[GitHub] [flink] zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors
zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors URL: https://github.com/apache/flink/pull/11400#discussion_r39241 ## File path: flink-runtime/src/main/java/org/apache/flink/runtime/io/network/NettyShuffleEnvironment.java ## @@ -217,6 +218,7 @@ public ShuffleIOOwnerContext createShuffleIOOwnerContext( for (InputGateDeploymentDescriptor igdd : inputGateDeploymentDescriptors) { SingleInputGate inputGate = singleInputGateFactory.create( ownerContext.getOwnerName(), + counter, Review comment: I thought you were referring to the unnecessary transformation in the logics before, sorry for misunderstanding. You are right that the current `Collection` actually does not provide any order guarantee which relies on the subclass interface (`List` or `Set`), and we indeed need the determined order after restarting. It is reasonable to adjust the `Collection` to `List` as you suggested, but the only trouble is that it would change the public interface of `ShuffleEnvironment` proposed in FLIP-31. So it might break the compatibility if some outside users already relied on it. I guess it needs to discussion in ML for finalization. Anyway, I already addressed this issue in a separate commit. 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: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [flink] zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors
zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors URL: https://github.com/apache/flink/pull/11400#discussion_r392247611 ## File path: flink-runtime/src/main/java/org/apache/flink/runtime/io/network/NettyShuffleEnvironment.java ## @@ -217,6 +218,7 @@ public ShuffleIOOwnerContext createShuffleIOOwnerContext( for (InputGateDeploymentDescriptor igdd : inputGateDeploymentDescriptors) { SingleInputGate inputGate = singleInputGateFactory.create( ownerContext.getOwnerName(), + counter, Review comment: I guess your suggestion is for unifying all the related data structures (collection, array) from `TaskDeploymentDescriptor.inputGates` to `List`? Actually I do not see any correctness or performance problems ATM. This unification really spans cross multiple components from scheduler to task stack, and I do think it is really necessary to refactor it now. Or if we want to do that, it should be a separate ticket out of this PR scope. We only make use of previous `counter` to give a correct gate index while creating in iteration for this PR motivation. WDYT? 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: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [flink] zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors
zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors URL: https://github.com/apache/flink/pull/11400#discussion_r392117548 ## File path: flink-runtime/src/test/java/org/apache/flink/runtime/io/network/partition/consumer/SingleInputGateTest.java ## @@ -60,10 +61,7 @@ import java.io.IOException; import java.nio.ByteBuffer; import java.nio.ByteOrder; -import java.util.Arrays; -import java.util.HashMap; -import java.util.Map; -import java.util.Optional; +import java.util.*; Review comment: Sure, I was not aware that my local IDE did it automatically. :( 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: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [flink] zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors
zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors URL: https://github.com/apache/flink/pull/11400#discussion_r392122559 ## File path: flink-runtime/src/test/java/org/apache/flink/runtime/io/network/partition/ResultPartitionBuilder.java ## @@ -40,6 +40,8 @@ private BoundedBlockingSubpartitionType blockingSubpartitionType = BoundedBlockingSubpartitionType.AUTO; + private int partitionIndex = 0; Review comment: This builder is only for testing purpose, and I do not think this property is more important or very necessary than other existing fields like `partitionId`, `numberOfSubpartitions` etc. Even I guess this new field is rare used for most of the testing cases, so I prefer to not making it as mandatory. 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: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [flink] zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors
zhijiangW commented on a change in pull request #11400: [FLINK-16586][network] Build ResultSubpartitionInfo and InputChannelInfo in respective constructors URL: https://github.com/apache/flink/pull/11400#discussion_r392117548 ## File path: flink-runtime/src/test/java/org/apache/flink/runtime/io/network/partition/consumer/SingleInputGateTest.java ## @@ -60,10 +61,7 @@ import java.io.IOException; import java.nio.ByteBuffer; import java.nio.ByteOrder; -import java.util.Arrays; -import java.util.HashMap; -import java.util.Map; -import java.util.Optional; +import java.util.*; Review comment: Sure, my local IDE did it automatically. 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: us...@infra.apache.org With regards, Apache Git Services