R David Murray | 3da240f | 2013-10-16 22:48:40 -0400 | [diff] [blame] | 1 | import binascii |
| 2 | import email.charset |
| 3 | import email.message |
| 4 | import email.errors |
| 5 | from email import quoprimime |
| 6 | |
| 7 | class ContentManager: |
| 8 | |
| 9 | def __init__(self): |
| 10 | self.get_handlers = {} |
| 11 | self.set_handlers = {} |
| 12 | |
| 13 | def add_get_handler(self, key, handler): |
| 14 | self.get_handlers[key] = handler |
| 15 | |
| 16 | def get_content(self, msg, *args, **kw): |
| 17 | content_type = msg.get_content_type() |
| 18 | if content_type in self.get_handlers: |
| 19 | return self.get_handlers[content_type](msg, *args, **kw) |
| 20 | maintype = msg.get_content_maintype() |
| 21 | if maintype in self.get_handlers: |
| 22 | return self.get_handlers[maintype](msg, *args, **kw) |
| 23 | if '' in self.get_handlers: |
| 24 | return self.get_handlers[''](msg, *args, **kw) |
| 25 | raise KeyError(content_type) |
| 26 | |
| 27 | def add_set_handler(self, typekey, handler): |
| 28 | self.set_handlers[typekey] = handler |
| 29 | |
| 30 | def set_content(self, msg, obj, *args, **kw): |
| 31 | if msg.get_content_maintype() == 'multipart': |
| 32 | # XXX: is this error a good idea or not? We can remove it later, |
| 33 | # but we can't add it later, so do it for now. |
| 34 | raise TypeError("set_content not valid on multipart") |
| 35 | handler = self._find_set_handler(msg, obj) |
| 36 | msg.clear_content() |
| 37 | handler(msg, obj, *args, **kw) |
| 38 | |
| 39 | def _find_set_handler(self, msg, obj): |
| 40 | full_path_for_error = None |
| 41 | for typ in type(obj).__mro__: |
| 42 | if typ in self.set_handlers: |
| 43 | return self.set_handlers[typ] |
| 44 | qname = typ.__qualname__ |
| 45 | modname = getattr(typ, '__module__', '') |
| 46 | full_path = '.'.join((modname, qname)) if modname else qname |
| 47 | if full_path_for_error is None: |
| 48 | full_path_for_error = full_path |
| 49 | if full_path in self.set_handlers: |
| 50 | return self.set_handlers[full_path] |
| 51 | if qname in self.set_handlers: |
| 52 | return self.set_handlers[qname] |
| 53 | name = typ.__name__ |
| 54 | if name in self.set_handlers: |
| 55 | return self.set_handlers[name] |
| 56 | if None in self.set_handlers: |
| 57 | return self.set_handlers[None] |
| 58 | raise KeyError(full_path_for_error) |
| 59 | |
| 60 | |
| 61 | raw_data_manager = ContentManager() |
| 62 | |
| 63 | |
| 64 | def get_text_content(msg, errors='replace'): |
| 65 | content = msg.get_payload(decode=True) |
| 66 | charset = msg.get_param('charset', 'ASCII') |
| 67 | return content.decode(charset, errors=errors) |
| 68 | raw_data_manager.add_get_handler('text', get_text_content) |
| 69 | |
| 70 | |
| 71 | def get_non_text_content(msg): |
| 72 | return msg.get_payload(decode=True) |
| 73 | for maintype in 'audio image video application'.split(): |
| 74 | raw_data_manager.add_get_handler(maintype, get_non_text_content) |
| 75 | |
| 76 | |
| 77 | def get_message_content(msg): |
| 78 | return msg.get_payload(0) |
| 79 | for subtype in 'rfc822 external-body'.split(): |
| 80 | raw_data_manager.add_get_handler('message/'+subtype, get_message_content) |
| 81 | |
| 82 | |
| 83 | def get_and_fixup_unknown_message_content(msg): |
| 84 | # If we don't understand a message subtype, we are supposed to treat it as |
| 85 | # if it were application/octet-stream, per |
| 86 | # tools.ietf.org/html/rfc2046#section-5.2.4. Feedparser doesn't do that, |
| 87 | # so do our best to fix things up. Note that it is *not* appropriate to |
| 88 | # model message/partial content as Message objects, so they are handled |
| 89 | # here as well. (How to reassemble them is out of scope for this comment :) |
| 90 | return bytes(msg.get_payload(0)) |
| 91 | raw_data_manager.add_get_handler('message', |
| 92 | get_and_fixup_unknown_message_content) |
| 93 | |
