[
https://issues.apache.org/jira/browse/FLINK-18651?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17188185#comment-17188185
]
Timo Walther commented on FLINK-18651:
--------------------------------------
A proper solution is described in FLINK-10211. I don't know if we want close
this issue as a duplicate. A solution is to move the time indicator converter
between logical and physical optimization. I had a branch where I started this
effort for the legacy planner (it was working for almost all operators) but was
interrupted by the Blink merge.
> implicitly cast the time attribute to regular TIMESTAMP type in regular join
> ----------------------------------------------------------------------------
>
> Key: FLINK-18651
> URL: https://issues.apache.org/jira/browse/FLINK-18651
> Project: Flink
> Issue Type: Improvement
> Components: Table SQL / Planner
> Reporter: godfrey he
> Priority: Major
>
> Currently, regular join does not accept rowtime attribute field as input, and
> requires users manually cast the time attribute as regular timestamp. Because
> time attribute will be out-of-order after regular join, and then we can't do
> window aggregate based on the time attribute.
> We can improve it that the planner can implicitly cast the time attribute to
> regular TIMESTAMP type, and throws exception there is an operator (after
> join) depended on time attribute, like window aggregate.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
