boyuanzz commented on a change in pull request #12572:
URL: https://github.com/apache/beam/pull/12572#discussion_r497131119
##########
File path:
sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaIO.java
##########
@@ -1620,14 +1635,43 @@ public void processElement(
CoderRegistry coderRegistry = input.getPipeline().getCoderRegistry();
Coder<K> keyCoder = getKeyCoder(coderRegistry);
Coder<V> valueCoder = getValueCoder(coderRegistry);
- Coder<KafkaRecord<K, V>> outputCoder = KafkaRecordCoder.of(keyCoder,
valueCoder);
- PCollection<KafkaRecord<K, V>> output =
- input.apply(ParDo.of(new ReadFromKafkaDoFn<K,
V>(this))).setCoder(outputCoder);
- // TODO(BEAM-10123): Add CommitOffsetTransform to expansion.
- if (isCommitOffsetEnabled() && !configuredKafkaCommit()) {
- throw new IllegalStateException("Offset committed is not supported
yet");
+ Coder<KafkaRecord<K, V>> recordCoder = KafkaRecordCoder.of(keyCoder,
valueCoder);
+
+ try {
+ PCollection<KV<KafkaSourceDescriptor, KafkaRecord<K, V>>>
outputWithDescriptor =
+ input
+ .apply(ParDo.of(new ReadFromKafkaDoFn<K, V>(this)))
+ .setCoder(
+ KvCoder.of(
+ input
+ .getPipeline()
+ .getSchemaRegistry()
+ .getSchemaCoder(KafkaSourceDescriptor.class),
+ recordCoder));
+ if (isCommitOffsetEnabled() && !configuredKafkaCommit()) {
+ outputWithDescriptor =
+ outputWithDescriptor
+ .apply(Reshuffle.viaRandomKey())
Review comment:
The `Reshuffle` is used as a persistent layer which help us to guarantee
that we will not re-read records priori to the committed offset.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
