[
https://issues.apache.org/jira/browse/FLINK-24054?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17407231#comment-17407231
]
Timo Walther edited comment on FLINK-24054 at 8/31/21, 10:39 AM:
-----------------------------------------------------------------
In the first version, the test was failing because I did not perform a look up
by key (why should I if it is just an insertion?). We recently exposed
`toChangelogStream` and our changelog contains +I/+U/-U/-D for a reason. We
could have discussed whether we only expose +I/+D when we introduced FLIP-95
but now we should adhere to our own semantics.
was (Author: twalthr):
In the first version, the test was failing because I did not perform a look up
by key (why should I if it is just an insertion?). We recently exposed
`toChangelogStream` and our changelog contains +I/+U/-U/-D for a reason. We
could have discussed whether we only expose only +I/+D when we introduced
FLIP-95 but now we should adhere to our own semantics.
> Let SinkUpsertMaterializer emit +U instead of only +I
> -----------------------------------------------------
>
> Key: FLINK-24054
> URL: https://issues.apache.org/jira/browse/FLINK-24054
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Runtime
> Reporter: Timo Walther
> Assignee: Timo Walther
> Priority: Blocker
> Fix For: 1.14.0
>
>
> Currently, {{SinkUpsertMaterializer}} is not able to emit +U's but will
> always emit +I's. Thus, resulting changelogs are incorrect strictly speaking
> and only valid when treating +U and +I as similar changes in downstream
> operators.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
