Re: WIP: Upper planner pathification

Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> writes:
> On latest commit-51c0f63e, I am seeing some issues w.r.t parallel query.
> Consider a below case:

> create table t1(c1 int, c2 char(1000));
> insert into t1 values(generate_series(1,300000),'aaaa');
> analyze t1;
> set max_parallel_degree=2;
> postgres=# explain select c1, count(c1) from t1 where c1 < 1000 group by c1;
> ERROR: ORDER/GROUP BY expression not found in targetlist

Hm. That does not speak well for the coverage of the "run the regression
tests with force_parallel_mode enabled" testing approach.

> Without setting max_parallel_degree, it works fine and generate the
> appropriate results. Here the issue seems to be that the code in
> grouping_planner doesn't apply the required PathTarget to Path below Gather
> Path due to which when we generate target list for scan plan,

Yeah, fixed. I had assumed that the existing coding in create_gather_plan
was OK, because it looked like it was right for a non-projecting node.
But actually Gather can project (why though?), so it's not right.

> Approach-2
> ------------------
> Always generate a tlist in Gather plan as we do for many other cases. I
> think this approach will also resolve the issue but haven't tried yet.

I think this is the correct fix.

regards, tom lane