You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Robert Metzger (Jira)" <ji...@apache.org> on 2022/06/29 07:14:00 UTC
[jira] [Commented] (FLINK-19830) Properly implements processing-time temporal table join
|  |
[Robert
Metzger](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=rmetzger)
**commented** on [
FLINK-19830](https://issues.apache.org/jira/browse/FLINK-19830)
---|---
|
---
| [Re: Properly implements processing-time temporal table
join](https://issues.apache.org/jira/browse/FLINK-19830)
---
|
What's the status of this ticket? The custom event mechanism discussion
mentioned above seem to be stalling since 2019, this doesn't seem to get
completed anytime soon.
Let me re-iterate the reason why processing time temporal joins have not been
implemented yet: The concern is that users might get incorrect results,
because there is currently no mechanism to make sure the build side of the
join is properly populated. E.g. when starting a query using this join there
might be records in the probe side which won't match on the build site,
because the matching record has not been received at the join operator yet.
I agree that this is a concern, but I believe it is something we can clearly
state in the documentation. Also, we have the same limitation with the Event
Time Temporal Join. So why is this a problem with the processing time join,
but not with the event time join?
I personally think the processing time join would be a very nice addition,
because it would allow implementing a lookup join based on Flink state,
instead of relying on an external database.
---
| | [ 
](https://issues.apache.org/jira/browse/FLINK-19830#add-comment "Add Comment")
| [Add Comment](https://issues.apache.org/jira/browse/FLINK-19830#add-comment
"Add Comment")
---|---
| This message was sent by Atlassian Jira (v8.20.10#820010-sha1:ace47f9) | |

---