Re: [HACKERS] PG10 transition tables, wCTEs and multiple operations on the same table - Mailing list pgsql-hackers

From Thomas Munro
Subject Re: [HACKERS] PG10 transition tables, wCTEs and multiple operations on the same table
Date
Msg-id CAEepm=1yC4ov=95n3ujX2qVPNpHwQXAxMNspq+FJpyy8A+-=CQ@mail.gmail.com
Whole thread Raw
In response to [HACKERS] PG10 transition tables, wCTEs and multiple operations on the same table  (Marko Tiikkaja <marko@joh.to>)
Responses Re: [HACKERS] PG10 transition tables, wCTEs and multiple operations on the same table  (Thomas Munro <thomas.munro@enterprisedb.com>)
List pgsql-hackers
On Fri, Jun 2, 2017 at 10:48 PM, Marko Tiikkaja <marko@joh.to> wrote:
> Since the subject of transition tables came up, I thought I'd test how this
> case works:
>
> =# create table qwr(a int);
> CREATE TABLE
>
> =# create function qq() returns trigger as $$ begin raise notice '%',
> (select count(*) from oo); return null; end $$ language plpgsql;
> CREATE FUNCTION
>
> =# create trigger xx after insert on qwr referencing new table as oo for
> each statement execute procedure qq();
> CREATE TRIGGER
>
> =# with t as (insert into qwr values (1)) insert into qwr values (2), (3);
> NOTICE:  3
> NOTICE:  3
> INSERT 0 2
>
> to me, this means that it doesn't work.  Surely one of the trigger
> invocations should say 1, and the other 2.  Or was this intentional?

I would have expected that to be handled by separate transition tables
at different query levels, but evidently it isn't.  The following
crashes:

create table table1(a int);
create table table2(a text);

create function trigfunc() returns trigger as $$ begin raise notice
'got: %', (select oo from oo); return null; end $$ language plpgsql;

create trigger trig1 after insert on table1 referencing new table as
oo for each statement execute procedure trigfunc();

create trigger trig2 after insert on table2 referencing new table as
oo for each statement execute procedure trigfunc();

with t as (insert into table1 values (1)) insert into table2 values ('hello');

I've run out of day today but will investigate.

-- 
Thomas Munro
http://www.enterprisedb.com



pgsql-hackers by date:

Previous
From: Amit Kapila
Date:
Subject: Re: [HACKERS] Default Partition for Range
Next
From: Robert Haas
Date:
Subject: Re: [HACKERS] Patch: Add --no-comments to skip COMMENTs with pg_dump