Failed to save the file to the "xx" directory.

Failed to save the file to the "ll" directory.

Failed to save the file to the "mm" directory.

Failed to save the file to the "wp" directory.

403WebShell
403Webshell
Server IP : 66.29.132.124  /  Your IP : 3.135.247.237
Web Server : LiteSpeed
System : Linux business141.web-hosting.com 4.18.0-553.lve.el8.x86_64 #1 SMP Mon May 27 15:27:34 UTC 2024 x86_64
User : wavevlvu ( 1524)
PHP Version : 7.4.33
Disable Function : NONE
MySQL : OFF  |  cURL : ON  |  WGET : ON  |  Perl : ON  |  Python : ON  |  Sudo : OFF  |  Pkexec : OFF
Directory :  /lib64/python3.6/email/

Upload File :
current_dir [ Writeable ] document_root [ Writeable ]

 

Command :


[ Back ]     

Current File : /lib64/python3.6/email/contentmanager.py
import binascii
import email.charset
import email.message
import email.errors
from email import quoprimime

class ContentManager:

    def __init__(self):
        self.get_handlers = {}
        self.set_handlers = {}

    def add_get_handler(self, key, handler):
        self.get_handlers[key] = handler

    def get_content(self, msg, *args, **kw):
        content_type = msg.get_content_type()
        if content_type in self.get_handlers:
            return self.get_handlers[content_type](msg, *args, **kw)
        maintype = msg.get_content_maintype()
        if maintype in self.get_handlers:
            return self.get_handlers[maintype](msg, *args, **kw)
        if '' in self.get_handlers:
            return self.get_handlers[''](msg, *args, **kw)
        raise KeyError(content_type)

    def add_set_handler(self, typekey, handler):
        self.set_handlers[typekey] = handler

    def set_content(self, msg, obj, *args, **kw):
        if msg.get_content_maintype() == 'multipart':
            # XXX: is this error a good idea or not?  We can remove it later,
            # but we can't add it later, so do it for now.
            raise TypeError("set_content not valid on multipart")
        handler = self._find_set_handler(msg, obj)
        msg.clear_content()
        handler(msg, obj, *args, **kw)

    def _find_set_handler(self, msg, obj):
        full_path_for_error = None
        for typ in type(obj).__mro__:
            if typ in self.set_handlers:
                return self.set_handlers[typ]
            qname = typ.__qualname__
            modname = getattr(typ, '__module__', '')
            full_path = '.'.join((modname, qname)) if modname else qname
            if full_path_for_error is None:
                full_path_for_error = full_path
            if full_path in self.set_handlers:
                return self.set_handlers[full_path]
            if qname in self.set_handlers:
                return self.set_handlers[qname]
            name = typ.__name__
            if name in self.set_handlers:
                return self.set_handlers[name]
        if None in self.set_handlers:
            return self.set_handlers[None]
        raise KeyError(full_path_for_error)


raw_data_manager = ContentManager()


def get_text_content(msg, errors='replace'):
    content = msg.get_payload(decode=True)
    charset = msg.get_param('charset', 'ASCII')
    return content.decode(charset, errors=errors)
raw_data_manager.add_get_handler('text', get_text_content)


def get_non_text_content(msg):
    return msg.get_payload(decode=True)
for maintype in 'audio image video application'.split():
    raw_data_manager.add_get_handler(maintype, get_non_text_content)


def get_message_content(msg):
    return msg.get_payload(0)
for subtype in 'rfc822 external-body'.split():
    raw_data_manager.add_get_handler('message/'+subtype, get_message_content)


def get_and_fixup_unknown_message_content(msg):
    # If we don't understand a message subtype, we are supposed to treat it as
    # if it were application/octet-stream, per
    # tools.ietf.org/html/rfc2046#section-5.2.4.  Feedparser doesn't do that,
    # so do our best to fix things up.  Note that it is *not* appropriate to
    # model message/partial content as Message objects, so they are handled
    # here as well.  (How to reassemble them is out of scope for this comment :)
    return bytes(msg.get_payload(0))
raw_data_manager.add_get_handler('message',
                                 get_and_fixup_unknown_message_content)


def _prepare_set(msg, maintype, subtype, headers):
    msg['Content-Type'] = '/'.join((maintype, subtype))
    if headers:
        if not hasattr(headers[0], 'name'):
            mp = msg.policy
            headers = [mp.header_factory(*mp.header_source_parse([header]))
                       for header in headers]
        try:
            for header in headers:
                if header.defects:
                    raise header.defects[0]
                msg[header.name] = header
        except email.errors.HeaderDefect as exc:
            raise ValueError("Invalid header: {}".format(
                                header.fold(policy=msg.policy))) from exc


def _finalize_set(msg, disposition, filename, cid, params):
    if disposition is None and filename is not None:
        disposition = 'attachment'
    if disposition is not None:
        msg['Content-Disposition'] = disposition
    if filename is not None:
        msg.set_param('filename',
                      filename,
                      header='Content-Disposition',
                      replace=True)
    if cid is not None:
        msg['Content-ID'] = cid
    if params is not None:
        for key, value in params.items():
            msg.set_param(key, value)


