def construct_message( self, scheduled_message: ScheduledMessage) -> Dict[str, Any]: message = Message() original_sender = scheduled_message.sender message.content = scheduled_message.content message.recipient = scheduled_message.recipient message.subject = scheduled_message.subject message.pub_date = timezone_now() message.sending_client = scheduled_message.sending_client delivery_type = scheduled_message.delivery_type if delivery_type == ScheduledMessage.SEND_LATER: message.sender = original_sender elif delivery_type == ScheduledMessage.REMIND: message.sender = get_user(settings.REMINDER_BOT, original_sender.realm) whos_reminding = ('%s asked me to do a reminder about:\n' % (original_sender.full_name)) message.content = whos_reminding + message.content return { 'message': message, 'stream': scheduled_message.stream, 'realm': scheduled_message.realm }
def render_message_backend(request, user_profile, content=REQ()): # type: (HttpRequest, UserProfile, Text) -> HttpResponse message = Message() message.sender = user_profile message.content = content message.sending_client = request.client rendered_content = render_markdown(message, content, realm=user_profile.realm) return json_success({"rendered": rendered_content})
def render_message_backend( request: HttpRequest, user_profile: UserProfile, content: str = REQ()) -> HttpResponse: message = Message() message.sender = user_profile message.content = content message.sending_client = request.client rendering_result = render_markdown(message, content, realm=user_profile.realm) return json_success({"rendered": rendering_result.rendered_content})
def render_message_backend( request: HttpRequest, user_profile: UserProfile, content: str = REQ() ) -> HttpResponse: message = Message() message.sender = user_profile message.content = content client = RequestNotes.get_notes(request).client assert client is not None message.sending_client = client rendering_result = render_markdown(message, content, realm=user_profile.realm) return json_success(request, data={"rendered": rendering_result.rendered_content})
def construct_message( self, scheduled_message: ScheduledMessage) -> Dict[str, Any]: message = Message() message.sender = scheduled_message.sender message.content = scheduled_message.content message.recipient = scheduled_message.recipient message.subject = scheduled_message.subject message.pub_date = timezone_now() message.sending_client = scheduled_message.sending_client return { 'message': message, 'stream': scheduled_message.stream, 'realm': scheduled_message.realm }
def construct_message(self, scheduled_message: ScheduledMessage) -> Dict[str, Any]: message = Message() original_sender = scheduled_message.sender message.content = scheduled_message.content message.recipient = scheduled_message.recipient message.subject = scheduled_message.subject message.pub_date = timezone_now() message.sending_client = scheduled_message.sending_client delivery_type = scheduled_message.delivery_type if delivery_type == ScheduledMessage.SEND_LATER: message.sender = original_sender elif delivery_type == ScheduledMessage.REMIND: message.sender = get_user(settings.REMINDER_BOT, original_sender.realm) return {'message': message, 'stream': scheduled_message.stream, 'realm': scheduled_message.realm}
def construct_message( self, scheduled_message: ScheduledMessage) -> SendMessageRequest: message = Message() original_sender = scheduled_message.sender message.content = scheduled_message.content message.recipient = scheduled_message.recipient message.subject = scheduled_message.subject message.date_sent = timezone_now() message.sending_client = scheduled_message.sending_client delivery_type = scheduled_message.delivery_type if delivery_type == ScheduledMessage.SEND_LATER: message.sender = original_sender elif delivery_type == ScheduledMessage.REMIND: message.sender = get_user_by_delivery_email( settings.NOTIFICATION_BOT, original_sender.realm) return build_message_send_dict(message=message, stream=scheduled_message.stream, realm=scheduled_message.realm)
def construct_message( self, scheduled_message: ScheduledMessage) -> Dict[str, Any]: message = Message() original_sender = scheduled_message.sender message.content = scheduled_message.content message.recipient = scheduled_message.recipient message.subject = scheduled_message.subject message.date_sent = timezone_now() message.sending_client = scheduled_message.sending_client delivery_type = scheduled_message.delivery_type if delivery_type == ScheduledMessage.SEND_LATER: message.sender = original_sender elif delivery_type == ScheduledMessage.REMIND: message.sender = get_user_by_delivery_email( settings.NOTIFICATION_BOT, original_sender.realm) return { 'message': message, 'stream': scheduled_message.stream, 'realm': scheduled_message.realm }
def generate_and_send_messages( data: Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int] ) -> int: (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open( os.path.join(get_or_create_dev_uuid_var_path("test-backend"), "test_messages.json"), "rb") as infile: dialog = orjson.loads(infile.read()) random.shuffle(dialog) texts = itertools.cycle(dialog) # We need to filter out streams from the analytics realm as we don't want to generate # messages to its streams - and they might also have no subscribers, which would break # our message generation mechanism below. stream_ids = Stream.objects.filter(realm=get_realm("zulip")).values_list( "id", flat=True) recipient_streams: List[int] = [ recipient.id for recipient in Recipient.objects.filter(type=Recipient.STREAM, type_id__in=stream_ids) ] recipient_huddles: List[int] = [ h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE) ] huddle_members: Dict[int, List[int]] = {} for h in recipient_huddles: huddle_members[h] = [ s.user_profile.id for s in Subscription.objects.filter(recipient_id=h) ] # Generate different topics for each stream possible_topics = {} for stream_id in recipient_streams: possible_topics[stream_id] = generate_topics(options["max_topics"]) message_batch_size = options["batch_size"] num_messages = 0 random_max = 1000000 recipients: Dict[int, Tuple[int, int, Dict[str, Any]]] = {} messages: List[Message] = [] while num_messages < tot_messages: saved_data: Dict[str, Any] = {} message = Message() message.sending_client = get_client("populate_db") message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100.0 / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data["personals_pair"] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data["subject"] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif randkey <= random_max * options["percent_huddles"] / 100.0: message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id( random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.0): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif randkey <= random_max * 1.0: message_type = Recipient.STREAM message.recipient = get_recipient_by_id( random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get( type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data["personals_pair"] = personals_pair elif message_type == Recipient.STREAM: # Pick a random subscriber to the stream message.sender = random.choice( Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = random.choice( possible_topics[message.recipient.id]) saved_data["subject"] = message.subject message.date_sent = choose_date_sent(num_messages, tot_messages, options["threads"]) messages.append(message) recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 if (num_messages % message_batch_size) == 0: # Send the batch and empty the list: send_messages(messages) messages = [] if len(messages) > 0: # If there are unsent messages after exiting the loop, send them: send_messages(messages) return tot_messages
