Lib/email/Header.py

   1 # Copyright (C) 2002 Python Software Foundation
   2 # Author: che@debian.org (Ben Gertzfield), barry@zope.com (Barry Warsaw)
   3
   4 """Header encoding and decoding functionality."""
   5
   6 import re
   7 from types import StringType, UnicodeType
   8
   9 import email.quopriMIME
  10 import email.base64MIME
  11 from email.Charset import Charset
  12
  13 try:
  14     from email._compat22 import _floordiv
  15 except SyntaxError:
  16     # Python 2.1 spells integer division differently
  17     from email._compat21 import _floordiv
  18
  19 try:
  20     True, False
  21 except NameError:
  22     True = 1
  23     False = 0
  24
  25 CRLFSPACE = '\r\n '
  26 CRLF = '\r\n'
  27 NL = '\n'
  28 SPACE8 = ' ' * 8
  29 EMPTYSTRING = ''
  30
  31 MAXLINELEN = 76
  32
  33 ENCODE = 1
  34 DECODE = 2
  35
  36 USASCII = Charset('us-ascii')
  37 UTF8 = Charset('utf-8')
  38
  39 # Match encoded-word strings in the form =?charset?q?Hello_World?=
  40 ecre = re.compile(r'''
  41   =\?                   # literal =?
  42   (?P<charset>[^?]*?)   # non-greedy up to the next ? is the charset
  43   \?                    # literal ?
  44   (?P<encoding>[qb])    # either a "q" or a "b", case insensitive
  45   \?                    # literal ?
  46   (?P<encoded>.*?)      # non-greedy up to the next ?= is the encoded string
  47   \?=                   # literal ?=
  48   ''', re.VERBOSE | re.IGNORECASE)
  49
  50
  51 \f
  52 # Helpers
  53 _max_append = email.quopriMIME._max_append
  54
  55
  56 \f
  57 def decode_header(header):
  58     """Decode a message header value without converting charset.
  59
  60     Returns a list of (decoded_string, charset) pairs containing each of the
  61     decoded parts of the header.  Charset is None for non-encoded parts of the
  62     header, otherwise a lower-case string containing the name of the character
  63     set specified in the encoded string.
  64     """
  65     # If no encoding, just return the header
  66     header = str(header)
  67     if not ecre.search(header):
  68         return [(header, None)]
  69     decoded = []
  70     dec = ''
  71     for line in header.splitlines():
  72         # This line might not have an encoding in it
  73         if not ecre.search(line):
  74             decoded.append((line, None))
  75             continue
  76         parts = ecre.split(line)
  77         while parts:
  78             unenc = parts.pop(0).strip()
  79             if unenc:
  80                 # Should we continue a long line?
  81                 if decoded and decoded[-1][1] is None:
  82                     decoded[-1] = (decoded[-1][0] + dec, None)
  83                 else:
  84                     decoded.append((unenc, None))
  85             if parts:
  86                 charset, encoding = [s.lower() for s in parts[0:2]]
  87                 encoded = parts[2]
  88                 dec = ''
  89                 if encoding == 'q':
  90                     dec = email.quopriMIME.header_decode(encoded)
  91                 elif encoding == 'b':
  92                     dec = email.base64MIME.decode(encoded)
  93                 else:
  94                     dec = encoded
  95
  96                 if decoded and decoded[-1][1] == charset:
  97                     decoded[-1] = (decoded[-1][0] + dec, decoded[-1][1])
  98                 else:
  99                     decoded.append((dec, charset))
 100             del parts[0:3]
 101     return decoded
 102
 103
 104 \f
 105 def make_header(decoded_seq, maxlinelen=None, header_name=None,
 106                 continuation_ws=' '):
 107     """Create a Header from a sequence of pairs as returned by decode_header()
 108
 109     decode_header() takes a header value string and returns a sequence of
 110     pairs of the format (decoded_string, charset) where charset is the string
 111     name of the character set.
 112
 113     This function takes one of those sequence of pairs and returns a Header
 114     instance.  Optional maxlinelen, header_name, and continuation_ws are as in
 115     the Header constructor.
 116     """
 117     h = Header(maxlinelen=maxlinelen, header_name=header_name,
 118                continuation_ws=continuation_ws)
 119     for s, charset in decoded_seq:
 120         # None means us-ascii but we can simply pass it on to h.append()
 121         if charset is not None and not isinstance(charset, Charset):
 122             charset = Charset(charset)
 123         h.append(s, charset)
 124     return h
 125
 126
 127 \f
 128 class Header:
 129     def __init__(self, s=None, charset=None, maxlinelen=None, header_name=None,
 130                  continuation_ws=' ', errors='strict'):
 131         """Create a MIME-compliant header that can contain many character sets.
