wsry commented on a change in pull request #11877:
URL: https://github.com/apache/flink/pull/11877#discussion_r665830690
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/BoundedBlockingSubpartitionDirectTransferReader.java
##########
@@ -91,10 +91,14 @@ public BufferAndBacklog getNextBuffer() throws IOException {
updateStatistics(current);
- // We simply assume all the data are non-events for batch jobs to
avoid pre-fetching the
- // next header
- Buffer.DataType nextDataType =
- numDataAndEventBuffers > 0 ? Buffer.DataType.DATA_BUFFER :
Buffer.DataType.NONE;
+ // We simply assume all the data except for the last one
(EndOfPartitionEvent)
+ // are non-events for batch jobs to avoid pre-fetching the next header
+ Buffer.DataType nextDataType = Buffer.DataType.NONE;
+ if (numDataBuffers > 0) {
+ nextDataType = Buffer.DataType.DATA_BUFFER;
+ } else if (numDataAndEventBuffers > 0) {
+ nextDataType = Buffer.DataType.EVENT_BUFFER;
+ }
Review comment:
After rethink about it, the first choice can support more events type in
the future and the second choice make the assumption that we only have one
event at the end of the data. Maybe the first choice is better?
--
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]
