Lib/email/charset.py

   1 # Copyright (C) 2001-2006 Python Software Foundation
   2 # Author: Ben Gertzfield, Barry Warsaw
   3 # Contact: email-sig@python.org
   4
   5 __all__ = [
   6     'Charset',
   7     'add_alias',
   8     'add_charset',
   9     'add_codec',
  10     ]
  11
  12 import email.base64mime
  13 import email.quoprimime
  14
  15 from email import errors
  16 from email.encoders import encode_7or8bit
  17
  18
  19 \f
  20 # Flags for types of header encodings
  21 QP          = 1 # Quoted-Printable
  22 BASE64      = 2 # Base64
  23 SHORTEST    = 3 # the shorter of QP and base64, but only for headers
  24
  25 # In "=?charset?q?hello_world?=", the =?, ?q?, and ?= add up to 7
  26 MISC_LEN = 7
  27
  28 DEFAULT_CHARSET = 'us-ascii'
  29
  30
  31 \f
  32 # Defaults
  33 CHARSETS = {
  34     # input        header enc  body enc output conv
  35     'iso-8859-1':  (QP,        QP,      None),
  36     'iso-8859-2':  (QP,        QP,      None),
  37     'iso-8859-3':  (QP,        QP,      None),
  38     'iso-8859-4':  (QP,        QP,      None),
  39     # iso-8859-5 is Cyrillic, and not especially used
  40     # iso-8859-6 is Arabic, also not particularly used
  41     # iso-8859-7 is Greek, QP will not make it readable
  42     # iso-8859-8 is Hebrew, QP will not make it readable
  43     'iso-8859-9':  (QP,        QP,      None),
  44     'iso-8859-10': (QP,        QP,      None),
  45     # iso-8859-11 is Thai, QP will not make it readable
  46     'iso-8859-13': (QP,        QP,      None),
  47     'iso-8859-14': (QP,        QP,      None),
  48     'iso-8859-15': (QP,        QP,      None),
  49     'iso-8859-16': (QP,        QP,      None),
  50     'windows-1252':(QP,        QP,      None),
  51     'viscii':      (QP,        QP,      None),
  52     'us-ascii':    (None,      None,    None),
  53     'big5':        (BASE64,    BASE64,  None),
  54     'gb2312':      (BASE64,    BASE64,  None),
  55     'euc-jp':      (BASE64,    None,    'iso-2022-jp'),
  56     'shift_jis':   (BASE64,    None,    'iso-2022-jp'),
  57     'iso-2022-jp': (BASE64,    None,    None),
  58     'koi8-r':      (BASE64,    BASE64,  None),
  59     'utf-8':       (SHORTEST,  BASE64, 'utf-8'),
  60     # We're making this one up to represent raw unencoded 8-bit
  61     '8bit':        (None,      BASE64, 'utf-8'),
  62     }
  63
  64 # Aliases for other commonly-used names for character sets.  Map
  65 # them to the real ones used in email.
  66 ALIASES = {
  67     'latin_1': 'iso-8859-1',
  68     'latin-1': 'iso-8859-1',
  69     'latin_2': 'iso-8859-2',
  70     'latin-2': 'iso-8859-2',
  71     'latin_3': 'iso-8859-3',
  72     'latin-3': 'iso-8859-3',
  73     'latin_4': 'iso-8859-4',
  74     'latin-4': 'iso-8859-4',
  75     'latin_5': 'iso-8859-9',
  76     'latin-5': 'iso-8859-9',
  77     'latin_6': 'iso-8859-10',
  78     'latin-6': 'iso-8859-10',
  79     'latin_7': 'iso-8859-13',
  80     'latin-7': 'iso-8859-13',
  81     'latin_8': 'iso-8859-14',
  82     'latin-8': 'iso-8859-14',
  83     'latin_9': 'iso-8859-15',
  84     'latin-9': 'iso-8859-15',
  85     'latin_10':'iso-8859-16',
  86     'latin-10':'iso-8859-16',
  87     'cp949':   'ks_c_5601-1987',
  88     'euc_jp':  'euc-jp',
  89     'euc_kr':  'euc-kr',
  90     'ascii':   'us-ascii',
  91     }
  92
  93
  94 # Map charsets to their Unicode codec strings.
