1 # Copyright (C) 2001-2010 Python Software Foundation 2 # Author: Barry Warsaw 3 # Contact: email-sig (at] python.org 4 5 """Classes to generate plain text from a message object tree.""" 6 7 __all__ = ['Generator', 'DecodedGenerator', 'BytesGenerator'] 8 9 import re 10 import sys 11 import time 12 import random 13 14 from copy import deepcopy 15 from io import StringIO, BytesIO 16 from email.utils import _has_surrogates 17 18 UNDERSCORE = '_' 19 NL = '\n' # XXX: no longer used by the code below. 20 21 NLCRE = re.compile(r'\r\n|\r|\n') 22 fcre = re.compile(r'^From ', re.MULTILINE) 23 24 25 27 class Generator: 28 """Generates output from a Message object tree. 29 30 This basic generator writes the message to the given file object as plain 31 text. 32 """ 33 # 34 # Public interface 35 # 36 37 def __init__(self, outfp, mangle_from_=None, maxheaderlen=None, *, 38 policy=None): 39 """Create the generator for message flattening. 40 41 outfp is the output file-like object for writing the message to. It 42 must have a write() method. 43 44 Optional mangle_from_ is a flag that, when True (the default if policy 45 is not set), escapes From_ lines in the body of the message by putting 46 a `>' in front of them. 47 48 Optional maxheaderlen specifies the longest length for a non-continued 49 header. When a header line is longer (in characters, with tabs 50 expanded to 8 spaces) than maxheaderlen, the header will split as 51 defined in the Header class. Set maxheaderlen to zero to disable 52 header wrapping. The default is 78, as recommended (but not required) 53 by RFC 2822. 54 55 The policy keyword specifies a policy object that controls a number of 56 aspects of the generator's operation. If no policy is specified, 57 the policy associated with the Message object passed to the 58 flatten method is used. 59 60 """ 61 62 if mangle_from_ is None: 63 mangle_from_ = True if policy is None else policy.mangle_from_ 64 self._fp = outfp 65 self._mangle_from_ = mangle_from_ 66 self.maxheaderlen = maxheaderlen 67 self.policy = policy 68 69 def write(self, s): 70 # Just delegate to the file object 71 self._fp.write(s) 72 73 def flatten(self, msg, unixfrom=False, linesep=None): 74 r"""Print the message object tree rooted at msg to the output file 75 specified when the Generator instance was created. 76 77 unixfrom is a flag that forces the printing of a Unix From_ delimiter 78 before the first object in the message tree. If the original message 79 has no From_ delimiter, a `standard' one is crafted. By default, this 80 is False to inhibit the printing of any From_ delimiter. 81 82 Note that for subobjects, no From_ line is printed. 83 84 linesep specifies the characters used to indicate a new line in 85 the output. The default value is determined by the policy specified 86 when the Generator instance was created or, if none was specified, 87 from the policy associated with the msg. 88 89 """ 90 # We use the _XXX constants for operating on data that comes directly 91 # from the msg, and _encoded_XXX constants for operating on data that 92 # has already been converted (to bytes in the BytesGenerator) and 93 # inserted into a temporary buffer. 94 policy = msg.policy if self.policy is None else self.policy 95 if linesep is not None: 96 policy = policy.clone(linesep=linesep) 97 if self.maxheaderlen is not None: 98 policy = policy.clone(max_line_length=self.maxheaderlen) 99 self._NL = policy.linesep 100 self._encoded_NL = self._encode(self._NL) 101 self._EMPTY = '' 102 self._encoded_EMPTY = self._encode(self._EMPTY) 103 # Because we use clone (below) when we recursively process message 104 # subparts, and because clone uses the computed policy (not None), 105 # submessages will automatically get set to the computed policy when 106 # they are processed by this code. 107 old_gen_policy = self.policy 108 old_msg_policy = msg.policy 109 try: 110 self.policy = policy 111 msg.policy = policy 112 if unixfrom: 113 ufrom = msg.get_unixfrom() 114 if not ufrom: 115 ufrom = 'From nobody ' + time.ctime(time.time()) 116 self.write(ufrom + self._NL) 117 self._write(msg) 118 finally: 119 self.policy = old_gen_policy 120 msg.policy = old_msg_policy 121 122 def clone(self, fp): 123 """Clone this generator with the exact same options.""" 124 return self.