2016-10-04 15:52:26 +02:00
|
|
|
|
|
|
|
import datetime
|
|
|
|
import ujson
|
|
|
|
import zlib
|
|
|
|
|
2016-10-12 02:14:08 +02:00
|
|
|
from django.utils.translation import ugettext as _
|
2017-07-16 09:41:38 +02:00
|
|
|
from django.utils.timezone import now as timezone_now
|
2018-01-22 21:50:22 +01:00
|
|
|
from django.db.models import Sum
|
|
|
|
|
|
|
|
from analytics.lib.counts import COUNT_STATS, RealmCount
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-10-10 05:23:53 +02:00
|
|
|
from zerver.lib.avatar import get_avatar_field
|
2016-10-04 15:52:26 +02:00
|
|
|
import zerver.lib.bugdown as bugdown
|
2017-11-07 17:36:29 +01:00
|
|
|
from zerver.lib.cache import (
|
|
|
|
cache_with_key,
|
|
|
|
generic_bulk_cached_fetch,
|
|
|
|
to_dict_cache_key,
|
|
|
|
to_dict_cache_key_id,
|
2018-01-17 15:04:27 +01:00
|
|
|
realm_first_visible_message_id_cache_key,
|
2018-01-22 21:50:22 +01:00
|
|
|
cache_get, cache_set,
|
2017-11-07 17:36:29 +01:00
|
|
|
)
|
2016-10-12 02:14:08 +02:00
|
|
|
from zerver.lib.request import JsonableError
|
2017-10-29 17:11:11 +01:00
|
|
|
from zerver.lib.stream_subscription import (
|
|
|
|
get_stream_subscriptions_for_user,
|
|
|
|
)
|
2016-10-04 15:52:26 +02:00
|
|
|
from zerver.lib.timestamp import datetime_to_timestamp
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
from zerver.lib.topic_mutes import (
|
|
|
|
build_topic_mute_checker,
|
|
|
|
topic_is_muted,
|
|
|
|
)
|
2016-10-04 15:52:26 +02:00
|
|
|
|
|
|
|
from zerver.models import (
|
|
|
|
get_display_recipient_by_id,
|
2017-02-03 23:21:56 +01:00
|
|
|
get_user_profile_by_id,
|
2017-10-10 20:10:57 +02:00
|
|
|
query_for_ids,
|
2016-10-04 15:52:26 +02:00
|
|
|
Message,
|
2017-01-18 23:19:18 +01:00
|
|
|
Realm,
|
2016-10-04 15:52:26 +02:00
|
|
|
Recipient,
|
2016-10-12 02:14:08 +02:00
|
|
|
Stream,
|
2017-08-09 02:22:00 +02:00
|
|
|
Subscription,
|
2016-10-04 18:32:46 +02:00
|
|
|
UserProfile,
|
2016-10-12 02:14:08 +02:00
|
|
|
UserMessage,
|
2016-12-06 07:19:34 +01:00
|
|
|
Reaction
|
2016-10-04 15:52:26 +02:00
|
|
|
)
|
|
|
|
|
2018-05-11 01:40:23 +02:00
|
|
|
from typing import Any, Dict, List, Optional, Set, Tuple, Union
|
2017-08-09 04:01:00 +02:00
|
|
|
from mypy_extensions import TypedDict
|
2016-10-04 18:32:46 +02:00
|
|
|
|
2018-05-11 01:40:23 +02:00
|
|
|
RealmAlertWords = Dict[int, List[str]]
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-11-10 15:57:43 +01:00
|
|
|
RawUnreadMessagesResult = TypedDict('RawUnreadMessagesResult', {
|
|
|
|
'pm_dict': Dict[int, Any],
|
|
|
|
'stream_dict': Dict[int, Any],
|
|
|
|
'huddle_dict': Dict[int, Any],
|
|
|
|
'mentions': Set[int],
|
|
|
|
'muted_stream_ids': List[int],
|
|
|
|
'unmuted_stream_msgs': Set[int],
|
|
|
|
})
|
|
|
|
|
2017-08-09 04:01:00 +02:00
|
|
|
UnreadMessagesResult = TypedDict('UnreadMessagesResult', {
|
|
|
|
'pms': List[Dict[str, Any]],
|
|
|
|
'streams': List[Dict[str, Any]],
|
|
|
|
'huddles': List[Dict[str, Any]],
|
|
|
|
'mentions': List[int],
|
|
|
|
'count': int,
|
|
|
|
})
|
|
|
|
|
2017-08-01 18:28:56 +02:00
|
|
|
MAX_UNREAD_MESSAGES = 5000
|
|
|
|
|
2017-11-07 17:36:29 +01:00
|
|
|
def messages_for_ids(message_ids: List[int],
|
|
|
|
user_message_flags: Dict[int, List[str]],
|
2018-05-11 01:40:23 +02:00
|
|
|
search_fields: Dict[int, Dict[str, str]],
|
2017-11-07 17:36:29 +01:00
|
|
|
apply_markdown: bool,
|
|
|
|
client_gravatar: bool,
|
|
|
|
allow_edit_history: bool) -> List[Dict[str, Any]]:
|
|
|
|
|
|
|
|
cache_transformer = MessageDict.build_dict_from_raw_db_row
|
|
|
|
id_fetcher = lambda row: row['id']
|
|
|
|
|
|
|
|
message_dicts = generic_bulk_cached_fetch(to_dict_cache_key_id,
|
|
|
|
MessageDict.get_raw_db_rows,
|
|
|
|
message_ids,
|
|
|
|
id_fetcher=id_fetcher,
|
|
|
|
cache_transformer=cache_transformer,
|
|
|
|
extractor=extract_message_dict,
|
|
|
|
setter=stringify_message_dict)
|
|
|
|
|
|
|
|
message_list = [] # type: List[Dict[str, Any]]
|
|
|
|
|
|
|
|
for message_id in message_ids:
|
|
|
|
msg_dict = message_dicts[message_id]
|
|
|
|
msg_dict.update({"flags": user_message_flags[message_id]})
|
|
|
|
if message_id in search_fields:
|
|
|
|
msg_dict.update(search_fields[message_id])
|
|
|
|
# Make sure that we never send message edit history to clients
|
|
|
|
# in realms with allow_edit_history disabled.
|
|
|
|
if "edit_history" in msg_dict and not allow_edit_history:
|
|
|
|
del msg_dict["edit_history"]
|
|
|
|
message_list.append(msg_dict)
|
|
|
|
|
|
|
|
MessageDict.post_process_dicts(message_list, apply_markdown, client_gravatar)
|
|
|
|
|
|
|
|
return message_list
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def sew_messages_and_reactions(messages: List[Dict[str, Any]],
|
|
|
|
reactions: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
|
2017-10-14 16:14:54 +02:00
|
|
|
"""Given a iterable of messages and reactions stitch reactions
|
|
|
|
into messages.