def send_messages(data): # type: (Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int]) -> int (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open("var/test_messages.json", "r") as infile: dialog = ujson.load(infile) random.shuffle(dialog) texts = itertools.cycle(dialog) recipient_streams = [ klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM) ] # type: List[int] recipient_huddles = [ h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE) ] # type: List[int] huddle_members = {} # type: Dict[int, List[int]] for h in recipient_huddles: huddle_members[h] = [ s.user_profile.id for s in Subscription.objects.filter(recipient_id=h) ] num_messages = 0 random_max = 1000000 recipients = {} # type: Dict[int, Tuple[int, int, Dict[str, Any]]] while num_messages < tot_messages: saved_data = {} # type: Dict[str, Any] message = Message() message.sending_client = get_client('populate_db') message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id( random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id( random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get( type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: stream = Stream.objects.get(id=message.recipient.type_id) # Pick a random subscriber to the stream message.sender = random.choice( Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = stream.name + Text(random.randint(1, 3)) saved_data['subject'] = message.subject message.pub_date = timezone_now() do_send_messages([{'message': message}]) recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 return tot_messages
def send_messages(data): # type: (Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any]]) -> int (tot_messages, personals_pairs, options, output) = data random.seed(os.getpid()) texts = open("zilencer/management/commands/test_messages.txt", "r").readlines() offset = random.randint(0, len(texts)) recipient_streams = [klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM)] # type: List[int] recipient_huddles = [h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE)] # type: List[int] huddle_members = {} # type: Dict[int, List[int]] for h in recipient_huddles: huddle_members[h] = [s.user_profile.id for s in Subscription.objects.filter(recipient_id=h)] num_messages = 0 random_max = 1000000 recipients = {} # type: Dict[int, Tuple[int, int, Dict[str, Any]]] while num_messages < tot_messages: saved_data = {} # type: Dict[str, Any] message = Message() message.sending_client = get_client('populate_db') length = random.randint(1, 5) lines = (t.strip() for t in texts[offset: offset + length]) message.content = '\n'.join(lines) offset += length offset = offset % len(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id(random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id(random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get(type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: stream = Stream.objects.get(id=message.recipient.type_id) # Pick a random subscriber to the stream message.sender = random.choice(Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = stream.name + text_type(random.randint(1, 3)) saved_data['subject'] = message.subject message.pub_date = now() do_send_message(message) recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 return tot_messages
def send_messages(data: Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int]) -> int: (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open("var/test_messages.json", "r") as infile: dialog = ujson.load(infile) random.shuffle(dialog) texts = itertools.cycle(dialog) recipient_streams = [klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM)] # type: List[int] recipient_huddles = [h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE)] # type: List[int] huddle_members = {} # type: Dict[int, List[int]] for h in recipient_huddles: huddle_members[h] = [s.user_profile.id for s in Subscription.objects.filter(recipient_id=h)] num_messages = 0 random_max = 1000000 recipients = {} # type: Dict[int, Tuple[int, int, Dict[str, Any]]] while num_messages < tot_messages: saved_data = {} # type: Dict[str, Any] message = Message() message.sending_client = get_client('populate_db') message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id(random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id(random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get(type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: stream = Stream.objects.get(id=message.recipient.type_id) # Pick a random subscriber to the stream message.sender = random.choice(Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = stream.name + str(random.randint(1, 3)) saved_data['subject'] = message.subject # Spoofing time not supported with threading if options['threads'] != 1: message.pub_date = timezone_now() else: # Distrubutes 80% of messages starting from 5 days ago, over a period # of 3 days. Then, distributes remaining messages over past 24 hours. spoofed_date = timezone_now() - timezone_timedelta(days = 5) if (num_messages < tot_messages * 0.8): # Maximum of 3 days ahead, convert to minutes time_ahead = 3 * 24 * 60 time_ahead //= int(tot_messages * 0.8) else: time_ahead = 24 * 60 time_ahead //= int(tot_messages * 0.2) spoofed_minute = random.randint(time_ahead * num_messages, time_ahead * (num_messages + 1)) spoofed_date += timezone_timedelta(minutes = spoofed_minute) message.pub_date = spoofed_date # We disable USING_RABBITMQ here, so that deferred work is # executed in do_send_message_messages, rather than being # queued. This is important, because otherwise, if run-dev.py # wasn't running when populate_db was run, a developer can end # up with queued events that reference objects from a previous # life of the database, which naturally throws exceptions. settings.USING_RABBITMQ = False do_send_messages([{'message': message}]) settings.USING_RABBITMQ = True recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 return tot_messages
def generate_and_send_messages(data: Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int]) -> int: (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open(os.path.join(get_or_create_dev_uuid_var_path('test-backend'), "test_messages.json")) as infile: dialog = ujson.load(infile) random.shuffle(dialog) texts = itertools.cycle(dialog) recipient_streams: List[int] = [ klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM) ] recipient_huddles: List[int] = [h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE)] huddle_members: Dict[int, List[int]] = {} for h in recipient_huddles: huddle_members[h] = [s.user_profile.id for s in Subscription.objects.filter(recipient_id=h)] # Generate different topics for each stream possible_topics = dict() for stream_id in recipient_streams: possible_topics[stream_id] = generate_topics(options["max_topics"]) message_batch_size = options['batch_size'] num_messages = 0 random_max = 1000000 recipients: Dict[int, Tuple[int, int, Dict[str, Any]]] = {} messages = [] messages_add_reaction = [] while num_messages < tot_messages: saved_data: Dict[str, Any] = {} message = Message() message.sending_client = get_client('populate_db') message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id(random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id(random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get(type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: # Pick a random subscriber to the stream message.sender = random.choice(Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = random.choice(possible_topics[message.recipient.id]) saved_data['subject'] = message.subject message.date_sent = choose_date_sent(num_messages, tot_messages, options['threads']) messages.append(message) messages_add_reaction.append(message) recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 if (num_messages % message_batch_size) == 0: # Send the batch and empty the list: send_messages(messages) messages = [] if len(messages) > 0: # If there are unsent messages after exiting the loop, send them: send_messages(messages) reactions_message = [] add_emojis = ["1f44d", "1f642", "1f60a"] for rmessage in messages_add_reaction: if random.random() > 0.9: reactedmessage = Reaction(user_profile=rmessage.sender, message=rmessage, emoji_name="+1", emoji_code=add_emojis[random.randint(0, 2)], reaction_type="unicode_emoji") reactions_message.append(reactedmessage) Reaction.objects.bulk_create(reactions_message) return tot_messages
