[
https://issues.apache.org/jira/browse/FLINK-24877?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17442336#comment-17442336
]
Timo Walther commented on FLINK-24877:
--------------------------------------
Alternatively or additionally, we could also make the operators more flexible
and allow time operation even if there is not time attribute if if all sources
are bounded.
> Implicitly add time attributes for VALUES clause
> ------------------------------------------------
>
> Key: FLINK-24877
> URL: https://issues.apache.org/jira/browse/FLINK-24877
> Project: Flink
> Issue Type: Improvement
> Components: Table SQL / API
> Reporter: Timo Walther
> Priority: Major
>
> This might be a big change that we cannot do due to backwards compatibility.
> But I would like to at least log the idea and hear the opinions of others.
> In theory, we could make all top-level timestamp columns of the {{VALUES}}
> clause time attributes. Since VALUES are always bounded, a watermark strategy
> is implicitly added with MAX WATERMARK at the end. This means that any
> timestamp can be used a time attribute.
> Given the following example:
> {code}
> VALUES
> (1, 'name 1', TIMESTAMP '2020-03-08 13:12:11.123', 100, 41, 'payload 1'),
> (2, 'name 2', TIMESTAMP '2020-03-09 13:12:11.123', 101, 42, 'payload 2'),
> (3, 'name 3', TIMESTAMP '2020-03-10 13:12:11.123', 102, 43, 'payload 3'),
> (2, 'name 2', TIMESTAMP '2020-03-11 13:12:11.123', 101, 42, 'payload')
> {code}
> We could define windows or interval joins for testing or examples. It would
> make the starting experience nicer. With UNION ALL a user can have a minimum
> set of records in an otherwise streaming pipeline (if checkpointing finished
> tasks is enabled).
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
