Python send email with "quoted-printable" transfer-encoding and "utf-8" content-encoding

Question

python's email.mime tends to use encoding base64 or 7bit and us-ascii. I would like to use quoted-printable and utf-8 as this is easier for humans to read and debug.

Currently, my emails look like

--===============6135350048414329636==
MIME-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: base64

IyEvYmluL2Jhc2gKCmZvciBpIGluIHs4Mjg4Li44N

or

--===============0756888342500148236==
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit

happy face =E2=98=BA

I would like the raw email to be in quoted-printable unicode so it is easier for humans to read.

--===============5610730199728027971==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"

happy face ☺

Answer 1

short answer

set content-transfer-encoding

When creating the MIMEText object, which will be attached to the MIMEMultipart object, set the content-transfer-encoding to value quoted-printable first, then do set_payload. The order of operations matters.

from email.mime.multipart import MIMEMultipart
from email.mime.text import MIMEText

# first create MIMEText, then set content-transfer-encoding, then set payload
mt = MIMEText(None, _subtype='plain')
mt.replace_header('content-transfer-encoding', 'quoted-printable')
mt.set_payload(u'happy face ☺', 'utf-8')

# create the parent email object and the MIMEMultipart extension to it
email = MIMEMultipart('mixed')
inline = MIMEMultipart('alternative')

# assemble the objects
inline.attach(mt)
email.attach(inline)

set email charset and various encodings

cs = charset.Charset('utf-8')
cs.header_encoding = charset.QP
cs.body_encoding = charset.QP
email.set_charset(cs)

Result

This creates a raw email that is human readable (except the base64 encoded file attachment)

>>> print(email)
--===============5610730199728027971==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"

happy face ☺

--===============5610730199728027971==--

--===============0985725891393820576==
Content-Type: text/x-sh
MIME-Version: 1.0
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="test.sh"

Zm9vYmFyc2RmYXNkZmtqaGFzZGZrbGhhc2ZrbGpoYXNma2xqaGFzZmtsaGZkYXNmCg==

--===============0985725891393820576==--

---

long answer

The following is a longer script to provide more context for the prior code snippets.

This script will send a text/plain section encoded in UTF-8. For fun, it will also attach a file. The raw email this produces will be human readable (except for the file attachment).

from __future__ import print_function

from email import charset    
from email.encoders import encode_base64
from email.mime.base import MIMEBase
from email.mime.multipart import MIMEMultipart
from email.mime.text import MIMEText
import mimetypes

# create the parent email object
email = MIMEMultipart('mixed')
# set email charset and email encodings
cs_ = charset.Charset('utf-8')
cs_.header_encoding = charset.QP
cs_.body_encoding = charset.QP
email.set_charset(cs_)

# create the 'text/plain' MIMEText
# first create MIMEText, then set content-transfer-encoding, then set payload
mt = MIMEText(None, _subtype='plain')
mt.replace_header('content-transfer-encoding', 'quoted-printable')
mt.set_payload(u'happy face ☺', 'utf-8')

# assemble the parts
inline = MIMEMultipart('alternative')
inline.attach(mt)
email.attach(inline)

# for fun, attach a file to the email
my_file = '/tmp/test.sh'
mimetype, encoding = mimetypes.guess_type(my_file)
mimetype = mimetype or 'application/octet-stream'
mimetype = mimetype.split('/', 1)
attachment = MIMEBase(mimetype[0], mimetype[1])
attachment.set_payload(open(my_file, 'rb').read())
encode_base64(attachment)
attachment.add_header('Content-Disposition', 'attachment', filename=os.path.basename(my_file))
email.attach(attachment)

Result

This creates a raw email that is human readable (except the base64 encoded file attachment)

>>> print(email)
--===============5610730199728027971==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"

happy face ☺

--===============5610730199728027971==--

--===============0985725891393820576==
Content-Type: text/x-sh
MIME-Version: 1.0
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="test.sh"

Zm9vYmFyc2RmYXNkZmtqaGFzZGZrbGhhc2ZrbGpoYXNma2xqaGFzZmtsaGZkYXNmCg==

--===============0985725891393820576==--

---

(bonus) send the email

Using smtplib, the email can be emailed.

import smtplib

# set email address headers
email['From'] = '[email protected]'
email['To'] = '[email protected]'
email['Subject'] = 'hello'

# send the email
smtp_srv = smtplib.SMTP('localhost')
smtp_srv.set_debuglevel(True)
print(mesg_html, end='\n\n')
print(email.as_string(), end='\n\n')
smtp_srv.sendmail('[email protected]', '[email protected]', email.as_string())
smtp_srv.quit()

Answer 2

In trying to alter the body of an existing message (email.Message object) and set its encoding to quoted-printable, I found that this problem took way more effort than I had anticipated.

import email
#... 'part' is the Message object
content = part.get_payload(decode=True)
#... Modify content
part['Content-Transfer-Encoding'] = '8bit'
part.set_payload(content, 'UTF-8')
del part['Content-Transfer-Encoding']
email.encoders.encode_quopri(part)

Now, why do I set and then delete the Content-Transfer-Encoding header? The set_payload call will set the Content-Transfer-Encoding header and encode the data (to Base64) if no header exists. Otherwise, the set_payload call will assume the caller has already encoded the data and will not alter it (by encoding). So, it actually doesn't matter the value to which I set the Content-Transfer-Encoding header, only that I don't leave it blank.

But then why do I need to delete the header? The email.encoders.encode_quopri call will only add a header, so the message will result with multiple Content-Transfer-Encoding headers.

So, just using set_payload then encode_quopri for a message with no Content-Transfer-Encoding header will result in a quoted-printable representation of a Base64 string, and for a message with an existing Content-Transfer-Encoding header will result in a message with duplicate headers. Using encode_quopri then set_payload may result in duplicate headers, but will not encode the message. Hence the add/delete rigamarole to avoid dipping into the quopri module.