def send_messages(data): # type: (Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int]) -> int (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open("var/test_messages.json", "r") as infile: dialog = ujson.load(infile) random.shuffle(dialog) texts = itertools.cycle(dialog) recipient_streams = [klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM)] # type: List[int] recipient_huddles = [h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE)] # type: List[int] huddle_members = {} # type: Dict[int, List[int]] for h in recipient_huddles: huddle_members[h] = [s.user_profile.id for s in Subscription.objects.filter(recipient_id=h)] num_messages = 0 random_max = 1000000 recipients = {} # type: Dict[int, Tuple[int, int, Dict[str, Any]]] while num_messages < tot_messages: saved_data = {} # type: Dict[str, Any] message = Message() message.sending_client = get_client('populate_db') message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id(random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id(random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get(type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: stream = Stream.objects.get(id=message.recipient.type_id) # Pick a random subscriber to the stream message.sender = random.choice(Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = stream.name + Text(random.randint(1, 3)) saved_data['subject'] = message.subject message.pub_date = timezone_now() do_send_messages([{'message': message}]) recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 return tot_messages
def send_messages( data: Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int] ) -> int: (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open("var/test_messages.json", "r") as infile: dialog = ujson.load(infile) random.shuffle(dialog) texts = itertools.cycle(dialog) recipient_streams = [ klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM) ] # type: List[int] recipient_huddles = [ h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE) ] # type: List[int] huddle_members = {} # type: Dict[int, List[int]] for h in recipient_huddles: huddle_members[h] = [ s.user_profile.id for s in Subscription.objects.filter(recipient_id=h) ] num_messages = 0 random_max = 1000000 recipients = {} # type: Dict[int, Tuple[int, int, Dict[str, Any]]] while num_messages < tot_messages: saved_data = {} # type: Dict[str, Any] message = Message() message.sending_client = get_client('populate_db') message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id( random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id( random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get( type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: stream = Stream.objects.get(id=message.recipient.type_id) # Pick a random subscriber to the stream message.sender = random.choice( Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = stream.name + str(random.randint(1, 3)) saved_data['subject'] = message.subject # Spoofing time not supported with threading if options['threads'] != 1: message.pub_date = timezone_now() else: # Distrubutes 80% of messages starting from 5 days ago, over a period # of 3 days. Then, distributes remaining messages over past 24 hours. spoofed_date = timezone_now() - timezone_timedelta(days=5) if (num_messages < tot_messages * 0.8): # Maximum of 3 days ahead, convert to minutes time_ahead = 3 * 24 * 60 time_ahead //= int(tot_messages * 0.8) else: time_ahead = 24 * 60 time_ahead //= int(tot_messages * 0.2) spoofed_minute = random.randint(time_ahead * num_messages, time_ahead * (num_messages + 1)) spoofed_date += timezone_timedelta(minutes=spoofed_minute) message.pub_date = spoofed_date # We disable USING_RABBITMQ here, so that deferred work is # executed in do_send_message_messages, rather than being # queued. This is important, because otherwise, if run-dev.py # wasn't running when populate_db was run, a developer can end # up with queued events that reference objects from a previous # life of the database, which naturally throws exceptions. settings.USING_RABBITMQ = False do_send_messages([{'message': message}]) settings.USING_RABBITMQ = True recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 return tot_messages
def do_update_message( user_profile: UserProfile, target_message: Message, new_stream: Optional[Stream], topic_name: Optional[str], propagate_mode: Optional[str], send_notification_to_old_thread: bool, send_notification_to_new_thread: bool, content: Optional[str], rendering_result: Optional[MessageRenderingResult], prior_mention_user_ids: Set[int], mention_data: Optional[MentionData] = None, ) -> int: """ The main function for message editing. A message edit event can modify: * the message's content (in which case the caller will have set both content and rendered_content), * the topic, in which case the caller will have set topic_name * or both message's content and the topic * or stream and/or topic, in which case the caller will have set new_stream and/or topic_name. With topic edits, propagate_mode determines whether other message also have their topics edited. """ timestamp = timezone_now() target_message.last_edit_time = timestamp event: Dict[str, Any] = { "type": "update_message", "user_id": user_profile.id, "edit_timestamp": datetime_to_timestamp(timestamp), "message_id": target_message.id, "rendering_only": False, } edit_history_event: EditHistoryEvent = { "user_id": user_profile.id, "timestamp": event["edit_timestamp"], } changed_messages = [target_message] realm = user_profile.realm stream_being_edited = None if target_message.is_stream_message(): stream_id = target_message.recipient.type_id stream_being_edited = get_stream_by_id_in_realm(stream_id, realm) event["stream_name"] = stream_being_edited.name event["stream_id"] = stream_being_edited.id ums = UserMessage.objects.filter(message=target_message.id) if content is not None: assert rendering_result is not None # mention_data is required if there's a content edit. assert mention_data is not None # add data from group mentions to mentions_user_ids. for group_id in rendering_result.mentions_user_group_ids: members = mention_data.get_group_members(group_id) rendering_result.mentions_user_ids.update(members) update_user_message_flags(rendering_result, ums) # One could imagine checking realm.allow_edit_history here and # modifying the events based on that setting, but doing so # doesn't really make sense. We need to send the edit event # to clients regardless, and a client already had access to # the original/pre-edit