source: python/trunk/Lib/email/header.py@ 383

Last change on this file since 383 was 2, checked in by Yuri Dario, 15 years ago

Initial import for vendor code.
  • Property svn:eol-style set to native
File size: 21.2 KB
Line 
1# Copyright (C) 2002-2006 Python Software Foundation
2# Author: Ben Gertzfield, Barry Warsaw
3# Contact: email-sig@python.org
4
5"""Header encoding and decoding functionality."""
6
7__all__ = [
8 'Header',
9 'decode_header',
10 'make_header',
11 ]
12
13import re
14import binascii
15
16import email.quoprimime
17import email.base64mime
18
19from email.errors import HeaderParseError
20from email.charset import Charset
21
22NL = '\n'
23SPACE = ' '
24USPACE = u' '
25SPACE8 = ' ' * 8
26UEMPTYSTRING = u''
27
28MAXLINELEN = 76
29
30USASCII = Charset('us-ascii')
31UTF8 = Charset('utf-8')
32
33# Match encoded-word strings in the form =?charset?q?Hello_World?=
34ecre = re.compile(r'''
35 =\? # literal =?
36 (?P<charset>[^?]*?) # non-greedy up to the next ? is the charset
37 \? # literal ?
38 (?P<encoding>[qb]) # either a "q" or a "b", case insensitive
39 \? # literal ?
40 (?P<encoded>.*?) # non-greedy up to the next ?= is the encoded string
41 \?= # literal ?=
42 (?=[ \t]|$) # whitespace or the end of the string
43 ''', re.VERBOSE | re.IGNORECASE | re.MULTILINE)
44
45# Field name regexp, including trailing colon, but not separating whitespace,
46# according to RFC 2822. Character range is from tilde to exclamation mark.
47# For use with .match()
48fcre = re.compile(r'[\041-\176]+:$')
49
50
51
52
53# Helpers
54_max_append = email.quoprimime._max_append
55
56
57
58
59def decode_header(header):
60 """Decode a message header value without converting charset.
61
62 Returns a list of (decoded_string, charset) pairs containing each of the
63 decoded parts of the header. Charset is None for non-encoded parts of the
64 header, otherwise a lower-case string containing the name of the character
65 set specified in the encoded string.
66
67 An email.errors.HeaderParseError may be raised when certain decoding error
68 occurs (e.g. a base64 decoding exception).
69 """
70 # If no encoding, just return the header
71 header = str(header)
72 if not ecre.search(header):
73 return [(header, None)]
74 decoded = []
75 dec = ''
76 for line in header.splitlines():
77 # This line might not have an encoding in it
78 if not ecre.search(line):
79 decoded.append((line, None))
80 continue
81 parts = ecre.split(line)
82 while parts:
83 unenc = parts.pop(0).strip()
84 if unenc:
85 # Should we continue a long line?
86 if decoded and decoded[-1][1] is None:
87 decoded[-1] = (decoded[-1][0] + SPACE + unenc, None)
88 else:
89 decoded.append((unenc, None))
90 if parts:
91 charset, encoding = [s.lower() for s in parts[0:2]]
92 encoded = parts[2]
93 dec = None
94 if encoding == 'q':
95 dec = email.quoprimime.header_decode(encoded)
96 elif encoding == 'b':
97 try:
98 dec = email.base64mime.decode(encoded)
99 except binascii.Error:
100 # Turn this into a higher level exception. BAW: Right
101 # now we throw the lower level exception away but
102 # when/if we get exception chaining, we'll preserve it.
103 raise HeaderParseError
104 if dec is None:
105 dec = encoded
106
107 if decoded and decoded[-1][1] == charset:
108 decoded[-1] = (decoded[-1][0] + dec, decoded[-1][1])
109 else:
110 decoded.append((dec, charset))
111 del parts[0:3]
112 return decoded
113
114
115
116
117def make_header(decoded_seq, maxlinelen=None, header_name=None,
118 continuation_ws=' '):
119 """Create a Header from a sequence of pairs as returned by decode_header()
120
121 decode_header() takes a header value string and returns a sequence of
122 pairs of the format (decoded_string, charset) where charset is the string
123 name of the character set.
