Re: 看了官方文档的Versioned Table，有一些小疑惑希望可以得到解答

[林影]

好的，明白了~，谢谢解答

Leonard Xu  于2022年8月9日周二 00:12写道：

>
>
> > 2022年8月8日 下午3:34，林影  写道：
> >
> > 先上链接， Versioned Table
> > <
> https://nightlies.apache.org/flink/flink-docs-release-1.15/docs/dev/table/concepts/versioned_tables/
> >
> > 从文档描述中可知，以Upsert-Kafka作为Source，以debezium或canal作为format时，可被认为是Versioned
> > Table Source。
> >
> > 1. 那么flink cdc所提供的connector下，是否也可以被认定为一种Versioned Table Source?
> 可以，cdc 流上定义了 pk 和 watermark就可以作为 versioned table
>
>
> > 2. Versioned Table 在转化成DataStream时，转化后是否必定是一个restract stream？
> 是的，所有cdc流（即changelog流）从SQL API转到Datastream时都是一个retractStream
>
> > 3. 是否所有的Versioned Table，都可以发送往带有撤销能力的sink（如MySQL/ES/Hudi等等）？
>
>
> 是的，只要sink支持回撤（retract），那么sink就支持消费changelog流
>
>
> 祝好，
> Leonard
>
>

Re: 看了官方文档的Versioned Table，有一些小疑惑希望可以得到解答

[Leonard Xu]

> 2022年8月8日 下午3:34，林影  写道：
> 
> 先上链接， Versioned Table
> 
> 从文档描述中可知，以Upsert-Kafka作为Source，以debezium或canal作为format时，可被认为是Versioned
> Table Source。
> 
> 1. 那么flink cdc所提供的connector下，是否也可以被认定为一种Versioned Table Source?
可以，cdc 流上定义了 pk 和 watermark就可以作为 versioned table


> 2. Versioned Table 在转化成DataStream时，转化后是否必定是一个restract stream？
是的，所有cdc流（即changelog流）从SQL API转到Datastream时都是一个retractStream

> 3. 是否所有的Versioned Table，都可以发送往带有撤销能力的sink（如MySQL/ES/Hudi等等）？


是的，只要sink支持回撤（retract），那么sink就支持消费changelog流


祝好，
Leonard

Re: Re:Does flink sql support UDTAGG

[Weihua Hu]

Hi, wang

Maybe you can take a look at
https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/sql/create/#create-function

Best,
Weihua


On Mon, Aug 8, 2022 at 3:52 PM wang <24248...@163.com> wrote:

>
>
>
> Hi,
>
>
> Thanks for your response,  I guess what I need should be this one
> (UDTAGG):
> https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/functions/udfs/#table-aggregate-functions
> As I want multiple rows as aggregate output. So my question: can we use
> UDTAGG in flink SQL?.  If so, is there some guide of how to use UDTAGG in
> flink SQL?  As there are only flink table api instructions of UDTAGG  in
> the page above.
>
>
>
>
> Thanks,
> Hunk
>
>
>
>
>
>
>
>
> At 2022-08-08 10:56:22, "Xuyang"  wrote:
> >Hi, what you want is UDAF? Please check whether this[1] is meet your
> requirement.
> >
> >[1]
> https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/functions/udfs/#aggregate-functions
> >
> >
> >
> >在 2022-08-07 22:06:29，"wang" <24248...@163.com> 写道：
> >
> >Hi dear engineers,
> >
> >
> >One small question:  does flink sql support UDTAGG? (user-defined table
> aggregate function), seems only supported in flink table api? If not
> supported in flink sql, how can I define an aggregated udf which could
> output multiple rows to kafka.
> >
> >
> >Thanks for your help!
> >
> >
> >
> >
> >Regards,
> >Hunk
>

Re:退订

[Howie Yang]

退订邮件需要发送邮件至 user-zh-unsubscr...@flink.apache.org










--

Best,
Howie





At 2022-08-08 17:09:50, "jack zhang"  wrote:
>

退订

[jack zhang]

Re:Re:Does flink sql support UDTAGG

[wang]

Hi,


Thanks for your response,  I guess what I need should be this one (UDTAGG): 
https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/functions/udfs/#table-aggregate-functions
As I want multiple rows as aggregate output. So my question: can we use UDTAGG 
in flink SQL?.  If so, is there some guide of how to use UDTAGG in flink SQL?  
As there are only flink table api instructions of UDTAGG  in the page above.




Thanks,
Hunk








At 2022-08-08 10:56:22, "Xuyang"  wrote:
>Hi, what you want is UDAF? Please check whether this[1] is meet your 
>requirement.
>
>[1] 
>https://nightlies.apache.org/flink/flink-docs-master/docs/dev/table/functions/udfs/#aggregate-functions
>
>
>
>在 2022-08-07 22:06:29，"wang" <24248...@163.com> 写道：
>
>Hi dear engineers,
>
>
>One small question:  does flink sql support UDTAGG? (user-defined table 
>aggregate function), seems only supported in flink table api? If not supported 
>in flink sql, how can I define an aggregated udf which could output multiple 
>rows to kafka.
>
>
>Thanks for your help!
>
>
>
>
>Regards,
>Hunk

看了官方文档的Versioned Table，有一些小疑惑希望可以得到解答

[林影]

先上链接， Versioned Table

从文档描述中可知，以Upsert-Kafka作为Source，以debezium或canal作为format时，可被认为是Versioned
Table Source。

1. 那么flink cdc所提供的connector下，是否也可以被认定为一种Versioned Table Source?
2. Versioned Table 在转化成DataStream时，转化后是否必定是一个restract stream？
3. 是否所有的Versioned Table，都可以发送往带有撤销能力的sink（如MySQL/ES/Hudi等等）？