content of the message anyway. That # setting must be enforced on the client side, and making a # change here simply complicates the logic for clients parsing # edit history events. event["orig_content"] = target_message.content event["orig_rendered_content"] = target_message.rendered_content edit_history_event["prev_content"] = target_message.content edit_history_event[ "prev_rendered_content"] = target_message.rendered_content edit_history_event[ "prev_rendered_content_version"] = target_message.rendered_content_version target_message.content = content target_message.rendered_content = rendering_result.rendered_content target_message.rendered_content_version = markdown_version event["content"] = content event["rendered_content"] = rendering_result.rendered_content event[ "prev_rendered_content_version"] = target_message.rendered_content_version event["is_me_message"] = Message.is_status_message( content, rendering_result.rendered_content) # target_message.has_image and target_message.has_link will have been # already updated by Markdown rendering in the caller. target_message.has_attachment = check_attachment_reference_change( target_message, rendering_result) if target_message.is_stream_message(): if topic_name is not None: new_topic_name = topic_name else: new_topic_name = target_message.topic_name() stream_topic: Optional[StreamTopicTarget] = StreamTopicTarget( stream_id=stream_id, topic_name=new_topic_name, ) else: stream_topic = None info = get_recipient_info( realm_id=realm.id, recipient=target_message.recipient, sender_id=target_message.sender_id, stream_topic=stream_topic, possible_wildcard_mention=mention_data.message_has_wildcards(), ) event["online_push_user_ids"] = list(info["online_push_user_ids"]) event["pm_mention_push_disabled_user_ids"] = list( info["pm_mention_push_disabled_user_ids"]) event["pm_mention_email_disabled_user_ids"] = list( info["pm_mention_email_disabled_user_ids"]) event["stream_push_user_ids"] = list(info["stream_push_user_ids"]) event["stream_email_user_ids"] = list(info["stream_email_user_ids"]) event["muted_sender_user_ids"] = list(info["muted_sender_user_ids"]) event["prior_mention_user_ids"] = list(prior_mention_user_ids) event["presence_idle_user_ids"] = filter_presence_idle_user_ids( info["active_user_ids"]) event["all_bot_user_ids"] = list(info["all_bot_user_ids"]) if rendering_result.mentions_wildcard: event["wildcard_mention_user_ids"] = list( info["wildcard_mention_user_ids"]) else: event["wildcard_mention_user_ids"] = [] do_update_mobile_push_notification( target_message, prior_mention_user_ids, rendering_result.mentions_user_ids, info["stream_push_user_ids"], ) if topic_name is not None or new_stream is not None: assert propagate_mode is not None orig_topic_name = target_message.topic_name() event["propagate_mode"] = propagate_mode if new_stream is not None: assert content is None assert target_message.is_stream_message() assert stream_being_edited is not None edit_history_event["prev_stream"] = stream_being_edited.id edit_history_event["stream"] = new_stream.id event[ORIG_TOPIC] = orig_topic_name assert new_stream.recipient_id is not None target_message.recipient_id = new_stream.recipient_id event["new_stream_id"] = new_stream.id event["propagate_mode"] = propagate_mode # When messages are moved from one stream to another, some # users may lose access to those messages, including guest # users and users not subscribed to the new stream (if it is a # private stream). For those users, their experience is as # though the messages were deleted, and we should send a # delete_message event to them instead. subs_to_old_stream = get_active_subscriptions_for_stream_id( stream_id, include_deactivated_users=True).select_related("user_profile") subs_to_new_stream = list( get_active_subscriptions_for_stream_id( new_stream.id, include_deactivated_users=True).select_related("user_profile")) old_stream_sub_ids = [ user.user_profile_id for user in subs_to_old_stream ] new_stream_sub_ids = [ user.user_profile_id for user in subs_to_new_stream ] # Get users who aren't subscribed to the new_stream. subs_losing_usermessages = [ sub for sub in subs_to_old_stream if sub.user_profile_id not in new_stream_sub_ids ] # Users who can longer access the message without some action # from administrators. subs_losing_access = [ sub for sub in subs_losing_usermessages if sub.user_profile.is_guest or not new_stream.is_public() ] ums = ums.exclude(user_profile_id__in=[ sub.user_profile_id for sub in subs_losing_usermessages ]) subs_gaining_usermessages = [] if not new_stream.is_history_public_to_subscribers(): # For private streams, with history not public to subscribers, # We find out users who are not present in the msgs' old stream # and create new UserMessage for these users so that they can # access this message. subs_gaining_usermessages += [ user_id for user_id in new_stream_sub_ids if user_id not in old_stream_sub_ids ] if topic_name is not None: topic_name = truncate_topic(topic_name) target_message.set_topic_name(topic_name) # These fields have legacy field names. event[ORIG_TOPIC] = orig_topic_name event[TOPIC_NAME] = topic_name event[TOPIC_LINKS] = topic_links(target_message.sender.realm_id, topic_name) edit_history_event["prev_topic"] = orig_topic_name edit_history_event["topic"] = topic_name update_edit_history(target_message, timestamp, edit_history_event) delete_event_notify_user_ids: List[int] = [] if propagate_mode in ["change_later", "change_all"]: assert topic_name is not None or new_stream is not None assert stream_being_edited is not None # Other messages should only get topic/stream fields in their edit history. topic_only_edit_history_event: EditHistoryEvent = { "user_id": edit_history_event["user_id"], "timestamp": edit_history_event["timestamp"], } if topic_name is not None: topic_only_edit_history_event["prev_topic"] = edit_history_event[ "prev_topic"] topic_only_edit_history_event["topic"] = edit_history_event[ "topic"] if new_stream is not None: topic_only_edit_history_event["prev_stream"] = edit_history_event[ "prev_stream"] topic_only_edit_history_event["stream"] = edit_history_event[ "stream"] messages_list = update_messages_for_topic_edit( acting_user=user_profile, edited_message=target_message, propagate_mode=propagate_mode, orig_topic_name=orig_topic_name, topic_name=topic_name, new_stream=new_stream, old_stream=stream_being_edited, edit_history_event=topic_only_edit_history_event, last_edit_time=timestamp, ) changed_messages += messages_list if new_stream is not None: assert stream_being_edited is not None changed_message_ids = [msg.id for msg in changed_messages] if subs_gaining_usermessages: ums_to_create = [] for message_id in changed_message_ids: for user_profile_id in subs_gaining_usermessages: # The fact that the user didn't have a UserMessage originally means we can infer that the user # was not mentioned in the original message (even if mention syntax was present, it would not # take effect for a user who was not subscribed). If we were editing the message's content, we # would rerender the message and then use the new stream's data to determine whether this is # a mention of a subscriber; but as we are not doing so, we choose to preserve the "was this # mention syntax an actual mention" decision made during the original rendering for implementation # simplicity. As a result, the only flag to consider applying here is read. um = UserMessageLite( user_profile_id=user_profile_id, message_id=message_id, flags=UserMessage.flags.read, ) ums_to_create.append(um) bulk_insert_ums(ums_to_create) # Delete