> If `order by` is
> not pushed down to the underlying database, then how does Calcite handle
> this operator?
It will generate a plan with sort operator on top of the data source.
> Will it read all data provided by the underlying database
> and enumerate it to sort?
I believe so.
> If so, how to guarantee the performance while a
> large amount of data?
It is not guaranteed. So better to push down sort into data source, if not, you
need to teach the Calcite-based optimizer to push it down.
On 2020/11/03 03:05:10, Xiangwei Wei wrote:
> Hi all,
>
> I'd like to ask some questions about the Calcite adapter. If `order by` is
> not pushed down to the underlying database, then how does Calcite handle
> this operator? Will it read all data provided by the underlying database
> and enumerate it to sort? If so, how to guarantee the performance while a
> large amount of data ?
>
> Sorry to trouble you guys. But look forward to hearing from you :D
>
> --
> Best,
> Xiangwei Wei
>