|
|
|
|
"""
|
|
|
|
# Add all messages with empty reaction item
|
|
|
|
for message in messages:
|
|
|
|
message['reactions'] = []
|
|
|
|
|
|
|
|
# Convert list of messages into dictionary to make reaction stitching easy
|
|
|
|
converted_messages = {message['id']: message for message in messages}
|
|
|
|
|
|
|
|
for reaction in reactions:
|
|
|
|
converted_messages[reaction['message_id']]['reactions'].append(
|
|
|
|
reaction)
|
|
|
|
|
|
|
|
return list(converted_messages.values())
|
|
|
|
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def extract_message_dict(message_bytes: bytes) -> Dict[str, Any]:
|
2017-11-04 19:22:54 +01:00
|
|
|
return ujson.loads(zlib.decompress(message_bytes).decode("utf-8"))
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def stringify_message_dict(message_dict: Dict[str, Any]) -> bytes:
|
2017-11-04 16:25:20 +01:00
|
|
|
return zlib.compress(ujson.dumps(message_dict).encode())
|
2016-10-04 15:52:26 +02:00
|
|
|
|
|
|
|
@cache_with_key(to_dict_cache_key, timeout=3600*24)
|
2017-11-05 11:15:10 +01:00
|
|
|
def message_to_dict_json(message: Message) -> bytes:
|
2017-10-20 20:29:49 +02:00
|
|
|
return MessageDict.to_dict_uncached(message)
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-11-05 11:37:41 +01:00
|
|
|
class MessageDict:
|
2017-10-20 21:34:05 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def wide_dict(message: Message) -> Dict[str, Any]:
|
2017-10-20 21:34:05 +02:00
|
|
|
'''
|
|
|
|
The next two lines get the cachable field related
|
|
|
|
to our message object, with the side effect of
|
|
|
|
populating the cache.
|
|
|
|
'''
|
|
|
|
json = message_to_dict_json(message)
|
|
|
|
obj = extract_message_dict(json)
|
|
|
|
|
|
|
|
'''
|
|
|
|
The steps below are similar to what we do in
|
|
|
|
post_process_dicts(), except we don't call finalize_payload(),
|
|
|
|
since that step happens later in the queue
|
|
|
|
processor.
|
|
|
|
'''
|
|
|
|
MessageDict.bulk_hydrate_sender_info([obj])
|
|
|
|
MessageDict.hydrate_recipient_info(obj)
|
|
|
|
|
|
|
|
return obj
|
|
|
|
|
2017-10-10 09:22:21 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def post_process_dicts(objs: List[Dict[str, Any]], apply_markdown: bool, client_gravatar: bool) -> None:
|
2017-10-10 20:10:57 +02:00
|
|
|
MessageDict.bulk_hydrate_sender_info(objs)
|
|
|
|
|
2017-10-10 09:22:21 +02:00
|
|
|
for obj in objs:
|
|
|
|
MessageDict.hydrate_recipient_info(obj)
|
2017-10-31 03:02:23 +01:00
|
|
|
MessageDict.finalize_payload(obj, apply_markdown, client_gravatar)
|
2017-10-14 02:01:20 +02:00
|
|
|
|
2017-10-20 21:27:26 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def finalize_payload(obj: Dict[str, Any],
|
|
|
|
apply_markdown: bool,
|
|
|
|
client_gravatar: bool) -> None:
|
2017-10-31 03:02:23 +01:00
|
|
|
MessageDict.set_sender_avatar(obj, client_gravatar)
|
2017-10-20 21:27:26 +02:00
|
|
|
if apply_markdown:
|
|
|
|
obj['content_type'] = 'text/html'
|
|
|
|
obj['content'] = obj['rendered_content']
|
|
|
|
else:
|
|
|
|
obj['content_type'] = 'text/x-markdown'
|
|
|
|
|
|
|
|
del obj['rendered_content']
|
|
|
|
del obj['sender_realm_id']
|
|
|
|
del obj['sender_avatar_source']
|
|
|
|
del obj['sender_avatar_version']
|
|
|
|
|
|
|
|
del obj['raw_display_recipient']
|
|
|
|
del obj['recipient_type']
|
|
|
|
del obj['recipient_type_id']
|
|
|
|
del obj['sender_is_mirror_dummy']
|
2017-10-14 15:44:59 +02:00
|
|
|
|
2016-10-04 15:52:26 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def to_dict_uncached(message: Message) -> bytes:
|
2017-10-20 20:29:49 +02:00
|
|
|
dct = MessageDict.to_dict_uncached_helper(message)
|
2016-10-04 15:52:26 +02:00
|
|
|
return stringify_message_dict(dct)
|
|
|
|
|
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def to_dict_uncached_helper(message: Message) -> Dict[str, Any]:
|
2016-10-04 15:52:26 +02:00
|
|
|
return MessageDict.build_message_dict(
|
2017-01-24 07:06:13 +01:00
|
|
|
message = message,
|
|
|
|
message_id = message.id,
|
|
|
|
last_edit_time = message.last_edit_time,
|
|
|
|
edit_history = message.edit_history,
|
|
|
|
content = message.content,
|
|
|
|
subject = message.subject,
|
|
|
|
pub_date = message.pub_date,
|
|
|
|
rendered_content = message.rendered_content,
|
|
|
|
rendered_content_version = message.rendered_content_version,
|
|
|
|
sender_id = message.sender.id,
|
|
|
|
sender_realm_id = message.sender.realm_id,
|
|
|
|
sending_client_name = message.sending_client.name,
|
|
|
|
recipient_id = message.recipient.id,
|
|
|
|
recipient_type = message.recipient.type,
|
|
|
|
recipient_type_id = message.recipient.type_id,
|
|
|
|
reactions = Reaction.get_raw_db_rows([message.id])
|
2016-10-04 15:52:26 +02:00
|
|
|
)
|
|
|
|
|
2017-10-14 16:14:54 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def get_raw_db_rows(needed_ids: List[int]) -> List[Dict[str, Any]]:
|
2017-10-14 16:14:54 +02:00
|
|
|
# This is a special purpose function optimized for
|
|
|
|
# callers like get_messages_backend().
|
|
|
|
fields = [
|
|
|
|
'id',
|
|
|
|
'subject',
|
|
|
|
'pub_date',
|
|
|
|
'last_edit_time',
|
|
|
|
'edit_history',
|
|
|
|
'content',
|
|
|
|
'rendered_content',
|
|
|
|
'rendered_content_version',
|
|
|
|
'recipient_id',
|
|
|
|
'recipient__type',
|
|
|
|
'recipient__type_id',
|
|
|
|
'sender_id',
|
|
|
|
'sending_client__name',
|
Remove sender-related fields from message cache.
This change makes the cache entries smaller for message
dictionaries. It also ensures we get valid data put into
message dictionaries if, for example, the sender's avatar
changes.
After this change, all of the attributes for a message
sender are only fetched during post-processing with two
exceptions:
* We get sender_id for "free" from the message,
and it's the primary key that we need to figure
out which data to fetch in post-processing.
* We need sender_realm_id to be able to cache topic
links, and a sender's realm id will never change,
so it's not a concern for invalidating cache rows.
All the other attributes are either likely to change (e.g.
sender avatar_version) and/or impact the size of cache
entries more severely than the two small id fields above.
This change should improve our overall system performance
by reducing the amount of memory used by every N message
rows we cache, and typically N will be in the thousands or
so on a large realm.
The other major implication of this change is that when
a user changes their avatar, and then later messages that
the user sent are fetched, all of the fields that go into
computing the avatar url will be pulled from the database,
not from cache.
