Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 1 | # Copyright (C) 2004-2006 Python Software Foundation |
| 2 | # Authors: Baxter, Wouters and Warsaw |
| 3 | # Contact: email-sig@python.org |
| 4 | |
| 5 | """FeedParser - An email feed parser. |
| 6 | |
| 7 | The feed parser implements an interface for incrementally parsing an email |
| 8 | message, line by line. This has advantages for certain applications, such as |
| 9 | those reading email messages off a socket. |
| 10 | |
| 11 | FeedParser.feed() is the primary interface for pushing new data into the |
| 12 | parser. It returns when there's nothing more it can do with the available |
| 13 | data. When you have no more data to push into the parser, call .close(). |
| 14 | This completes the parsing and returns the root message object. |
| 15 | |
| 16 | The other advantage of this parser is that it will never throw a parsing |
| 17 | exception. Instead, when it finds something unexpected, it adds a 'defect' to |
| 18 | the current message. Defects are just instances that live on the message |
| 19 | object's .defects attribute. |
| 20 | """ |
| 21 | |
R David Murray | 1b6c724 | 2012-03-16 22:43:05 -0400 | [diff] [blame] | 22 | __all__ = ['FeedParser', 'BytesFeedParser'] |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 23 | |
| 24 | import re |
| 25 | |
| 26 | from email import errors |
| 27 | from email import message |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 28 | from email._policybase import compat32 |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 29 | |
| 30 | NLCRE = re.compile('\r\n|\r|\n') |
| 31 | NLCRE_bol = re.compile('(\r\n|\r|\n)') |
R. David Murray | 45e0e14 | 2010-06-16 02:19:40 +0000 | [diff] [blame] | 32 | NLCRE_eol = re.compile('(\r\n|\r|\n)\Z') |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 33 | NLCRE_crack = re.compile('(\r\n|\r|\n)') |
| 34 | # RFC 2822 $3.6.8 Optional fields. ftext is %d33-57 / %d59-126, Any character |
| 35 | # except controls, SP, and ":". |
| 36 | headerRE = re.compile(r'^(From |[\041-\071\073-\176]{1,}:|[\t ])') |
| 37 | EMPTYSTRING = '' |
| 38 | NL = '\n' |
| 39 | |
| 40 | NeedMoreData = object() |
| 41 | |
| 42 | |
| 43 | |
| 44 | class BufferedSubFile(object): |
| 45 | """A file-ish object that can have new data loaded into it. |
| 46 | |
| 47 | You can also push and pop line-matching predicates onto a stack. When the |
| 48 | current predicate matches the current line, a false EOF response |
| 49 | (i.e. empty string) is returned instead. This lets the parser adhere to a |
| 50 | simple abstraction -- it parses until EOF closes the current message. |
| 51 | """ |
| 52 | def __init__(self): |
| 53 | # The last partial line pushed into this object. |
| 54 | self._partial = '' |
| 55 | # The list of full, pushed lines, in reverse order |
| 56 | self._lines = [] |
| 57 | # The stack of false-EOF checking predicates. |
| 58 | self._eofstack = [] |
| 59 | # A flag indicating whether the file has been closed or not. |
| 60 | self._closed = False |
| 61 | |
| 62 | def push_eof_matcher(self, pred): |
| 63 | self._eofstack.append(pred) |
| 64 | |
| 65 | def pop_eof_matcher(self): |
| 66 | return self._eofstack.pop() |
| 67 | |
| 68 | def close(self): |
| 69 | # Don't forget any trailing partial line. |
| 70 | self._lines.append(self._partial) |
| 71 | self._partial = '' |
| 72 | self._closed = True |
| 73 | |
| 74 | def readline(self): |
| 75 | if not self._lines: |
| 76 | if self._closed: |
| 77 | return '' |
| 78 | return NeedMoreData |
| 79 | # Pop the line off the stack and see if it matches the current |
| 80 | # false-EOF predicate. |
| 81 | line = self._lines.pop() |
| 82 | # RFC 2046, section 5.1.2 requires us to recognize outer level |
| 83 | # boundaries at any level of inner nesting. Do this, but be sure it's |
| 84 | # in the order of most to least nested. |
