tree 3563c226dd6fdbba81c36d780d99258f8108aa44
parent 706427457f405afbce61199fd57fae3afef20e2e
author Edwin Kempin <ekempin@google.com> 1632469071 +0200
committer Edwin Kempin <ekempin@google.com> 1632482797 +0200

PostReview: Send batch event when multiple reviewers are added at once

So far when multiple reviewers are added at once via the PostReview
REST endpoint we've sent out one event per each newly added reviewer.
Now we only send out a single event with all newly added reviewers. This
follows the example of the email notifications, where we also do not
send out one email notification for each newly added reviewers, but only
one batch email at the end.

The event for removing reviewers does not allow to provided multiple
removed reviewers, hence for reviewer removals we continue to send out
one event for each removed reviewer.

Adding CCs and adding/removing reviewers by email doesn't trigger any
event, hence we do not need to do anything for those.

Sending out one batch event instead of single events per reviewers
provides more flexibility for the listeners of the event, as they can
decide to do an operation per reviewer or a batch operation for all
reviewers. An example of such a listener is the code-owners plugin that
posts the paths that are owned by the reviewers as change message. So
far it was forced to post one change message per reviewer, but now it
can post a single change message that contains the owned paths for all
newly added reviewers. This improve the latency of the PostReview REST
endpoint because now we will do only a single NoteDb update instead of
one NoteDb updater per reviewer.

Speaking of the code-owners plugin this will also allow us to post the
change message asynchronously and thus reduce the latency for PostReview
calls (computing the owned paths is rather expensive and can contribute
significantly to the latency of PostReview). Posting the change messages
asynchronously when the event is received once per reviewer would not be
working that well, as it would mean posting a change message for each
new reviewer in parallel, which is prone to fail with LOCK_FAILURES.

Signed-off-by: Edwin Kempin <ekempin@google.com>
Change-Id: I22b82d35d7c4becebdda6ae8b6bc3f67ec75bc96