 132
 133         Optional s is the initial header value.  If None, the initial header
 134         value is not set.  You can later append to the header with .append()
 135         method calls.  s may be a byte string or a Unicode string, but see the
 136         .append() documentation for semantics.
 137
 138         Optional charset serves two purposes: it has the same meaning as the
 139         charset argument to the .append() method.  It also sets the default
 140         character set for all subsequent .append() calls that omit the charset
 141         argument.  If charset is not provided in the constructor, the us-ascii
 142         charset is used both as s's initial charset and as the default for
 143         subsequent .append() calls.
 144
 145         The maximum line length can be specified explicit via maxlinelen.  For
 146         splitting the first line to a shorter value (to account for the field
 147         header which isn't included in s, e.g. `Subject') pass in the name of
 148         the field in header_name.  The default maxlinelen is 76.
 149
 150         continuation_ws must be RFC 2822 compliant folding whitespace (usually
 151         either a space or a hard tab) which will be prepended to continuation
 152         lines.
 153
 154         errors is passed through to the .append() call.
 155         """
 156         if charset is None:
 157             charset = USASCII
 158         if not isinstance(charset, Charset):
 159             charset = Charset(charset)
 160         self._charset = charset
 161         self._continuation_ws = continuation_ws
 162         cws_expanded_len = len(continuation_ws.replace('\t', SPACE8))
 163         # BAW: I believe `chunks' and `maxlinelen' should be non-public.
 164         self._chunks = []
 165         if s is not None:
 166             self.append(s, charset, errors)
 167         if maxlinelen is None:
 168             maxlinelen = MAXLINELEN
 169         if header_name is None:
 170             # We don't know anything about the field header so the first line
 171             # is the same length as subsequent lines.
 172             self._firstlinelen = maxlinelen
 173         else:
 174             # The first line should be shorter to take into account the field
 175             # header.  Also subtract off 2 extra for the colon and space.
 176             self._firstlinelen = maxlinelen - len(header_name) - 2
 177         # Second and subsequent lines should subtract off the length in
 178         # columns of the continuation whitespace prefix.
 179         self._maxlinelen = maxlinelen - cws_expanded_len
 180
 181     def __str__(self):
 182         """A synonym for self.encode()."""
 183         return self.encode()
 184
 185     def __unicode__(self):
 186         """Helper for the built-in unicode function."""
 187         # charset item is a Charset instance so we need to stringify it.
 188         uchunks = [unicode(s, str(charset)) for s, charset in self._chunks]
 189         return u''.join(uchunks)
 190
 191     # Rich comparison operators for equality only.  BAW: does it make sense to
 192     # have or explicitly disable <, <=, >, >= operators?
 193     def __eq__(self, other):
 194         # other may be a Header or a string.  Both are fine so coerce
 195         # ourselves to a string, swap the args and do another comparison.
 196         return other == self.encode()
 197
 198     def __ne__(self, other):
 199         return not self == other
 200
 201     def append(self, s, charset=None, errors='strict'):
 202         """Append a string to the MIME header.
 203
 204         Optional charset, if given, should be a Charset instance or the name
 205         of a character set (which will be converted to a Charset instance).  A
 206         value of None (the default) means that the charset given in the
 207         constructor is used.
 208
 209         s may be a byte string or a Unicode string.  If it is a byte string
 210         (i.e. isinstance(s, StringType) is true), then charset is the encoding
 211         of that byte string, and a UnicodeError will be raised if the string
 212         cannot be decoded with that charset.  If s is a Unicode string, then
 213         charset is a hint specifying the character set of the characters in
 214         the string.  In this case, when producing an RFC 2822 compliant header
 215         using RFC 2047 rules, the Unicode string will be encoded using the
 216         following charsets in order: us-ascii, the charset hint, utf-8.  The
 217         first character set not to provoke a UnicodeError is used.
 218
 219         Optional `errors' is passed as the third argument to any unicode() or
 220         ustr.encode() call.
 221         """
 222         if charset is None:
 223             charset = self._charset
 224         elif not isinstance(charset, Charset):
 225             charset = Charset(charset)
 226         # If the charset is our faux 8bit charset, leave the string unchanged
 227         if charset <> '8bit':
 228             # We need to test that the string can be converted to unicode and
 229             # back to a byte string, given the input and output codecs of the
 230             # charset.