| 85 | for ateof in self._eofstack[::-1]: |
| 86 | if ateof(line): |
| 87 | # We're at the false EOF. But push the last line back first. |
| 88 | self._lines.append(line) |
| 89 | return '' |
| 90 | return line |
| 91 | |
| 92 | def unreadline(self, line): |
| 93 | # Let the consumer push a line back into the buffer. |
| 94 | assert line is not NeedMoreData |
| 95 | self._lines.append(line) |
| 96 | |
| 97 | def push(self, data): |
| 98 | """Push some new data into this object.""" |
| 99 | # Handle any previous leftovers |
| 100 | data, self._partial = self._partial + data, '' |
| 101 | # Crack into lines, but preserve the newlines on the end of each |
| 102 | parts = NLCRE_crack.split(data) |
| 103 | # The *ahem* interesting behaviour of re.split when supplied grouping |
| 104 | # parentheses is that the last element of the resulting list is the |
| 105 | # data after the final RE. In the case of a NL/CR terminated string, |
| 106 | # this is the empty string. |
| 107 | self._partial = parts.pop() |
R. David Murray | 45bf773f | 2010-07-17 01:19:57 +0000 | [diff] [blame] | 108 | #GAN 29Mar09 bugs 1555570, 1721862 Confusion at 8K boundary ending with \r: |
| 109 | # is there a \n to follow later? |
| 110 | if not self._partial and parts and parts[-1].endswith('\r'): |
| 111 | self._partial = parts.pop(-2)+parts.pop() |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 112 | # parts is a list of strings, alternating between the line contents |
| 113 | # and the eol character(s). Gather up a list of lines after |
| 114 | # re-attaching the newlines. |
| 115 | lines = [] |
| 116 | for i in range(len(parts) // 2): |
| 117 | lines.append(parts[i*2] + parts[i*2+1]) |
| 118 | self.pushlines(lines) |
| 119 | |
| 120 | def pushlines(self, lines): |
| 121 | # Reverse and insert at the front of the lines. |
| 122 | self._lines[:0] = lines[::-1] |
| 123 | |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 124 | def __iter__(self): |
| 125 | return self |
| 126 | |
| 127 | def __next__(self): |
| 128 | line = self.readline() |
| 129 | if line == '': |
| 130 | raise StopIteration |
| 131 | return line |
| 132 | |
| 133 | |
| 134 | |
| 135 | class FeedParser: |
| 136 | """A feed-style parser of email.""" |
| 137 | |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 138 | def __init__(self, _factory=message.Message, *, policy=compat32): |
R David Murray | 3edd22a | 2011-04-18 13:59:37 -0400 | [diff] [blame] | 139 | """_factory is called with no arguments to create a new message obj |
| 140 | |
| 141 | The policy keyword specifies a policy object that controls a number of |
| 142 | aspects of the parser's operation. The default policy maintains |
| 143 | backward compatibility. |
| 144 | |
| 145 | """ |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 146 | self._factory = _factory |
R David Murray | 3edd22a | 2011-04-18 13:59:37 -0400 | [diff] [blame] | 147 | self.policy = policy |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 148 | try: |
| 149 | _factory(policy=self.policy) |
| 150 | self._factory_kwds = lambda: {'policy': self.policy} |
| 151 | except TypeError: |
| 152 | # Assume this is an old-style factory |
| 153 | self._factory_kwds = lambda: {} |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 154 | self._input = BufferedSubFile() |
| 155 | self._msgstack = [] |
| 156 | self._parse = self._parsegen().__next__ |
| 157 | self._cur = None |
| 158 | self._last = None |
| 159 | self._headersonly = False |
| 160 | |
| 161 | # Non-public interface for supporting Parser's headersonly flag |
| 162 | def _set_headersonly(self): |
| 163 | self._headersonly = True |
| 164 | |
| 165 | def feed(self, data): |
| 166 | """Push more data into the parser.""" |
| 167 | self._input.push(data) |
| 168 | self._call_parse() |
| 169 | |
| 170 | def _call_parse(self): |
| 171 | try: |
| 172 | self._parse() |
| 173 | except StopIteration: |
