[
https://issues.apache.org/jira/browse/FLINK-9870?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16695665#comment-16695665
]
Timo Walther edited comment on FLINK-9870 at 11/22/18 8:40 AM:
---------------------------------------------------------------
[~junz] time attributes also matter for sinks for certain use cases. For
example, in case of event-time, when you write out data to Kafka, a timestamp
is added to the message header. In case of multiple rowtime columns in a row,
the sink needs to know which timestamp to ingest. For processing time, we can
simply ignore this information. This Jira issue might be invalid soon depending
on the outcome of the [design
document|https://docs.google.com/document/d/1Yaxp1UJUFW-peGLt8EIidwKIZEWrrA-pznWLuvaH39Y/edit#heading=h.41fd6rs7b3cf]
that is still in progress and needs a major update from my side.
was (Author: twalthr):
[~junz] time attributes also matter for sinks for certain use cases. For
example, in case of event-time, when you write out data to Kafka, a timestamp
is added to the message header. In case of multiple rowtime columns in a row,
the sink needs to know which timestamp to ingest. For processing time, we can
simply ignore this information. This issue might be invalid soon depending on
the outcome of the [design
document|https://docs.google.com/document/d/1Yaxp1UJUFW-peGLt8EIidwKIZEWrrA-pznWLuvaH39Y/edit#heading=h.41fd6rs7b3cf]
that is still in progress and needs a major update from my side.
> Support field mapping and time attributes for table sinks
> ---------------------------------------------------------
>
> Key: FLINK-9870
> URL: https://issues.apache.org/jira/browse/FLINK-9870
> Project: Flink
> Issue Type: Improvement
> Components: Table API & SQL
> Affects Versions: 1.6.0
> Reporter: Timo Walther
> Assignee: Timo Walther
> Priority: Major
> Fix For: 1.8.0
>
>
> FLINK-7548 reworked the table source design and implemented the interfaces
> {{DefinedFieldMapping}}, {{DefinedProctimeAttribute}}, and
> {{DefinedRowtimeAttributes}}.
> However, these interfaces need to be implemented by table sinks as well in
> order to map a table back into a sink similar how source do it for reading
> input data.
> The current unified sink design assumes that this is possible.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