2017-10-14 18:28:43 +02:00
|
|
|
'sender__realm_id',
|
2017-10-14 16:14:54 +02:00
|
|
|
]
|
|
|
|
messages = Message.objects.filter(id__in=needed_ids).values(*fields)
|
|
|
|
"""Adding one-many or Many-Many relationship in values results in N X
|
|
|
|
results.
|
|
|
|
|
|
|
|
Link: https://docs.djangoproject.com/en/1.8/ref/models/querysets/#values
|
|
|
|
"""
|
|
|
|
reactions = Reaction.get_raw_db_rows(needed_ids)
|
|
|
|
return sew_messages_and_reactions(messages, reactions)
|
|
|
|
|
2016-10-04 15:52:26 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def build_dict_from_raw_db_row(row: Dict[str, Any]) -> Dict[str, Any]:
|
2016-10-04 15:52:26 +02:00
|
|
|
'''
|
|
|
|
row is a row from a .values() call, and it needs to have
|
|
|
|
all the relevant fields populated
|
|
|
|
'''
|
|
|
|
return MessageDict.build_message_dict(
|
2017-01-24 07:06:13 +01:00
|
|
|
message = None,
|
|
|
|
message_id = row['id'],
|
|
|
|
last_edit_time = row['last_edit_time'],
|
|
|
|
edit_history = row['edit_history'],
|
|
|
|
content = row['content'],
|
|
|
|
subject = row['subject'],
|
|
|
|
pub_date = row['pub_date'],
|
|
|
|
rendered_content = row['rendered_content'],
|
|
|
|
rendered_content_version = row['rendered_content_version'],
|
|
|
|
sender_id = row['sender_id'],
|
Remove sender-related fields from message cache.
This change makes the cache entries smaller for message
dictionaries. It also ensures we get valid data put into
message dictionaries if, for example, the sender's avatar
changes.
After this change, all of the attributes for a message
sender are only fetched during post-processing with two
exceptions:
* We get sender_id for "free" from the message,
and it's the primary key that we need to figure
out which data to fetch in post-processing.
* We need sender_realm_id to be able to cache topic
links, and a sender's realm id will never change,
so it's not a concern for invalidating cache rows.
All the other attributes are either likely to change (e.g.
sender avatar_version) and/or impact the size of cache
entries more severely than the two small id fields above.
This change should improve our overall system performance
by reducing the amount of memory used by every N message
rows we cache, and typically N will be in the thousands or
so on a large realm.
The other major implication of this change is that when
a user changes their avatar, and then later messages that
the user sent are fetched, all of the fields that go into
computing the avatar url will be pulled from the database,
not from cache.
2017-10-14 18:28:43 +02:00
|
|
|
sender_realm_id = row['sender__realm_id'],
|
2017-01-24 07:06:13 +01:00
|
|
|
sending_client_name = row['sending_client__name'],
|
|
|
|
recipient_id = row['recipient_id'],
|
|
|
|
recipient_type = row['recipient__type'],
|
|
|
|
recipient_type_id = row['recipient__type_id'],
|
|
|
|
reactions=row['reactions']
|
2016-10-04 15:52:26 +02:00
|
|
|
)
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def build_message_dict(
|
2018-03-12 02:47:49 +01:00
|
|
|
message: Optional[Message],
|
|
|
|
message_id: int,
|
|
|
|
last_edit_time: Optional[datetime.datetime],
|
2018-05-11 01:40:23 +02:00
|
|
|
edit_history: Optional[str],
|
|
|
|
content: str,
|
|
|
|
subject: str,
|
2018-03-12 02:47:49 +01:00
|
|
|
pub_date: datetime.datetime,
|
2018-05-11 01:40:23 +02:00
|
|
|
rendered_content: Optional[str],
|
2018-03-12 02:47:49 +01:00
|
|
|
rendered_content_version: Optional[int],
|
|
|
|
sender_id: int,
|
|
|
|
sender_realm_id: int,
|
2018-05-11 01:40:23 +02:00
|
|
|
sending_client_name: str,
|
2018-03-12 02:47:49 +01:00
|
|
|
recipient_id: int,
|
|
|
|
recipient_type: int,
|
|
|
|
recipient_type_id: int,
|
|
|
|
reactions: List[Dict[str, Any]]
|
|
|
|
) -> Dict[str, Any]:
|
2016-10-04 15:52:26 +02:00
|
|
|
|
|
|
|
obj = dict(
|
|
|
|
id = message_id,
|
|
|
|
sender_id = sender_id,
|
2017-10-20 20:29:49 +02:00
|
|
|
content = content,
|
2017-10-10 08:12:03 +02:00
|
|
|
recipient_type_id = recipient_type_id,
|
|
|
|
recipient_type = recipient_type,
|
2016-10-04 15:52:26 +02:00
|
|
|
recipient_id = recipient_id,
|
|
|
|
subject = subject,
|
|
|
|
timestamp = datetime_to_timestamp(pub_date),
|
|
|
|
client = sending_client_name)
|
|
|
|
|
2017-10-14 02:01:20 +02:00
|
|
|
obj['sender_realm_id'] = sender_realm_id
|
|
|
|
|
2017-10-10 08:12:03 +02:00
|
|
|
obj['raw_display_recipient'] = get_display_recipient_by_id(
|
|
|
|
recipient_id,
|
|
|
|
recipient_type,
|
|
|
|
recipient_type_id
|
|
|
|
)
|
|
|
|
|
2016-12-31 03:08:43 +01:00
|
|
|
obj['subject_links'] = bugdown.subject_links(sender_realm_id, subject)
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-01-24 06:11:49 +01:00
|
|
|
if last_edit_time is not None:
|
2016-10-04 15:52:26 +02:00
|
|
|
obj['last_edit_timestamp'] = datetime_to_timestamp(last_edit_time)
|
2017-05-24 21:28:26 +02:00
|
|
|
assert edit_history is not None
|
2016-10-04 15:52:26 +02:00
|
|
|
obj['edit_history'] = ujson.loads(edit_history)
|
|
|
|
|
2017-10-20 20:29:49 +02:00
|
|
|
if Message.need_to_render_content(rendered_content, rendered_content_version, bugdown.version):
|
|
|
|
if message is None:
|
|
|
|
# We really shouldn't be rendering objects in this method, but there is
|
|
|
|
# a scenario where we upgrade the version of bugdown and fail to run
|
|
|
|
# management commands to re-render historical messages, and then we
|
|
|
|
# need to have side effects. This method is optimized to not need full
|
|
|
|
# blown ORM objects, but the bugdown renderer is unfortunately highly
|
|
|
|
# coupled to Message, and we also need to persist the new rendered content.
|
|
|
|
# If we don't have a message object passed in, we get one here. The cost
|
|
|
|
# of going to the DB here should be overshadowed by the cost of rendering
|
|
|
|
# and updating the row.
|
|
|
|
# TODO: see #1379 to eliminate bugdown dependencies
|
|
|
|
message = Message.objects.select_related().get(id=message_id)
|
|
|
|
|
|
|
|
assert message is not None # Hint for mypy.