__class__(fp, 125 self._mangle_from_, 126 None, # Use policy setting, which we've adjusted 127 policy=self.policy) 128 129 # 130 # Protected interface - undocumented ;/ 131 # 132 133 # Note that we use 'self.write' when what we are writing is coming from 134 # the source, and self._fp.write when what we are writing is coming from a 135 # buffer (because the Bytes subclass has already had a chance to transform 136 # the data in its write method in that case). This is an entirely 137 # pragmatic split determined by experiment; we could be more general by 138 # always using write and having the Bytes subclass write method detect when 139 # it has already transformed the input; but, since this whole thing is a 140 # hack anyway this seems good enough. 141 142 def _new_buffer(self): 143 # BytesGenerator overrides this to return BytesIO. 144 return StringIO() 145 146 def _encode(self, s): 147 # BytesGenerator overrides this to encode strings to bytes. 148 return s 149 150 def _write_lines(self, lines): 151 # We have to transform the line endings. 152 if not lines: 153 return 154 lines = NLCRE.split(lines) 155 for line in lines[:-1]: 156 self.write(line) 157 self.write(self._NL) 158 if lines[-1]: 159 self.write(lines[-1]) 160 # XXX logic tells me this else should be needed, but the tests fail 161 # with it and pass without it. (NLCRE.split ends with a blank element 162 # if and only if there was a trailing newline.) 163 #else: 164 # self.write(self._NL) 165 166 def _write(self, msg): 167 # We can't write the headers yet because of the following scenario: 168 # say a multipart message includes the boundary string somewhere in 169 # its body. We'd have to calculate the new boundary /before/ we write 170 # the headers so that we can write the correct Content-Type: 171 # parameter. 172 # 173 # The way we do this, so as to make the _handle_*() methods simpler, 174 # is to cache any subpart writes into a buffer. The we write the 175 # headers and the buffer contents. That way, subpart handlers can 176 # Do The Right Thing, and can still modify the Content-Type: header if 177 # necessary. 178 oldfp = self._fp 179 try: 180 self._munge_cte = None 181 self._fp = sfp = self._new_buffer() 182 self._dispatch(msg) 183 finally: 184 self._fp = oldfp 185 munge_cte = self._munge_cte 186 del self._munge_cte 187 # If we munged the cte, copy the message again and re-fix the CTE. 188 if munge_cte: 189 msg = deepcopy(msg) 190 msg.replace_header('content-transfer-encoding', munge_cte[0]) 191 msg.replace_header('content-type', munge_cte[1]) 192 # Write the headers. First we see if the message object wants to 193 # handle that itself. If not, we'll do it generically. 194 meth = getattr(msg, '_write_headers', None) 195 if meth is None: 196 self._write_headers(msg) 197 else: 198 meth(self) 199 self._fp.write(sfp.getvalue()) 200 201 def _dispatch(self, msg): 202 # Get the Content-Type: for the message, then try to dispatch to 203 # self._handle_<maintype>_<subtype>(). If there's no handler for the 204 # full MIME type, then dispatch to self._handle_<maintype>(). If 205 # that's missing too, then dispatch to self._writeBody(). 206 main = msg.get_content_maintype() 207 sub = msg.get_content_subtype() 208 specific = UNDERSCORE.join((main, sub)).replace('-', '_') 209 meth = getattr(self, '_handle_' + specific, None) 210 if meth is None: 211 generic = main.replace('-', '_') 212 meth = getattr(self, '_handle_' + generic, None) 213 if meth is None: 214 meth = self._writeBody 215 meth(msg) 216 217 # 218 # Default handlers 219 # 220 221 def _write_headers(self, msg): 222 for h, v in msg.raw_items(): 223 self.write(self.policy.fold(h, v)) 224 # A blank line always separates headers from body 225 self.write(self._NL) 226 227 # 228 # Handlers for writing types and subtypes 229 # 230 231 def _handle_text(self, msg): 232 payload = msg.get_payload() 233 if payload is None: 234 return 235 if not isinstance(payload, str): 236 raise TypeError('string payload expected: %s' % type(payload)) 237 if _has_surrogates(msg._payload): 238 charset = msg.get_param('charset') 239 if charset is not None: 240 # XXX: This copy stuff is an ugly hack to avoid modifying the 241 # existing message. 242 msg = deepcopy(msg) 243 del msg['content-transfer-encoding'] 244 msg.set_payload(payload, charset) 245 payload = msg.get_payload() 246 self._munge_cte = (msg['content-transfer-encoding'], 247 msg['content-type']) 248 if self._mangle_from_: 249 payload = fcre.sub('>From ', payload) 250 self._write_lines(payload) 251 252 # Default body handler 253 _writeBody = _handle_text 254 255 def _handle_multipart(self, msg): 256 # The trick here is to write out each part separately, merge them all 257 # together, and then make sure that the boundary we've chosen isn't 258 # present in the payload. 