Re: EXPLAIN VERBOSE with parallel Aggregate - Mailing list pgsql-hackers

From David Rowley
Subject Re: EXPLAIN VERBOSE with parallel Aggregate
Date
Msg-id CAKJS1f_b27rxwYheeE+b35OxX-A=XbYFhr3SHrOo5ENOQ9ZDSw@mail.gmail.com
Whole thread Raw
In response to Re: EXPLAIN VERBOSE with parallel Aggregate  (Robert Haas <robertmhaas@gmail.com>)
List pgsql-hackers
On 23 April 2016 at 06:35, Robert Haas <robertmhaas@gmail.com> wrote:
> 2. You're using the term "combine agg", but as far as the EXPLAIN
> ANALYZE output is concerned, that's not a thing.  There is
> PartialAggregate, Aggregate, and FinalizeAggregate.  I think you mean
> FinalizeAggregate when you say "combine aggregate", since we identify
> a node as a FinalizeAggregate by observing that combineStates = true.

I really do mean combine when I say combine. I'm pretty sure that I've
coded everything properly to work even if there was 10 stages of
aggregation, in this case 9 of those would be combine nodes, and only
1 of them a finalize node. I can see why you'd say that at a glace of
explain.c, but it's more relevant that agg->finalizeAggs is true,
giving the combine test is in the else if condition. The test for
combineStates is only there so we don't output Finalize Aggregate for
normal 1 stage aggregates.

I simply don't think FILTER should be shown if combineStates is true.
This means that it will only be shown at the first aggregate node, the
one with combineStates == false. I think this is a good idea due to
the fact that this is the only place where FILTERing is done.

-- David Rowley                   http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services



pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: EXPLAIN VERBOSE with parallel Aggregate
Next
From: Robert Haas
Date:
Subject: Re: [BUGS] Breakage with VACUUM ANALYSE + partitions