UserMessage objects for users who will no # longer have access to these messages. Note: This could be # very expensive, since it's N guest users x M messages. UserMessage.objects.filter( user_profile_id__in=[ sub.user_profile_id for sub in subs_losing_usermessages ], message_id__in=changed_message_ids, ).delete() delete_event: DeleteMessagesEvent = { "type": "delete_message", "message_ids": changed_message_ids, "message_type": "stream", "stream_id": stream_being_edited.id, "topic": orig_topic_name, } delete_event_notify_user_ids = [ sub.user_profile_id for sub in subs_losing_access ] send_event(user_profile.realm, delete_event, delete_event_notify_user_ids) # Reset the Attachment.is_*_public caches for all messages # moved to another stream with different access permissions. if new_stream.invite_only != stream_being_edited.invite_only: Attachment.objects.filter( messages__in=changed_message_ids).update( is_realm_public=None, ) ArchivedAttachment.objects.filter( messages__in=changed_message_ids).update( is_realm_public=None, ) if new_stream.is_web_public != stream_being_edited.is_web_public: Attachment.objects.filter( messages__in=changed_message_ids).update( is_web_public=None, ) ArchivedAttachment.objects.filter( messages__in=changed_message_ids).update( is_web_public=None, ) # This does message.save(update_fields=[...]) save_message_for_edit_use_case(message=target_message) realm_id: Optional[int] = None if stream_being_edited is not None: realm_id = stream_being_edited.realm_id event["message_ids"] = update_to_dict_cache(changed_messages, realm_id) def user_info(um: UserMessage) -> Dict[str, Any]: return { "id": um.user_profile_id, "flags": um.flags_list(), } # The following blocks arranges that users who are subscribed to a # stream and can see history from before they subscribed get # live-update when old messages are edited (e.g. if the user does # a topic edit themself). # # We still don't send an update event to users who are not # subscribed to this stream and don't have a UserMessage row. This # means if a non-subscriber is viewing the narrow, they won't get # a real-time updates. This is a balance between sending # message-edit notifications for every public stream to every user # in the organization (too expansive, and also not what we do for # newly sent messages anyway) and having magical live-updates # where possible. users_to_be_notified = list(map(user_info, ums)) if stream_being_edited is not None: if stream_being_edited.is_history_public_to_subscribers(): subscriptions = get_active_subscriptions_for_stream_id( stream_id, include_deactivated_users=False) # We exclude long-term idle users, since they by # definition have no active clients. subscriptions = subscriptions.exclude( user_profile__long_term_idle=True) # Remove duplicates by excluding the id of users already # in users_to_be_notified list. This is the case where a # user both has a UserMessage row and is a current # Subscriber subscriptions = subscriptions.exclude( user_profile_id__in=[um.user_profile_id for um in ums]) if new_stream is not None: assert delete_event_notify_user_ids is not None subscriptions = subscriptions.exclude( user_profile_id__in=delete_event_notify_user_ids) # All users that are subscribed to the stream must be # notified when a message is edited subscriber_ids = set( subscriptions.values_list("user_profile_id", flat=True)) if new_stream is not None: # TODO: Guest users don't see the new moved topic # unless breadcrumb message for new stream is # enabled. Excluding these users from receiving this # event helps us avoid a error traceback for our # clients. We should figure out a way to inform the # guest users of this new topic if sending a 'message' # event for these messages is not an option. # # Don't send this event to guest subs who are not # subscribers of the old stream but are subscribed to # the new stream; clients will be confused. old_stream_unsubbed_guests = [ sub for sub in subs_to_new_stream if sub.user_profile.is_guest and sub.user_profile_id not in subscriber_ids ] subscriptions = subscriptions.exclude(user_profile_id__in=[ sub.user_profile_id for sub in old_stream_unsubbed_guests ]) subscriber_ids = set( subscriptions.values_list("user_profile_id", flat=True)) users_to_be_notified += list( map(subscriber_info, sorted(list(subscriber_ids)))) # UserTopic updates and the content of notifications depend on # whether we've moved the entire topic, or just part of it. We # make that determination here. moved_all_visible_messages = False if topic_name is not None or new_stream is not None: assert stream_being_edited is not None if propagate_mode == "change_all": moved_all_visible_messages = True else: # With other propagate modes, if the user in fact moved # all messages in the stream, we want to explain it was a # full-topic move. # # For security model reasons, we don't want to allow a # user to take any action that would leak information # about older messages they cannot access (E.g. the only # remaining messages are in a stream without shared # history). The bulk_access_messages call below addresses # that concern. # # bulk_access_messages is inefficient for this task, since # we just want to do the exists() version of this # query. But it's nice to reuse code, and this bulk # operation is likely cheaper than a `GET /messages` # unless the topic has thousands of messages of history. assert stream_being_edited.recipient_id is not None unmoved_messages = messages_for_topic( stream_being_edited.recipient_id, orig_topic_name, ) visible_unmoved_messages = bulk_access_messages( user_profile, unmoved_messages, stream=stream_being_edited) moved_all_visible_messages = len(visible_unmoved_messages) == 0 # Migrate muted topic configuration in the following circumstances: # # * If propagate_mode is change_all, do so unconditionally. # # * If propagate_mode is change_later or change_one, do so when # the acting user has moved the entire topic (as visible to them). # # This rule corresponds to checking moved_all_visible_messages. # # We may want more complex behavior in cases where one appears to # be merging topics (E.g. there are existing messages in the # target topic). if moved_all_visible_messages: assert stream_being_edited is not None assert topic_name is not None or new_stream is not None for muting_user in get_users_muting_topic(stream_being_edited.id, orig_topic_name): # TODO: Ideally, this would be a bulk update operation, # because we are doing database operations in a loop here. # # This loop is only acceptable in production because it is # rare for more than a few users to have muted an # individual topic that is being moved; as of this # writing, no individual topic in Zulip Cloud had been # muted by more than 100 users. if new_stream is not None and muting_user.id in delete_event_notify_user_ids: # If the messages are being moved to a stream the user # cannot access, then we treat this as the # messages/topic being deleted for this user. This is # important for security reasons; we don't want to # give users a UserTopic row in a stream they cannot # access. Unmute the topic for such users. do_unmute_topic(muting_user, stream_being_edited, orig_topic_name) else: # Otherwise, we move the muted topic record for the user. # We call remove_topic_mute rather than do_unmute_topic to # avoid sending two events with new muted topics in # immediate succession; this is correct only because # muted_topics events always send the full set of topics. remove_topic_mute(muting_user, stream_being_edited.id, orig_topic_name) do_mute_topic( muting_user, new_stream