[
https://issues.apache.org/jira/browse/FLINK-1127?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14599466#comment-14599466
]
ASF GitHub Bot commented on FLINK-1127:
---------------------------------------
Github user thvasilo commented on the pull request:
https://github.com/apache/flink/pull/861#issuecomment-114881148
So in order to make this a feature that we can merge, we should add some
functionality that goes beyond the decision tree use case.
My thoughts were that we use the continuous histogram code in order to add
a pimp-your-class style histogram(numBuckets: Int) function to a
DataSet[Double].
This will cover in part
[FLINK-1127](https://issues.apache.org/jira/browse/FLINK-1127). The design
should be similar to
[this](https://github.com/apache/flink/pull/832/files#diff-04a913384132b325b844f41b9545b2cf)
pending PR, where the histogram function would be added to DataSetUtils, for
DataSets[Double]. How does that sound?
> Utility library
> ---------------
>
> Key: FLINK-1127
> URL: https://issues.apache.org/jira/browse/FLINK-1127
> Project: Flink
> Issue Type: Improvement
> Components: Java API
> Reporter: Ufuk Celebi
> Priority: Minor
>
> This is for discussion. We should think about adding some utility
> implementations for different use cases. Data sets of type
> {{DataSet<Double>}} (like histogram) for example are open to further utility
> methods, which don't make sense in general {{DataSet}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