  95 CODEC_MAP = {
  96     'gb2312':      'eucgb2312_cn',
  97     'big5':        'big5_tw',
  98     # Hack: We don't want *any* conversion for stuff marked us-ascii, as all
  99     # sorts of garbage might be sent to us in the guise of 7-bit us-ascii.
 100     # Let that stuff pass through without conversion to/from Unicode.
 101     'us-ascii':    None,
 102     }
 103
 104
 105 \f
 106 # Convenience functions for extending the above mappings
 107 def add_charset(charset, header_enc=None, body_enc=None, output_charset=None):
 108     """Add character set properties to the global registry.
 109
 110     charset is the input character set, and must be the canonical name of a
 111     character set.
 112
 113     Optional header_enc and body_enc is either Charset.QP for
 114     quoted-printable, Charset.BASE64 for base64 encoding, Charset.SHORTEST for
 115     the shortest of qp or base64 encoding, or None for no encoding.  SHORTEST
 116     is only valid for header_enc.  It describes how message headers and
 117     message bodies in the input charset are to be encoded.  Default is no
 118     encoding.
 119
 120     Optional output_charset is the character set that the output should be
 121     in.  Conversions will proceed from input charset, to Unicode, to the
 122     output charset when the method Charset.convert() is called.  The default
 123     is to output in the same character set as the input.
 124
 125     Both input_charset and output_charset must have Unicode codec entries in
 126     the module's charset-to-codec mapping; use add_codec(charset, codecname)
 127     to add codecs the module does not know about.  See the codecs module's
 128     documentation for more information.
 129     """
 130     if body_enc == SHORTEST:
 131         raise ValueError('SHORTEST not allowed for body_enc')
 132     CHARSETS[charset] = (header_enc, body_enc, output_charset)
 133
 134
 135 def add_alias(alias, canonical):
 136     """Add a character set alias.
 137
 138     alias is the alias name, e.g. latin-1
 139     canonical is the character set's canonical name, e.g. iso-8859-1
 140     """
 141     ALIASES[alias] = canonical
 142
 143
 144 def add_codec(charset, codecname):
 145     """Add a codec that map characters in the given charset to/from Unicode.
 146
 147     charset is the canonical name of a character set.  codecname is the name
 148     of a Python codec, as appropriate for the second argument to the unicode()
 149     built-in, or to the encode() method of a Unicode string.
 150     """
 151     CODEC_MAP[charset] = codecname
 152
 153
 154 \f
 155 class Charset:
 156     """Map character sets to their email properties.
 157
 158     This class provides information about the requirements imposed on email
 159     for a specific character set.  It also provides convenience routines for
 160     converting between character sets, given the availability of the
 161     applicable codecs.  Given a character set, it will do its best to provide
 162     information on how to use that character set in an email in an
 163     RFC-compliant way.
 164
 165     Certain character sets must be encoded with quoted-printable or base64
 166     when used in email headers or bodies.  Certain character sets must be
 167     converted outright, and are not allowed in email.  Instances of this
 168     module expose the following information about a character set:
 169
 170     input_charset: The initial character set specified.  Common aliases
 171                    are converted to their `official' email names (e.g. latin_1
 172                    is converted to iso-8859-1).  Defaults to 7-bit us-ascii.
 173
 174     header_encoding: If the character set must be encoded before it can be
 175                      used in an email header, this attribute will be set to
 176                      Charset.QP (for quoted-printable), Charset.BASE64 (for
 177                      base64 encoding), or Charset.SHORTEST for the shortest of
 178                      QP or BASE64 encoding.  Otherwise, it will be None.
 179
 180     body_encoding: Same as header_encoding, but describes the encoding for the
 181                    mail message's body, which indeed may be different than the
 182                    header encoding.  Charset.SHORTEST is not allowed for
 183                    body_encoding.
 184
 185     output_charset: Some character sets must be converted before the can be
 186                     used in email headers or bodies.  If the input_charset is
 187                     one of them, this attribute will contain the name of the
 188                     charset output will be converted to.  Otherwise, it will
 189                     be None.