if new_stream is not None else stream_being_edited, topic_name if topic_name is not None else orig_topic_name, ignore_duplicate=True, ) send_event(user_profile.realm, event, users_to_be_notified) if len( changed_messages ) > 0 and new_stream is not None and stream_being_edited is not None: # Notify users that the topic was moved. changed_messages_count = len(changed_messages) old_thread_notification_string = None if send_notification_to_old_thread: if moved_all_visible_messages: old_thread_notification_string = gettext_lazy( "This topic was moved to {new_location} by {user}.") elif changed_messages_count == 1: old_thread_notification_string = gettext_lazy( "A message was moved from this topic to {new_location} by {user}." ) else: old_thread_notification_string = gettext_lazy( "{changed_messages_count} messages were moved from this topic to {new_location} by {user}." ) new_thread_notification_string = None if send_notification_to_new_thread: if moved_all_visible_messages: new_thread_notification_string = gettext_lazy( "This topic was moved here from {old_location} by {user}.") elif changed_messages_count == 1: new_thread_notification_string = gettext_lazy( "A message was moved here from {old_location} by {user}.") else: new_thread_notification_string = gettext_lazy( "{changed_messages_count} messages were moved here from {old_location} by {user}." ) send_message_moved_breadcrumbs( user_profile, stream_being_edited, orig_topic_name, old_thread_notification_string, new_stream, topic_name, new_thread_notification_string, changed_messages_count, ) if (topic_name is not None and new_stream is None and content is None and len(changed_messages) > 0): assert stream_being_edited is not None maybe_send_resolve_topic_notifications( user_profile=user_profile, stream=stream_being_edited, old_topic=orig_topic_name, new_topic=topic_name, changed_messages=changed_messages, ) return len(changed_messages)
def generate_and_send_messages( data: Tuple[int, Sequence[Sequence[int]], Mapping[str, Any], Callable[[str], Any], int] ) -> int: (tot_messages, personals_pairs, options, output, random_seed) = data random.seed(random_seed) with open( os.path.join(get_or_create_dev_uuid_var_path('test-backend'), "test_messages.json"), "r") as infile: dialog = ujson.load(infile) random.shuffle(dialog) texts = itertools.cycle(dialog) recipient_streams = [ klass.id for klass in Recipient.objects.filter(type=Recipient.STREAM) ] # type: List[int] recipient_huddles = [ h.id for h in Recipient.objects.filter(type=Recipient.HUDDLE) ] # type: List[int] huddle_members = {} # type: Dict[int, List[int]] for h in recipient_huddles: huddle_members[h] = [ s.user_profile.id for s in Subscription.objects.filter(recipient_id=h) ] message_batch_size = options['batch_size'] num_messages = 0 random_max = 1000000 recipients = {} # type: Dict[int, Tuple[int, int, Dict[str, Any]]] messages = [] while num_messages < tot_messages: saved_data = {} # type: Dict[str, Any] message = Message() message.sending_client = get_client('populate_db') message.content = next(texts) randkey = random.randint(1, random_max) if (num_messages > 0 and random.randint(1, random_max) * 100. / random_max < options["stickyness"]): # Use an old recipient message_type, recipient_id, saved_data = recipients[num_messages - 1] if message_type == Recipient.PERSONAL: personals_pair = saved_data['personals_pair'] random.shuffle(personals_pair) elif message_type == Recipient.STREAM: message.subject = saved_data['subject'] message.recipient = get_recipient_by_id(recipient_id) elif message_type == Recipient.HUDDLE: message.recipient = get_recipient_by_id(recipient_id) elif (randkey <= random_max * options["percent_huddles"] / 100.): message_type = Recipient.HUDDLE message.recipient = get_recipient_by_id( random.choice(recipient_huddles)) elif (randkey <= random_max * (options["percent_huddles"] + options["percent_personals"]) / 100.): message_type = Recipient.PERSONAL personals_pair = random.choice(personals_pairs) random.shuffle(personals_pair) elif (randkey <= random_max * 1.0): message_type = Recipient.STREAM message.recipient = get_recipient_by_id( random.choice(recipient_streams)) if message_type == Recipient.HUDDLE: sender_id = random.choice(huddle_members[message.recipient.id]) message.sender = get_user_profile_by_id(sender_id) elif message_type == Recipient.PERSONAL: message.recipient = Recipient.objects.get( type=Recipient.PERSONAL, type_id=personals_pair[0]) message.sender = get_user_profile_by_id(personals_pair[1]) saved_data['personals_pair'] = personals_pair elif message_type == Recipient.STREAM: stream = Stream.objects.get(id=message.recipient.type_id) # Pick a random subscriber to the stream message.sender = random.choice( Subscription.objects.filter( recipient=message.recipient)).user_profile message.subject = stream.name + str(random.randint(1, 3)) saved_data['subject'] = message.subject message.date_sent = choose_date_sent(num_messages, tot_messages, options['threads']) messages.append(message) recipients[num_messages] = (message_type, message.recipient.id, saved_data) num_messages += 1 if (num_messages % message_batch_size) == 0: # Send the batch and empty the list: send_messages(messages) messages = [] if len(messages) > 0: # If there are unsent messages after exiting the loop, send them: send_messages(messages) return tot_messages
def restore_saved_messages(): # type: () -> None old_messages = [] # type: List[Dict[str, Any]] duplicate_suppression_hash = {} # type: Dict[str, bool] stream_dict = {} # type: Dict[Tuple[text_type, text_type], Tuple[text_type, text_type]] user_set = set() # type: Set[Tuple[text_type, text_type, text_type, bool]] email_set = set([u.email for u in UserProfile.objects.all()]) # type: Set[text_type] realm_set = set() # type: Set[text_type] # Initial client_set is nonempty temporarily because we don't have # clients in logs at all right now -- later we can start with nothing. client_set = set(["populate_db", "website", "zephyr_mirror"]) huddle_user_set = set() # type: Set[Tuple[text_type, ...]] # First, determine all the objects our messages will need. print(datetime.datetime.now(), "Creating realms/streams/etc...") def process_line(line): # type: (str) -> None old_message_json = line.strip() # Due to populate_db's shakespeare mode, we have a lot of # duplicate messages in our log that only differ in their # logged ID numbers (same timestamp, content, etc.). With # sqlite, bulk creating those messages won't work properly: in # particular, the first 100 messages will actually only result # in 20 rows ending up in the target table, which screws up # the below accounting where for handling changing # subscriptions, we assume that the Nth row populate_db # created goes with the Nth non-subscription row of the input # So suppress the duplicates when using sqlite. if "sqlite" in settings.DATABASES["default"]["ENGINE"]: tmp_message = ujson.loads(old_message_json) tmp_message['id'] = '1' duplicate_suppression_key = ujson.dumps(tmp_message) if duplicate_suppression_key in duplicate_suppression_hash: return duplicate_suppression_hash[duplicate_suppression_key] = True old_message = ujson.loads(old_message_json) message_type = old_message["type"] # Lower case emails and domains; it will screw up # deduplication if we don't def fix_email(email): # type: (text_type) -> text_type return email.strip().lower() if message_type in ["stream", "huddle", "personal"]: old_message["sender_email"] = fix_email(old_message["sender_email"]) # Fix the length on too-long messages before we start processing