124
125 This function takes one of those sequence of pairs and returns a Header
126 instance. Optional maxlinelen, header_name, and continuation_ws are as in
127 the Header constructor.
128 """
129 h = Header(maxlinelen=maxlinelen, header_name=header_name,
130 continuation_ws=continuation_ws)
131 for s, charset in decoded_seq:
132 # None means us-ascii but we can simply pass it on to h.append()
133 if charset is not None and not isinstance(charset, Charset):
134 charset = Charset(charset)
135 h.append(s, charset)
136 return h
137
138
139
140
141class Header:
142 def __init__(self, s=None, charset=None,
143 maxlinelen=None, header_name=None,
144 continuation_ws=' ', errors='strict'):
145 """Create a MIME-compliant header that can contain many character sets.
146
147 Optional s is the initial header value. If None, the initial header
148 value is not set. You can later append to the header with .append()
149 method calls. s may be a byte string or a Unicode string, but see the
150 .append() documentation for semantics.
151
152 Optional charset serves two purposes: it has the same meaning as the
153 charset argument to the .append() method. It also sets the default
154 character set for all subsequent .append() calls that omit the charset
155 argument. If charset is not provided in the constructor, the us-ascii
156 charset is used both as s's initial charset and as the default for
157 subsequent .append() calls.
158
159 The maximum line length can be specified explicit via maxlinelen. For
160 splitting the first line to a shorter value (to account for the field
161 header which isn't included in s, e.g. `Subject') pass in the name of
162 the field in header_name. The default maxlinelen is 76.
163
164 continuation_ws must be RFC 2822 compliant folding whitespace (usually
165 either a space or a hard tab) which will be prepended to continuation
166 lines.
167
168 errors is passed through to the .append() call.
169 """
170 if charset is None:
171 charset = USASCII
172 if not isinstance(charset, Charset):
173 charset = Charset(charset)
174 self._charset = charset
175 self._continuation_ws = continuation_ws
176 cws_expanded_len = len(continuation_ws.replace('\t', SPACE8))
177 # BAW: I believe `chunks' and `maxlinelen' should be non-public.
178 self._chunks = []
179 if s is not None:
180 self.append(s, charset, errors)
181 if maxlinelen is None:
182 maxlinelen = MAXLINELEN
183 if header_name is None:
184 # We don't know anything about the field header so the first line
185 # is the same length as subsequent lines.
186 self._firstlinelen = maxlinelen
187 else:
188 # The first line should be shorter to take into account the field
189 # header. Also subtract off 2 extra for the colon and space.
190 self._firstlinelen = maxlinelen - len(header_name) - 2
191 # Second and subsequent lines should subtract off the length in
192 # columns of the continuation whitespace prefix.
193 self._maxlinelen = maxlinelen - cws_expanded_len
194
195 def __str__(self):
196 """A synonym for self.encode()."""
197 return self.encode()
198
199 def __unicode__(self):
200 """Helper for the built-in unicode function."""
201 uchunks = []
202 lastcs = None
203 for s, charset in self._chunks:
204 # We must preserve spaces between encoded and non-encoded word
205 # boundaries, which means for us we need to add a space when we go
206 # from a charset to None/us-ascii, or from None/us-ascii to a
207 # charset. Only do this for the second and subsequent chunks.
208 nextcs = charset
209 if uchunks:
210 if lastcs not in (None, 'us-ascii'):
211 if nextcs in (None, 'us-ascii'):
212 uchunks.append(USPACE)
213 nextcs = None
214 elif nextcs not in (None, 'us-ascii'):
215 uchunks.append(USPACE)
216 lastcs = nextcs
217 uchunks.append(unicode(s, str(charset)))
218 return UEMPTYSTRING.join(uchunks)
219
220 # Rich comparison operators for equality only. BAW: does it make sense to
221 # have or explicitly disable <, <=, >, >= operators?
