Flink SQL ordering issues ?

[Etienne Chauchot]

Hi all,

I've been running TPCDS benchmarks lately and I noticed something: as
global sorted output is not supported (1), if we run a SQL query that
specifies an ORDER clause, then the output of such query is ordered per
partition only which makes the output incorrect.

WDYT ? Should I open a ticket ?

[1]
https://ci.apache.org/projects/flink/flink-docs-master/docs/dev/dataset/overview/#locally-sorted-output 


Best

Etienne Chauchot

[Jark Wu-2]

Hi Etienne,

AFAIK, only blink planner (and batch mode) support TPCDS benchmarks
and blink planner does support global ORDER BY.

The docs link you mentioned above refers to the DataSet API,
blink planner implements batch mode on DataStream (actually low-level
StreamOperator)
instead of DataSet API.

Best,
Jark

On Thu, 22 Apr 2021 at 22:26, Etienne Chauchot <[hidden email]> wrote:

> Hi all,
>
> I've been running TPCDS benchmarks lately and I noticed something: as
> global sorted output is not supported (1), if we run a SQL query that
> specifies an ORDER clause, then the output of such query is ordered per
> partition only which makes the output incorrect.
>
> WDYT ? Should I open a ticket ?
>
> [1]
>
> https://ci.apache.org/projects/flink/flink-docs-master/docs/dev/dataset/overview/#locally-sorted-output
>
>
> Best
>
> Etienne Chauchot
>
>