|
|
|
|
# It's unfortunate that we need to have side effects on the message
|
|
|
|
# in some cases.
|
|
|
|
rendered_content = render_markdown(message, content, realm=message.get_realm())
|
|
|
|
message.rendered_content = rendered_content
|
|
|
|
message.rendered_content_version = bugdown.version
|
|
|
|
message.save_rendered_content()
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-10-20 20:29:49 +02:00
|
|
|
if rendered_content is not None:
|
|
|
|
obj['rendered_content'] = rendered_content
|
2016-10-04 15:52:26 +02:00
|
|
|
else:
|
2017-11-03 03:12:25 +01:00
|
|
|
obj['rendered_content'] = ('<p>[Zulip note: Sorry, we could not ' +
|
|
|
|
'understand the formatting of your message]</p>')
|
2016-10-04 15:52:26 +02:00
|
|
|
|
2017-09-25 04:24:13 +02:00
|
|
|
if rendered_content is not None:
|
|
|
|
obj['is_me_message'] = Message.is_status_message(content, rendered_content)
|
|
|
|
else:
|
|
|
|
obj['is_me_message'] = False
|
|
|
|
|
2016-12-06 07:19:34 +01:00
|
|
|
obj['reactions'] = [ReactionDict.build_dict_from_raw_db_row(reaction)
|
|
|
|
for reaction in reactions]
|
2016-10-04 15:52:26 +02:00
|
|
|
return obj
|
|
|
|
|
2017-10-10 20:10:57 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def bulk_hydrate_sender_info(objs: List[Dict[str, Any]]) -> None:
|
2017-10-10 20:10:57 +02:00
|
|
|
|
|
|
|
sender_ids = list({
|
|
|
|
obj['sender_id']
|
|
|
|
for obj in objs
|
|
|
|
})
|
|
|
|
|
|
|
|
if not sender_ids:
|
|
|
|
return
|
|
|
|
|
|
|
|
query = UserProfile.objects.values(
|
|
|
|
'id',
|
|
|
|
'full_name',
|
|
|
|
'short_name',
|
Remove sender-related fields from message cache.
This change makes the cache entries smaller for message
dictionaries. It also ensures we get valid data put into
message dictionaries if, for example, the sender's avatar
changes.
After this change, all of the attributes for a message
sender are only fetched during post-processing with two
exceptions:
* We get sender_id for "free" from the message,
and it's the primary key that we need to figure
out which data to fetch in post-processing.
* We need sender_realm_id to be able to cache topic
links, and a sender's realm id will never change,
so it's not a concern for invalidating cache rows.
All the other attributes are either likely to change (e.g.
sender avatar_version) and/or impact the size of cache
entries more severely than the two small id fields above.
This change should improve our overall system performance
by reducing the amount of memory used by every N message
rows we cache, and typically N will be in the thousands or
so on a large realm.
The other major implication of this change is that when
a user changes their avatar, and then later messages that
the user sent are fetched, all of the fields that go into
computing the avatar url will be pulled from the database,
not from cache.
2017-10-14 18:28:43 +02:00
|
|
|
'email',
|
|
|
|
'realm__string_id',
|
|
|
|
'avatar_source',
|
|
|
|
'avatar_version',
|
|
|
|
'is_mirror_dummy',
|
2017-10-10 20:10:57 +02:00
|
|
|
)
|
|
|
|
|
Remove sender-related fields from message cache.
This change makes the cache entries smaller for message
dictionaries. It also ensures we get valid data put into
message dictionaries if, for example, the sender's avatar
changes.
After this change, all of the attributes for a message
sender are only fetched during post-processing with two
exceptions:
* We get sender_id for "free" from the message,
and it's the primary key that we need to figure
out which data to fetch in post-processing.
* We need sender_realm_id to be able to cache topic
links, and a sender's realm id will never change,
so it's not a concern for invalidating cache rows.
All the other attributes are either likely to change (e.g.
sender avatar_version) and/or impact the size of cache
entries more severely than the two small id fields above.
This change should improve our overall system performance
by reducing the amount of memory used by every N message
rows we cache, and typically N will be in the thousands or
so on a large realm.
The other major implication of this change is that when
a user changes their avatar, and then later messages that
the user sent are fetched, all of the fields that go into
computing the avatar url will be pulled from the database,
not from cache.
2017-10-14 18:28:43 +02:00
|
|
|
rows = query_for_ids(query, sender_ids, 'zerver_userprofile.id')
|
2017-10-10 20:10:57 +02:00
|
|
|
|
|
|
|
sender_dict = {
|
|
|
|
row['id']: row
|
|
|
|
for row in rows
|
|
|
|
}
|
|
|
|
|
|
|
|
for obj in objs:
|
|
|
|
sender_id = obj['sender_id']
|
|
|
|
user_row = sender_dict[sender_id]
|
|
|
|
obj['sender_full_name'] = user_row['full_name']
|
|
|
|
obj['sender_short_name'] = user_row['short_name']
|
Remove sender-related fields from message cache.
This change makes the cache entries smaller for message
dictionaries. It also ensures we get valid data put into
message dictionaries if, for example, the sender's avatar
changes.
After this change, all of the attributes for a message
sender are only fetched during post-processing with two
exceptions:
* We get sender_id for "free" from the message,
and it's the primary key that we need to figure
out which data to fetch in post-processing.
* We need sender_realm_id to be able to cache topic
links, and a sender's realm id will never change,
so it's not a concern for invalidating cache rows.
All the other attributes are either likely to change (e.g.
sender avatar_version) and/or impact the size of cache
entries more severely than the two small id fields above.
This change should improve our overall system performance
by reducing the amount of memory used by every N message
rows we cache, and typically N will be in the thousands or
so on a large realm.
The other major implication of this change is that when
a user changes their avatar, and then later messages that
the user sent are fetched, all of the fields that go into
computing the avatar url will be pulled from the database,
not from cache.
2017-10-14 18:28:43 +02:00
|
|
|
obj['sender_email'] = user_row['email']
|
|
|
|
obj['sender_realm_str'] = user_row['realm__string_id']
|
|
|
|
obj['sender_avatar_source'] = user_row['avatar_source']
|
|
|
|
obj['sender_avatar_version'] = user_row['avatar_version']
|
|
|
|
obj['sender_is_mirror_dummy'] = user_row['is_mirror_dummy']
|
2017-10-10 20:10:57 +02:00
|
|
|
|
2017-10-10 08:12:03 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def hydrate_recipient_info(obj: Dict[str, Any]) -> None:
|
2017-10-10 08:12:03 +02:00
|
|
|
'''
|
|
|
|
This method hyrdrates recipient info with things
|
|
|
|
like full names and emails of senders. Eventually
|
|
|
|
our clients should be able to hyrdrate these fields
|
|
|
|
themselves with info they already have on users.
