[
https://issues.apache.org/jira/browse/FLINK-4832?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15652028#comment-15652028
]
Fabian Hueske commented on FLINK-4832:
--------------------------------------
It is true that Flink does not support {{null}} fields in record types, i.e.,
Flink's Java Tuples and Scala's Tuples.
However, the {{Row}} type used for aggregations does support {{null}} fields.
Have a look at how the {{DataSetValues}} class creates a {{DataSet<Row>}} for a
given set of row values.
This technique and the {{ValuesInputFormat}} can be used to create a DataSet
with a single row with all fields being {{null}}.
> Count/Sum 0 elements
> --------------------
>
> Key: FLINK-4832
> URL: https://issues.apache.org/jira/browse/FLINK-4832
> Project: Flink
> Issue Type: Improvement
> Components: Table API & SQL
> Reporter: Timo Walther
> Assignee: Anton Mushin
>
> Currently, the Table API is unable to count or sum up 0 elements. We should
> improve DataSet aggregations for this. Maybe by union the original DataSet
> with a dummy record or by using a MapPartition function. Coming up with a
> good design for this is also part of this issue.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
