mirror of
https://github.com/python/cpython.git
synced 2024-11-23 18:04:37 +08:00
49fd7fa443
number of tests, all because of the codecs/_multibytecodecs issue described here (it's not a Py3K issue, just something Py3K discovers): http://mail.python.org/pipermail/python-dev/2006-April/064051.html Hye-Shik Chang promised to look for a fix, so no need to fix it here. The tests that are expected to break are: test_codecencodings_cn test_codecencodings_hk test_codecencodings_jp test_codecencodings_kr test_codecencodings_tw test_codecs test_multibytecodec This merge fixes an actual test failure (test_weakref) in this branch, though, so I believe merging is the right thing to do anyway.
242 lines
6.7 KiB
Python
242 lines
6.7 KiB
Python
"""Various tools used by MIME-reading or MIME-writing programs."""
|
|
|
|
|
|
import os
|
|
import rfc822
|
|
import tempfile
|
|
|
|
__all__ = ["Message","choose_boundary","encode","decode","copyliteral",
|
|
"copybinary"]
|
|
|
|
class Message(rfc822.Message):
|
|
"""A derived class of rfc822.Message that knows about MIME headers and
|
|
contains some hooks for decoding encoded and multipart messages."""
|
|
|
|
def __init__(self, fp, seekable = 1):
|
|
rfc822.Message.__init__(self, fp, seekable)
|
|
self.encodingheader = \
|
|
self.getheader('content-transfer-encoding')
|
|
self.typeheader = \
|
|
self.getheader('content-type')
|
|
self.parsetype()
|
|
self.parseplist()
|
|
|
|
def parsetype(self):
|
|
str = self.typeheader
|
|
if str is None:
|
|
str = 'text/plain'
|
|
if ';' in str:
|
|
i = str.index(';')
|
|
self.plisttext = str[i:]
|
|
str = str[:i]
|
|
else:
|
|
self.plisttext = ''
|
|
fields = str.split('/')
|
|
for i in range(len(fields)):
|
|
fields[i] = fields[i].strip().lower()
|
|
self.type = '/'.join(fields)
|
|
self.maintype = fields[0]
|
|
self.subtype = '/'.join(fields[1:])
|
|
|
|
def parseplist(self):
|
|
str = self.plisttext
|
|
self.plist = []
|
|
while str[:1] == ';':
|
|
str = str[1:]
|
|
if ';' in str:
|
|
# XXX Should parse quotes!
|
|
end = str.index(';')
|
|
else:
|
|
end = len(str)
|
|
f = str[:end]
|
|
if '=' in f:
|
|
i = f.index('=')
|
|
f = f[:i].strip().lower() + \
|
|
'=' + f[i+1:].strip()
|
|
self.plist.append(f.strip())
|
|
str = str[end:]
|
|
|
|
def getplist(self):
|
|
return self.plist
|
|
|
|
def getparam(self, name):
|
|
name = name.lower() + '='
|
|
n = len(name)
|
|
for p in self.plist:
|
|
if p[:n] == name:
|
|
return rfc822.unquote(p[n:])
|
|
return None
|
|
|
|
def getparamnames(self):
|
|
result = []
|
|
for p in self.plist:
|
|
i = p.find('=')
|
|
if i >= 0:
|
|
result.append(p[:i].lower())
|
|
return result
|
|
|
|
def getencoding(self):
|
|
if self.encodingheader is None:
|
|
return '7bit'
|
|
return self.encodingheader.lower()
|
|
|
|
def gettype(self):
|
|
return self.type
|
|
|
|
def getmaintype(self):
|
|
return self.maintype
|
|
|
|
def getsubtype(self):
|
|
return self.subtype
|
|
|
|
|
|
|
|
|
|
# Utility functions
|
|
# -----------------
|
|
|
|
try:
|
|
import thread
|
|
except ImportError:
|
|
import dummy_thread as thread
|
|
_counter_lock = thread.allocate_lock()
|
|
del thread
|
|
|
|
_counter = 0
|
|
def _get_next_counter():
|
|
global _counter
|
|
_counter_lock.acquire()
|
|
_counter += 1
|
|
result = _counter
|
|
_counter_lock.release()
|
|
return result
|
|
|
|
_prefix = None
|
|
|
|
def choose_boundary():
|
|
"""Return a string usable as a multipart boundary.
|
|
|
|
The string chosen is unique within a single program run, and
|
|
incorporates the user id (if available), process id (if available),
|
|
and current time. So it's very unlikely the returned string appears
|
|
in message text, but there's no guarantee.
