diff --git a/synapse/storage/room.py b/synapse/storage/room.py
index 08e13f3a3b..9f67c1862c 100644
--- a/synapse/storage/room.py
+++ b/synapse/storage/room.py
@@ -1,5 +1,6 @@
# -*- coding: utf-8 -*-
# Copyright 2014-2016 OpenMarket Ltd
+# Copyright 2019 The Matrix.org Foundation C.I.C.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@@ -16,14 +17,19 @@
import collections
import logging
import re
+from typing import Optional, Tuple
+
+from six import integer_types
from canonicaljson import json
from twisted.internet import defer
+from synapse.api.constants import EventTypes
from synapse.api.errors import StoreError
from synapse.storage._base import SQLBaseStore
from synapse.storage.search import SearchStore
+from synapse.types import ThirdPartyInstanceID
from synapse.util.caches.descriptors import cached, cachedInlineCallbacks
logger = logging.getLogger(__name__)
@@ -63,103 +69,196 @@ class RoomWorkerStore(SQLBaseStore):
desc="get_public_room_ids",
)
- @cached(num_args=2, max_entries=100)
- def get_public_room_ids_at_stream_id(self, stream_id, network_tuple):
- """Get pulbic rooms for a particular list, or across all lists.
+ def count_public_rooms(self, network_tuple, ignore_non_federatable):
+ """Counts the number of public rooms as tracked in the room_stats_current
+ and room_stats_state table.
Args:
- stream_id (int)
- network_tuple (ThirdPartyInstanceID): The list to use (None, None)
- means the main list, None means all lsits.
+ network_tuple (ThirdPartyInstanceID|None)
+ ignore_non_federatable (bool): If true filters out non-federatable rooms
"""
- return self.runInteraction(
- "get_public_room_ids_at_stream_id",
- self.get_public_room_ids_at_stream_id_txn,
- stream_id,
- network_tuple=network_tuple,
- )
-
- def get_public_room_ids_at_stream_id_txn(self, txn, stream_id, network_tuple):
- return {
- rm
- for rm, vis in self.get_published_at_stream_id_txn(
- txn, stream_id, network_tuple=network_tuple
- ).items()
- if vis
- }
- def get_published_at_stream_id_txn(self, txn, stream_id, network_tuple):
- if network_tuple:
- # We want to get from a particular list. No aggregation required.
+ def _count_public_rooms_txn(txn):
+ query_args = []
+
+ if network_tuple:
+ if network_tuple.appservice_id:
+ published_sql = """
+ SELECT room_id from appservice_room_list
+ WHERE appservice_id = ? AND network_id = ?
+ """
+ query_args.append(network_tuple.appservice_id)
+ query_args.append(network_tuple.network_id)
+ else:
+ published_sql = """
+ SELECT room_id FROM rooms WHERE is_public
+ """
+ else:
+ published_sql = """
+ SELECT room_id FROM rooms WHERE is_public
+ UNION SELECT room_id from appservice_room_list
+ """
sql = """
- SELECT room_id, visibility FROM public_room_list_stream
- INNER JOIN (
- SELECT room_id, max(stream_id) AS stream_id
- FROM public_room_list_stream
- WHERE stream_id <= ? %s
- GROUP BY room_id
- ) grouped USING (room_id, stream_id)
+ SELECT
+ COALESCE(COUNT(*), 0)
+ FROM (
+ %(published_sql)s
+ ) published
+ INNER JOIN room_stats_state USING (room_id)
+ INNER JOIN room_stats_current USING (room_id)
+ WHERE
+ (
+ join_rules = 'public' OR history_visibility = 'world_readable'
+ )
+ AND joined_members > 0
+ """ % {
+ "published_sql": published_sql
+ }
+
+ txn.execute(sql, query_args)
+ return txn.fetchone()[0]
+
+ return self.runInteraction("count_public_rooms", _count_public_rooms_txn)
+
+ @defer.inlineCallbacks
+ def get_largest_public_rooms(
+ self,
+ network_tuple: Optional[ThirdPartyInstanceID],
+ search_filter: Optional[dict],
+ limit: Optional[int],
+ bounds: Optional[Tuple[int, str]],
+ forwards: bool,
+ ignore_non_federatable: bool = False,
+ ):
+ """Gets the largest public rooms (where largest is in terms of joined
+ members, as tracked in the statistics table).
