2363d6de43
git-svn-id: http://google-refine.googlecode.com/svn/trunk@517 7d457c2a-affb-35e4-300a-418c747d4874
185 lines
5.7 KiB
Python
185 lines
5.7 KiB
Python
# Copyright (C) 2002-2006 Python Software Foundation
|
||
# Author: Ben Gertzfield
|
||
# Contact: email-sig@python.org
|
||
|
||
"""Base64 content transfer encoding per RFCs 2045-2047.
|
||
|
||
This module handles the content transfer encoding method defined in RFC 2045
|
||
to encode arbitrary 8-bit data using the three 8-bit bytes in four 7-bit
|
||
characters encoding known as Base64.
|
||
|
||
It is used in the MIME standards for email to attach images, audio, and text
|
||
using some 8-bit character sets to messages.
|
||
|
||
This module provides an interface to encode and decode both headers and bodies
|
||
with Base64 encoding.
|
||
|
||
RFC 2045 defines a method for including character set information in an
|
||
`encoded-word' in a header. This method is commonly used for 8-bit real names
|
||
in To:, From:, Cc:, etc. fields, as well as Subject: lines.
|
||
|
||
This module does not do the line wrapping or end-of-line character conversion
|
||
necessary for proper internationalized headers; it only does dumb encoding and
|
||
decoding. To deal with the various line wrapping issues, use the email.Header
|
||
module.
|
||
"""
|
||
|
||
__all__ = [
|
||
'base64_len',
|
||
'body_decode',
|
||
'body_encode',
|
||
'decode',
|
||
'decodestring',
|
||
'encode',
|
||
'encodestring',
|
||
'header_encode',
|
||
]
|
||
|
||
import re
|
||
|
||
from binascii import b2a_base64, a2b_base64
|
||
from email.utils import fix_eols
|
||
|
||
CRLF = '\r\n'
|
||
NL = '\n'
|
||
EMPTYSTRING = ''
|
||
|
||
# See also Charset.py
|
||
MISC_LEN = 7
|
||
|
||
|
||
|
||
# Helpers
|
||
def base64_len(s):
|
||
"""Return the length of s when it is encoded with base64."""
|
||
groups_of_3, leftover = divmod(len(s), 3)
|
||
# 4 bytes out for each 3 bytes (or nonzero fraction thereof) in.
|
||
# Thanks, Tim!
|
||
n = groups_of_3 * 4
|
||
if leftover:
|
||
n += 4
|
||
return n
|
||
|
||
|
||
|
||
def header_encode(header, charset='iso-8859-1', keep_eols=False,
|
||
maxlinelen=76, eol=NL):
|
||
"""Encode a single header line with Base64 encoding in a given charset.
|
||
|
||
Defined in RFC 2045, this Base64 encoding is identical to normal Base64
|
||
encoding, except that each line must be intelligently wrapped (respecting
|
||
the Base64 encoding), and subsequent lines must start with a space.
|
||
|
||
charset names the character set to use to encode the header. It defaults
|
||
to iso-8859-1.
|
||
|
||
End-of-line characters (\\r, \\n, \\r\\n) will be automatically converted
|
||
to the canonical email line separator \\r\\n unless the keep_eols
|
||
parameter is True (the default is False).
|
||
|
||
Each line of the header will be terminated in the value of eol, which
|
||
defaults to "\\n". Set this to "\\r\\n" if you are using the result of
|
||
this function directly in email.
|
||
|
||
The resulting string will be in the form:
|
||
|
||
"=?charset?b?WW/5ciBtYXp66XLrIHf8eiBhIGhhbXBzdGHuciBBIFlv+XIgbWF6euly?=\\n
|
||
=?charset?b?6yB3/HogYSBoYW1wc3Rh7nIgQkMgWW/5ciBtYXp66XLrIHf8eiBhIGhh?="
|
||
|
||
with each line wrapped at, at most, maxlinelen characters (defaults to 76
|
||
characters).
|
||
"""
|
||
# Return empty headers unchanged
|
||
if not header:
|
||
return header
|
||
|
||
if not keep_eols:
|
||
header = fix_eols(header)
|
||
|
||
# Base64 encode each line, in encoded chunks no greater than maxlinelen in
|
||
# length, after the RFC chrome is added in.
|
||
base64ed = []
|
||
max_encoded = maxlinelen - len(charset) - MISC_LEN
|
||
max_unencoded = max_encoded * 3 // 4
|
||
|
||
for i in range(0, len(header), max_unencoded):
|
||
base64ed.append(b2a_base64(header[i:i+max_unencoded]))
|
||
|
||
# Now add the RFC chrome to each encoded chunk
|
||
lines = []
|
||
for line in base64ed:
|
||
# Ignore the last character of each line if it is a newline
|
||
if line.endswith(NL):
|
||
line = line[:-1]
|
||
# Add the chrome
|
||
lines.append('=?%s?b?%s?=' % (charset, line))
|
||
# Glue the lines together and return it. BAW: should we be able to
|
||
# specify the leading whitespace in the joiner?
|
||
joiner = eol + ' '
|
||
return joiner.join(lines)
|
||
|
||
|
||
|
||
def encode(s, binary=True, maxlinelen=76, eol=NL):
|
||
"""Encode a string with base64.
|
||
|
||
Each line will be wrapped at, at most, maxlinelen characters (defaults to
|
||
76 characters).
|
||
|
||
If binary is False, end-of-line characters will be converted to the
|
||
canonical email end-of-line sequence \\r\\n. Otherwise they will be left
|
||
verbatim (this is the default).
|
||
|
||
Each line of encoded text will end with eol, which defaults to "\\n". Set
|
||
this to "\r\n" if you will be using the result of this function directly
|
||
in an email.
|
||
"""
|
||
if not s:
|
||
return s
|
||
|
||
if not binary:
|
||
s = fix_eols(s)
|
||
|
||
encvec = []
|
||
max_unencoded = maxlinelen * 3 // 4
|
||
for i in range(0, len(s), max_unencoded):
|
||
# BAW: should encode() inherit b2a_base64()'s dubious behavior in
|
||
# adding a newline to the encoded string?
|
||
enc = b2a_base64(s[i:i + max_unencoded])
|
||
if enc.endswith(NL) and eol <> NL:
|
||
enc = enc[:-1] + eol
|
||
encvec.append(enc)
|
||
return EMPTYSTRING.join(encvec)
|
||
|
||
|
||
# For convenience and backwards compatibility w/ standard base64 module
|
||
body_encode = encode
|
||
encodestring = encode
|
||
|
||
|
||
|
||
def decode(s, convert_eols=None):
|
||
"""Decode a raw base64 string.
|
||
|
||
If convert_eols is set to a string value, all canonical email linefeeds,
|
||
e.g. "\\r\\n", in the decoded text will be converted to the value of
|
||
convert_eols. os.linesep is a good choice for convert_eols if you are
|
||
decoding a text attachment.
|
||
|
||
This function does not parse a full MIME header value encoded with
|
||
base64 (like =?iso-8895-1?b?bmloISBuaWgh?=) -- please use the high
|
||
level email.Header class for that functionality.
|
||
"""
|
||
if not s:
|
||
return s
|
||
|
||
dec = a2b_base64(s)
|
||
if convert_eols:
|
||
return dec.replace(CRLF, convert_eols)
|
||
return dec
|
||
|
||
|
||
# For convenience and backwards compatibility w/ standard base64 module
|
||
body_decode = decode
|
||
decodestring = decode
|