# charset is ASCII, as the standard (RFC XXX) requires.
        try:
            if isinstance(input_charset, str):
                input_charset.encode('ascii')
            else:
                input_charset = str(input_charset, 'ascii')
        except UnicodeError:
            raise errors.CharsetError(input_charset)
        input_charset = input_charset.lower()
        # Set the input charset after filtering through the aliases
        self.input_charset = ALIASES.get(input_charset, input_charset)
        # We can try to guess which encoding and conversion to use by the
        # charset_map dictionary.  Try that first, but let the user override
        # it.
        henc, benc, conv = CHARSETS.get(self.input_charset,
                                        (SHORTEST, BASE64, None))
        if not conv:
            conv = self.input_charset
        # Set the attributes, allowing the arguments to override the default.
        self.header_encoding = henc
        self.body_encoding = benc
        self.output_charset = ALIASES.get(conv, conv)
        # Now set the codecs.  If one isn't defined for input_charset,
        # guess and try a Unicode codec with the same name as input_codec.
        self.input_codec = CODEC_MAP.get(self.input_charset,
                                         self.input_charset)
        self.output_codec = CODEC_MAP.get(self.output_charset,
                                          self.output_charset)

    def __repr__(self):
        return self.input_charset.lower()

    def __eq__(self, other):
        return str(self) == str(other).lower()

    def get_body_encoding(self):
        """Return the content-transfer-encoding used for body encoding.

        This is either the string `quoted-printable' or `base64' depending on
        the encoding used, or it is a function in which case you should call
        the function with a single argument, the Message object being
        encoded.  The function should then set the Content-Transfer-Encoding
        header itself to whatever is appropriate.

        Returns "quoted-printable" if self.body_encoding is QP.
        Returns "base64" if self.body_encoding is BASE64.
        Returns conversion function otherwise.
        """
        assert self.body_encoding != SHORTEST
        if self.body_encoding == QP:
            return 'quoted-printable'
        elif self.body_encoding == BASE64:
            return 'base64'
        else:
            return encode_7or8bit

    def get_output_charset(self):
        """Return the output character set.

        This is self.output_charset if that is not None, otherwise it is
        self.input_charset.
        """
        return self.output_charset or self.input_charset

    def header_encode(self, string):
        """Header-encode a string by converting it first to bytes.

        The type of encoding (base64 or quoted-printable) will be based on
        this charset's `header_encoding`.

        :param string: A unicode string for the header.  It must be possible
            to encode this string to bytes using the character set's
            output codec.
        :return: The encoded string, with RFC 2047 chrome.
        """
        codec = self.output_codec or 'us-ascii'
        header_bytes = _encode(string, codec)
        # 7bit/8bit encodings return the string unchanged (modulo conversions)
        encoder_module = self._get_encoder(header_bytes)
        if encoder_module is None:
            return string
        return encoder_module.header_encode(header_bytes, codec)

    def header_encode_lines(self, string, maxlengths):
        """Header-encode a string by converting it first to bytes.

        This is similar to `header_encode()` except that the string is fit
        into maximum line lengths as given by the argument.

        :param string: A unicode string for the header.  It must be possible
            to encode this string to bytes using the character set's
            output codec.
        :param maxlengths: Maximum line length iterator.  Each element
            returned from this iterator will provide the next maximum line
            length.  This parameter is used as an argument to built-in next()
            and should never be exhausted.  The maximum line lengths should
            not count the RFC 2047 chrome.  These line lengths are only a
            hint; the splitter does the best it can.
        :return: Lines of encoded strings, each with RFC 2047 chrome.
        """
        # See which encoding we should use.
        codec = self.output_codec or 'us-ascii'
        header_bytes = _encode(string, codec)
        encoder_module = self._get_encoder(header_bytes)
        encoder = partial(encoder_module.header_encode, charset=codec)
        # Calculate the number of characters that the RFC 2047 chrome will
        # contribute to each line.
        charset = self.get_output_charset()
        extra = len(charset) + RFC2047_CHROME_LEN
        # Now comes the hard part.  We must encode bytes but we can't split on
        # bytes because some character sets are variable length and each
        # encoded word must stand on its own.  So the problem is you have to
        # encode to bytes to figure out this word's length, but you must split
        # on characters.  This causes two problems: first, we don't know how
        # many octets a specific substring of unicode characters will get
        # encoded to, and second, we don't know how many ASCII characters
        # those octets will get encoded to.  Unless we try it.  Which seems
        # inefficient.  In the interest of being correct rather than fast (and
        # in the hope that there will be few encoded headers in any such
        # message), brute force it. :(
        lines = []
        current_line = []
        maxlen = next(maxlengths) - extra
        for character in string:
            current_line.append(character)
            this_line = EMPTYSTRING.join(current_line)
            length = encoder_module.header_length(_encode(this_line, charset))
            if length > maxlen:
                # This last character doesn't fit so pop it off.
                current_line.pop()
                # Does nothing fit on the first line?
                if not lines and not current_line:
                    lines.append(None)
                else:
                    separator = (' ' if lines else '')
                    joined_line = EMPTYSTRING.join(current_line)
                    header_bytes = _encode(joined_line, codec)
                    lines.append(encoder(header_bytes))
                current_line = [character]
                maxlen = next(maxlengths) - extra
        joined_line = EMPTYSTRING.join(current_line)
        header_bytes = _encode(joined_line, codec)
        lines.append(encoder(header_bytes))
        return lines

    def _get_encoder(self, header_bytes):
        if self.header_encoding == BASE64:
            return email.base64mime
        elif self.header_encoding == QP:
            return email.quoprimime
        elif self.header_encoding == SHORTEST:
            len64 = email.base64mime.header_length(header_bytes)
            lenqp = email.quoprimime.header_length(header_bytes)
            if len64 < lenqp:
                return email.base64mime
            else:
                return email.quoprimime
        else:
            return None

    def body_encode(self, string):
        """Body-encode a string by converting it first to bytes.

        The type of encoding (base64 or quoted-printable) will be based on
        self.body_encoding.  If body_encoding is None, we assume the
        output charset is a 7bit encoding, so re-encoding the decoded
        string using the ascii codec produces the correct string version
        of the content.
        """
        if not string:
            return string
        if self.body_encoding is BASE64:
            if isinstance(string, str):
                string = string.encode(self.output_charset)
            return email.base64mime.body_encode(string)
        elif self.body_encoding is QP:
            # quopromime.body_encode takes a string, but operates on it as if
            # it were a list of byte codes.  For a (minimal) history on why
            # this is so, see changeset 0cf700464177.  To correctly encode a
            # character set, then, we must turn it into pseudo bytes via the
            # latin1 charset, which will encode any byte as a single code point
            # between 0 and 255, which is what body_encode is expecting.
            if isinstance(string, str):
                string = string.encode(self.output_charset)
            string = string.decode('latin1')
            return email.quoprimime.body_encode(string)
        else:
            if isinstance(string, str):
                string = string.encode(self.output_charset).decode('ascii')
            return string