222 def __eq__(self, other):
223 # other may be a Header or a string. Both are fine so coerce
224 # ourselves to a string, swap the args and do another comparison.
225 return other == self.encode()
226
227 def __ne__(self, other):
228 return not self == other
229
230 def append(self, s, charset=None, errors='strict'):
231 """Append a string to the MIME header.
232
233 Optional charset, if given, should be a Charset instance or the name
234 of a character set (which will be converted to a Charset instance). A
235 value of None (the default) means that the charset given in the
236 constructor is used.
237
238 s may be a byte string or a Unicode string. If it is a byte string
239 (i.e. isinstance(s, str) is true), then charset is the encoding of
240 that byte string, and a UnicodeError will be raised if the string
241 cannot be decoded with that charset. If s is a Unicode string, then
242 charset is a hint specifying the character set of the characters in
243 the string. In this case, when producing an RFC 2822 compliant header
244 using RFC 2047 rules, the Unicode string will be encoded using the
245 following charsets in order: us-ascii, the charset hint, utf-8. The
246 first character set not to provoke a UnicodeError is used.
247
248 Optional `errors' is passed as the third argument to any unicode() or
249 ustr.encode() call.
250 """
251 if charset is None:
252 charset = self._charset
253 elif not isinstance(charset, Charset):
254 charset = Charset(charset)
255 # If the charset is our faux 8bit charset, leave the string unchanged
256 if charset != '8bit':
257 # We need to test that the string can be converted to unicode and
258 # back to a byte string, given the input and output codecs of the
259 # charset.
260 if isinstance(s, str):
261 # Possibly raise UnicodeError if the byte string can't be
262 # converted to a unicode with the input codec of the charset.
263 incodec = charset.input_codec or 'us-ascii'
264 ustr = unicode(s, incodec, errors)
265 # Now make sure that the unicode could be converted back to a
266 # byte string with the output codec, which may be different
267 # than the iput coded. Still, use the original byte string.
268 outcodec = charset.output_codec or 'us-ascii'
269 ustr.encode(outcodec, errors)
270 elif isinstance(s, unicode):
271 # Now we have to be sure the unicode string can be converted
272 # to a byte string with a reasonable output codec. We want to
273 # use the byte string in the chunk.
274 for charset in USASCII, charset, UTF8:
275 try:
276 outcodec = charset.output_codec or 'us-ascii'
277 s = s.encode(outcodec, errors)
278 break
279 except UnicodeError:
280 pass
281 else:
282 assert False, 'utf-8 conversion failed'
283 self._chunks.append((s, charset))
284
285 def _split(self, s, charset, maxlinelen, splitchars):
286 # Split up a header safely for use with encode_chunks.
287 splittable = charset.to_splittable(s)
288 encoded = charset.from_splittable(splittable, True)
289 elen = charset.encoded_header_len(encoded)
290 # If the line's encoded length first, just return it
291 if elen <= maxlinelen:
292 return [(encoded, charset)]
293 # If we have undetermined raw 8bit characters sitting in a byte
294 # string, we really don't know what the right thing to do is. We
295 # can't really split it because it might be multibyte data which we
296 # could break if we split it between pairs. The least harm seems to
297 # be to not split the header at all, but that means they could go out
298 # longer than maxlinelen.
299 if charset == '8bit':
300 return [(s, charset)]
301 # BAW: I'm not sure what the right test here is. What we're trying to
302 # do is be faithful to RFC 2822's recommendation that ($2.2.3):
303 #
304 # "Note: Though structured field bodies are defined in such a way that
305 # folding can take place between many of the lexical tokens (and even
306 # within some of the lexical tokens), folding SHOULD be limited to
307 # placing the CRLF at higher-level syntactic breaks."
308 #
309 # For now, I can only imagine doing this when the charset is us-ascii,
310 # although it's possible that other charsets may also benefit from the
311 # higher-level syntactic breaks.
