from datetime import datetime from typing import Any, Dict, List, Optional, Tuple import orjson from django.db import connection from django.db.models.query import Q, QuerySet from sqlalchemy import Text from sqlalchemy.sql import ColumnElement, column, func, literal from zerver.lib.request import REQ from zerver.models import Message, Stream, UserMessage, UserProfile # Only use these constants for events. ORIG_TOPIC = "orig_subject" TOPIC_NAME = "subject" TOPIC_LINKS = "topic_links" MATCH_TOPIC = "match_subject" # This constant is actually embedded into # the JSON data for message edit history, # so we'll always need to handle legacy data # unless we do a pretty tricky migration. LEGACY_PREV_TOPIC = "prev_subject" # This constant is pretty closely coupled to the # database, but it's the JSON field. EXPORT_TOPIC_NAME = "subject" """ The following functions are for user-facing APIs where we'll want to support "subject" for a while. """ def get_topic_from_message_info(message_info: Dict[str, Any]) -> str: """ Use this where you are getting dicts that are based off of messages that may come from the outside world, especially from third party APIs and bots. We prefer 'topic' to 'subject' here. We expect at least one field to be present (or the caller must know how to handle KeyError). """ if "topic" in message_info: return message_info["topic"] return message_info["subject"] def REQ_topic() -> Optional[str]: # REQ handlers really return a REQ, but we # lie to make the rest of the type matching work. return REQ( whence="topic", aliases=["subject"], converter=lambda x: x.strip(), default=None, ) """ TRY TO KEEP THIS DIVIDING LINE. Below this line we want to make it so that functions are only using "subject" in the DB sense, and nothing customer facing. """ # This is used in low-level message functions in # zerver/lib/message.py, and it's not user facing. DB_TOPIC_NAME = "subject" MESSAGE__TOPIC = "message__subject" def topic_match_sa(topic_name: str) -> "ColumnElement[bool]": # _sa is short for SQLAlchemy, which we use mostly for # queries that search messages topic_cond = func.upper(column("subject", Text)) == func.upper(literal(topic_name)) return topic_cond def topic_column_sa() -> "ColumnElement[str]": return column("subject", Text) def filter_by_exact_message_topic(query: QuerySet, message: Message) -> QuerySet: topic_name = message.topic_name() return query.filter(subject=topic_name) def filter_by_topic_name_via_message(query: QuerySet, topic_name: str) -> QuerySet: return query.filter(message__subject__iexact=topic_name) def messages_for_topic(stream_recipient_id: int, topic_name: str) -> QuerySet: return Message.objects.filter( recipient_id=stream_recipient_id, subject__iexact=topic_name, ) def save_message_for_edit_use_case(message: Message) -> None: message.save( update_fields=[ TOPIC_NAME, "content", "rendered_content", "rendered_content_version", "last_edit_time", "edit_history", "has_attachment", "has_image", "has_link", "recipient_id", ] ) def user_message_exists_for_topic( user_profile: UserProfile, recipient_id: int, topic_name: str ) -> bool: return UserMessage.objects.filter( user_profile=user_profile, message__recipient_id=recipient_id, message__subject__iexact=topic_name, ).exists() def update_edit_history( message: Message, last_edit_time: datetime, edit_history_event: Dict[str, Any] ) -> None: message.last_edit_time = last_edit_time if message.edit_history is not None: edit_history = orjson.loads(message.edit_history) edit_history.insert(0, edit_history_event) else: edit_history = [edit_history_event] message.edit_history = orjson.dumps(edit_history).decode() def update_messages_for_topic_edit( edited_message: Message, propagate_mode: str, orig_topic_name: str, topic_name: Optional[str], new_stream: Optional[Stream], old_recipient_id: Optional[int], edit_history_event: Dict[str, Any], last_edit_time: datetime, ) -> List[Message]: assert (new_stream and old_recipient_id) or (not new_stream and not old_recipient_id) if old_recipient_id is not None: recipient_id = old_recipient_id else: recipient_id = edited_message.recipient_id propagate_query = Q(recipient_id=recipient_id, subject__iexact=orig_topic_name) if propagate_mode == "change_all": propagate_query = propagate_query & ~Q(id=edited_message.id) if propagate_mode == "change_later": propagate_query = propagate_query & Q(id__gt=edited_message.id) messages = Message.objects.filter(propagate_query).select_related() update_fields = ["edit_history", "last_edit_time"] # Evaluate the query before running the update messages_list = list(messages) # The cached ORM objects are not changed by the upcoming # messages.update(), and the remote cache update (done by the # caller) requires the new value, so we manually update the # objects in addition to sending a bulk query to the database. if new_stream is not None: update_fields.append("recipient") for m in messages_list: m.recipient = new_stream.recipient if topic_name is not None: update_fields.append("subject") for m in messages_list: m.set_topic_name(topic_name) for message in messages_list: update_edit_history(message, last_edit_time, edit_history_event) Message.objects.bulk_update(messages_list, update_fields) return messages_list def generate_topic_history_from_db_rows(rows: List[Tuple[str, int]]) -> List[Dict[str, Any]]: canonical_topic_names: Dict[str, Tuple[int, str]] = {} # Sort rows by max_message_id so that if a topic # has many different casings, we use the most # recent row. rows = sorted(rows, key=lambda tup: tup[1]) for (topic_name, max_message_id) in rows: canonical_name = topic_name.lower() canonical_topic_names[canonical_name] = (max_message_id, topic_name) history = [] for canonical_topic, (max_message_id, topic_name) in canonical_topic_names.items(): history.append( dict(name=topic_name, max_id=max_message_id), ) return sorted(history, key=lambda x: -x["max_id"]) def get_topic_history_for_public_stream(recipient_id: int) -> List[Dict[str, Any]]: cursor = connection.cursor() query = """ SELECT "zerver_message"."subject" as topic, max("zerver_message".id) as max_message_id FROM "zerver_message" WHERE ( "zerver_message"."recipient_id" = %s ) GROUP BY ( "zerver_message"."subject" ) ORDER BY max("zerver_message".id) DESC """ cursor.execute(query, [recipient_id]) rows = cursor.fetchall() cursor.close() return generate_topic_history_from_db_rows(rows) def get_topic_history_for_stream( user_profile: UserProfile, recipient_id: int, public_history: bool ) -> List[Dict[str, Any]]: if public_history: return get_topic_history_for_public_stream(recipient_id) cursor = connection.cursor() query = """ SELECT "zerver_message"."subject" as topic, max("zerver_message".id) as max_message_id FROM "zerver_message" INNER JOIN "zerver_usermessage" ON ( "zerver_usermessage"."message_id" = "zerver_message"."id" ) WHERE ( "zerver_usermessage"."user_profile_id" = %s AND "zerver_message"."recipient_id" = %s ) GROUP BY ( "zerver_message"."subject" ) ORDER BY max("zerver_message".id) DESC """ cursor.execute(query, [user_profile.id, recipient_id]) rows = cursor.fetchall() cursor.close() return generate_topic_history_from_db_rows(rows)