def test_does_not_split_links(self):
# token that causes html2text to make a line break if wrap_links option is not set
token = '4522e1fa-9827-44ec-bf76-f7fa8fb132ff'
url = 'http://localhost:8080/#/signup?invite={}&email=please%40join.com'.format(
token)
html = '<a href="{}">some link</a>'.format(url)
text = generate_plaintext_from_html(html)
self.assertIn(url, text)
def inbound_received(sender, event, esp_name, **kwargs):
incoming_message = event.message
# check local part of reply-to and extract conversation and user (fail if they don't exist)
local_part = incoming_message.to[0].username
try:
conversation_id, user_id, thread_id = parse_local_part(local_part)
except (UnicodeDecodeError, binascii.Error):
sentry_client.captureException()
return
user = get_user_model().objects.get(id=user_id)
thread = None
if thread_id is not None:
thread = ConversationMessage.objects.get(id=thread_id)
conversation = thread.conversation
else:
conversation = Conversation.objects.get(id=conversation_id)
# get email content as plain text
if incoming_message.text is not None:
text_content = incoming_message.text
elif incoming_message.html is not None:
# let's just make HTML into plain text
html_content = incoming_message.html
text_content = generate_plaintext_from_html(html_content)
else:
# Inform the user if we couldn't find any content
notify_about_rejected_email(user,
'Karrot could not find any reply text')
return
# extract email reply text
# Try out both talon and discourse email_reply_trimmer
# Trimmers are conservative and sometimes keep more lines, leading to bloated replies.
# We choose the trimmed reply that has fewer lines.
trimmed_talon, line_count_talon = trim_with_talon(text_content)
trimmed_discourse, line_count_discourse = trim_with_discourse(text_content)
reply_plain = trimmed_discourse if line_count_discourse <= line_count_talon else trimmed_talon
stats.incoming_email_trimmed({
'line_count_original':
len(text_content.splitlines()),
'line_count_talon':
line_count_talon,
'line_count_discourse':
line_count_discourse,
'from_html':
1 if incoming_message.text is None else 0
})
# add reply to conversation
if conversation.is_closed:
notify_about_rejected_email(user, reply_plain)
return
if not conversation.participants.filter(id=user.id).exists():
notify_about_rejected_email(user, reply_plain)
return
created_message = ConversationMessage.objects.create(
author=user,
conversation=conversation,
thread=thread,
content=reply_plain,
received_via='email',
)
incoming_message_serialized = dict(incoming_message)
incoming_message_serialized['text'] = incoming_message.text
incoming_message_serialized['html'] = incoming_message.html
IncomingEmail.objects.create(
user=user,
message=created_message,
payload=incoming_message_serialized,
version=2,
)
def post(self, request):
"""
Receive conversation replies via e-mail
Request payload spec: https://developers.sparkpost.com/api/relay-webhooks/#header-relay-webhook-payload
"""
auth_key = request.META.get('HTTP_X_MESSAGESYSTEMS_WEBHOOK_TOKEN')
if auth_key is None or auth_key != settings.SPARKPOST_RELAY_SECRET:
return Response(
status=status.HTTP_403_FORBIDDEN,
data={
'message':
'Invalid HTTP_X_MESSAGESYSTEMS_WEBHOOK_TOKEN header'
})
for messages in [e['msys'].values() for e in request.data]:
for incoming_message in messages:
# 1. get email content and reply-to
reply_to = parseaddr(incoming_message['rcpt_to'])[1]
content = incoming_message['content']
# 2. check local part of reply-to and extract conversation and user (fail if they don't exist)
local_part = reply_to.split('@')[0]
try:
conversation_id, user_id, thread_id = parse_local_part(
local_part)
except (UnicodeDecodeError, binascii.Error):
sentry_client.captureException()
continue
user = get_user_model().objects.get(id=user_id)
thread = None
if thread_id is not None:
thread = ConversationMessage.objects.get(id=thread_id)
conversation = thread.conversation
else:
conversation = Conversation.objects.get(id=conversation_id)
# 3. extract the email reply text
# Try plain text first, most emails have that.
if 'text' in content:
# Try out both talon and discourse email_reply_trimmer
# Trimmers are conservative and sometimes keep more lines, leading to bloated replies.
# We choose the trimmed reply that has fewer lines.
text_content = content['text']
trimmed_talon, line_count_talon = trim_with_talon(
text_content)
trimmed_discourse, line_count_discourse = trim_with_discourse(
text_content)
reply_plain = trimmed_discourse if line_count_discourse <= line_count_talon else trimmed_talon
stats.incoming_email_trimmed({
'line_count_original':
len(text_content.splitlines()),
'line_count_talon':
line_count_talon,
'line_count_discourse':
line_count_discourse,
})
else:
# Fall back to html
try:
html_content = content['html']
except KeyError:
# Inform the user if we couldn't find any content
sentry_client.captureException()
notify_about_rejected_email(
user, 'Karrot could not find any reply text')
continue
reply_html = trim_html_with_talon(html_content)
reply_plain = generate_plaintext_from_html(reply_html)
stats.incoming_html_email_trimmed({
'length_original':
len(html_content.splitlines()),
'length_html_talon':
len(reply_html),
'length_plain_talon':
len(reply_plain),
})
# 4. add reply to conversation
if conversation.is_closed:
notify_about_rejected_email(user, reply_plain)
continue
if not conversation.participants.filter(id=user.id).exists():
notify_about_rejected_email(user, reply_plain)
continue
created_message = ConversationMessage.objects.create(
author=user,
conversation=conversation,
thread=thread,
content=reply_plain,
received_via='email',
)
IncomingEmail.objects.create(
user=user,
message=created_message,
payload=incoming_message,
)
return Response(status=status.HTTP_200_OK, data={})