Hi Tomas,
I don't think there is much work left to do here.
Did you have a look at the test case? Did it make sense to you?
And I am sorry. I had another look at this and it seems I was confused (again).
From: Arne Roland
Sent: Monday, April 26, 2021 13:00
To: Tomas Vondra; pgsql-hackers
Subject: Re: PATCH: generate fractional cheapest paths in
generate_orderedappend_path
> I think it should. We have a ParallelAppend node after all.
> I'm not really familiar with the way
> get_cheapest_fractional_path_for_pathkeys is used, but a quick search
> suggests to
> me, that build_minmax_path was thus far the only one using it. And minmax
> paths are never parallel safe anyway. I think that is the reason it doesn't
> do that already.
The whole segment were are talking about obviously assumes
require_parallel_safe is not needed. I wasn't aware that in
set_append_rel_size. And I just realized there is a great comment explaining
why it rightfully does so:
/*
* If any live child is not parallel-safe, treat the whole appendrel
* as not parallel-safe. In future we might be able to generate plans
* in which some children are farmed out to workers while others are
* not; but we don't have that today, so it's a waste to consider
* partial paths anywhere in the appendrel unless it's all safe.
* (Child rels visited before this one will be unmarked in
* set_append_rel_pathlist().)
*/
So afaik we don't need to think further about this.
From: Tomas Vondra <tomas.von...@enterprisedb.com>
Sent: Thursday, June 3, 2021 22:57
To: Zhihong Yu
Cc: Arne Roland; pgsql-hackers
Subject: Re: PATCH: generate fractional cheapest paths in
generate_orderedappend_path
> Actually, there are two comments
>
> /* XXX maybe we should have startup_new_fractional? */
>
> in the patch I posted - I completely forgot about that. But I think
> that's a typo, I think - it should be
>
> /* XXX maybe we should have startup_neq_fractional? */
>
> and the new flag would work similarly to startup_neq_total, i.e. it's
> pointless to add paths where startup == fractional cost.
>
> At least I think that was the idea when I wrote the patch, it way too
> long ago.
> Sorry, I almost forgot about this myself. I only got reminded upon seeing
> that again with different queries/tables.
> Just to be sure I get this correctly: You mean startup_gt_fractional (cost)
> as an additional condition, right?
Could you clarify that for me?
Regards
Arne
