[jira] [Commented] (FLINK-1001) Aggregate Min/Max return unexpected values.

["Stephan Ewen (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-1001?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14053531#comment-14053531 ] 

Stephan Ewen commented on FLINK-1001:
-------------------------------------

Aggregate.MIN uses the {{java.util.Comparable}} interface to find the min value. Can you make sure that it is implemented correctly?

If you are sure that this is all correct, can you extract a data sample where the min aggregation returns something different than the min reducer? (for us to reproduce the issue)

> Aggregate Min/Max return unexpected values.
> -------------------------------------------
>
>                 Key: FLINK-1001
>                 URL: https://issues.apache.org/jira/browse/FLINK-1001
>             Project: Flink
>          Issue Type: Bug
>            Reporter: Bastian Köcher
>
> I wanted to replace my simple MinReducer
> 	public static class MinSimilarityReducer extends ReduceFunction<ClusterPair> {
> 		@Override
> 		public ClusterPair reduce(ClusterPair value1, ClusterPair value2) throws Exception {
> 			if (value1.getSimilarity().doubleValue() < value2.getSimilarity().doubleValue()) {
> 				return value1;
> 			}
> 			return value2;
> 		}
> 	}
> With Aggregate.Min, but Aggregate.Min delivers a value which should already be removed in my WorkSet.
> If I use my own Reducer it works as expected, but with an Aggregate this doesn't work anymore :/ 



--
This message was sent by Atlassian JIRA
(v6.2#6252)