+
+ Args:
+ network_tuple
+ search_filter
+ limit: Maxmimum number of rows to return, unlimited otherwise.
+ bounds: An uppoer or lower bound to apply to result set if given,
+ consists of a joined member count and room_id (these are
+ excluded from result set).
+ forwards: true iff going forwards, going backwards otherwise
+ ignore_non_federatable: If true filters out non-federatable rooms.
+
+ Returns:
+ Rooms in order: biggest number of joined users first.
+ We then arbitrarily use the room_id as a tie breaker.
+
+ """
+
+ where_clauses = []
+ query_args = []
+
+ if network_tuple:
+ if network_tuple.appservice_id:
+ published_sql = """
+ SELECT room_id from appservice_room_list
+ WHERE appservice_id = ? AND network_id = ?
+ """
+ query_args.append(network_tuple.appservice_id)
+ query_args.append(network_tuple.network_id)
+ else:
+ published_sql = """
+ SELECT room_id FROM rooms WHERE is_public
+ """
+ else:
+ published_sql = """
+ SELECT room_id FROM rooms WHERE is_public
+ UNION SELECT room_id from appservice_room_list
"""
- if network_tuple.appservice_id is not None:
- txn.execute(
- sql % ("AND appservice_id = ? AND network_id = ?",),
- (stream_id, network_tuple.appservice_id, network_tuple.network_id),
+ # Work out the bounds if we're given them, these bounds look slightly
+ # odd, but are designed to help query planner use indices by pulling
+ # out a common bound.
+ if bounds:
+ last_joined_members, last_room_id = bounds
+ if forwards:
+ where_clauses.append(
+ """
+ joined_members <= ? AND (
+ joined_members < ? OR room_id < ?
+ )
+ """
)
else:
- txn.execute(sql % ("AND appservice_id IS NULL",), (stream_id,))
- return dict(txn)
- else:
- # We want to get from all lists, so we need to aggregate the results
+ where_clauses.append(
+ """
+ joined_members >= ? AND (
+ joined_members > ? OR room_id > ?
+ )
+ """
+ )
- logger.info("Executing full list")
+ query_args += [last_joined_members, last_joined_members, last_room_id]
- sql = """
- SELECT room_id, visibility
- FROM public_room_list_stream
- INNER JOIN (
- SELECT
- room_id, max(stream_id) AS stream_id, appservice_id,
- network_id
- FROM public_room_list_stream
- WHERE stream_id <= ?
- GROUP BY room_id, appservice_id, network_id
- ) grouped USING (room_id, stream_id)
- """
+ if ignore_non_federatable:
+ where_clauses.append("is_federatable")
- txn.execute(sql, (stream_id,))
+ if search_filter and search_filter.get("generic_search_term", None):
+ search_term = "%" + search_filter["generic_search_term"] + "%"
- results = {}
- # A room is visible if its visible on any list.
- for room_id, visibility in txn:
- results[room_id] = bool(visibility) or results.get(room_id, False)
+ where_clauses.append(
+ """
+ (
+ name LIKE ?
+ OR topic LIKE ?
+ OR canonical_alias LIKE ?