|
|
|
|
'''
|
|
|
|
|
|
|
|
display_recipient = obj['raw_display_recipient']
|
|
|
|
recipient_type = obj['recipient_type']
|
|
|
|
recipient_type_id = obj['recipient_type_id']
|
|
|
|
sender_is_mirror_dummy = obj['sender_is_mirror_dummy']
|
|
|
|
sender_email = obj['sender_email']
|
|
|
|
sender_full_name = obj['sender_full_name']
|
|
|
|
sender_short_name = obj['sender_short_name']
|
|
|
|
sender_id = obj['sender_id']
|
|
|
|
|
|
|
|
if recipient_type == Recipient.STREAM:
|
|
|
|
display_type = "stream"
|
|
|
|
elif recipient_type in (Recipient.HUDDLE, Recipient.PERSONAL):
|
2018-05-11 01:40:23 +02:00
|
|
|
assert not isinstance(display_recipient, str)
|
2017-10-10 08:12:03 +02:00
|
|
|
display_type = "private"
|
|
|
|
if len(display_recipient) == 1:
|
|
|
|
# add the sender in if this isn't a message between
|
|
|
|
# someone and themself, preserving ordering
|
|
|
|
recip = {'email': sender_email,
|
|
|
|
'full_name': sender_full_name,
|
|
|
|
'short_name': sender_short_name,
|
|
|
|
'id': sender_id,
|
|
|
|
'is_mirror_dummy': sender_is_mirror_dummy}
|
|
|
|
if recip['email'] < display_recipient[0]['email']:
|
|
|
|
display_recipient = [recip, display_recipient[0]]
|
|
|
|
elif recip['email'] > display_recipient[0]['email']:
|
|
|
|
display_recipient = [display_recipient[0], recip]
|
|
|
|
else:
|
|
|
|
raise AssertionError("Invalid recipient type %s" % (recipient_type,))
|
|
|
|
|
|
|
|
obj['display_recipient'] = display_recipient
|
|
|
|
obj['type'] = display_type
|
|
|
|
if obj['type'] == 'stream':
|
|
|
|
obj['stream_id'] = recipient_type_id
|
2016-12-06 07:19:34 +01:00
|
|
|
|
2017-10-14 02:01:20 +02:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def set_sender_avatar(obj: Dict[str, Any], client_gravatar: bool) -> None:
|
2017-10-14 02:01:20 +02:00
|
|
|
sender_id = obj['sender_id']
|
|
|
|
sender_realm_id = obj['sender_realm_id']
|
|
|
|
sender_email = obj['sender_email']
|
|
|
|
sender_avatar_source = obj['sender_avatar_source']
|
|
|
|
sender_avatar_version = obj['sender_avatar_version']
|
|
|
|
|
|
|
|
obj['avatar_url'] = get_avatar_field(
|
|
|
|
user_id=sender_id,
|
|
|
|
realm_id=sender_realm_id,
|
|
|
|
email=sender_email,
|
|
|
|
avatar_source=sender_avatar_source,
|
|
|
|
avatar_version=sender_avatar_version,
|
|
|
|
medium=False,
|
|
|
|
client_gravatar=client_gravatar,
|
|
|
|
)
|
|
|
|
|
2017-11-05 11:37:41 +01:00
|
|
|
class ReactionDict:
|
2016-12-06 07:19:34 +01:00
|
|
|
@staticmethod
|
2017-11-05 11:15:10 +01:00
|
|
|
def build_dict_from_raw_db_row(row: Dict[str, Any]) -> Dict[str, Any]:
|
2017-06-07 14:09:15 +02:00
|
|
|
return {'emoji_name': row['emoji_name'],
|
2017-05-01 07:29:56 +02:00
|
|
|
'emoji_code': row['emoji_code'],
|
|
|
|
'reaction_type': row['reaction_type'],
|
2017-06-07 14:09:15 +02:00
|
|
|
'user': {'email': row['user_profile__email'],
|
|
|
|
'id': row['user_profile__id'],
|
|
|
|
'full_name': row['user_profile__full_name']}}
|
2016-12-06 07:19:34 +01:00
|
|
|
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def access_message(user_profile: UserProfile, message_id: int) -> Tuple[Message, UserMessage]:
|
2016-10-12 02:14:08 +02:00
|
|
|
"""You can access a message by ID in our APIs that either:
|
|
|
|
(1) You received or have previously accessed via starring
|
|
|
|
(aka have a UserMessage row for).
|
|
|
|
(2) Was sent to a public stream in your realm.
|
|
|
|
|
|
|
|
We produce consistent, boring error messages to avoid leaking any
|
|
|
|
information from a security perspective.
|
|
|
|
"""
|
|
|
|
try:
|
|
|
|
message = Message.objects.select_related().get(id=message_id)
|
|
|
|
except Message.DoesNotExist:
|
|
|
|
raise JsonableError(_("Invalid message(s)"))
|
|
|
|
|
|
|
|
try:
|
|
|
|
user_message = UserMessage.objects.select_related().get(user_profile=user_profile,
|
|
|
|
message=message)
|
|
|
|
except UserMessage.DoesNotExist:
|
|
|
|
user_message = None
|
|
|
|
|
|
|
|
if user_message is None:
|
|
|
|
if message.recipient.type != Recipient.STREAM:
|
|
|
|
# You can't access private messages you didn't receive
|
|
|
|
raise JsonableError(_("Invalid message(s)"))
|
|
|
|
stream = Stream.objects.get(id=message.recipient.type_id)
|
|
|
|
if stream.realm != user_profile.realm:
|
|
|
|
# You can't access public stream messages in other realms
|
|
|
|
raise JsonableError(_("Invalid message(s)"))
|
|
|
|
|
2018-04-05 01:12:30 +02:00
|
|
|
if not stream.is_public():
|
|
|
|
if not stream.is_history_public_to_subscribers():
|
|
|
|
# You can't access messages sent to invite-only streams
|
|
|
|
# that you didn't receive
|
|
|
|
raise JsonableError(_("Invalid message(s)"))
|
|
|
|
|
|
|
|
# This stream is an invite-only stream where message
|
|
|
|
# history is available to subscribers. So we check if
|
|
|
|
# you're subscribed.
|
|
|
|
if not Subscription.objects.filter(user_profile=user_profile, active=True,
|
|
|
|
recipient=message.recipient).exists():
|
|
|
|
raise JsonableError(_("Invalid message(s)"))
|
|
|
|
|
|
|
|
# You are subscribed, so let this fall through to the public stream case.
|
|
|
|
else:
|
|
|
|
# Otherwise, the message was sent to a public stream in
|
|
|
|
# your realm, so return the message, user_message pair
|
|
|
|
pass
|
2016-10-12 02:14:08 +02:00
|
|
|
return (message, user_message)
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def render_markdown(message: Message,
|
2018-05-11 01:40:23 +02:00
|
|
|
content: str,
|
2017-11-05 11:15:10 +01:00
|
|
|
realm: Optional[Realm]=None,
|
|
|
|
realm_alert_words: Optional[RealmAlertWords]=None,
|
|
|
|
user_ids: Optional[Set[int]]=None,
|
|
|
|
mention_data: Optional[bugdown.MentionData]=None,
|
2018-05-11 01:40:23 +02:00
|
|
|
email_gateway: Optional[bool]=False) -> str:
|
2016-10-04 18:32:46 +02:00
|
|
|
"""Return HTML for given markdown. Bugdown may add properties to the
|
2017-09-25 09:47:15 +02:00
|
|
|
message object such as `mentions_user_ids`, `mentions_user_group_ids`, and
|
|
|
|
`mentions_wildcard`. These are only on this Django object and are not
|
|
|
|
saved in the database.
