Optimisation for filter_events_for_server

We're better off hashing just the event_id than the whole ((type, state_key),
event_id) tuple - so use a dict instead of a set.

Also, iteritems > items.

2 files changed, 7 insertions, 7 deletions

diff --git a/changelog.d/4017.misc b/changelog.d/4017.misc
new file mode 100644

index 0000000000..b1ceb06560
--- /dev/null
+++ b/changelog.d/4017.misc
@@ -0,0 +1 @@
+Optimisation for serving federation requests
\ No newline at end of file
diff --git a/synapse/visibility.py b/synapse/visibility.py

index d4680863d3..c64ad2144c 100644
--- a/synapse/visibility.py
+++ b/synapse/visibility.py
@@ -324,14 +324,13 @@ def filter_events_for_server(store, server_name, events):
     # server's domain.
     #
     # event_to_state_ids contains lots of duplicates, so it turns out to be
-    # cheaper to build a complete set of unique
-    # ((type, state_key), event_id) tuples, and then filter out the ones we
-    # don't want.
+    # cheaper to build a complete event_id => (type, state_key) dict, and then
+    # filter out the ones we don't want
     #
-    state_key_to_event_id_set = {
-        e
+    event_id_to_state_key = {
+        event_id: key
         for key_to_eid in itervalues(event_to_state_ids)
-        for e in key_to_eid.items()
+        for key, event_id in iteritems(key_to_eid)
     }
 
     def include(typ, state_key):
@@ -346,7 +345,7 @@ def filter_events_for_server(store, server_name, events):
 
     event_map = yield store.get_events([
         e_id
-        for key, e_id in state_key_to_event_id_set
+        for e_id, key in iteritems(event_id_to_state_key)
         if include(key[0], key[1])
     ])