259 msgtexts = [] 260 subparts = msg.get_payload() 261 if subparts is None: 262 subparts = [] 263 elif isinstance(subparts, str): 264 # e.g. a non-strict parse of a message with no starting boundary. 265 self.write(subparts) 266 return 267 elif not isinstance(subparts, list): 268 # Scalar payload 269 subparts = [subparts] 270 for part in subparts: 271 s = self._new_buffer() 272 g = self.clone(s) 273 g.flatten(part, unixfrom=False, linesep=self._NL) 274 msgtexts.append(s.getvalue()) 275 # BAW: What about boundaries that are wrapped in double-quotes? 276 boundary = msg.get_boundary() 277 if not boundary: 278 # Create a boundary that doesn't appear in any of the 279 # message texts. 280 alltext = self._encoded_NL.join(msgtexts) 281 boundary = self._make_boundary(alltext) 282 msg.set_boundary(boundary) 283 # If there's a preamble, write it out, with a trailing CRLF 284 if msg.preamble is not None: 285 if self._mangle_from_: 286 preamble = fcre.sub('>From ', msg.preamble) 287 else: 288 preamble = msg.preamble 289 self._write_lines(preamble) 290 self.write(self._NL) 291 # dash-boundary transport-padding CRLF 292 self.write('--' + boundary + self._NL) 293 # body-part 294 if msgtexts: 295 self._fp.write(msgtexts.pop(0)) 296 # *encapsulation 297 # --> delimiter transport-padding 298 # --> CRLF body-part 299 for body_part in msgtexts: 300 # delimiter transport-padding CRLF 301 self.write(self._NL + '--' + boundary + self._NL) 302 # body-part 303 self._fp.write(body_part) 304 # close-delimiter transport-padding 305 self.write(self._NL + '--' + boundary + '--' + self._NL) 306 if msg.epilogue is not None: 307 if self._mangle_from_: 308 epilogue = fcre.sub('>From ', msg.epilogue) 309 else: 310 epilogue = msg.epilogue 311 self._write_lines(epilogue) 312 313 def _handle_multipart_signed(self, msg): 314 # The contents of signed parts has to stay unmodified in order to keep 315 # the signature intact per RFC1847 2.1, so we disable header wrapping. 316 # RDM: This isn't enough to completely preserve the part, but it helps. 317 p = self.policy 318 self.policy = p.clone(max_line_length=0) 319 try: 320 self._handle_multipart(msg) 321 finally: 322 self.policy = p 323 324 def _handle_message_delivery_status(self, msg): 325 # We can't just write the headers directly to self's file object 326 # because this will leave an extra newline between the last header 327 # block and the boundary. Sigh. 328 blocks = [] 329 for part in msg.get_payload(): 330 s = self._new_buffer() 331 g = self.clone(s) 332 g.flatten(part, unixfrom=False, linesep=self._NL) 333 text = s.getvalue() 334 lines = text.split(self._encoded_NL) 335 # Strip off the unnecessary trailing empty line 336 if lines and lines[-1] == self._encoded_EMPTY: 337 blocks.append(self._encoded_NL.join(lines[:-1])) 338 else: 339 blocks.append(text) 340 # Now join all the blocks with an empty line. This has the lovely 341 # effect of separating each block with an empty line, but not adding 342 # an extra one after the last one. 343 self._fp.write(self._encoded_NL.join(blocks)) 344 345 def _handle_message(self, msg): 346 s = self._new_buffer() 347 g = self.clone(s) 348 # The payload of a message/rfc822 part should be a multipart sequence 349 # of length 1. The zeroth element of the list should be the Message 350 # object for the subpart. Extract that object, stringify it, and 351 # write it out. 352 # Except, it turns out, when it's a string instead, which happens when 353 # and only when HeaderParser is used on a message of mime type 354 # message/rfc822. Such messages are generated by, for example, 355 # Groupwise when forwarding unadorned messages. (Issue 7970.) So 356 # in that case we just emit the string body. 357 payload = msg._payload 358 if isinstance(payload, list): 359 g.flatten(msg.get_payload(0), unixfrom=False, linesep=self._NL) 360 payload = s.getvalue() 361 else: 362 payload = self._encode(payload) 363 self._fp.write(payload) 364 365 # This used to be a module level function; we use a classmethod for this 366 # and _compile_re so we can continue to provide the module level function 367 # for backward compatibility by doing 368 # _make_boundary = Generator._make_boundary 369 # at the end of the module. It *is* internal, so we could drop that... 370 @classmethod 371 def _make_boundary(cls, text=None): 372 # Craft a random boundary. If text is given, ensure that the chosen 373 # boundary doesn't appear in the text. 374 token = random.randrange(sys.maxsize) 375 boundary = ('=' * 15) + (_fmt % token) + '==' 376 if text is None: 377 return boundary 378 b = boundary 379 counter = 0 380 while True: 381 cre = cls._compile_re('^--' + re.escape(b) + '(--)?