|
2016-10-04 18:32:46 +02:00
|
|
|
"""
|
|
|
|
|
2017-09-09 02:50:57 +02:00
|
|
|
if user_ids is None:
|
2017-05-17 21:11:03 +02:00
|
|
|
message_user_ids = set() # type: Set[int]
|
2016-10-04 18:32:46 +02:00
|
|
|
else:
|
2017-09-09 02:50:57 +02:00
|
|
|
message_user_ids = user_ids
|
2016-10-04 18:32:46 +02:00
|
|
|
|
2018-03-18 20:37:47 +01:00
|
|
|
message.mentions_wildcard = False
|
|
|
|
message.mentions_user_ids = set()
|
|
|
|
message.mentions_user_group_ids = set()
|
|
|
|
message.alert_words = set()
|
|
|
|
message.links_for_preview = set()
|
2017-01-05 11:07:18 +01:00
|
|
|
|
2018-03-18 20:37:47 +01:00
|
|
|
if realm is None:
|
|
|
|
realm = message.get_realm()
|
2016-10-04 18:32:46 +02:00
|
|
|
|
2018-05-11 01:40:23 +02:00
|
|
|
possible_words = set() # type: Set[str]
|
2016-10-04 18:32:46 +02:00
|
|
|
if realm_alert_words is not None:
|
|
|
|
for user_id, words in realm_alert_words.items():
|
|
|
|
if user_id in message_user_ids:
|
|
|
|
possible_words.update(set(words))
|
|
|
|
|
2018-03-18 20:37:47 +01:00
|
|
|
sent_by_bot = get_user_profile_by_id(message.sender_id).is_bot
|
2017-02-03 23:21:56 +01:00
|
|
|
|
2016-10-04 18:32:46 +02:00
|
|
|
# DO MAIN WORK HERE -- call bugdown to convert
|
2017-10-24 02:47:09 +02:00
|
|
|
rendered_content = bugdown.convert(
|
|
|
|
content,
|
|
|
|
message=message,
|
|
|
|
message_realm=realm,
|
|
|
|
possible_words=possible_words,
|
|
|
|
sent_by_bot=sent_by_bot,
|
|
|
|
mention_data=mention_data,
|
2017-11-03 12:13:17 +01:00
|
|
|
email_gateway=email_gateway
|
2017-10-24 02:47:09 +02:00
|
|
|
)
|
2016-10-04 18:32:46 +02:00
|
|
|
|
2017-01-05 11:07:18 +01:00
|
|
|
if message is not None:
|
|
|
|
message.user_ids_with_alert_words = set()
|
2016-10-04 18:32:46 +02:00
|
|
|
|
2017-01-05 11:07:18 +01:00
|
|
|
if realm_alert_words is not None:
|
|
|
|
for user_id, words in realm_alert_words.items():
|
|
|
|
if user_id in message_user_ids:
|
|
|
|
if set(words).intersection(message.alert_words):
|
|
|
|
message.user_ids_with_alert_words.add(user_id)
|
2016-10-04 18:32:46 +02:00
|
|
|
|
|
|
|
return rendered_content
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def huddle_users(recipient_id: int) -> str:
|
2017-05-23 03:02:01 +02:00
|
|
|
display_recipient = get_display_recipient_by_id(recipient_id,
|
|
|
|
Recipient.HUDDLE,
|
2018-05-11 01:40:23 +02:00
|
|
|
None) # type: Union[str, List[Dict[str, Any]]]
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2018-05-11 01:40:23 +02:00
|
|
|
# str is for streams.
|
|
|
|
assert not isinstance(display_recipient, str)
|
2017-05-23 03:02:01 +02:00
|
|
|
|
|
|
|
user_ids = [obj['id'] for obj in display_recipient] # type: List[int]
|
|
|
|
user_ids = sorted(user_ids)
|
|
|
|
return ','.join(str(uid) for uid in user_ids)
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def aggregate_message_dict(input_dict: Dict[int, Dict[str, Any]],
|
|
|
|
lookup_fields: List[str],
|
|
|
|
collect_senders: bool) -> List[Dict[str, Any]]:
|
2017-11-02 06:28:31 +01:00
|
|
|
lookup_dict = dict() # type: Dict[Tuple[Any, ...], Dict[str, Any]]
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-10-04 18:13:04 +02:00
|
|
|
'''
|
|
|
|
A concrete example might help explain the inputs here:
|
|
|
|
|
|
|
|
input_dict = {
|
2017-10-05 18:35:34 +02:00
|
|
|
1002: dict(stream_id=5, topic='foo', sender_id=40),
|
|
|
|
1003: dict(stream_id=5, topic='foo', sender_id=41),
|
|
|
|
1004: dict(stream_id=6, topic='baz', sender_id=99),
|
2017-10-04 18:13:04 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
lookup_fields = ['stream_id', 'topic']
|
|
|
|
|
|
|
|
The first time through the loop:
|
2017-10-05 18:35:34 +02:00
|
|
|
attribute_dict = dict(stream_id=5, topic='foo', sender_id=40)
|
|
|
|
lookup_dict = (5, 'foo')
|
2017-10-04 18:13:04 +02:00
|
|
|
|
|
|
|
lookup_dict = {
|
2017-10-05 18:35:34 +02:00
|
|
|
(5, 'foo'): dict(stream_id=5, topic='foo',
|
|
|
|
unread_message_ids=[1002, 1003],
|
|
|
|
sender_ids=[40, 41],
|
|
|
|
),
|
2017-10-04 18:13:04 +02:00
|
|
|
...
|
|
|
|
}
|
|
|
|
|
|
|
|
result = [
|
2017-10-05 18:35:34 +02:00
|
|
|
dict(stream_id=5, topic='foo',
|
|
|
|
unread_message_ids=[1002, 1003],
|
|
|
|
sender_ids=[40, 41],
|
|
|
|
),
|
2017-10-04 18:13:04 +02:00
|
|
|
...
|
|
|
|
]
|
|
|
|
'''
|
|
|
|
|
2017-10-05 18:35:34 +02:00
|
|
|
for message_id, attribute_dict in input_dict.items():
|
2017-10-04 18:13:04 +02:00
|
|
|
lookup_key = tuple([attribute_dict[f] for f in lookup_fields])
|
2017-05-23 03:02:01 +02:00
|
|
|
if lookup_key not in lookup_dict:
|
|
|
|
obj = {}
|
|
|
|
for f in lookup_fields:
|
2017-10-04 18:13:04 +02:00
|
|
|
obj[f] = attribute_dict[f]
|
2017-10-05 18:35:34 +02:00
|
|
|
obj['unread_message_ids'] = []
|
|
|
|
if collect_senders:
|
|
|
|
obj['sender_ids'] = set()
|
2017-05-23 03:02:01 +02:00
|
|
|
lookup_dict[lookup_key] = obj
|
|
|
|
|
2017-10-05 18:35:34 +02:00
|
|
|
bucket = lookup_dict[lookup_key]
|
|
|
|
bucket['unread_message_ids'].append(message_id)
|
|
|
|
if collect_senders:
|
|
|
|
bucket['sender_ids'].add(attribute_dict['sender_id'])
|
2017-10-04 18:13:04 +02:00
|
|
|
|
|
|
|
for dct in lookup_dict.values():
|
2017-10-05 18:35:34 +02:00
|
|
|
dct['unread_message_ids'].sort()
|
|
|
|
if collect_senders:
|
|
|
|
dct['sender_ids'] = sorted(list(dct['sender_ids']))
|
2017-05-23 03:02:01 +02:00
|
|
|
|
|
|
|
sorted_keys = sorted(lookup_dict.keys())
|
|
|
|
|
|
|
|
return [lookup_dict[k] for k in sorted_keys]
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def get_inactive_recipient_ids(user_profile: UserProfile) -> List[int]:
|
2017-10-29 17:11:11 +01:00
|
|
|
rows = get_stream_subscriptions_for_user(user_profile).filter(
|
2017-08-09 02:22:00 +02:00
|
|
|
active=False,
|
|
|
|
).values(
|
|
|
|
'recipient_id'
|
|
|
|
)
|
|
|
|
inactive_recipient_ids = [
|
|
|
|
row['recipient_id']
|
|
|
|
for row in rows]
|
|
|
|
return inactive_recipient_ids
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def get_muted_stream_ids(user_profile: UserProfile) -> List[int]:
|
2017-10-29 17:11:11 +01:00
|
|
|
rows = get_stream_subscriptions_for_user(user_profile).filter(
|
2017-08-23 22:45:50 +02:00
|
|
|
active=True,
|
|
|
|
in_home_view=False,
|
|
|
|
).values(
|
2017-10-05 16:18:13 +02:00
|
|
|
'recipient__type_id'
|
2017-08-23 22:45:50 +02:00
|
|
|
)
|
2017-10-05 16:18:13 +02:00
|
|
|
muted_stream_ids = [
|
|
|
|
row['recipient__type_id']
|
2017-08-23 22:45:50 +02:00
|
|
|
for row in rows]
|
2017-10-05 16:18:13 +02:00
|
|
|
return muted_stream_ids
|
2017-08-23 22:45:50 +02:00
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def get_raw_unread_data(user_profile: UserProfile) -> RawUnreadMessagesResult:
|
2017-08-09 02:22:00 +02:00
|
|
|
|
|
|
|
excluded_recipient_ids = get_inactive_recipient_ids(user_profile)
|
|
|
|
|
2017-05-23 03:02:01 +02:00
|
|
|
user_msgs = UserMessage.objects.filter(
|
|
|
|
user_profile=user_profile
|
2017-08-09 02:22:00 +02:00
|
|
|
).exclude(
|
|
|
|
message__recipient_id__in=excluded_recipient_ids
|
2017-05-23 03:02:01 +02:00
|
|
|
).extra(
|
|
|
|
where=[UserMessage.where_unread()]
|
|
|
|
).values(
|
|
|
|
'message_id',
|
|
|
|
'message__sender_id',
|
|
|
|
'message__subject',
|
|
|
|
'message__recipient_id',
|
|
|
|
'message__recipient__type',
|
|
|
|
'message__recipient__type_id',
|
2017-07-21 20:31:25 +02:00
|
|
|
'flags',
|
2017-08-01 18:28:56 +02:00
|
|
|
).order_by("-message_id")
