[ 
https://issues.apache.org/jira/browse/BEAM-3247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16267617#comment-16267617
 ] 

ASF GitHub Bot commented on BEAM-3247:
--------------------------------------

jkff commented on a change in pull request #4175: [BEAM-3247] fix Sample.any 
performance
URL: https://github.com/apache/beam/pull/4175#discussion_r153332162
 
 

 ##########
 File path: 
sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/Sample.java
 ##########
 @@ -209,29 +202,67 @@ public void populateDisplayData(DisplayData.Builder 
builder) {
   }
 
   /**
-   * A {@link DoFn} that returns up to limit elements from the side input 
PCollection.
+   * A {@link DoFn} that outputs up to limit elements.
    */
-  private static class SampleAnyDoFn<T> extends DoFn<Void, T> {
-    long limit;
-    final PCollectionView<Iterable<T>> iterableView;
+  private static class SampleAnyDoFn<T> extends DoFn<T, T> {
 
 Review comment:
   Not sure why you say that: views are also per-window, so I think it 
shouldn't matter whether the collection is bounded or unbounded. (though, of 
course, it'll be behaving weirdly in case of multiple trigger firings - see 
also https://issues.apache.org/jira/browse/BEAM-2305, maybe similar issues 
apply here too)

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


> Sample.any memory constraint
> ----------------------------
>
>                 Key: BEAM-3247
>                 URL: https://issues.apache.org/jira/browse/BEAM-3247
>             Project: Beam
>          Issue Type: Improvement
>          Components: sdk-java-core
>    Affects Versions: 2.1.0
>            Reporter: Neville Li
>            Assignee: Neville Li
>            Priority: Minor
>
> Right now {{Sample.any}} converts the collection to an iterable view and take 
> first n in a side input. This may require materializing the entire collection 
> to disk and is potentially inefficient.
> https://github.com/apache/beam/blob/v2.1.0/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/Sample.java#L74
> It can be fixed by applying a truncating `DoFn` first, then a combine into 
> `List<T>` which limits the list size, and finally flattening the list.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to