| 174 | pass |
| 175 | |
| 176 | def close(self): |
| 177 | """Parse all remaining data and return the root message object.""" |
| 178 | self._input.close() |
| 179 | self._call_parse() |
| 180 | root = self._pop_message() |
| 181 | assert not self._msgstack |
| 182 | # Look for final set of defects |
| 183 | if root.get_content_maintype() == 'multipart' \ |
| 184 | and not root.is_multipart(): |
R David Murray | 3edd22a | 2011-04-18 13:59:37 -0400 | [diff] [blame] | 185 | defect = errors.MultipartInvariantViolationDefect() |
| 186 | self.policy.handle_defect(root, defect) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 187 | return root |
| 188 | |
| 189 | def _new_message(self): |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 190 | msg = self._factory(**self._factory_kwds()) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 191 | if self._cur and self._cur.get_content_type() == 'multipart/digest': |
| 192 | msg.set_default_type('message/rfc822') |
| 193 | if self._msgstack: |
| 194 | self._msgstack[-1].attach(msg) |
| 195 | self._msgstack.append(msg) |
| 196 | self._cur = msg |
| 197 | self._last = msg |
| 198 | |
| 199 | def _pop_message(self): |
| 200 | retval = self._msgstack.pop() |
| 201 | if self._msgstack: |
| 202 | self._cur = self._msgstack[-1] |
| 203 | else: |
| 204 | self._cur = None |
| 205 | return retval |
| 206 | |
| 207 | def _parsegen(self): |
| 208 | # Create a new message and start by parsing headers. |
| 209 | self._new_message() |
| 210 | headers = [] |
| 211 | # Collect the headers, searching for a line that doesn't match the RFC |
| 212 | # 2822 header or continuation pattern (including an empty line). |
| 213 | for line in self._input: |
| 214 | if line is NeedMoreData: |
| 215 | yield NeedMoreData |
| 216 | continue |
| 217 | if not headerRE.match(line): |
| 218 | # If we saw the RFC defined header/body separator |
| 219 | # (i.e. newline), just throw it away. Otherwise the line is |
| 220 | # part of the body so push it back. |
| 221 | if not NLCRE.match(line): |
R David Murray | adbdcdb | 2012-05-27 20:45:01 -0400 | [diff] [blame^] | 222 | defect = errors.MissingHeaderBodySeparatorDefect() |
| 223 | self.policy.handle_defect(self._cur, defect) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 224 | self._input.unreadline(line) |
| 225 | break |
| 226 | headers.append(line) |
| 227 | # Done with the headers, so parse them and figure out what we're |
| 228 | # supposed to see in the body of the message. |
| 229 | self._parse_headers(headers) |
| 230 | # Headers-only parsing is a backwards compatibility hack, which was |
| 231 | # necessary in the older parser, which could throw errors. All |
| 232 | # remaining lines in the input are thrown into the message body. |
| 233 | if self._headersonly: |
| 234 | lines = [] |
| 235 | while True: |
| 236 | line = self._input.readline() |
| 237 | if line is NeedMoreData: |
| 238 | yield NeedMoreData |
| 239 | continue |
| 240 | if line == '': |
| 241 | break |
| 242 | lines.append(line) |
| 243 | self._cur.set_payload(EMPTYSTRING.join(lines)) |
| 244 | return |
| 245 | if self._cur.get_content_type() == 'message/delivery-status': |
| 246 | # message/delivery-status contains blocks of headers separated by |
| 247 | # a blank line. We'll represent each header block as a separate |
| 248 | # nested message object, but the processing is a bit different |
| 249 | # than standard message/* types because there is no body for the |
| 250 | # nested messages. A blank line separates the subparts. |
| 251 | while True: |
| 252 | self._input.push_eof_matcher(NLCRE.match) |
| 253 | for retval in self._parsegen(): |
| 254 | if retval is NeedMoreData: |
| 255 | yield NeedMoreData |
| 256 | continue |
| 257 | break |
| 258 | msg = self._pop_message() |
| 259 | # We need to pop the EOF matcher in order to tell if we're at |
