[
https://issues.apache.org/jira/browse/FLINK-17397?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17183715#comment-17183715
]
Jingsong Lee commented on FLINK-17397:
--------------------------------------
Conceptually, support {{ScanTableSource}} for processing-time temporal join
looks better.
But, in terms of requirements, what do users want, or what scenarios do we want
to support?
Join a static bounded table, which has no lookup capability.
* If we really map it to "processing-time temporal join". First, we should
deal with batch source and streaming source in a Job. Second, the temporal join
operator needs wait batch source for receiving all static records.
Join a slowly changing bounded table, which has no lookup capability.
* If we really map it to "processing-time temporal join", actually, it is hard
to get change log from a filesystem bounded table, so how to support it?
Optimize it to lookup table source in plan? Send a flag to downstream operator
to tell it that it is time to reload, you should delete all old records?
It seems that we still lack a lot of design and implementation.
> FileSystemTableFactory supports LookupableTableSource
> ------------------------------------------------------
>
> Key: FLINK-17397
> URL: https://issues.apache.org/jira/browse/FLINK-17397
> Project: Flink
> Issue Type: Sub-task
> Components: Connectors / FileSystem
> Reporter: Jingsong Lee
> Priority: Major
> Labels: starter
> Fix For: 1.12.0
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
