Optimise state resolution

[erikjohnston]

No description provided.

[NegativeMjark]

Maybe call this membership rather than e_type?

[NegativeMjark]

Doesn't this clobber the auth_events argument?

[NegativeMjark]

Maybe the argument should be called auth_event_ids?

[NegativeMjark]

I wonder if state_map should be called state_events_by_id to make it clearer that it's a map from event_id to event.

[NegativeMjark]

Put some comments here to explain what's going on please.
(Also you could be super awful and use the unconflicted_state and the conflicted_state dict to track if you've seen the key before...)
((Which would mean that you could defer constructing the set until you had an actual conflict...))

[NegativeMjark]

for state_set in state_sets[1:]:
  for key, value in state_set.iter_items():
    # Check if there is an unconflicted entry for the state key.
    unconflicted_value = unconflicted_state.get(key)
    if unconflicted_value is None:
        # There isn't an unconflicted entry so check if there is a conflicted entry.
        ls = conflicted_state.get(key)
        if ls is None:
          # There wasn't a conflicted entry so haven't seen this key before.
          # Therefore it isn't conflicted yet.
          unconflicted_state[key] = value
        else:
          # This key is already conflicted, add our value to the conflict set.
          ls.add(value)
    elif unconflicted_value != value:
        # If the unconflicted value is not the same as our value then we have a new conflict.
        # So move the key from the unconflicted_state to the conflicted state.
        conflicted_state[key] = {value, unconflicted_value}
        unconflicted_state.pop(key, None)

[NegativeMjark]

LGTM