| 260 | # the end of the current file, not the end of the last block |
| 261 | # of message headers. |
| 262 | self._input.pop_eof_matcher() |
| 263 | # The input stream must be sitting at the newline or at the |
| 264 | # EOF. We want to see if we're at the end of this subpart, so |
| 265 | # first consume the blank line, then test the next line to see |
| 266 | # if we're at this subpart's EOF. |
| 267 | while True: |
| 268 | line = self._input.readline() |
| 269 | if line is NeedMoreData: |
| 270 | yield NeedMoreData |
| 271 | continue |
| 272 | break |
| 273 | while True: |
| 274 | line = self._input.readline() |
| 275 | if line is NeedMoreData: |
| 276 | yield NeedMoreData |
| 277 | continue |
| 278 | break |
| 279 | if line == '': |
| 280 | break |
| 281 | # Not at EOF so this is a line we're going to need. |
| 282 | self._input.unreadline(line) |
| 283 | return |
| 284 | if self._cur.get_content_maintype() == 'message': |
| 285 | # The message claims to be a message/* type, then what follows is |
| 286 | # another RFC 2822 message. |
| 287 | for retval in self._parsegen(): |
| 288 | if retval is NeedMoreData: |
| 289 | yield NeedMoreData |
| 290 | continue |
| 291 | break |
| 292 | self._pop_message() |
| 293 | return |
| 294 | if self._cur.get_content_maintype() == 'multipart': |
| 295 | boundary = self._cur.get_boundary() |
| 296 | if boundary is None: |
| 297 | # The message /claims/ to be a multipart but it has not |
| 298 | # defined a boundary. That's a problem which we'll handle by |
| 299 | # reading everything until the EOF and marking the message as |
| 300 | # defective. |
R David Murray | 3edd22a | 2011-04-18 13:59:37 -0400 | [diff] [blame] | 301 | defect = errors.NoBoundaryInMultipartDefect() |
| 302 | self.policy.handle_defect(self._cur, defect) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 303 | lines = [] |
| 304 | for line in self._input: |
| 305 | if line is NeedMoreData: |
| 306 | yield NeedMoreData |
| 307 | continue |
| 308 | lines.append(line) |
| 309 | self._cur.set_payload(EMPTYSTRING.join(lines)) |
| 310 | return |
R David Murray | 749073a | 2011-06-22 13:47:53 -0400 | [diff] [blame] | 311 | # Make sure a valid content type was specified per RFC 2045:6.4. |
| 312 | if (self._cur.get('content-transfer-encoding', '8bit').lower() |
| 313 | not in ('7bit', '8bit', 'binary')): |
| 314 | defect = errors.InvalidMultipartContentTransferEncodingDefect() |
| 315 | self.policy.handle_defect(self._cur, defect) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 316 | # Create a line match predicate which matches the inter-part |
| 317 | # boundary as well as the end-of-multipart boundary. Don't push |
| 318 | # this onto the input stream until we've scanned past the |
| 319 | # preamble. |
| 320 | separator = '--' + boundary |
| 321 | boundaryre = re.compile( |
| 322 | '(?P<sep>' + re.escape(separator) + |
| 323 | r')(?P<end>--)?(?P<ws>[ \t]*)(?P<linesep>\r\n|\r|\n)?$') |
| 324 | capturing_preamble = True |
| 325 | preamble = [] |
| 326 | linesep = False |
| 327 | while True: |
| 328 | line = self._input.readline() |
| 329 | if line is NeedMoreData: |
| 330 | yield NeedMoreData |
| 331 | continue |
| 332 | if line == '': |
| 333 | break |
| 334 | mo = boundaryre.match(line) |
| 335 | if mo: |
| 336 | # If we're looking at the end boundary, we're done with |
| 337 | # this multipart. If there was a newline at the end of |
| 338 | # the closing boundary, then we need to initialize the |
| 339 | # epilogue with the empty string (see below). |
| 340 | if mo.group('end'): |
| 341 | linesep = mo.group('linesep') |
| 342 | break |
| 343 | # We saw an inter-part boundary. Were we in the preamble? |
| 344 | if capturing_preamble: |
| 345 | if preamble: |
| 346 | # According to RFC 2046, the last newline belongs |
| 347 | # to the boundary. |
| 348 | lastline = preamble[-1] |
| 349 | eolmo = NLCRE_eol.search(lastline) |
