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