[
https://issues.apache.org/jira/browse/FLINK-6462?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16010212#comment-16010212
]
ASF GitHub Bot commented on FLINK-6462:
---------------------------------------
Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3851#discussion_r116442782
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/plan/rules/dataSet/DataSetAggregateRule.scala
---
@@ -50,6 +51,13 @@ class DataSetAggregateRule
agg.getGroupSets.size() == 1) {
return false
}
+ // check if we have over aggregates
+ val overAggNames =
--- End diff --
I thought about this again.
The optimization rules are not the right place for this check and should
not be modified.
Instead the check should be moved to the Table API validation phase.
SQL queries are automatically checked by Calcite's `SqlValidator`.
> Add requiresOver interface for AggregateFunction
> -------------------------------------------------
>
> Key: FLINK-6462
> URL: https://issues.apache.org/jira/browse/FLINK-6462
> Project: Flink
> Issue Type: Improvement
> Components: Table API & SQL
> Reporter: sunjincheng
> Assignee: sunjincheng
> Fix For: 1.3.0
>
>
> The window function in the standard database only supports the "over" window,
> such as `LAG`,` LEAD`, `FIRST_VALUE`,` LAST_VALUE`. These window functions do
> not apply to `Slide`,` Tumble`, `Session` window. So when the user-defined
> AGG need to be clearly defined. In calcite `SqlAggFunction` will using
> `requiresOver` to distinguish.
> In this JIRA. will deal with this feature.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