| 350 | if eolmo: |
| 351 | preamble[-1] = lastline[:-len(eolmo.group(0))] |
| 352 | self._cur.preamble = EMPTYSTRING.join(preamble) |
| 353 | capturing_preamble = False |
| 354 | self._input.unreadline(line) |
| 355 | continue |
| 356 | # We saw a boundary separating two parts. Consume any |
| 357 | # multiple boundary lines that may be following. Our |
| 358 | # interpretation of RFC 2046 BNF grammar does not produce |
| 359 | # body parts within such double boundaries. |
| 360 | while True: |
| 361 | line = self._input.readline() |
| 362 | if line is NeedMoreData: |
| 363 | yield NeedMoreData |
| 364 | continue |
| 365 | mo = boundaryre.match(line) |
| 366 | if not mo: |
| 367 | self._input.unreadline(line) |
| 368 | break |
| 369 | # Recurse to parse this subpart; the input stream points |
| 370 | # at the subpart's first line. |
| 371 | self._input.push_eof_matcher(boundaryre.match) |
| 372 | for retval in self._parsegen(): |
| 373 | if retval is NeedMoreData: |
| 374 | yield NeedMoreData |
| 375 | continue |
| 376 | break |
| 377 | # Because of RFC 2046, the newline preceding the boundary |
| 378 | # separator actually belongs to the boundary, not the |
| 379 | # previous subpart's payload (or epilogue if the previous |
| 380 | # part is a multipart). |
| 381 | if self._last.get_content_maintype() == 'multipart': |
| 382 | epilogue = self._last.epilogue |
| 383 | if epilogue == '': |
| 384 | self._last.epilogue = None |
| 385 | elif epilogue is not None: |
| 386 | mo = NLCRE_eol.search(epilogue) |
| 387 | if mo: |
| 388 | end = len(mo.group(0)) |
| 389 | self._last.epilogue = epilogue[:-end] |
| 390 | else: |
R David Murray | c5c1472 | 2011-04-06 08:13:02 -0400 | [diff] [blame] | 391 | payload = self._last._payload |
Guido van Rossum | 3172c5d | 2007-10-16 18:12:55 +0000 | [diff] [blame] | 392 | if isinstance(payload, str): |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 393 | mo = NLCRE_eol.search(payload) |
| 394 | if mo: |
| 395 | payload = payload[:-len(mo.group(0))] |
R David Murray | c5c1472 | 2011-04-06 08:13:02 -0400 | [diff] [blame] | 396 | self._last._payload = payload |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 397 | self._input.pop_eof_matcher() |
| 398 | self._pop_message() |
| 399 | # Set the multipart up for newline cleansing, which will |
| 400 | # happen if we're in a nested multipart. |
| 401 | self._last = self._cur |
| 402 | else: |
| 403 | # I think we must be in the preamble |
| 404 | assert capturing_preamble |
| 405 | preamble.append(line) |
| 406 | # We've seen either the EOF or the end boundary. If we're still |
| 407 | # capturing the preamble, we never saw the start boundary. Note |
| 408 | # that as a defect and store the captured text as the payload. |
| 409 | # Everything from here to the EOF is epilogue. |
| 410 | if capturing_preamble: |
R David Murray | 3edd22a | 2011-04-18 13:59:37 -0400 | [diff] [blame] | 411 | defect = errors.StartBoundaryNotFoundDefect() |
| 412 | self.policy.handle_defect(self._cur, defect) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 413 | self._cur.set_payload(EMPTYSTRING.join(preamble)) |
| 414 | epilogue = [] |
| 415 | for line in self._input: |
| 416 | if line is NeedMoreData: |
| 417 | yield NeedMoreData |
| 418 | continue |
| 419 | self._cur.epilogue = EMPTYSTRING.join(epilogue) |
| 420 | return |
| 421 | # If the end boundary ended in a newline, we'll need to make sure |
| 422 | # the epilogue isn't None |
| 423 | if linesep: |
| 424 | epilogue = [''] |
| 425 | else: |
| 426 | epilogue = [] |
| 427 | for line in self._input: |
| 428 | if line is NeedMoreData: |
| 429 | yield NeedMoreData |
| 430 | continue |
| 431 | epilogue.append(line) |
| 432 | # Any CRLF at the front of the epilogue is not technically part of |
| 433 | # the epilogue. Also, watch out for an empty string epilogue, |
| 434 | # which means a single newline. |