 190
 191     input_codec: The name of the Python codec used to convert the
 192                  input_charset to Unicode.  If no conversion codec is
 193                  necessary, this attribute will be None.
 194
 195     output_codec: The name of the Python codec used to convert Unicode
 196                   to the output_charset.  If no conversion codec is necessary,
 197                   this attribute will have the same value as the input_codec.
 198     """
 199     def __init__(self, input_charset=DEFAULT_CHARSET):
 200         # RFC 2046, $4.1.2 says charsets are not case sensitive.  We coerce to
 201         # unicode because its .lower() is locale insensitive.  If the argument
 202         # is already a unicode, we leave it at that, but ensure that the
 203         # charset is ASCII, as the standard (RFC XXX) requires.
 204         try:
 205             if isinstance(input_charset, unicode):
 206                 input_charset.encode('ascii')
 207             else:
 208                 input_charset = unicode(input_charset, 'ascii')
 209         except UnicodeError:
 210             raise errors.CharsetError(input_charset)
 211         input_charset = input_charset.lower()
 212         # Set the input charset after filtering through the aliases
 213         self.input_charset = ALIASES.get(input_charset, input_charset)
 214         # We can try to guess which encoding and conversion to use by the
 215         # charset_map dictionary.  Try that first, but let the user override
 216         # it.
 217         henc, benc, conv = CHARSETS.get(self.input_charset,
 218                                         (SHORTEST, BASE64, None))
 219         if not conv:
 220             conv = self.input_charset
 221         # Set the attributes, allowing the arguments to override the default.
 222         self.header_encoding = henc
 223         self.body_encoding = benc
 224         self.output_charset = ALIASES.get(conv, conv)
 225         # Now set the codecs.  If one isn't defined for input_charset,
 226         # guess and try a Unicode codec with the same name as input_codec.
 227         self.input_codec = CODEC_MAP.get(self.input_charset,
 228                                          self.input_charset)
 229         self.output_codec = CODEC_MAP.get(self.output_charset,
 230                                           self.output_charset)
 231
 232     def __str__(self):
 233         return self.input_charset.lower()
 234
 235     __repr__ = __str__
 236
 237     def __eq__(self, other):
 238         return str(self) == str(other).lower()
 239
 240     def __ne__(self, other):
 241         return not self.__eq__(other)
 242
 243     def get_body_encoding(self):
 244         """Return the content-transfer-encoding used for body encoding.
 245
 246         This is either the string `quoted-printable' or `base64' depending on
 247         the encoding used, or it is a function in which case you should call
 248         the function with a single argument, the Message object being
 249         encoded.  The function should then set the Content-Transfer-Encoding
 250         header itself to whatever is appropriate.
 251
 252         Returns "quoted-printable" if self.body_encoding is QP.
 253         Returns "base64" if self.body_encoding is BASE64.
 254         Returns "7bit" otherwise.
 255         """
 256         assert self.body_encoding != SHORTEST
 257         if self.body_encoding == QP:
 258             return 'quoted-printable'
 259         elif self.body_encoding == BASE64:
 260             return 'base64'
 261         else:
 262             return encode_7or8bit
 263
 264     def convert(self, s):
 265         """Convert a string from the input_codec to the output_codec."""
 266         if self.input_codec != self.output_codec:
 267             return unicode(s, self.input_codec).encode(self.output_codec)
 268         else:
 269             return s
 270
 271     def to_splittable(self, s):
 272         """Convert a possibly multibyte string to a safely splittable format.
 273
 274         Uses the input_codec to try and convert the string to Unicode, so it
 275         can be safely split on character boundaries (even for multibyte
 276         characters).
 277
 278         Returns the string as-is if it isn't known how to convert it to
 279         Unicode with the input_charset.
 280
 281         Characters that could not be converted to Unicode will be replaced
 282         with the Unicode replacement character U+FFFD.
 283         """
 284         if isinstance(s, unicode) or self.input_codec is None:
 285             return s
 286         try:
 287             return unicode(s, self.input_codec, 'replace')
 288         except LookupError:
 289             # Input codec not installed on system, so return the original
 290             # string unchanged.