$', re.MULTILINE) 382 if not cre.search(text): 383 break 384 b = boundary + '.' + str(counter) 385 counter += 1 386 return b 387 388 @classmethod 389 def _compile_re(cls, s, flags): 390 return re.compile(s, flags) 391 392 394 class BytesGenerator(Generator): 395 """Generates a bytes version of a Message object tree. 396 397 Functionally identical to the base Generator except that the output is 398 bytes and not string. When surrogates were used in the input to encode 399 bytes, these are decoded back to bytes for output. If the policy has 400 cte_type set to 7bit, then the message is transformed such that the 401 non-ASCII bytes are properly content transfer encoded, using the charset 402 unknown-8bit. 403 404 The outfp object must accept bytes in its write method. 405 """ 406 407 def write(self, s): 408 self._fp.write(s.encode('ascii', 'surrogateescape')) 409 410 def _new_buffer(self): 411 return BytesIO() 412 413 def _encode(self, s): 414 return s.encode('ascii') 415 416 def _write_headers(self, msg): 417 # This is almost the same as the string version, except for handling 418 # strings with 8bit bytes. 419 for h, v in msg.raw_items(): 420 self._fp.write(self.policy.fold_binary(h, v)) 421 # A blank line always separates headers from body 422 self.write(self._NL) 423 424 def _handle_text(self, msg): 425 # If the string has surrogates the original source was bytes, so 426 # just write it back out. 427 if msg._payload is None: 428 return 429 if _has_surrogates(msg._payload) and not self.policy.cte_type=='7bit': 430 if self._mangle_from_: 431 msg._payload = fcre.sub(">From ", msg._payload) 432 self._write_lines(msg._payload) 433 else: 434 super(BytesGenerator,self)._handle_text(msg) 435 436 # Default body handler 437 _writeBody = _handle_text 438 439 @classmethod 440 def _compile_re(cls, s, flags): 441 return re.compile(s.encode('ascii'), flags) 442 443 444 446 _FMT = '[Non-text (%(type)s) part of message omitted, filename %(filename)s]' 447 448 class DecodedGenerator(Generator): 449 """Generates a text representation of a message. 450 451 Like the Generator base class, except that non-text parts are substituted 452 with a format string representing the part. 453 """ 454 def __init__(self, outfp, mangle_from_=None, maxheaderlen=None, fmt=None, *, 455 policy=None): 456 """Like Generator.__init__() except that an additional optional 457 argument is allowed. 458 459 Walks through all subparts of a message. If the subpart is of main 460 type `text', then it prints the decoded payload of the subpart. 461 462 Otherwise, fmt is a format string that is used instead of the message 463 payload. fmt is expanded with the following keywords (in 464 %(keyword)s format): 465 466 type : Full MIME type of the non-text part 467 maintype : Main MIME type of the non-text part 468 subtype : Sub-MIME type of the non-text part 469 filename : Filename of the non-text part 470 description: Description associated with the non-text part 471 encoding : Content transfer encoding of the non-text part 472 473 The default value for fmt is None, meaning 474 475 [Non-text (%(type)s) part of message omitted, filename %(filename)s] 476 """ 477 Generator.__init__(self, outfp, mangle_from_, maxheaderlen, 478 policy=policy) 479 if fmt is None: 480 self._fmt = _FMT 481 else: 482 self._fmt = fmt 483 484 def _dispatch(self, msg): 485 for part in msg.walk(): 486 maintype = part.get_content_maintype() 487 if maintype == 'text': 488 print(part.get_payload(decode=False), file=self) 489 elif maintype == 'multipart': 490 # Just skip this 491 pass 492 else: 493 print(self._fmt % { 494 'type' : part.get_content_type(), 495 'maintype' : part.get_content_maintype(), 496 'subtype' : part.get_content_subtype(), 497 'filename' : part.get_filename('[no filename]'), 498 'description': part.get('Content-Description', 499 '[no description]'), 500 'encoding' : part.get('Content-Transfer-Encoding', 501 '[no encoding]'), 502 }, file=self) 503 504 505 507 # Helper used by Generator._make_boundary 508 _width = len(repr(sys.maxsize-1)) 509 _fmt = '%%0%dd' % _width 510 511 # Backward compatibility 512 _make_boundary = Generator._make_boundary 513