312 elif charset == 'us-ascii':
313 return self._split_ascii(s, charset, maxlinelen, splitchars)
314 # BAW: should we use encoded?
315 elif elen == len(s):
316 # We can split on _maxlinelen boundaries because we know that the
317 # encoding won't change the size of the string
318 splitpnt = maxlinelen
319 first = charset.from_splittable(splittable[:splitpnt], False)
320 last = charset.from_splittable(splittable[splitpnt:], False)
321 else:
322 # Binary search for split point
323 first, last = _binsplit(splittable, charset, maxlinelen)
324 # first is of the proper length so just wrap it in the appropriate
325 # chrome. last must be recursively split.
326 fsplittable = charset.to_splittable(first)
327 fencoded = charset.from_splittable(fsplittable, True)
328 chunk = [(fencoded, charset)]
329 return chunk + self._split(last, charset, self._maxlinelen, splitchars)
330
331 def _split_ascii(self, s, charset, firstlen, splitchars):
332 chunks = _split_ascii(s, firstlen, self._maxlinelen,
333 self._continuation_ws, splitchars)
334 return zip(chunks, [charset]*len(chunks))
335
336 def _encode_chunks(self, newchunks, maxlinelen):
337 # MIME-encode a header with many different charsets and/or encodings.
338 #
339 # Given a list of pairs (string, charset), return a MIME-encoded
340 # string suitable for use in a header field. Each pair may have
341 # different charsets and/or encodings, and the resulting header will
342 # accurately reflect each setting.
343 #
344 # Each encoding can be email.utils.QP (quoted-printable, for
345 # ASCII-like character sets like iso-8859-1), email.utils.BASE64
346 # (Base64, for non-ASCII like character sets like KOI8-R and
347 # iso-2022-jp), or None (no encoding).
348 #
349 # Each pair will be represented on a separate line; the resulting
350 # string will be in the format:
351 #
352 # =?charset1?q?Mar=EDa_Gonz=E1lez_Alonso?=\n
353 # =?charset2?b?SvxyZ2VuIEL2aW5n?="
354 chunks = []
355 for header, charset in newchunks:
356 if not header:
357 continue
358 if charset is None or charset.header_encoding is None:
359 s = header
360 else:
361 s = charset.header_encode(header)
362 # Don't add more folding whitespace than necessary
363 if chunks and chunks[-1].endswith(' '):
364 extra = ''
365 else:
366 extra = ' '
367 _max_append(chunks, s, maxlinelen, extra)
368 joiner = NL + self._continuation_ws
369 return joiner.join(chunks)
370
371 def encode(self, splitchars=';, '):
372 """Encode a message header into an RFC-compliant format.
373
374 There are many issues involved in converting a given string for use in
375 an email header. Only certain character sets are readable in most
376 email clients, and as header strings can only contain a subset of
377 7-bit ASCII, care must be taken to properly convert and encode (with
378 Base64 or quoted-printable) header strings. In addition, there is a
379 75-character length limit on any given encoded header field, so
380 line-wrapping must be performed, even with double-byte character sets.
381
382 This method will do its best to convert the string to the correct
383 character set used in email, and encode and line wrap it safely with
384 the appropriate scheme for that character set.
385
386 If the given charset is not known or an error occurs during
387 conversion, this function will return the header untouched.
388
389 Optional splitchars is a string containing characters to split long
390 ASCII lines on, in rough support of RFC 2822's `highest level
391 syntactic breaks'. This doesn't affect RFC 2047 encoded lines.
392 """
393 newchunks = []
394 maxlinelen = self._firstlinelen
395 lastlen = 0
396 for s, charset in self._chunks:
397 # The first bit of the next chunk should be just long enough to
398 # fill the next line. Don't forget the space separating the
399 # encoded words.
