tillrohrmann commented on a change in pull request #10362:
[FLINK-14792][coordination] Implement TE cluster partition release
URL: https://github.com/apache/flink/pull/10362#discussion_r394483007
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/TaskExecutorPartitionTrackerImpl.java
##########
@@ -76,39 +78,58 @@ public void
promoteJobPartitions(Collection<ResultPartitionID> partitionsToPromo
final Collection<PartitionTrackerEntry<JobID,
TaskExecutorPartitionInfo>> partitionTrackerEntries =
stopTrackingPartitions(partitionsToPromote);
- final Map<TaskExecutorPartitionInfo, Set<ResultPartitionID>>
newClusterPartitions = partitionTrackerEntries.stream()
- .collect(Collectors.groupingBy(
- PartitionTrackerEntry::getMetaInfo,
-
Collectors.mapping(PartitionTrackerEntry::getResultPartitionId,
Collectors.toSet())));
-
- newClusterPartitions.forEach(
- (dataSetMetaInfo, newPartitionEntries) ->
clusterPartitions.compute(dataSetMetaInfo, (ignored, existingPartitions) -> {
- if (existingPartitions == null) {
- return newPartitionEntries;
+ partitionTrackerEntries.forEach(
+ partitionTrackerEntry ->
clusterPartitions.compute(partitionTrackerEntry.getMetaInfo().getIntermediateDataSetId(),
(key, existingEntry) -> {
+ if (existingEntry == null) {
+ final Set<ResultPartitionID> newSet =
new HashSet<>();
+
newSet.add(partitionTrackerEntry.getResultPartitionId());
+ return new PartitionEntry(newSet ,
partitionTrackerEntry.getMetaInfo().getNumberOfPartitions());
} else {
-
existingPartitions.addAll(newPartitionEntries);
- return existingPartitions;
+
existingEntry.addPartition(partitionTrackerEntry.getResultPartitionId());
+ return existingEntry;
}
- }));
+ })
+ );
Review comment:
Maybe this whole block could be simplified via
```
for (PartitionTrackerEntry<JobID, TaskExecutorPartitionInfo>
partitionTrackerEntry : partitionTrackerEntries) {
final TaskExecutorPartitionInfo metaInfo =
partitionTrackerEntry.getMetaInfo();
clusterPartitions.computeIfAbsent(
metaInfo.getIntermediateDataSetId(),
ignored -> new
PartitionEntry(metaInfo.getNumberOfPartitions()))
.addPartition(partitionTrackerEntry.getResultPartitionId());
}
```
If doing computations with side effects I would choose explicit loops
instead of the stream API. For the stream API, computations should be side
effect free.
----------------------------------------------------------------
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]
With regards,
Apache Git Services
