diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py
index c24e35362a..0c21ac2c77 100644
--- a/synapse/handlers/sync.py
+++ b/synapse/handlers/sync.py
@@ -1,5 +1,6 @@
# -*- coding: utf-8 -*-
-# Copyright 2015 - 2016 OpenMarket Ltd
+# Copyright 2015, 2016 OpenMarket Ltd
+# Copyright 2018 New Vector Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@@ -416,29 +417,38 @@ class SyncHandler(object):
))
@defer.inlineCallbacks
- def get_state_after_event(self, event):
+ def get_state_after_event(self, event, types=None):
"""
Get the room state after the given event
Args:
event(synapse.events.EventBase): event of interest
-
+ types(list[(str|None, str|None)]|None): List of (type, state_key) tuples
+ which are used to filter the state fetched. If `state_key` is None,
+ all events are returned of the given type. Presence of type of `None`
+ indicates that types not in the list should not be filtered out.
+ May be None, which matches any key.
Returns:
A Deferred map from ((type, state_key)->Event)
"""
- state_ids = yield self.store.get_state_ids_for_event(event.event_id)
+ state_ids = yield self.store.get_state_ids_for_event(event.event_id, types)
if event.is_state():
state_ids = state_ids.copy()
state_ids[(event.type, event.state_key)] = event.event_id
defer.returnValue(state_ids)
@defer.inlineCallbacks
- def get_state_at(self, room_id, stream_position):
+ def get_state_at(self, room_id, stream_position, types=None):
""" Get the room state at a particular stream position
Args:
room_id(str): room for which to get state
stream_position(StreamToken): point at which to get state
+ types(list[(str|None, str|None)]|None): List of (type, state_key) tuples
+ which are used to filter the state fetched. If `state_key` is None,
+ all events are returned of the given type. Presence of type of `None`
+ indicates that types not in the list should not be filtered out.
+ May be None, which matches any key.
Returns:
A Deferred map from ((type, state_key)->Event)
@@ -453,7 +463,7 @@ class SyncHandler(object):
if last_events:
last_event = last_events[-1]
- state = yield self.get_state_after_event(last_event)
+ state = yield self.get_state_after_event(last_event, types)
else:
# no events in this room - so presumably no state
@@ -485,22 +495,60 @@ class SyncHandler(object):
# TODO(mjark) Check for new redactions in the state events.
with Measure(self.clock, "compute_state_delta"):
+
+ types = None
+ member_state_ids = {}
+ lazy_load_members = sync_config.filter_collection.lazy_load_members()
+
+ if lazy_load_members:
+ # We only request state for the members needed to display the
+ # timeline:
+
+ types = [
+ (EventTypes.Member, state_key)
+ for state_key in set(
+ event.sender # FIXME: we also care about invite targets etc.
+ for event in batch.events
+ )
+ ]
+
+ # We can't remove redundant member types at this stage as it has
+ # to be done based on event_id, and we don't have the member
+ # event ids until we've pulled them out of the DB.
+
+ if not types:
+ # an optimisation to stop needlessly trying to calculate
+ # member_state_ids
+ #
+ # XXX: i can't remember what this trying to do. why would
+ # types ever be []? --matthew
+ lazy_load_members = False
+
+ types.append((None, None)) # don't just filter to room members
+
if full_state:
if batch:
current_state_ids = yield self.store.get_state_ids_for_event(
- batch.events[-1].event_id
+ batch.events[-1].event_id, types=types
)
state_ids = yield self.store.get_state_ids_for_event(
- batch.events[0].event_id
+ batch.events[0].event_id, types=types
)
+
else:
current_state_ids = yield self.get_state_at(
- room_id, stream_position=now_token
+ room_id, stream_position=now_token, types=types
)
state_ids = current_state_ids
+ if lazy_load_members:
+ member_state_ids = {
+ t: state_ids[t]
+ for t in state_ids if t[0] == EventTypes.Member
+ }
+
timeline_state = {
(event.type, event.state_key): event.event_id
for event in batch.events if event.is_state()
@@ -509,22 +557,33 @@ class SyncHandler(object):
state_ids = _calculate_state(
timeline_contains=timeline_state,
timeline_start=state_ids,
+ timeline_start_members=member_state_ids,
previous={},
current=current_state_ids,
)
elif batch.limited:
state_at_previous_sync = yield self.get_state_at(
- room_id, stream_position=since_token
+ room_id, stream_position=since_token, types=types
)
current_state_ids = yield self.store.get_state_ids_for_event(
- batch.events[-1].event_id
+ batch.events[-1].event_id, types=types
)
state_at_timeline_start = yield self.store.get_state_ids_for_event(
- batch.events[0].event_id
+ batch.events[0].event_id, types=types
)
+ if lazy_load_members:
+ # TODO: filter out redundant members based on their event_ids
+ # (not mxids) at this point. In practice, limited syncs are
+ # relatively rare so it's not a total disaster to send redundant
+ # members down at this point.
+ member_state_ids = {
+ t: state_at_timeline_start[t]
+ for t in state_at_timeline_start if t[0] == EventTypes.Member
+ }
+
timeline_state = {
(event.type, event.state_key): event.event_id
for event in batch.events if event.is_state()
@@ -533,11 +592,23 @@ class SyncHandler(object):
state_ids = _calculate_state(
timeline_contains=timeline_state,
timeline_start=state_at_timeline_start,
+ timeline_start_members=member_state_ids,
previous=state_at_previous_sync,
current=current_state_ids,
)
else:
state_ids = {}
+ if lazy_load_members:
+ # TODO: filter out redundant members based on their mxids (not their
+ # event_ids) at this point. We know we can do it based on mxid as this
+ # is an non-gappy incremental sync.
+
+ # strip off the (None, None) and filter to just room members
+ types = types[:-1]
+ if types:
+ state_ids = yield self.store.get_state_ids_for_event(
+ batch.events[0].event_id, types=types
+ )
state = {}
if state_ids:
@@ -1448,12 +1519,16 @@ def _action_has_highlight(actions):
return False
-def _calculate_state(timeline_contains, timeline_start, previous, current):
+def _calculate_state(timeline_contains, timeline_start, timeline_start_members,
+ previous, current):
"""Works out what state to include in a sync response.
Args:
timeline_contains (dict): state in the timeline
timeline_start (dict): state at the start of the timeline
+ timeline_start_members (dict): state at the start of the timeline
+ for room members who participate in this chunk of timeline.
+ Should always be a subset of timeline_start.
previous (dict): state at the end of the previous sync (or empty dict
if this is an initial sync)
current (dict): state at the end of the timeline
@@ -1472,11 +1547,12 @@ def _calculate_state(timeline_contains, timeline_start, previous, current):
}
c_ids = set(e for e in current.values())
- tc_ids = set(e for e in timeline_contains.values())
- p_ids = set(e for e in previous.values())
ts_ids = set(e for e in timeline_start.values())
+ tsm_ids = set(e for e in timeline_start_members.values())
+ p_ids = set(e for e in previous.values())
+ tc_ids = set(e for e in timeline_contains.values())
- state_ids = ((c_ids | ts_ids) - p_ids) - tc_ids
+ state_ids = (((c_ids | ts_ids) - p_ids) - tc_ids) | tsm_ids
return {
event_id_to_key[e]: e for e in state_ids
|