+ )
+ """
+ )
+ query_args += [search_term, search_term, search_term]
+
+ where_clause = ""
+ if where_clauses:
+ where_clause = " AND " + " AND ".join(where_clauses)
+
+ sql = """
+ SELECT
+ room_id, name, topic, canonical_alias, joined_members,
+ avatar, history_visibility, joined_members, guest_access
+ FROM (
+ %(published_sql)s
+ ) published
+ INNER JOIN room_stats_state USING (room_id)
+ INNER JOIN room_stats_current USING (room_id)
+ WHERE
+ (
+ join_rules = 'public' OR history_visibility = 'world_readable'
+ )
+ AND joined_members > 0
+ %(where_clause)s
+ ORDER BY joined_members %(dir)s, room_id %(dir)s
+ """ % {
+ "published_sql": published_sql,
+ "where_clause": where_clause,
+ "dir": "DESC" if forwards else "ASC",
+ }
- return results
+ if limit is not None:
+ query_args.append(limit)
- def get_public_room_changes(self, prev_stream_id, new_stream_id, network_tuple):
- def get_public_room_changes_txn(txn):
- then_rooms = self.get_public_room_ids_at_stream_id_txn(
- txn, prev_stream_id, network_tuple
- )
+ sql += """
+ LIMIT ?
+ """
- now_rooms_dict = self.get_published_at_stream_id_txn(
- txn, new_stream_id, network_tuple
- )
+ def _get_largest_public_rooms_txn(txn):
+ txn.execute(sql, query_args)
- now_rooms_visible = set(rm for rm, vis in now_rooms_dict.items() if vis)
- now_rooms_not_visible = set(
- rm for rm, vis in now_rooms_dict.items() if not vis
- )
+ results = self.cursor_to_dict(txn)
- newly_visible = now_rooms_visible - then_rooms
- newly_unpublished = now_rooms_not_visible & then_rooms
+ if not forwards:
+ results.reverse()
- return newly_visible, newly_unpublished
+ return results
- return self.runInteraction(
- "get_public_room_changes", get_public_room_changes_txn
+ ret_val = yield self.runInteraction(
+ "get_largest_public_rooms", _get_largest_public_rooms_txn
)
+ defer.returnValue(ret_val)
@cached(max_entries=10000)
def is_room_blocked(self, room_id):
@@ -171,6 +270,24 @@ class RoomWorkerStore(SQLBaseStore):
desc="is_room_blocked",
)
+ @defer.inlineCallbacks
+ def is_room_published(self, room_id):
+ """Check whether a room has been published in the local public room
+ directory.
+
+ Args:
+ room_id (str)
+ Returns:
+ bool: Whether the room is currently published in the room directory
+ """
+ # Get room information
+ room_info = yield self.get_room(room_id)
+ if not room_info:
+ defer.returnValue(False)
+
+ # Check the is_public value
+ defer.returnValue(room_info.get("is_public", False))
+
@cachedInlineCallbacks(max_entries=10000)
def get_ratelimit_for_user(self, user_id):
"""Check if there are any overrides for ratelimiting for the given
@@ -200,8 +317,146 @@ class RoomWorkerStore(SQLBaseStore):
else:
return None
+ @cachedInlineCallbacks()
+ def get_retention_policy_for_room(self, room_id):
+ """Get the retention policy for a given room.
+
+ If no retention policy has been found for this room, returns a policy defined
+ by the configured default policy (which has None as both the 'min_lifetime' and
+ the 'max_lifetime' if no default policy has been defined in the server's
+ configuration).
+
+ Args:
+ room_id (str): The ID of the room to get the retention policy of.
+
+ Returns:
+ dict[int, int]: "min_lifetime" and "max_lifetime" for this room.
+ """
+ # If the room retention feature is disabled, return a policy with no minimum nor
+ # maximum, in order not to filter out events we should filter out when sending to
+ # the client.
+ if not self.config.retention_enabled:
+ defer.returnValue({"min_lifetime": None, "max_lifetime": None})
+
+ def get_retention_policy_for_room_txn(txn):
+ txn.execute(
+ """
+ SELECT min_lifetime, max_lifetime FROM room_retention
+ INNER JOIN current_state_events USING (event_id, room_id)
+ WHERE room_id = ?;
+ """,
+ (room_id,),
+ )
+
+ return self.cursor_to_dict(txn)
+
+ ret = yield self.runInteraction(
+ "get_retention_policy_for_room", get_retention_policy_for_room_txn
+ )
+
+ # If we don't know this room ID, ret will be None, in this case return the default
+ # policy.