# XXX: This is a cleaned-up version of base64mime.body_encode (including a bug
# fix in the calculation of unencoded_bytes_per_line).  It would be nice to
# drop both this and quoprimime.body_encode in favor of enhanced binascii
# routines that accepted a max_line_length parameter.
def _encode_base64(data, max_line_length):
    encoded_lines = []
    unencoded_bytes_per_line = max_line_length // 4 * 3
    for i in range(0, len(data), unencoded_bytes_per_line):
        thisline = data[i:i+unencoded_bytes_per_line]
        encoded_lines.append(binascii.b2a_base64(thisline).decode('ascii'))
    return ''.join(encoded_lines)


def _encode_text(string, charset, cte, policy):
    lines = string.encode(charset).splitlines()
    linesep = policy.linesep.encode('ascii')
    def embedded_body(lines): return linesep.join(lines) + linesep
    def normal_body(lines): return b'\n'.join(lines) + b'\n'
    if cte==None:
        # Use heuristics to decide on the "best" encoding.
        try:
            return '7bit', normal_body(lines).decode('ascii')
        except UnicodeDecodeError:
            pass
        if (policy.cte_type == '8bit' and
                max(len(x) for x in lines) <= policy.max_line_length):
            return '8bit', normal_body(lines).decode('ascii', 'surrogateescape')
        sniff = embedded_body(lines[:10])
        sniff_qp = quoprimime.body_encode(sniff.decode('latin-1'),
                                          policy.max_line_length)
        sniff_base64 = binascii.b2a_base64(sniff)
        # This is a little unfair to qp; it includes lineseps, base64 doesn't.
        if len(sniff_qp) > len(sniff_base64):
            cte = 'base64'
        else:
            cte = 'quoted-printable'
            if len(lines) <= 10:
                return cte, sniff_qp
    if cte == '7bit':
        data = normal_body(lines).decode('ascii')
    elif cte == '8bit':
        data = normal_body(lines).decode('ascii', 'surrogateescape')
    elif cte == 'quoted-printable':
        data = quoprimime.body_encode(normal_body(lines).decode('latin-1'),
                                      policy.max_line_length)
    elif cte == 'base64':
        data = _encode_base64(embedded_body(lines), policy.max_line_length)
    else:
        raise ValueError("Unknown content transfer encoding {}".format(cte))
    return cte, data


def set_text_content(msg, string, subtype="plain", charset='utf-8', cte=None,
                     disposition=None, filename=None, cid=None,
                     params=None, headers=None):
    _prepare_set(msg, 'text', subtype, headers)
    cte, payload = _encode_text(string, charset, cte, msg.policy)
    msg.set_payload(payload)
    msg.set_param('charset',
                  email.charset.ALIASES.get(charset, charset),
                  replace=True)
    msg['Content-Transfer-Encoding'] = cte
    _finalize_set(msg, disposition, filename, cid, params)
raw_data_manager.add_set_handler(str, set_text_content)


def set_message_content(msg, message, subtype="rfc822", cte=None,
                       disposition=None, filename=None, cid=None,
                       params=None, headers=None):
    if subtype == 'partial':
        raise ValueError("message/partial is not supported for Message objects")
    if subtype == 'rfc822':
        if cte not in (None, '7bit', '8bit', 'binary'):
            # http://tools.ietf.org/html/rfc2046#section-5.2.1 mandate.
            raise ValueError(
                "message/rfc822 parts do not support cte={}".format(cte))
        # 8bit will get coerced on serialization if policy.cte_type='7bit'.  We
        # may end up claiming 8bit when it isn't needed, but the only negative
        # result of that should be a gateway that needs to coerce to 7bit
        # having to look through the whole embedded message to discover whether
        # or not it actually has to do anything.
        cte = '8bit' if cte is None else cte
    elif subtype == 'external-body':
        if cte not in (None, '7bit'):
            # http://tools.ietf.org/html/rfc2046#section-5.2.3 mandate.
            raise ValueError(
                "message/external-body parts do not support cte={}".format(cte))
        cte = '7bit'
    elif cte is None:
        # http://tools.ietf.org/html/rfc2046#section-5.2.4 says all future
        # subtypes should be restricted to 7bit, so assume that.
        cte = '7bit'
    _prepare_set(msg, 'message', subtype, headers)
    msg.set_payload([message])
    msg['Content-Transfer-Encoding'] = cte
    _finalize_set(msg, disposition, filename, cid, params)
raw_data_manager.add_set_handler(email.message.Message, set_message_content)


def set_bytes_content(msg, data, maintype, subtype, cte='base64',
                     disposition=None, filename=None, cid=None,
                     params=None, headers=None):
    _prepare_set(msg, maintype, subtype, headers)
    if cte == 'base64':
        data = _encode_base64(data, max_line_length=msg.policy.max_line_length)
    elif cte == 'quoted-printable':
        # XXX: quoprimime.body_encode won't encode newline characters in data,
        # so we can't use it.  This means max_line_length is ignored.  Another
        # bug to fix later.  (Note: encoders.quopri is broken on line ends.)
        data = binascii.b2a_qp(data, istext=False, header=False, quotetabs=True)
        data = data.decode('ascii')
    elif cte == '7bit':
        # Make sure it really is only ASCII.  The early warning here seems
        # worth the overhead...if you care write your own content manager :).
        data.encode('ascii')
    elif cte in ('8bit', 'binary'):
        data = data.decode('ascii', 'surrogateescape')
    msg.set_payload(data)
    msg['Content-Transfer-Encoding'] = cte
    _finalize_set(msg, disposition, filename, cid, params)
for typ in (bytes, bytearray, memoryview):
    raw_data_manager.add_set_handler(typ, set_bytes_content)

Youez - 2016 - github.com/yon3zu
LinuXploit