| 435 | if epilogue: |
| 436 | firstline = epilogue[0] |
| 437 | bolmo = NLCRE_bol.match(firstline) |
| 438 | if bolmo: |
| 439 | epilogue[0] = firstline[len(bolmo.group(0)):] |
| 440 | self._cur.epilogue = EMPTYSTRING.join(epilogue) |
| 441 | return |
| 442 | # Otherwise, it's some non-multipart type, so the entire rest of the |
| 443 | # file contents becomes the payload. |
| 444 | lines = [] |
| 445 | for line in self._input: |
| 446 | if line is NeedMoreData: |
| 447 | yield NeedMoreData |
| 448 | continue |
| 449 | lines.append(line) |
| 450 | self._cur.set_payload(EMPTYSTRING.join(lines)) |
| 451 | |
| 452 | def _parse_headers(self, lines): |
| 453 | # Passed a list of lines that make up the headers for the current msg |
| 454 | lastheader = '' |
| 455 | lastvalue = [] |
| 456 | for lineno, line in enumerate(lines): |
| 457 | # Check for continuation |
| 458 | if line[0] in ' \t': |
| 459 | if not lastheader: |
| 460 | # The first line of the headers was a continuation. This |
| 461 | # is illegal, so let's note the defect, store the illegal |
| 462 | # line, and ignore it for purposes of headers. |
| 463 | defect = errors.FirstHeaderLineIsContinuationDefect(line) |
R David Murray | 3edd22a | 2011-04-18 13:59:37 -0400 | [diff] [blame] | 464 | self.policy.handle_defect(self._cur, defect) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 465 | continue |
| 466 | lastvalue.append(line) |
| 467 | continue |
| 468 | if lastheader: |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 469 | self._cur.set_raw(*self.policy.header_source_parse(lastvalue)) |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 470 | lastheader, lastvalue = '', [] |
| 471 | # Check for envelope header, i.e. unix-from |
| 472 | if line.startswith('From '): |
| 473 | if lineno == 0: |
| 474 | # Strip off the trailing newline |
| 475 | mo = NLCRE_eol.search(line) |
| 476 | if mo: |
| 477 | line = line[:-len(mo.group(0))] |
| 478 | self._cur.set_unixfrom(line) |
| 479 | continue |
| 480 | elif lineno == len(lines) - 1: |
| 481 | # Something looking like a unix-from at the end - it's |
| 482 | # probably the first line of the body, so push back the |
| 483 | # line and stop. |
| 484 | self._input.unreadline(line) |
| 485 | return |
| 486 | else: |
| 487 | # Weirdly placed unix-from line. Note this as a defect |
| 488 | # and ignore it. |
| 489 | defect = errors.MisplacedEnvelopeHeaderDefect(line) |
| 490 | self._cur.defects.append(defect) |
| 491 | continue |
| 492 | # Split the line on the colon separating field name from value. |
R David Murray | adbdcdb | 2012-05-27 20:45:01 -0400 | [diff] [blame^] | 493 | # There will always be a colon, because if there wasn't the part of |
| 494 | # the parser that calls us would have started parsing the body. |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 495 | i = line.find(':') |
R David Murray | adbdcdb | 2012-05-27 20:45:01 -0400 | [diff] [blame^] | 496 | assert i>0, "_parse_headers fed line with no : and no leading WS" |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 497 | lastheader = line[:i] |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 498 | lastvalue = [line] |
Guido van Rossum | 8b3febe | 2007-08-30 01:15:14 +0000 | [diff] [blame] | 499 | # Done with all the lines, so handle the last header. |
| 500 | if lastheader: |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 501 | self._cur.set_raw(*self.policy.header_source_parse(lastvalue)) |
R. David Murray | 96fd54e | 2010-10-08 15:55:28 +0000 | [diff] [blame] | 502 | |
R David Murray | c27e522 | 2012-05-25 15:01:48 -0400 | [diff] [blame] | 503 | |
R. David Murray | 96fd54e | 2010-10-08 15:55:28 +0000 | [diff] [blame] | 504 | class BytesFeedParser(FeedParser): |
| 505 | """Like FeedParser, but feed accepts bytes.""" |
| 506 | |
| 507 | def feed(self, data): |
| 508 | super().feed(data.decode('ascii', 'surrogateescape')) |