The original MIME tree structure of the incoming message is as follows (using email.iterators._structure(msg)):
multipart/mixed
text/html (message)
application/octet-stream (attachment 1)
application/octet-stream (attachment 2)
Replying via GMail results in the following structure:
multipart/alternative
text/plain
text/html
I.e. they aren't being as smart as I thought, just discarding the attachments (good) and providing text and HTML versions that explicitly restructure the "quoted content."
I'm beginning to think that's all I should do too, just reply with a simple message as after discarding the attachments there's not much point in keeping the original message.
Still, might as well answer my original question since I've figured out how to now anyway.
First, replace all the attachments in the original message with text/plain placeholders:
import email
original = email.message_from_string( ... )
for part in original.walk():
if (part.get('Content-Disposition')
and part.get('Content-Disposition').startswith("attachment")):
part.set_type("text/plain")
part.set_payload("Attachment removed: %s (%s, %d bytes)"
%(part.get_filename(),
part.get_content_type(),
len(part.get_payload(decode=True))))
del part["Content-Disposition"]
del part["Content-Transfer-Encoding"]
Then create a reply message:
from email.mime.multipart import MIMEMultipart
from email.mime.text import MIMEText
from email.mime.message import MIMEMessage
new = MIMEMultipart("mixed")
body = MIMEMultipart("alternative")
body.attach( MIMEText("reply body text", "plain") )
body.attach( MIMEText("<html>reply body text</html>", "html") )
new.attach(body)
new["Message-ID"] = email.utils.make_msgid()
new["In-Reply-To"] = original["Message-ID"]
new["References"] = original["Message-ID"]
new["Subject"] = "Re: "+original["Subject"]
new["To"] = original["Reply-To"] or original["From"]
new["From"] = "[email protected]"
Then attach the original MIME message object and send:
new.attach( MIMEMessage(original) )
s = smtplib.SMTP()
s.sendmail("[email protected]", [new["To"]], new.as_string())
s.quit()
The resulting structure is:
multipart/mixed
multipart/alternative
text/plain
text/html
message/rfc822
multipart/mixed
text/html
text/plain
text/plain
Or it's a bit simpler using Django:
from django.core.mail import EmailMultiAlternatives
from email.mime.message import MIMEMessage
new = EmailMultiAlternatives("Re: "+original["Subject"],
"reply body text",
"[email protected]", # from
[original["Reply-To"] or original["From"]], # to
headers = {'Reply-To': "[email protected]",
"In-Reply-To": original["Message-ID"],
"References": original["Message-ID"]})
new.attach_alternative("<html>reply body text</html>", "text/html")
new.attach( MIMEMessage(original) ) # attach original message
new.send()
The result ends (in GMail at least) showing the original message as "---- Forwarded message ----" which isn't quite what I was after, but the general idea works and I hope this answer helps someone trying to figure out how to fiddle with MIME messages.