ahmedabu98 commented on code in PR #31608:
URL: https://github.com/apache/beam/pull/31608#discussion_r1779098644
##########
sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/pubsub/PubsubUnboundedSink.java:
##########
@@ -201,8 +206,15 @@ public void processElement(
break;
}
+ // TODO(sjvanrossum): https://github.com/apache/beam/issues/31828
+ // NOTE: Null and empty ordering keys are treated as equivalent.
@Nullable String topic = dynamicTopicFn.apply(element);
- K key =
keyFunction.apply(ThreadLocalRandom.current().nextInt(numShards), topic);
+ @Nullable String orderingKey = message.getOrderingKey();
+ int shard =
+ Strings.isNullOrEmpty(orderingKey)
+ ? ThreadLocalRandom.current().nextInt(numShards)
+ : Hashing.murmur3_32_fixed().hashString(orderingKey,
StandardCharsets.UTF_8).asInt();
Review Comment:
Makes sense. I see we already group messages into different batches based on
ordering key down in `WriterFn`. In other words, downstream steps don't rely on
bundles being grouped by ordering key.
Do we necessarily need to shard by ordering key here? I wonder if we can
revert this section and keep to `numShards` CC @sjvanrossum
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