+ if not ret:
+ defer.returnValue(
+ {
+ "min_lifetime": self.config.retention_default_min_lifetime,
+ "max_lifetime": self.config.retention_default_max_lifetime,
+ }
+ )
+
+ row = ret[0]
+
+ # If one of the room's policy's attributes isn't defined, use the matching
+ # attribute from the default policy.
+ # The default values will be None if no default policy has been defined, or if one
+ # of the attributes is missing from the default policy.
+ if row["min_lifetime"] is None:
+ row["min_lifetime"] = self.config.retention_default_min_lifetime
+
+ if row["max_lifetime"] is None:
+ row["max_lifetime"] = self.config.retention_default_max_lifetime
+
+ defer.returnValue(row)
+
class RoomStore(RoomWorkerStore, SearchStore):
+ def __init__(self, db_conn, hs):
+ super(RoomStore, self).__init__(db_conn, hs)
+
+ self.config = hs.config
+
+ self.register_background_update_handler(
+ "insert_room_retention", self._background_insert_retention
+ )
+
+ @defer.inlineCallbacks
+ def _background_insert_retention(self, progress, batch_size):
+ """Retrieves a list of all rooms within a range and inserts an entry for each of
+ them into the room_retention table.
+ NULLs the property's columns if missing from the retention event in the room's
+ state (or NULLs all of them if there's no retention event in the room's state),
+ so that we fall back to the server's retention policy.
+ """
+
+ last_room = progress.get("room_id", "")
+
+ def _background_insert_retention_txn(txn):
+ txn.execute(
+ """
+ SELECT state.room_id, state.event_id, events.json
+ FROM current_state_events as state
+ LEFT JOIN event_json AS events ON (state.event_id = events.event_id)
+ WHERE state.room_id > ? AND state.type = '%s'
+ ORDER BY state.room_id ASC
+ LIMIT ?;
+ """
+ % EventTypes.Retention,
+ (last_room, batch_size),
+ )
+
+ rows = self.cursor_to_dict(txn)
+
+ if not rows:
+ return True
+
+ for row in rows:
+ if not row["json"]:
+ retention_policy = {}
+ else:
+ ev = json.loads(row["json"])
+ retention_policy = json.dumps(ev["content"])
+
+ self._simple_insert_txn(
+ txn=txn,
+ table="room_retention",
+ values={
+ "room_id": row["room_id"],
+ "event_id": row["event_id"],
+ "min_lifetime": retention_policy.get("min_lifetime"),
+ "max_lifetime": retention_policy.get("max_lifetime"),
+ },
+ )
+
+ logger.info("Inserted %d rows into room_retention", len(rows))
+
+ self._background_update_progress_txn(
+ txn, "insert_room_retention", {"room_id": rows[-1]["room_id"]}
+ )
+
+ if batch_size > len(rows):
+ return True
+ else:
+ return False
+
+ end = yield self.runInteraction(
+ "insert_room_retention", _background_insert_retention_txn
+ )
+
+ if end:
+ yield self._end_background_update("insert_room_retention")
+
+ defer.returnValue(batch_size)
+
@defer.inlineCallbacks
def store_room(self, room_id, room_creator_user_id, is_public):
"""Stores a room.
@@ -402,6 +657,35 @@ class RoomStore(RoomWorkerStore, SearchStore):
txn, event, "content.body", event.content["body"]
)
+ def _store_retention_policy_for_room_txn(self, txn, event):
+ if hasattr(event, "content") and (
+ "min_lifetime" in event.content or "max_lifetime" in event.content
+ ):
+ if (
+ "min_lifetime" in event.content
+ and not isinstance(event.content.get("min_lifetime"), integer_types)
+ ) or (
+ "max_lifetime" in event.content
+ and not isinstance(event.content.get("max_lifetime"), integer_types)
+ ):
+ # Ignore the event if one of the value isn't an integer.
