[
https://issues.apache.org/jira/browse/FLINK-2039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14558324#comment-14558324
]
ASF GitHub Bot commented on FLINK-2039:
---------------------------------------
Github user ggevay commented on the pull request:
https://github.com/apache/flink/pull/684#issuecomment-105249478
I coded the grouped case and did the refactoring of SumAggregator and
ComparableAggregator to use the new FieldAccessor class to access the
user-specified field. The first commit does the refactoring and the second adds
the median calculation.
As part of the refactoring I also fixed a small bug:
SimpleComparableAggregator.reduce was not handling the byAggregate case.
I also moved the logic of DataStream.getClassAtPos and checkFieldRange to
FieldAccessor.
The refactoring also solves FLINK-2039. (by the first two lines of the
second overload of FieldAccessor.create)
There are three things left to do: deciding where will the median be
placed, creating a Jira, and possibly adding the median to the Scala API. I
will do these after speaking with @mbalassi tomorrow.
> Streaming API's sum(String) does not work for Tuples
> ----------------------------------------------------
>
> Key: FLINK-2039
> URL: https://issues.apache.org/jira/browse/FLINK-2039
> Project: Flink
> Issue Type: Bug
> Components: Streaming
> Reporter: Timo Walther
> Priority: Minor
>
> It is confusing for the user if the following code works:
> {code}
> .window(...)
> .every(...)
> .groupBy("f0")
> .sum(1)
> {code}
> but this does not work:
> {code}
> .window(...)
> .every(...)
> .groupBy("f0")
> .sum("f1")
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