|
|
|
|
|
|
|
|
# Limit unread messages for performance reasons.
|
|
|
|
user_msgs = list(user_msgs[:MAX_UNREAD_MESSAGES])
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-08-01 18:28:56 +02:00
|
|
|
rows = list(reversed(user_msgs))
|
2017-08-23 22:45:50 +02:00
|
|
|
|
2017-10-05 16:18:13 +02:00
|
|
|
muted_stream_ids = get_muted_stream_ids(user_profile)
|
2017-08-31 23:19:05 +02:00
|
|
|
|
|
|
|
topic_mute_checker = build_topic_mute_checker(user_profile)
|
|
|
|
|
2018-05-11 01:40:23 +02:00
|
|
|
def is_row_muted(stream_id: int, recipient_id: int, topic: str) -> bool:
|
2017-10-05 16:18:13 +02:00
|
|
|
if stream_id in muted_stream_ids:
|
2017-08-31 23:19:05 +02:00
|
|
|
return True
|
|
|
|
|
2017-10-04 18:13:04 +02:00
|
|
|
if topic_mute_checker(recipient_id, topic):
|
2017-08-31 23:19:05 +02:00
|
|
|
return True
|
|
|
|
|
|
|
|
return False
|
|
|
|
|
2017-10-04 18:13:04 +02:00
|
|
|
huddle_cache = {} # type: Dict[int, str]
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def get_huddle_users(recipient_id: int) -> str:
|
2017-10-04 18:13:04 +02:00
|
|
|
if recipient_id in huddle_cache:
|
|
|
|
return huddle_cache[recipient_id]
|
2017-08-23 22:45:50 +02:00
|
|
|
|
2017-10-04 18:13:04 +02:00
|
|
|
user_ids_string = huddle_users(recipient_id)
|
|
|
|
huddle_cache[recipient_id] = user_ids_string
|
|
|
|
return user_ids_string
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-10-04 18:13:04 +02:00
|
|
|
pm_dict = {}
|
|
|
|
stream_dict = {}
|
|
|
|
unmuted_stream_msgs = set()
|
|
|
|
huddle_dict = {}
|
|
|
|
mentions = set()
|
|
|
|
|
|
|
|
for row in rows:
|
|
|
|
message_id = row['message_id']
|
|
|
|
msg_type = row['message__recipient__type']
|
|
|
|
recipient_id = row['message__recipient_id']
|
2017-10-05 18:35:34 +02:00
|
|
|
sender_id = row['message__sender_id']
|
2017-10-04 18:13:04 +02:00
|
|
|
|
|
|
|
if msg_type == Recipient.STREAM:
|
|
|
|
stream_id = row['message__recipient__type_id']
|
|
|
|
topic = row['message__subject']
|
|
|
|
stream_dict[message_id] = dict(
|
|
|
|
stream_id=stream_id,
|
|
|
|
topic=topic,
|
2017-10-05 18:35:34 +02:00
|
|
|
sender_id=sender_id,
|
2017-10-04 18:13:04 +02:00
|
|
|
)
|
2017-10-05 16:18:13 +02:00
|
|
|
if not is_row_muted(stream_id, recipient_id, topic):
|
2017-10-04 18:13:04 +02:00
|
|
|
unmuted_stream_msgs.add(message_id)
|
|
|
|
|
|
|
|
elif msg_type == Recipient.PERSONAL:
|
|
|
|
pm_dict[message_id] = dict(
|
|
|
|
sender_id=sender_id,
|
|
|
|
)
|
|
|
|
|
|
|
|
elif msg_type == Recipient.HUDDLE:
|
|
|
|
user_ids_string = get_huddle_users(recipient_id)
|
|
|
|
huddle_dict[message_id] = dict(
|
|
|
|
user_ids_string=user_ids_string,
|
|
|
|
)
|
|
|
|
|
|
|
|
is_mentioned = (row['flags'] & UserMessage.flags.mentioned) != 0
|
|
|
|
if is_mentioned:
|
|
|
|
mentions.add(message_id)
|
|
|
|
|
|
|
|
return dict(
|
|
|
|
pm_dict=pm_dict,
|
|
|
|
stream_dict=stream_dict,
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
muted_stream_ids=muted_stream_ids,
|
2017-10-04 18:13:04 +02:00
|
|
|
unmuted_stream_msgs=unmuted_stream_msgs,
|
|
|
|
huddle_dict=huddle_dict,
|
|
|
|
mentions=mentions,
|
|
|
|
)
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def aggregate_unread_data(raw_data: RawUnreadMessagesResult) -> UnreadMessagesResult:
|
2017-10-04 18:13:04 +02:00
|
|
|
|
|
|
|
pm_dict = raw_data['pm_dict']
|
|
|
|
stream_dict = raw_data['stream_dict']
|
|
|
|
unmuted_stream_msgs = raw_data['unmuted_stream_msgs']
|
|
|
|
huddle_dict = raw_data['huddle_dict']
|
|
|
|
mentions = list(raw_data['mentions'])
|
|
|
|
|
|
|
|
count = len(pm_dict) + len(unmuted_stream_msgs) + len(huddle_dict)
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-10-05 18:35:34 +02:00
|
|
|
pm_objects = aggregate_message_dict(
|
2017-10-04 18:13:04 +02:00
|
|
|
input_dict=pm_dict,
|
2017-05-23 03:02:01 +02:00
|
|
|
lookup_fields=[
|
|
|
|
'sender_id',
|
|
|
|
],
|
2017-10-05 18:35:34 +02:00
|
|
|
collect_senders=False,
|
2017-05-23 03:02:01 +02:00
|
|
|
)
|
|
|
|
|
2017-10-05 18:35:34 +02:00
|
|
|
stream_objects = aggregate_message_dict(
|
2017-10-04 18:13:04 +02:00
|
|
|
input_dict=stream_dict,
|
2017-05-23 03:02:01 +02:00
|
|
|
lookup_fields=[
|
|
|
|
'stream_id',
|
|
|
|
'topic',
|
|
|
|
],
|
2017-10-05 18:35:34 +02:00
|
|
|
collect_senders=True,
|
2017-05-23 03:02:01 +02:00
|
|
|
)
|
|
|
|
|
2017-10-05 18:35:34 +02:00
|
|
|
huddle_objects = aggregate_message_dict(
|
2017-10-04 18:13:04 +02:00
|
|
|
input_dict=huddle_dict,
|
2017-05-23 03:02:01 +02:00
|
|
|
lookup_fields=[
|
2017-10-04 18:13:04 +02:00
|
|
|
'user_ids_string',
|
2017-05-23 03:02:01 +02:00
|
|
|
],
|
2017-10-05 18:35:34 +02:00
|
|
|
collect_senders=False,
|
2017-05-23 03:02:01 +02:00
|
|
|
)
|
|
|
|
|
|
|
|
result = dict(
|
|
|
|
pms=pm_objects,
|
|
|
|
streams=stream_objects,
|
|
|
|
huddles=huddle_objects,
|
2017-10-04 18:13:04 +02:00
|
|
|
mentions=mentions,
|
2017-08-09 04:01:00 +02:00
|
|
|
count=count) # type: UnreadMessagesResult
|
2017-05-23 03:02:01 +02:00
|
|
|
|
|
|
|
return result
|
|
|
|
|
2017-11-05 11:15:10 +01:00
|
|
|
def apply_unread_message_event(user_profile: UserProfile,
|
|
|
|
state: Dict[str, Any],
|
|