 231             if isinstance(s, StringType):
 232                 # Possibly raise UnicodeError if the byte string can't be
 233                 # converted to a unicode with the input codec of the charset.
 234                 incodec = charset.input_codec or 'us-ascii'
 235                 ustr = unicode(s, incodec, errors)
 236                 # Now make sure that the unicode could be converted back to a
 237                 # byte string with the output codec, which may be different
 238                 # than the iput coded.  Still, use the original byte string.
 239                 outcodec = charset.output_codec or 'us-ascii'
 240                 ustr.encode(outcodec, errors)
 241             elif isinstance(s, UnicodeType):
 242                 # Now we have to be sure the unicode string can be converted
 243                 # to a byte string with a reasonable output codec.  We want to
 244                 # use the byte string in the chunk.
 245                 for charset in USASCII, charset, UTF8:
 246                     try:
 247                         outcodec = charset.output_codec or 'us-ascii'
 248                         s = s.encode(outcodec, errors)
 249                         break
 250                     except UnicodeError:
 251                         pass
 252                 else:
 253                     assert False, 'utf-8 conversion failed'
 254         self._chunks.append((s, charset))
 255
 256     def _split(self, s, charset, firstline=False):
 257         # Split up a header safely for use with encode_chunks.
 258         splittable = charset.to_splittable(s)
 259         encoded = charset.from_splittable(splittable)
 260         elen = charset.encoded_header_len(encoded)
 261
 262         if elen <= self._maxlinelen:
 263             return [(encoded, charset)]
 264         # If we have undetermined raw 8bit characters sitting in a byte
 265         # string, we really don't know what the right thing to do is.  We
 266         # can't really split it because it might be multibyte data which we
 267         # could break if we split it between pairs.  The least harm seems to
 268         # be to not split the header at all, but that means they could go out
 269         # longer than maxlinelen.
 270         elif charset == '8bit':
 271             return [(s, charset)]
 272         # BAW: I'm not sure what the right test here is.  What we're trying to
 273         # do is be faithful to RFC 2822's recommendation that ($2.2.3):
 274         #
 275         # "Note: Though structured field bodies are defined in such a way that
 276         #  folding can take place between many of the lexical tokens (and even
 277         #  within some of the lexical tokens), folding SHOULD be limited to
 278         #  placing the CRLF at higher-level syntactic breaks."
 279         #
 280         # For now, I can only imagine doing this when the charset is us-ascii,
 281         # although it's possible that other charsets may also benefit from the
 282         # higher-level syntactic breaks.
 283         #
 284         elif charset == 'us-ascii':
 285             return self._ascii_split(s, charset, firstline)
 286         # BAW: should we use encoded?
 287         elif elen == len(s):
 288             # We can split on _maxlinelen boundaries because we know that the
 289             # encoding won't change the size of the string
 290             splitpnt = self._maxlinelen
 291             first = charset.from_splittable(splittable[:splitpnt], False)
 292             last = charset.from_splittable(splittable[splitpnt:], False)
 293         else:
 294             # Divide and conquer.
 295             halfway = _floordiv(len(splittable), 2)
 296             first = charset.from_splittable(splittable[:halfway], False)
 297             last = charset.from_splittable(splittable[halfway:], False)
 298         # Do the split
 299         return self._split(first, charset, firstline) + \
 300                self._split(last, charset)
 301
 302     def _ascii_split(self, s, charset, firstline):
 303         # Attempt to split the line at the highest-level syntactic break
 304         # possible.  Note that we don't have a lot of smarts about field
 305         # syntax; we just try to break on semi-colons, then whitespace.
 306         rtn = []
 307         lines = s.splitlines()
 308         while lines:
 309             line = lines.pop(0)
 310             if firstline:
 311                 maxlinelen = self._firstlinelen
 312                 firstline = False
 313             else:
 314                 #line = line.lstrip()
 315                 maxlinelen = self._maxlinelen
 316             # Short lines can remain unchanged
 317             if len(line.replace('\t', SPACE8)) <= maxlinelen:
 318                 rtn.append(line)
 319             else:
 320                 oldlen = len(line)
 321                 # Try to break the line on semicolons, but if that doesn't
 322                 # work, try to split on folding whitespace.
 323                 while len(line) > maxlinelen:
 324                     i = line.rfind(';', 0, maxlinelen)
 325                     if i < 0:
 326                         break
 327                     rtn.append(line[:i] + ';')
 328                     line = line[i+1:]
 329                 # Is the remaining stuff still longer than maxlinelen?
