digest: Eliminate unneeded queries for hot convos.

We can easily keep track of messages by bucket from the
original loop through messages.

tools/linter_lib/custom_check.py

@@ -40,7 +40,6 @@ FILES_WITH_LEGACY_SUBJECT = {
     # TRY TO FIX THESE! If you can't fix them, try to
     # add comments here and/or in the file itself about
     # why sweeping subject is tricky.
-    'zerver/lib/digest.py',
     'zerver/lib/onboarding.py',
     'zerver/lib/stream_topic.py',
 

zerver/lib/digest.py

@@ -16,7 +16,7 @@ from zerver.lib.send_email import send_future_email, FromAddress
 from zerver.lib.url_encoding import encode_stream
 from zerver.models import UserProfile, UserMessage, Recipient, Stream, \
     Subscription, UserActivity, get_active_streams, get_user_profile_by_id, \
-    Realm
+    Realm, Message
 from zerver.context_processors import common_context
 from zerver.lib.queue import queue_json_publish
 from zerver.lib.logging_util import log_to_file
@@ -94,11 +94,14 @@ def gather_hot_conversations(user_profile: UserProfile, stream_ums: QuerySet) ->
     messages = [um.message for um in stream_ums]
 
     conversation_length = defaultdict(int)  # type: Dict[Tuple[int, str], int]
+    conversation_messages = defaultdict(list)  # type: Dict[Tuple[int, str], List[Message]]
     conversation_diversity = defaultdict(set)  # type: Dict[Tuple[int, str], Set[str]]
     for message in messages:
         key = (message.recipient.type_id,
                message.topic_name())
 
+        conversation_messages[key].append(message)
+
         if not message.sent_by_human():
             # Don't include automated messages in the count.
             continue
@@ -131,15 +134,12 @@ def gather_hot_conversations(user_profile: UserProfile, stream_ums: QuerySet) ->
 
     hot_conversation_render_payloads = []
     for h in hot_conversations:
-        stream_id, subject = h
         users = list(conversation_diversity[h])
         count = conversation_length[h]
+        messages = conversation_messages[h]
 
         # We'll display up to 2 messages from the conversation.
-        first_few_messages = [user_message.message for user_message in
-                              stream_ums.filter(
-                                  message__recipient__type_id=stream_id,
-                                  message__subject=subject)[:2]]
+        first_few_messages = messages[:2]
 
         teaser_data = {"participants": users,
                        "count": count - len(first_few_messages),

zerver/tests/test_digest.py

@@ -121,7 +121,7 @@ class TestDigestEmailMessages(ZulipTestCase):
         with queries_captured() as queries:
             handle_digest_email(othello.id, cutoff)
 
-        self.assertTrue(34 <= len(queries) <= 35)
+        self.assertTrue(29 <= len(queries) <= 30)
 
         self.assertEqual(mock_send_future_email.call_count, 1)
         kwargs = mock_send_future_email.call_args[1]