them if len(old_message["content"]) > MAX_MESSAGE_LENGTH: old_message["content"] = "[ This message was deleted because it was too long ]" if message_type in ["subscription_added", "subscription_removed"]: old_message["domain"] = old_message["domain"].lower() old_message["user"] = fix_email(old_message["user"]) elif message_type == "subscription_property": old_message["user"] = fix_email(old_message["user"]) elif message_type == "user_email_changed": old_message["old_email"] = fix_email(old_message["old_email"]) old_message["new_email"] = fix_email(old_message["new_email"]) elif message_type.startswith("user_"): old_message["user"] = fix_email(old_message["user"]) elif message_type.startswith("enable_"): old_message["user"] = fix_email(old_message["user"]) if message_type == 'personal': old_message["recipient"][0]["email"] = fix_email(old_message["recipient"][0]["email"]) elif message_type == "huddle": for i in range(len(old_message["recipient"])): old_message["recipient"][i]["email"] = fix_email(old_message["recipient"][i]["email"]) old_messages.append(old_message) if message_type in ["subscription_added", "subscription_removed"]: stream_name = old_message["name"].strip() # type: text_type canon_stream_name = stream_name.lower() if canon_stream_name not in stream_dict: stream_dict[(old_message["domain"], canon_stream_name)] = \ (old_message["domain"], stream_name) elif message_type == "user_created": user_set.add((old_message["user"], old_message["full_name"], old_message["short_name"], False)) elif message_type == "realm_created": realm_set.add(old_message["domain"]) if message_type not in ["stream", "huddle", "personal"]: return sender_email = old_message["sender_email"] domain = split_email_to_domain(sender_email) realm_set.add(domain) if old_message["sender_email"] not in email_set: user_set.add((old_message["sender_email"], old_message["sender_full_name"], old_message["sender_short_name"], False)) if 'sending_client' in old_message: client_set.add(old_message['sending_client']) if message_type == 'stream': stream_name = old_message["recipient"].strip() canon_stream_name = stream_name.lower() if canon_stream_name not in stream_dict: stream_dict[(domain, canon_stream_name)] = (domain, stream_name) elif message_type == 'personal': u = old_message["recipient"][0] if u["email"] not in email_set: user_set.add((u["email"], u["full_name"], u["short_name"], False)) email_set.add(u["email"]) elif message_type == 'huddle': for u in old_message["recipient"]: user_set.add((u["email"], u["full_name"], u["short_name"], False)) if u["email"] not in email_set: user_set.add((u["email"], u["full_name"], u["short_name"], False)) email_set.add(u["email"]) huddle_user_set.add(tuple(sorted(set(u["email"] for u in old_message["recipient"])))) else: raise ValueError('Bad message type') event_glob = os.path.join(settings.EVENT_LOG_DIR, 'events.*') for filename in sorted(glob.glob(event_glob)): with open(filename, "r") as message_log: for line in message_log.readlines(): process_line(line) stream_recipients = {} # type: Dict[Tuple[int, text_type], Recipient] user_recipients = {} # type: Dict[text_type, Recipient] huddle_recipients = {} # type: Dict[text_type, Recipient] # Then, create the objects our messages need. print(datetime.datetime.now(), "Creating realms...") bulk_create_realms(realm_set) realms = {} # type: Dict[text_type, Realm] for realm in Realm.objects.all(): realms[realm.domain] = realm print(datetime.datetime.now(), "Creating clients...") bulk_create_clients(client_set) clients = {} # type: Dict[text_type, Client] for client in Client.objects.all(): clients[client.name] = client print(datetime.datetime.now(), "Creating streams...") bulk_create_streams(realms, list(stream_dict.values())) streams = {} # type: Dict[int, Stream] for stream in Stream.objects.all(): streams[stream.id] = stream for recipient in Recipient.objects.filter(type=Recipient.STREAM): stream_recipients[(streams[recipient.type_id].realm_id, streams[recipient.type_id].name.lower())] = recipient print(datetime.datetime.now(), "Creating users...") bulk_create_users(realms, user_set, tos_version=settings.TOS_VERSION) users = {} # type: Dict[text_type, UserProfile] users_by_id = {} # type: Dict[int, UserProfile] for user_profile in UserProfile.objects.select_related().all(): users[user_profile.email] = user_profile users_by_id[user_profile.id] = user_profile for recipient in Recipient.objects.filter(type=Recipient.PERSONAL): user_recipients[users_by_id[recipient.type_id].email] = recipient print(datetime.datetime.now(), "Creating huddles...") bulk_create_huddles(users, huddle_user_set) huddles_by_id = {} # type: Dict[int, Huddle] for huddle in Huddle.objects.all(): huddles_by_id[huddle.id] = huddle for recipient in Recipient.objects.filter(type=Recipient.HUDDLE): huddle_recipients[huddles_by_id[recipient.type_id].huddle_hash] = recipient # TODO: Add a special entry type in the log that is a subscription # change and import those as we go to make subscription changes # take effect! print(datetime.datetime.now(), "Importing subscriptions...") subscribers = {} # type: Dict[int, Set[int]] for s in Subscription.objects.select_related().all(): if s.active: subscribers.setdefault(s.recipient.id, set()).add(s.user_profile.id) # Then create all the messages, without talking to the DB! print(datetime.datetime.now(), "Importing messages, part 1...") first_message_id = None if Message.objects.exists(): first_message_id = Message.objects.all().order_by("-id")[0].id + 1 messages_to_create = [] # type: List[Message] for idx, old_message in enumerate(old_messages): message_type = old_message["type"] if message_type not in ["stream", "huddle", "personal"]: continue message = Message() sender_email = old_message["sender_email"] domain = split_email_to_domain(sender_email) realm = realms[domain] message.sender = users[sender_email] type_hash = {"stream": Recipient.STREAM, "huddle": Recipient.HUDDLE, "personal": Recipient.PERSONAL} if 'sending_client' in old_message: message.sending_client = clients[old_message['sending_client']] elif sender_email in ["*****@*****.**", "*****@*****.**", "*****@*****.**", "*****@*****.**", "*****@*****.**"]: message.sending_client = clients['populate_db'] elif realm.domain == "zulip.com": message.sending_client = clients["website"] elif realm.domain == "mit.edu": message.sending_client = clients['zephyr_mirror'] else: message.sending_client = clients['populate_db'] message.type = type_hash[message_type] message.content = old_message["content"] message.subject = old_message["subject"] message.pub_date = timestamp_to_datetime(old_message["timestamp"]) if message.type == Recipient.PERSONAL: message.recipient = user_recipients[old_message["recipient"][0]["email"]] elif message.type == Recipient.STREAM: message.recipient = stream_recipients[(realm.id, old_message["recipient"].lower())] elif message.type == Recipient.HUDDLE: huddle_hash = get_huddle_hash([users[u["email"]].id for u in old_message["recipient"]]) message.recipient = huddle_recipients[huddle_hash] else: raise ValueError('Bad message type') messages_to_create.append(message) print(datetime.datetime.now(), "Importing messages, part 2...") Message.objects.bulk_create(messages_to_create) messages_to_create = [] # Finally, create all the UserMessage objects print(datetime.datetime.now(), "Importing usermessages, part 1...") personal_recipients = {} # type: Dict[int, bool] for r in Recipient.objects.filter(type = Recipient.PERSONAL): personal_recipients[r.id] = True all_messages = Message.objects.all() # type: