Re: NOT IN subquery optimization

On Sat, Jun 29, 2019 at 4:19 AM Li, Zheng <zhelli(at)amazon(dot)com> wrote:>
> Resending patch v2.2, looks like the previous submission did not get attached to the original thread.
>
> This version fixed an issue that involves CTE. Because we call subquery_planner before deciding whether to proceed with the transformation, we need to setup access to upper level CTEs at this point if the subquery contains any CTE RangeTblEntry.
>
> Also added more test cases of NOT IN accessing CTEs, including recursive CTE. It's nice that CTE can use index now!

Hi Zheng, Jim,

With my Commitfest doozer hat on, I have moved this entry to the
September 'fest. I noticed in passing that it needs to be adjusted
for the new pg_list.h API. It'd be good to get some feedback from
reviewers on these two competing proposals:

https://commitfest.postgresql.org/24/2020/
https://commitfest.postgresql.org/24/2023/

--
Thomas Munro
https://enterprisedb.com