| 94 | |
| 95 | def _prepare_set(msg, maintype, subtype, headers): |
| 96 | msg['Content-Type'] = '/'.join((maintype, subtype)) |
| 97 | if headers: |
| 98 | if not hasattr(headers[0], 'name'): |
| 99 | mp = msg.policy |
| 100 | headers = [mp.header_factory(*mp.header_source_parse([header])) |
| 101 | for header in headers] |
| 102 | try: |
| 103 | for header in headers: |
| 104 | if header.defects: |
| 105 | raise header.defects[0] |
| 106 | msg[header.name] = header |
| 107 | except email.errors.HeaderDefect as exc: |
| 108 | raise ValueError("Invalid header: {}".format( |
| 109 | header.fold(policy=msg.policy))) from exc |
| 110 | |
| 111 | |
| 112 | def _finalize_set(msg, disposition, filename, cid, params): |
| 113 | if disposition is None and filename is not None: |
| 114 | disposition = 'attachment' |
| 115 | if disposition is not None: |
| 116 | msg['Content-Disposition'] = disposition |
| 117 | if filename is not None: |
| 118 | msg.set_param('filename', |
| 119 | filename, |
| 120 | header='Content-Disposition', |
| 121 | replace=True) |
| 122 | if cid is not None: |
| 123 | msg['Content-ID'] = cid |
| 124 | if params is not None: |
| 125 | for key, value in params.items(): |
| 126 | msg.set_param(key, value) |
| 127 | |
| 128 | |
R David Murray | 94a7927 | 2016-09-09 15:00:09 -0400 | [diff] [blame^] | 129 | # XXX: This is a cleaned-up version of base64mime.body_encode (including a bug |
| 130 | # fix in the calculation of unencoded_bytes_per_line). It would be nice to |
| 131 | # drop both this and quoprimime.body_encode in favor of enhanced binascii |
| 132 | # routines that accepted a max_line_length parameter. |
R David Murray | 3da240f | 2013-10-16 22:48:40 -0400 | [diff] [blame] | 133 | def _encode_base64(data, max_line_length): |
| 134 | encoded_lines = [] |
R David Murray | 94a7927 | 2016-09-09 15:00:09 -0400 | [diff] [blame^] | 135 | unencoded_bytes_per_line = max_line_length // 4 * 3 |
R David Murray | 3da240f | 2013-10-16 22:48:40 -0400 | [diff] [blame] | 136 | for i in range(0, len(data), unencoded_bytes_per_line): |
| 137 | thisline = data[i:i+unencoded_bytes_per_line] |
| 138 | encoded_lines.append(binascii.b2a_base64(thisline).decode('ascii')) |
| 139 | return ''.join(encoded_lines) |
| 140 | |
| 141 | |
| 142 | def _encode_text(string, charset, cte, policy): |
| 143 | lines = string.encode(charset).splitlines() |
| 144 | linesep = policy.linesep.encode('ascii') |
| 145 | def embeded_body(lines): return linesep.join(lines) + linesep |
| 146 | def normal_body(lines): return b'\n'.join(lines) + b'\n' |
| 147 | if cte==None: |
| 148 | # Use heuristics to decide on the "best" encoding. |
| 149 | try: |
| 150 | return '7bit', normal_body(lines).decode('ascii') |
| 151 | except UnicodeDecodeError: |
| 152 | pass |
| 153 | if (policy.cte_type == '8bit' and |
| 154 | max(len(x) for x in lines) <= policy.max_line_length): |
| 155 | return '8bit', normal_body(lines).decode('ascii', 'surrogateescape') |
| 156 | sniff = embeded_body(lines[:10]) |
| 157 | sniff_qp = quoprimime.body_encode(sniff.decode('latin-1'), |
| 158 | policy.max_line_length) |
| 159 | sniff_base64 = binascii.b2a_base64(sniff) |
| 160 | # This is a little unfair to qp; it includes lineseps, base64 doesn't. |
| 161 | if len(sniff_qp) > len(sniff_base64): |
| 162 | cte = 'base64' |
| 163 | else: |
| 164 | cte = 'quoted-printable' |
| 165 | if len(lines) <= 10: |
| 166 | return cte, sniff_qp |
| 167 | if cte == '7bit': |
| 168 | data = normal_body(lines).decode('ascii') |
| 169 | elif cte == '8bit': |
| 170 | data = normal_body(lines).decode('ascii', 'surrogateescape') |
| 171 | elif cte == 'quoted-printable': |
| 172 | data = quoprimime.body_encode(normal_body(lines).decode('latin-1'), |
| 173 | policy.max_line_length) |
| 174 | elif cte == 'base64': |
| 175 | data = _encode_base64(embeded_body(lines), policy.max_line_length) |