 330                 if len(line) <= maxlinelen:
 331                     # Splitting on semis worked
 332                     rtn.append(line)
 333                     continue
 334                 # Splitting on semis didn't finish the job.  If it did any
 335                 # work at all, stick the remaining junk on the front of the
 336                 # `lines' sequence and let the next pass do its thing.
 337                 if len(line) <> oldlen:
 338                     lines.insert(0, line)
 339                     continue
 340                 # Otherwise, splitting on semis didn't help at all.
 341                 parts = re.split(r'(\s+)', line)
 342                 if len(parts) == 1 or (len(parts) == 3 and
 343                                        parts[0].endswith(':')):
 344                     # This line can't be split on whitespace.  There's now
 345                     # little we can do to get this into maxlinelen.  BAW:
 346                     # We're still potentially breaking the RFC by possibly
 347                     # allowing lines longer than the absolute maximum of 998
 348                     # characters.  For now, let it slide.
 349                     #
 350                     # len(parts) will be 1 if this line has no `Field: '
 351                     # prefix, otherwise it will be len(3).
 352                     rtn.append(line)
 353                     continue
 354                 # There is whitespace we can split on.
 355                 first = parts.pop(0)
 356                 sublines = [first]
 357                 acc = len(first)
 358                 while parts:
 359                     len0 = len(parts[0])
 360                     len1 = len(parts[1])
 361                     if acc + len0 + len1 <= maxlinelen:
 362                         sublines.append(parts.pop(0))
 363                         sublines.append(parts.pop(0))
 364                         acc += len0 + len1
 365                     else:
 366                         # Split it here, but don't forget to ignore the
 367                         # next whitespace-only part
 368                         if first <> '':
 369                             rtn.append(EMPTYSTRING.join(sublines))
 370                         del parts[0]
 371                         first = parts.pop(0)
 372                         sublines = [first]
 373                         acc = len(first)
 374                 rtn.append(EMPTYSTRING.join(sublines))
 375         return [(chunk, charset) for chunk in rtn]
 376
 377     def _encode_chunks(self, newchunks):
 378         # MIME-encode a header with many different charsets and/or encodings.
 379         #
 380         # Given a list of pairs (string, charset), return a MIME-encoded
 381         # string suitable for use in a header field.  Each pair may have
 382         # different charsets and/or encodings, and the resulting header will
 383         # accurately reflect each setting.
 384         #
 385         # Each encoding can be email.Utils.QP (quoted-printable, for
 386         # ASCII-like character sets like iso-8859-1), email.Utils.BASE64
 387         # (Base64, for non-ASCII like character sets like KOI8-R and
 388         # iso-2022-jp), or None (no encoding).
 389         #
 390         # Each pair will be represented on a separate line; the resulting
 391         # string will be in the format:
 392         #
 393         # =?charset1?q?Mar=EDa_Gonz=E1lez_Alonso?=\n
 394         #  =?charset2?b?SvxyZ2VuIEL2aW5n?="
 395         #
 396         chunks = []
 397         for header, charset in newchunks:
 398             if charset is None or charset.header_encoding is None:
 399                 # There's no encoding for this chunk's charsets
 400                 _max_append(chunks, header, self._maxlinelen)
 401             else:
 402                 _max_append(chunks, charset.header_encode(header),
 403                             self._maxlinelen, ' ')
 404         joiner = NL + self._continuation_ws
 405         return joiner.join(chunks)
 406
 407     def encode(self):
 408         """Encode a message header into an RFC-compliant format.
 409
 410         There are many issues involved in converting a given string for use in
 411         an email header.  Only certain character sets are readable in most
 412         email clients, and as header strings can only contain a subset of
 413         7-bit ASCII, care must be taken to properly convert and encode (with
 414         Base64 or quoted-printable) header strings.  In addition, there is a
 415         75-character length limit on any given encoded header field, so
 416         line-wrapping must be performed, even with double-byte character sets.
 417
 418         This method will do its best to convert the string to the correct
 419         character set used in email, and encode and line wrap it safely with
 420         the appropriate scheme for that character set.
 421
 422         If the given charset is not known or an error occurs during
 423         conversion, this function will return the header untouched.
 424         """
 425         newchunks = []
 426         for s, charset in self._chunks:
 427             newchunks += self._split(s, charset, True)
 428         return self._encode_chunks(newchunks)