Re: Re: parallel distinct union and aggregate support patch

 

From: David Rowley

Date: 2021-07-06 10:48

To: bucoo@sohu.com

CC: David Steele; pgsql-hackers; tgl; Dilip Kumar; Thomas Munro; Tomas Vondra; hlinnaka; robertmhaas; pgsql

Subject: Re: Re: parallel distinct union and aggregate support patch

On Tue, 30 Mar 2021 at 22:33, bucoo@sohu.com <bucoo@sohu.com> wrote:

> I have written a plan with similar functions, It is known that the following two situations do not work well.

 

I read through this thread and also wondered about a Parallel

Partition type operator.  It also seems to me that if it could be done

this way then you could just plug in existing nodes to get Sorting and

Aggregation rather than having to modify existing nodes to get them to

do what you need.

 

From what I've seen looking over the thread, a few people suggested

this and I didn't see anywhere where you responded to them about the

idea.  Just so you're aware, contributing to PostgreSQL is not a case

of throwing code at a wall and seeing which parts stick.  You need to

interact and respond to people reviewing your work. This is especially

true for the people who actually have the authority to merge any of

your work with the main code repo.

 

It seems to me you might be getting off to a bad start and you might

not be aware of this process. So I hope this email will help put you

on track.

 

Some of the people that you've not properly responded to include:

 

Thomas Munro:  PostgreSQL committer. Wrote Parallel Hash Join.

Robert Hass: PostgreSQL committer. Wrote much of the original parallel

query code

Heikki Linnakangas: PostgreSQL committer. Worked on many parts of the

planner and executor. Also works for the company that develops

Greenplum, a massively parallel processing RDBMS, based on Postgres.

 

You might find other information in [1].

 

If I wanted to do what you want to do, I think those 3 people might be

some of the last people I'd pick to ignore questions from! :-)

 

Also, I'd say also copying in Tom Lane randomly when he's not shown

any interest in the patch here is likely not a good way of making

forward progress.  You might find that it might bubble up on his radar

if you start constructively interacting with the people who have

questioned your design.  I'd say that should be your next step.

 

The probability of anyone merging any of your code without properly

discussing the design with the appropriate people are either very

close to zero or actually zero.

 

I hope this email helps you get on track.

 

David

 

[1] https://www.postgresql.org/community/contributors/