|
|
message: Dict[str, Any],
|
|
|
|
flags: List[str]) -> None:
|
2017-05-23 03:02:01 +02:00
|
|
|
message_id = message['id']
|
|
|
|
if message['type'] == 'stream':
|
|
|
|
message_type = 'stream'
|
|
|
|
elif message['type'] == 'private':
|
|
|
|
others = [
|
|
|
|
recip for recip in message['display_recipient']
|
|
|
|
if recip['id'] != message['sender_id']
|
|
|
|
]
|
|
|
|
if len(others) <= 1:
|
|
|
|
message_type = 'private'
|
|
|
|
else:
|
|
|
|
message_type = 'huddle'
|
2017-08-25 09:39:36 +02:00
|
|
|
else:
|
|
|
|
raise AssertionError("Invalid message type %s" % (message['type'],))
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-10-05 18:35:34 +02:00
|
|
|
sender_id = message['sender_id']
|
|
|
|
|
2017-05-23 03:02:01 +02:00
|
|
|
if message_type == 'stream':
|
|
|
|
stream_id = message['stream_id']
|
|
|
|
topic = message['subject']
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
new_row = dict(
|
2017-05-23 03:02:01 +02:00
|
|
|
stream_id=stream_id,
|
|
|
|
topic=topic,
|
2017-10-05 18:35:34 +02:00
|
|
|
sender_id=sender_id,
|
2017-05-23 03:02:01 +02:00
|
|
|
)
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
state['stream_dict'][message_id] = new_row
|
|
|
|
|
|
|
|
if stream_id not in state['muted_stream_ids']:
|
|
|
|
# This next check hits the database.
|
|
|
|
if not topic_is_muted(user_profile, stream_id, topic):
|
|
|
|
state['unmuted_stream_msgs'].add(message_id)
|
|
|
|
|
2017-05-23 03:02:01 +02:00
|
|
|
elif message_type == 'private':
|
|
|
|
sender_id = message['sender_id']
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
new_row = dict(
|
2017-05-23 03:02:01 +02:00
|
|
|
sender_id=sender_id,
|
|
|
|
)
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
state['pm_dict'][message_id] = new_row
|
|
|
|
|
2017-05-23 03:02:01 +02:00
|
|
|
else:
|
|
|
|
display_recipient = message['display_recipient']
|
|
|
|
user_ids = [obj['id'] for obj in display_recipient]
|
|
|
|
user_ids = sorted(user_ids)
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
user_ids_string = ','.join(str(uid) for uid in user_ids)
|
|
|
|
new_row = dict(
|
|
|
|
user_ids_string=user_ids_string,
|
2017-05-23 03:02:01 +02:00
|
|
|
)
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
state['huddle_dict'][message_id] = new_row
|
2017-05-23 03:02:01 +02:00
|
|
|
|
2017-10-12 01:21:34 +02:00
|
|
|
if 'mentioned' in flags:
|
Simplify how we apply events for unread messages.
The logic to apply events to page_params['unread_msgs'] was
complicated due to the aggregated data structures that we pass
down to the client.
Now we defer the aggregation logic until after we apply the
events. This leads to some simplifications in that codepath,
as well as some performance enhancements.
The intermediate data structure has sets and dictionaries that
generally are keyed by message_id, so most message-related
updates are O(1) in nature.
Also, by waiting to compute the counts until the end, it's a
bit less messy to try to keep track of increments/decrements.
Instead, we just update the dictionaries and sets during the
event-apply phase.
This change also fixes some corner cases:
* We now respect mutes when updating counts.
* For message updates, instead of bluntly updating
the whole topic bucket, we update individual
message ids.
Unfortunately, this change doesn't seem to address the pesky
test that fails sporadically on Travis, related to mention
updates. It will change the symptom, slightly, though.
2017-10-05 00:34:19 +02:00
|
|
|
state['mentions'].add(message_id)
|
2018-01-02 18:33:28 +01:00
|
|
|
|
2018-01-22 21:50:22 +01:00
|
|
|
def estimate_recent_messages(realm: Realm, hours: int) -> int:
|
|
|
|
stat = COUNT_STATS['messages_sent:is_bot:hour']
|
|
|
|
d = timezone_now() - datetime.timedelta(hours=hours)
|
|
|
|
return RealmCount.objects.filter(property=stat.property, end_time__gt=d,
|
|
|
|
realm=realm).aggregate(Sum('value'))['value__sum'] or 0
|
2018-01-04 13:49:39 +01:00
|
|
|
|
2018-01-22 21:50:22 +01:00
|
|
|
def get_first_visible_message_id(realm: Realm) -> int:
|
|
|
|
val = cache_get(realm_first_visible_message_id_cache_key(realm))
|
|
|
|
if val is not None:
|
|
|
|
return val[0]
|
|
|
|
return 0
|
|
|
|
|
|
|
|
def maybe_update_first_visible_message_id(realm: Realm, lookback_hours: int) -> None:
|
|
|
|
cache_empty = cache_get(realm_first_visible_message_id_cache_key(realm)) is None
|
|
|
|
recent_messages_count = estimate_recent_messages(realm, lookback_hours)
|
|
|
|
if realm.message_visibility_limit is not None and (recent_messages_count > 0 or cache_empty):
|
|
|
|
update_first_visible_message_id(realm)
|
|
|
|
|
|
|
|
def update_first_visible_message_id(realm: Realm) -> None:
|
2018-01-04 13:49:39 +01:00
|
|
|
try:
|
2018-01-22 21:50:22 +01:00
|
|
|
first_visible_message_id = Message.objects.filter(sender__realm=realm).values('id').\
|
2018-01-04 13:49:39 +01:00
|
|
|
order_by('-id')[realm.message_visibility_limit - 1]["id"]
|
|
|
|
except IndexError:
|
2018-01-22 21:50:22 +01:00
|
|
|
first_visible_message_id = 0
|
|
|
|
cache_set(realm_first_visible_message_id_cache_key(realm), first_visible_message_id)
|