Sequence[Message] user_messages_to_create = [] # type: List[UserMessage] messages_by_id = {} # type: Dict[int, Message] for message in all_messages: messages_by_id[message.id] = message if len(messages_by_id) == 0: print(datetime.datetime.now(), "No old messages to replay") return if first_message_id is None: first_message_id = min(messages_by_id.keys()) tot_user_messages = 0 pending_subs = {} # type: Dict[Tuple[int, int], bool] current_message_id = first_message_id pending_colors = {} # type: Dict[Tuple[text_type, text_type], text_type] for old_message in old_messages: message_type = old_message["type"] if message_type == 'subscription_added': stream_key = (realms[old_message["domain"]].id, old_message["name"].strip().lower()) subscribers.setdefault(stream_recipients[stream_key].id, set()).add(users[old_message["user"]].id) pending_subs[(stream_recipients[stream_key].id, users[old_message["user"]].id)] = True continue elif message_type == "subscription_removed": stream_key = (realms[old_message["domain"]].id, old_message["name"].strip().lower()) user_id = users[old_message["user"]].id subscribers.setdefault(stream_recipients[stream_key].id, set()) try: subscribers[stream_recipients[stream_key].id].remove(user_id) except KeyError: print("Error unsubscribing %s from %s: not subscribed" % ( old_message["user"], old_message["name"])) pending_subs[(stream_recipients[stream_key].id, users[old_message["user"]].id)] = False continue elif message_type == "user_activated" or message_type == "user_created": # These are rare, so just handle them the slow way user_profile = users[old_message["user"]] join_date = timestamp_to_datetime(old_message['timestamp']) do_activate_user(user_profile, log=False, join_date=join_date) # Update the cache of users to show this user as activated users_by_id[user_profile.id] = user_profile users[old_message["user"]] = user_profile continue elif message_type == "user_deactivated": user_profile = users[old_message["user"]] do_deactivate_user(user_profile, log=False) continue elif message_type == "user_change_password": # Just handle these the slow way user_profile = users[old_message["user"]] do_change_password(user_profile, old_message["pwhash"], log=False, hashed_password=True) continue elif message_type == "user_change_full_name": # Just handle these the slow way user_profile = users[old_message["user"]] user_profile.full_name = old_message["full_name"] user_profile.save(update_fields=["full_name"]) continue elif message_type == "enable_desktop_notifications_changed": # Just handle these the slow way user_profile = users[old_message["user"]] user_profile.enable_desktop_notifications = (old_message["enable_desktop_notifications"] != "false") user_profile.save(update_fields=["enable_desktop_notifications"]) continue elif message_type == "enable_sounds_changed": user_profile = users[old_message["user"]] user_profile.enable_sounds = (old_message["enable_sounds"] != "false") user_profile.save(update_fields=["enable_sounds"]) elif message_type == "enable_offline_email_notifications_changed": user_profile = users[old_message["user"]] user_profile.enable_offline_email_notifications = ( old_message["enable_offline_email_notifications"] != "false") user_profile.save(update_fields=["enable_offline_email_notifications"]) continue elif message_type == "enable_offline_push_notifications_changed": user_profile = users[old_message["user"]] user_profile.enable_offline_push_notifications = ( old_message["enable_offline_push_notifications"] != "false") user_profile.save(update_fields=["enable_offline_push_notifications"]) continue elif message_type == "default_streams": set_default_streams(get_realm(old_message["domain"]), old_message["streams"]) continue elif message_type == "subscription_property": property_name = old_message.get("property") if property_name == "stream_color" or property_name == "color": color = old_message.get("color", old_message.get("value")) pending_colors[(old_message["user"], old_message["stream_name"].lower())] = color elif property_name in ["in_home_view", "notifications"]: # TODO: Handle this continue else: raise RuntimeError("Unknown property %s" % (property_name,)) continue elif message_type == "realm_created": # No action required continue elif message_type in ["user_email_changed", "update_onboarding", "update_message"]: # TODO: Handle these continue if message_type not in ["stream", "huddle", "personal"]: raise RuntimeError("Unexpected message type %s" % (message_type,)) message = messages_by_id[current_message_id] current_message_id += 1 if message.recipient_id not in subscribers: # Nobody received this message -- probably due to our # subscriptions being out-of-date. continue recipient_user_ids = set() # type: Set[int] for user_profile_id in subscribers[message.recipient_id]: recipient_user_ids.add(user_profile_id) if message.recipient_id in personal_recipients: # Include the sender in huddle recipients recipient_user_ids.add(message.sender_id) for user_profile_id in recipient_user_ids: if users_by_id[user_profile_id].is_active: um = UserMessage(user_profile_id=user_profile_id, message=message) user_messages_to_create.append(um) if len(user_messages_to_create) > 100000: tot_user_messages += len(user_messages_to_create) UserMessage.objects.bulk_create(user_messages_to_create) user_messages_to_create = [] print(datetime.datetime.now(), "Importing usermessages, part 2...") tot_user_messages += len(user_messages_to_create) UserMessage.objects.bulk_create(user_messages_to_create) print(datetime.datetime.now(), "Finalizing subscriptions...") current_subs = {} # type: Dict[Tuple[int, int], bool] current_subs_obj = {} # type: Dict[Tuple[int, int], Subscription] for s in Subscription.objects.select_related().all(): current_subs[(s.recipient_id, s.user_profile_id)] = s.active current_subs_obj[(s.recipient_id, s.user_profile_id)] = s subscriptions_to_add = [] # type: List[Subscription] subscriptions_to_change = [] # type: List[Tuple[Tuple[int, int], bool]] for pending_sub in pending_subs.keys(): (recipient_id, user_profile_id) = pending_sub current_state = current_subs.get(pending_sub) if pending_subs[pending_sub] == current_state: # Already correct in the database continue elif current_state is not None: subscriptions_to_change.append((pending_sub, pending_subs[pending_sub])) continue s = Subscription(recipient_id=recipient_id, user_profile_id=user_profile_id, active=pending_subs[pending_sub]) subscriptions_to_add.append(s) Subscription.objects.bulk_create(subscriptions_to_add) for (sub_tuple, active) in subscriptions_to_change: current_subs_obj[sub_tuple].active = active current_subs_obj[sub_tuple].save(update_fields=["active"]) subs = {} # type: Dict[Tuple[int, int], Subscription] for sub in Subscription.objects.all(): subs[(sub.user_profile_id, sub.recipient_id)] = sub # TODO: do restore of subscription colors -- we're currently not # logging changes so there's little point in having the code :( print(datetime.datetime.now(), "Finished importing %s messages (%s usermessages)" % \ (len(all_messages), tot_user_messages)) site = Site.objects.get_current() site.domain = 'zulip.com' site.save() print(datetime.datetime.now(), "Filling in user pointers...") # Set restored pointers to the very latest messages for user_profile in UserProfile.objects.all(): try: top = UserMessage.objects.filter( user_profile_id=user_profile.id).order_by("-message")[0] user_profile.pointer = top.message_id except IndexError: user_profile.pointer = -1 user_profile.save(update_fields=["pointer"]) print(datetime.datetime.now(), "Done replaying old messages")