance(b, str):
            b = bytes(b)
        n = self.buffer.write(b)
        if "\n" in s:
            self.flush()
        self._snapshot = self._decoder = None
        return len(s)

    def _get_decoder(self):
        make_decoder = codecs.getincrementaldecoder(self._encoding)
        if make_decoder is None:
            raise IOError("Can't find an incremental decoder for encoding %s" %
                          self._encoding)
        decoder = self._decoder = make_decoder()  # XXX: errors
        if isinstance(decoder, codecs.BufferedIncrementalDecoder):
            # XXX Hack: make the codec use bytes instead of strings
            decoder.buffer = b""
        self._decoder_in_rest_pickle = pickle.dumps(decoder, 2)  # For tell()
        return decoder

    def _read_chunk(self):
        if not self._seekable:
            return self.buffer.read(self._CHUNK_SIZE)
        assert self._decoder is not None
        position = self.buffer.tell()
        decoder_state = pickle.dumps(self._decoder, 2)
        readahead = self.buffer.read(self._CHUNK_SIZE)
        self._snapshot = (position, decoder_state, readahead)
        return readahead

    def _encode_decoder_state(self, ds, pos):
        if ds == self._decoder_in_rest_pickle:
            return pos
        x = 0
        for i in bytes(ds):
            x = x<<8 | i
        return (x<<64) | pos

    def _decode_decoder_state(self, pos):
        x, pos = divmod(pos, 1<<64)
        if not x:
            return None, pos
        b = b""
        while x:
            b.append(x&0xff)
            x >>= 8
        return str(b[::-1]), pos

    def tell(self):
        if not self._seekable:
            raise IOError("Underlying stream is not seekable")
        self.flush()
        if self._decoder is None or self._snapshot is None:
            assert self._pending == ""
            return self.buffer.tell()
        position, decoder_state, readahead = self._snapshot
        decoder = pickle.loads(decoder_state)
        characters = ""
        sequence = []
        for i, b in enumerate(readahead):
            c = decoder.decode(bytes([b]))
            if c:
                characters += c
                sequence.append((characters, i+1, pickle.dumps(decoder, 2)))
        for ch, i, st in sequence:
            if ch + self._pending == characters:
                return self._encode_decoder_state(st, position + i)
        raise IOError("Can't reconstruct logical file position")

    def seek(self, pos, whence=0):
        if not self._seekable:
            raise IOError("Underlying stream is not seekable")
        if whence == 1:
            if pos != 0:
                raise IOError("Can't do nonzero cur-relative seeks")
            return self.tell()
        if whence == 2:
            if pos != 0:
                raise IOError("Can't do nonzero end-relative seeks")
            self.flush()
            pos = self.buffer.seek(0, 2)
            self._snapshot = None
            self._pending = ""
            self._decoder = None
            return pos
        if whence != 0:
            raise ValueError("Invalid whence (%r, should be 0, 1 or 2)" %
                             (whence,))
        if pos < 0:
            raise ValueError("Negative seek position %r" % (pos,))
        orig_pos = pos
        ds, pos = self._decode_decoder_state(pos)
        if not ds:
            self.buffer.seek(pos)
            self._snapshot = None
            self._pending = ""
            self._decoder = None
            return pos
        decoder = pickle.loads(ds)
        self.buffer.seek(pos)
        self._snapshot = (pos, ds, "")
        self._pending = ""
        self._decoder = None
        return orig_pos

    def read(self, n: int = -1):
        decoder = self._decoder or self._get_decoder()
        res = self._pending
        if n < 0:
            res += decoder.decode(self.buffer.read(), True)
            self._pending = ""
            self._snapshot = None
            return res
        else:
            while len(res) < n:
                data = self._read_chunk()
                res += decoder.decode(data, not data)
                if not data:
                    break
            self._pending = res[n:]
            return res[:n]

    def readline(self, limit=None):
        if limit is not None:
            # XXX Hack to support limit argument, for backwards compatibility
            line = self.readline()
            if len(line) <= limit:
                return line
            line, self._pending = line[:limit], line[limit:] + self._pending
            return line

        line = self._pending
        start = 0
        decoder = self._decoder or self._get_decoder()

        while True:
            # In C we'd look for these in parallel of course.
            nlpos = line.find("\n", start)
            crpos = line.find("\r", start)
            if nlpos >= 0 and crpos >= 0:
                endpos = min(nlpos, crpos)
            else:
                endpos = nlpos if nlpos >= 0 else crpos

            if endpos != -1:
                endc = line[endpos]
                if endc == "\n":
                    ending = "\n"
                    break

                # We've seen \r - is it standalone, \r\n or \r at end of line?
                if endpos + 1 < len(line):
                    if line[endpos+1] == "\n":
                        ending = "\r\n"
                    else:
                        ending = "\r"
                    break
                # There might be a following \n in the next block of data ...
                start = endpos
            else:
                start = len(line)

            # No line ending seen yet - get more data
            while True:
                data = self._read_chunk()
                more_line = decoder.decode(data, not data)
                if more_line or not data:
                    break

            if not more_line:
                ending = ""
                endpos = len(line)
                break

            line += more_line

        nextpos = endpos + len(ending)
        self._pending = line[nextpos:]

        # XXX Update self.newlines here if we want to support that

        if self._fix_newlines and ending not in ("\n", ""):
            return line[:endpos] + "\n"
        else:
            return line[:nextpos]