|
|
|
|
The boundary contains dots so you have to quote it in the header."""
|
|
|
|
global _prefix
|
|
import time
|
|
if _prefix is None:
|
|
import socket
|
|
try:
|
|
hostid = socket.gethostbyname(socket.gethostname())
|
|
except socket.gaierror:
|
|
hostid = '127.0.0.1'
|
|
try:
|
|
uid = repr(os.getuid())
|
|
except AttributeError:
|
|
uid = '1'
|
|
try:
|
|
pid = repr(os.getpid())
|
|
except AttributeError:
|
|
pid = '1'
|
|
_prefix = hostid + '.' + uid + '.' + pid
|
|
return "%s.%.3f.%d" % (_prefix, time.time(), _get_next_counter())
|
|
|
|
|
|
# Subroutines for decoding some common content-transfer-types
|
|
|
|
def decode(input, output, encoding):
|
|
"""Decode common content-transfer-encodings (base64, quopri, uuencode)."""
|
|
if encoding == 'base64':
|
|
import base64
|
|
return base64.decode(input, output)
|
|
if encoding == 'quoted-printable':
|
|
import quopri
|
|
return quopri.decode(input, output)
|
|
if encoding in ('uuencode', 'x-uuencode', 'uue', 'x-uue'):
|
|
import uu
|
|
return uu.decode(input, output)
|
|
if encoding in ('7bit', '8bit'):
|
|
return output.write(input.read())
|
|
if encoding in decodetab:
|
|
pipethrough(input, decodetab[encoding], output)
|
|
else:
|
|
raise ValueError, \
|
|
'unknown Content-Transfer-Encoding: %s' % encoding
|
|
|
|
def encode(input, output, encoding):
|
|
"""Encode common content-transfer-encodings (base64, quopri, uuencode)."""
|
|
if encoding == 'base64':
|
|
import base64
|
|
return base64.encode(input, output)
|
|
if encoding == 'quoted-printable':
|
|
import quopri
|
|
return quopri.encode(input, output, 0)
|
|
if encoding in ('uuencode', 'x-uuencode', 'uue', 'x-uue'):
|
|
import uu
|
|
return uu.encode(input, output)
|
|
if encoding in ('7bit', '8bit'):
|
|
return output.write(input.read())
|
|
if encoding in encodetab:
|
|
pipethrough(input, encodetab[encoding], output)
|
|
else:
|
|
raise ValueError, \
|
|
'unknown Content-Transfer-Encoding: %s' % encoding
|
|
|
|
# The following is no longer used for standard encodings
|
|
|
|
# XXX This requires that uudecode and mmencode are in $PATH
|
|
|
|
uudecode_pipe = '''(
|
|
TEMP=/tmp/@uu.$$
|
|
sed "s%^begin [0-7][0-7]* .*%begin 600 $TEMP%" | uudecode
|
|
cat $TEMP
|
|
rm $TEMP
|
|
)'''
|
|
|
|
decodetab = {
|
|
'uuencode': uudecode_pipe,
|
|
'x-uuencode': uudecode_pipe,
|
|
'uue': uudecode_pipe,
|
|
'x-uue': uudecode_pipe,
|
|
'quoted-printable': 'mmencode -u -q',
|
|
'base64': 'mmencode -u -b',
|
|
}
|
|
|
|
encodetab = {
|
|
'x-uuencode': 'uuencode tempfile',
|
|
'uuencode': 'uuencode tempfile',
|
|
'x-uue': 'uuencode tempfile',
|
|
'uue': 'uuencode tempfile',
|
|
'quoted-printable': 'mmencode -q',
|
|
'base64': 'mmencode -b',
|
|
}
|
|
|
|
def pipeto(input, command):
|
|
pipe = os.popen(command, 'w')
|
|
copyliteral(input, pipe)
|
|
pipe.close()
|
|
|
|
def pipethrough(input, command, output):
|
|
(fd, tempname) = tempfile.mkstemp()
|
|
temp = os.fdopen(fd, 'w')
|
|
copyliteral(input, temp)
|
|
temp.close()
|
|
pipe = os.popen(command + ' <' + tempname, 'r')
|
|
copybinary(pipe, output)
|
|
pipe.close()
|
|
os.unlink(tempname)
|
|
|
|
def copyliteral(input, output):
|
|
while 1:
|
|
line = input.readline()
|
|
if not line: break
|
|
output.write(line)
|
|
|
|
def copybinary(input, output):
|
|
BUFSIZE = 8192
|
|
while 1:
|
|
line = input.read(BUFSIZE)
|
|
if not line: break
|
|
output.write(line)
|