 291             return s
 292
 293     def from_splittable(self, ustr, to_output=True):
 294         """Convert a splittable string back into an encoded string.
 295
 296         Uses the proper codec to try and convert the string from Unicode back
 297         into an encoded format.  Return the string as-is if it is not Unicode,
 298         or if it could not be converted from Unicode.
 299
 300         Characters that could not be converted from Unicode will be replaced
 301         with an appropriate character (usually '?').
 302
 303         If to_output is True (the default), uses output_codec to convert to an
 304         encoded format.  If to_output is False, uses input_codec.
 305         """
 306         if to_output:
 307             codec = self.output_codec
 308         else:
 309             codec = self.input_codec
 310         if not isinstance(ustr, unicode) or codec is None:
 311             return ustr
 312         try:
 313             return ustr.encode(codec, 'replace')
 314         except LookupError:
 315             # Output codec not installed
 316             return ustr
 317
 318     def get_output_charset(self):
 319         """Return the output character set.
 320
 321         This is self.output_charset if that is not None, otherwise it is
 322         self.input_charset.
 323         """
 324         return self.output_charset or self.input_charset
 325
 326     def encoded_header_len(self, s):
 327         """Return the length of the encoded header string."""
 328         cset = self.get_output_charset()
 329         # The len(s) of a 7bit encoding is len(s)
 330         if self.header_encoding == BASE64:
 331             return email.base64mime.base64_len(s) + len(cset) + MISC_LEN
 332         elif self.header_encoding == QP:
 333             return email.quoprimime.header_quopri_len(s) + len(cset) + MISC_LEN
 334         elif self.header_encoding == SHORTEST:
 335             lenb64 = email.base64mime.base64_len(s)
 336             lenqp = email.quoprimime.header_quopri_len(s)
 337             return min(lenb64, lenqp) + len(cset) + MISC_LEN
 338         else:
 339             return len(s)
 340
 341     def header_encode(self, s, convert=False):
 342         """Header-encode a string, optionally converting it to output_charset.
 343
 344         If convert is True, the string will be converted from the input
 345         charset to the output charset automatically.  This is not useful for
 346         multibyte character sets, which have line length issues (multibyte
 347         characters must be split on a character, not a byte boundary); use the
 348         high-level Header class to deal with these issues.  convert defaults
 349         to False.
 350
 351         The type of encoding (base64 or quoted-printable) will be based on
 352         self.header_encoding.
 353         """
 354         cset = self.get_output_charset()
 355         if convert:
 356             s = self.convert(s)
 357         # 7bit/8bit encodings return the string unchanged (modulo conversions)
 358         if self.header_encoding == BASE64:
 359             return email.base64mime.header_encode(s, cset)
 360         elif self.header_encoding == QP:
 361             return email.quoprimime.header_encode(s, cset, maxlinelen=None)
 362         elif self.header_encoding == SHORTEST:
 363             lenb64 = email.base64mime.base64_len(s)
 364             lenqp = email.quoprimime.header_quopri_len(s)
 365             if lenb64 < lenqp:
 366                 return email.base64mime.header_encode(s, cset)
 367             else:
 368                 return email.quoprimime.header_encode(s, cset, maxlinelen=None)
 369         else:
 370             return s
 371
 372     def body_encode(self, s, convert=True):
 373         """Body-encode a string and convert it to output_charset.
 374
 375         If convert is True (the default), the string will be converted from
 376         the input charset to output charset automatically.  Unlike
 377         header_encode(), there are no issues with byte boundaries and
 378         multibyte charsets in email bodies, so this is usually pretty safe.
 379
 380         The type of encoding (base64 or quoted-printable) will be based on
 381         self.body_encoding.
 382         """
 383         if convert:
 384             s = self.convert(s)
 385         # 7bit/8bit encodings return the string unchanged (module conversions)
 386         if self.body_encoding is BASE64:
 387             return email.base64mime.body_encode(s)
 388         elif self.body_encoding is QP:
 389             return email.quoprimime.body_encode(s)
 390         else:
 391             return s