summaryrefslogtreecommitdiff
path: root/src/leap/mail/mail.py
diff options
context:
space:
mode:
Diffstat (limited to 'src/leap/mail/mail.py')
-rw-r--r--src/leap/mail/mail.py62
1 files changed, 44 insertions, 18 deletions
diff --git a/src/leap/mail/mail.py b/src/leap/mail/mail.py
index 47c8cba..9a32483 100644
--- a/src/leap/mail/mail.py
+++ b/src/leap/mail/mail.py
@@ -33,6 +33,7 @@ from leap.mail.adaptors.soledad import SoledadMailAdaptor
from leap.mail.constants import INBOX_NAME
from leap.mail.constants import MessageFlags
from leap.mail.mailbox_indexer import MailboxIndexer
+from leap.mail.utils import find_charset
logger = logging.getLogger(name=__name__)
@@ -58,6 +59,45 @@ def _write_and_rewind(payload):
return fd
+def _encode_payload(payload, ctype=""):
+ """
+ Properly encode an unicode payload (which can be string or unicode) as a
+ string.
+
+ :param payload: the payload to encode. currently soledad returns unicode
+ strings.
+ :type payload: basestring
+ :param ctype: optional, the content of the content-type header for this
+ payload.
+ :type ctype: str
+ :rtype: str
+ """
+ # TODO Related, it's proposed that we're able to pass
+ # the encoding to the soledad documents. Better to store the charset there?
+ # FIXME -----------------------------------------------
+ # this need a dedicated test-suite
+ charset = find_charset(ctype)
+
+ # XXX get from mail headers if not multipart!
+ # Beware also that we should pass the proper encoding to
+ # soledad when it's creating the documents.
+ # if not charset:
+ # charset = get_email_charset(payload)
+ #------------------------------------------------------
+
+ if not charset:
+ charset = "utf-8"
+
+ try:
+ if isinstance(payload, unicode):
+ payload = payload.encode(charset)
+ except UnicodeError as exc:
+ logger.error(
+ "Unicode error, using 'replace'. {0!r}".format(exc))
+ payload = payload.encode(charset, 'replace')
+ return payload
+
+
class MessagePart(object):
# TODO This class should be better abstracted from the data model.
# TODO support arbitrarily nested multiparts (right now we only support
@@ -107,7 +147,7 @@ class MessagePart(object):
# XXX uh, multi also... should recurse"
raise NotImplementedError
if payload:
- payload = self._format_payload(payload)
+ payload = _encode_payload(payload)
return _write_and_rewind(payload)
def get_headers(self):
@@ -134,23 +174,6 @@ class MessagePart(object):
return cdoc_wrapper.raw
return ""
- def _format_payload(self, payload):
- # FIXME -----------------------------------------------
- # Test against unicode payloads...
- # content_type = self._get_ctype_from_document(phash)
- # charset = find_charset(content_type)
- charset = None
- if charset is None:
- charset = get_email_charset(payload)
- try:
- if isinstance(payload, unicode):
- payload = payload.encode(charset)
- except UnicodeError as exc:
- logger.error(
- "Unicode error, using 'replace'. {0!r}".format(exc))
- payload = payload.encode(charset, 'replace')
- return payload
-
class Message(object):
"""
@@ -218,6 +241,9 @@ class Message(object):
"""
def write_and_rewind_if_found(cdoc):
payload = cdoc.raw if cdoc else ""
+ # XXX pass ctype from headers if not multipart?
+ if payload:
+ payload = _encode_payload(payload, ctype=cdoc.content_type)
return _write_and_rewind(payload)
d = defer.maybeDeferred(self._wrapper.get_body, store)