+ return
+
+ self._simple_insert_txn(
+ txn=txn,
+ table="room_retention",
+ values={
+ "room_id": event.room_id,
+ "event_id": event.event_id,
+ "min_lifetime": event.content.get("min_lifetime"),
+ "max_lifetime": event.content.get("max_lifetime"),
+ },
+ )
+
+ self._invalidate_cache_and_stream(
+ txn, self.get_retention_policy_for_room, (event.room_id,)
+ )
+
def add_event_report(
self, room_id, event_id, user_id, reason, content, received_ts
):
@@ -583,3 +867,89 @@ class RoomStore(RoomWorkerStore, SearchStore):
remote_media_mxcs.append((hostname, media_id))
return local_media_mxcs, remote_media_mxcs
+
+ @defer.inlineCallbacks
+ def get_rooms_for_retention_period_in_range(
+ self, min_ms, max_ms, include_null=False
+ ):
+ """Retrieves all of the rooms within the given retention range.
+
+ Optionally includes the rooms which don't have a retention policy.
+
+ Args:
+ min_ms (int|None): Duration in milliseconds that define the lower limit of
+ the range to handle (exclusive). If None, doesn't set a lower limit.
+ max_ms (int|None): Duration in milliseconds that define the upper limit of
+ the range to handle (inclusive). If None, doesn't set an upper limit.
+ include_null (bool): Whether to include rooms which retention policy is NULL
+ in the returned set.
+
+ Returns:
+ dict[str, dict]: The rooms within this range, along with their retention
+ policy. The key is "room_id", and maps to a dict describing the retention
+ policy associated with this room ID. The keys for this nested dict are
+ "min_lifetime" (int|None), and "max_lifetime" (int|None).
+ """
+
+ def get_rooms_for_retention_period_in_range_txn(txn):
+ range_conditions = []
+ args = []
+
+ if min_ms is not None:
+ range_conditions.append("max_lifetime > ?")
+ args.append(min_ms)
+
+ if max_ms is not None:
+ range_conditions.append("max_lifetime <= ?")
+ args.append(max_ms)
+
+ # Do a first query which will retrieve the rooms that have a retention policy
+ # in their current state.
+ sql = """
+ SELECT room_id, min_lifetime, max_lifetime FROM room_retention
+ INNER JOIN current_state_events USING (event_id, room_id)
+ """
+
+ if len(range_conditions):
+ sql += " WHERE (" + " AND ".join(range_conditions) + ")"
+
+ if include_null:
+ sql += " OR max_lifetime IS NULL"
+
+ txn.execute(sql, args)
+
+ rows = self.cursor_to_dict(txn)
+ rooms_dict = {}
+
+ for row in rows:
+ rooms_dict[row["room_id"]] = {
+ "min_lifetime": row["min_lifetime"],
+ "max_lifetime": row["max_lifetime"],
+ }
+
+ if include_null:
+ # If required, do a second query that retrieves all of the rooms we know
+ # of so we can handle rooms with no retention policy.
+ sql = "SELECT DISTINCT room_id FROM current_state_events"
+
+ txn.execute(sql)
+
+ rows = self.cursor_to_dict(txn)
+
+ # If a room isn't already in the dict (i.e. it doesn't have a retention
+ # policy in its state), add it with a null policy.
+ for row in rows:
+ if row["room_id"] not in rooms_dict:
+ rooms_dict[row["room_id"]] = {
+ "min_lifetime": None,
+ "max_lifetime": None,
+ }
+
+ return rooms_dict
+
+ rooms = yield self.runInteraction(
+ "get_rooms_for_retention_period_in_range",
+ get_rooms_for_retention_period_in_range_txn,
+ )
+
+ defer.returnValue(rooms)
|