| 176 | else: |
| 177 | raise ValueError("Unknown content transfer encoding {}".format(cte)) |
| 178 | return cte, data |
| 179 | |
| 180 | |
| 181 | def set_text_content(msg, string, subtype="plain", charset='utf-8', cte=None, |
| 182 | disposition=None, filename=None, cid=None, |
| 183 | params=None, headers=None): |
| 184 | _prepare_set(msg, 'text', subtype, headers) |
| 185 | cte, payload = _encode_text(string, charset, cte, msg.policy) |
| 186 | msg.set_payload(payload) |
| 187 | msg.set_param('charset', |
| 188 | email.charset.ALIASES.get(charset, charset), |
| 189 | replace=True) |
| 190 | msg['Content-Transfer-Encoding'] = cte |
| 191 | _finalize_set(msg, disposition, filename, cid, params) |
| 192 | raw_data_manager.add_set_handler(str, set_text_content) |
| 193 | |
| 194 | |
| 195 | def set_message_content(msg, message, subtype="rfc822", cte=None, |
| 196 | disposition=None, filename=None, cid=None, |
| 197 | params=None, headers=None): |
| 198 | if subtype == 'partial': |
| 199 | raise ValueError("message/partial is not supported for Message objects") |
| 200 | if subtype == 'rfc822': |
| 201 | if cte not in (None, '7bit', '8bit', 'binary'): |
| 202 | # http://tools.ietf.org/html/rfc2046#section-5.2.1 mandate. |
| 203 | raise ValueError( |
| 204 | "message/rfc822 parts do not support cte={}".format(cte)) |
| 205 | # 8bit will get coerced on serialization if policy.cte_type='7bit'. We |
| 206 | # may end up claiming 8bit when it isn't needed, but the only negative |
| 207 | # result of that should be a gateway that needs to coerce to 7bit |
| 208 | # having to look through the whole embedded message to discover whether |
| 209 | # or not it actually has to do anything. |
| 210 | cte = '8bit' if cte is None else cte |
| 211 | elif subtype == 'external-body': |
| 212 | if cte not in (None, '7bit'): |
| 213 | # http://tools.ietf.org/html/rfc2046#section-5.2.3 mandate. |
| 214 | raise ValueError( |
| 215 | "message/external-body parts do not support cte={}".format(cte)) |
| 216 | cte = '7bit' |
| 217 | elif cte is None: |
| 218 | # http://tools.ietf.org/html/rfc2046#section-5.2.4 says all future |
| 219 | # subtypes should be restricted to 7bit, so assume that. |
| 220 | cte = '7bit' |
| 221 | _prepare_set(msg, 'message', subtype, headers) |
| 222 | msg.set_payload([message]) |
| 223 | msg['Content-Transfer-Encoding'] = cte |
| 224 | _finalize_set(msg, disposition, filename, cid, params) |
| 225 | raw_data_manager.add_set_handler(email.message.Message, set_message_content) |
| 226 | |
| 227 | |
| 228 | def set_bytes_content(msg, data, maintype, subtype, cte='base64', |
| 229 | disposition=None, filename=None, cid=None, |
| 230 | params=None, headers=None): |
| 231 | _prepare_set(msg, maintype, subtype, headers) |
| 232 | if cte == 'base64': |
| 233 | data = _encode_base64(data, max_line_length=msg.policy.max_line_length) |
| 234 | elif cte == 'quoted-printable': |
| 235 | # XXX: quoprimime.body_encode won't encode newline characters in data, |
| 236 | # so we can't use it. This means max_line_length is ignored. Another |
| 237 | # bug to fix later. (Note: encoders.quopri is broken on line ends.) |
| 238 | data = binascii.b2a_qp(data, istext=False, header=False, quotetabs=True) |
| 239 | data = data.decode('ascii') |
| 240 | elif cte == '7bit': |
| 241 | # Make sure it really is only ASCII. The early warning here seems |
| 242 | # worth the overhead...if you care write your own content manager :). |
| 243 | data.encode('ascii') |
| 244 | elif cte in ('8bit', 'binary'): |
| 245 | data = data.decode('ascii', 'surrogateescape') |
| 246 | msg.set_payload(data) |
| 247 | msg['Content-Transfer-Encoding'] = cte |
| 248 | _finalize_set(msg, disposition, filename, cid, params) |
| 249 | for typ in (bytes, bytearray, memoryview): |
| 250 | raw_data_manager.add_set_handler(typ, set_bytes_content) |