400 targetlen = maxlinelen - lastlen - 1
401 if targetlen < charset.encoded_header_len(''):
402 # Stick it on the next line
403 targetlen = maxlinelen
404 newchunks += self._split(s, charset, targetlen, splitchars)
405 lastchunk, lastcharset = newchunks[-1]
406 lastlen = lastcharset.encoded_header_len(lastchunk)
407 return self._encode_chunks(newchunks, maxlinelen)
408
409
410
411
412def _split_ascii(s, firstlen, restlen, continuation_ws, splitchars):
413 lines = []
414 maxlen = firstlen
415 for line in s.splitlines():
416 # Ignore any leading whitespace (i.e. continuation whitespace) already
417 # on the line, since we'll be adding our own.
418 line = line.lstrip()
419 if len(line) < maxlen:
420 lines.append(line)
421 maxlen = restlen
422 continue
423 # Attempt to split the line at the highest-level syntactic break
424 # possible. Note that we don't have a lot of smarts about field
425 # syntax; we just try to break on semi-colons, then commas, then
426 # whitespace.
427 for ch in splitchars:
428 if ch in line:
429 break
430 else:
431 # There's nothing useful to split the line on, not even spaces, so
432 # just append this line unchanged
433 lines.append(line)
434 maxlen = restlen
435 continue
436 # Now split the line on the character plus trailing whitespace
437 cre = re.compile(r'%s\s*' % ch)
438 if ch in ';,':
439 eol = ch
440 else:
441 eol = ''
442 joiner = eol + ' '
443 joinlen = len(joiner)
444 wslen = len(continuation_ws.replace('\t', SPACE8))
445 this = []
446 linelen = 0
447 for part in cre.split(line):
448 curlen = linelen + max(0, len(this)-1) * joinlen
449 partlen = len(part)
450 onfirstline = not lines
451 # We don't want to split after the field name, if we're on the
452 # first line and the field name is present in the header string.
453 if ch == ' ' and onfirstline and \
454 len(this) == 1 and fcre.match(this[0]):
455 this.append(part)
456 linelen += partlen
457 elif curlen + partlen > maxlen:
458 if this:
459 lines.append(joiner.join(this) + eol)
460 # If this part is longer than maxlen and we aren't already
461 # splitting on whitespace, try to recursively split this line
462 # on whitespace.
463 if partlen > maxlen and ch != ' ':
464 subl = _split_ascii(part, maxlen, restlen,
465 continuation_ws, ' ')
466 lines.extend(subl[:-1])
467 this = [subl[-1]]
468 else:
469 this = [part]
470 linelen = wslen + len(this[-1])
471 maxlen = restlen
472 else:
473 this.append(part)
474 linelen += partlen
475 # Put any left over parts on a line by themselves
476 if this:
477 lines.append(joiner.join(this))
478 return lines
479
480
481
482
483def _binsplit(splittable, charset, maxlinelen):
484 i = 0
485 j = len(splittable)
486 while i < j:
487 # Invariants:
488 # 1. splittable[:k] fits for all k <= i (note that we *assume*,
489 # at the start, that splittable[:0] fits).
490 # 2. splittable[:k] does not fit for any k > j (at the start,
491 # this means we shouldn't look at any k > len(splittable)).
492 # 3. We don't know about splittable[:k] for k in i+1..j.
493 # 4. We want to set i to the largest k that fits, with i <= k <= j.
494 #
495 m = (i+j+1) >> 1 # ceiling((i+j)/2); i < m <= j
496 chunk = charset.from_splittable(splittable[:m], True)
497 chunklen = charset.encoded_header_len(chunk)
498 if chunklen <= maxlinelen:
499 # m is acceptable, so is a new lower bound.
500 i = m
501 else:
502 # m is not acceptable, so final i must be < m.
503 j = m - 1
504 # i == j. Invariant #1 implies that splittable[:i] fits, and
505 # invariant #2 implies that splittable[:i+1] does not fit, so i
506 # is what we're looking for.
507 first = charset.from_splittable(splittable[:i], False)
508 last = charset.from_splittable(splittable[i:], False)
509 return first, last
Note: See TracBrowser for help on using the repository browser.