agora inbox for [email protected]  

help / color / mirror / Atom feed

what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
288+ messages / 10 participants
[nested] [flat]

* what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-01 21:31  Pavel Stehule <[email protected]>
  0 siblings, 1 reply; 288+ messages in thread

From: Pavel Stehule @ 2007-07-01 21:31 UTC (permalink / raw)
  To: pgsql-hackers

Hello

if I understand well, there isn't any difference between local and
global temp tables in postgresql.

I have question. Is correct implementation of global temp in Oracle or
Firebird, where content of glob.temp table is session visible and
metadata of g.t.t is persistent? Standard is unclean and speak more
about local temp tables.

"The materialization of a temporary table does not persist beyond the
end of the SQL-session in which the table was materialized. Temporary
tables are effectively empty at the start of an SQL-session.' -- It
means so temp table exists on the start of session.

What is your opinion about implementation this feature into postgresql
(persistent temp tables)?

Regards
Pavel Stehule



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-01 21:46  Tom Lane <[email protected]>
  parent: Pavel Stehule <[email protected]>
  0 siblings, 2 replies; 288+ messages in thread

From: Tom Lane @ 2007-07-01 21:46 UTC (permalink / raw)
  To: Pavel Stehule <[email protected]>; +Cc: pgsql-hackers

"Pavel Stehule" <[email protected]> writes:
> if I understand well, there isn't any difference between local and
> global temp tables in postgresql.

See the archives; some time ago we determined that the correct reading
of the spec is that global/local determines visibility of temp tables
across modules, but still within a single session.  Since we don't have
modules there is no difference for us.

> I have question. Is correct implementation of global temp in Oracle or
> Firebird, where content of glob.temp table is session visible and
> metadata of g.t.t is persistent?

It's correct per spec.  Whether it's more useful than what we do is
highly debatable --- it forces all sessions to use the same definition
of any given temp table name, which is a bit silly for something that's
supposed to support session-local data.

			regards, tom lane



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-02 05:09  Pavel Stehule <[email protected]>
  parent: Tom Lane <[email protected]>
  1 sibling, 0 replies; 288+ messages in thread

From: Pavel Stehule @ 2007-07-02 05:09 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: pgsql-hackers

> > I have question. Is correct implementation of global temp in Oracle or
> > Firebird, where content of glob.temp table is session visible and
> > metadata of g.t.t is persistent?
>
> It's correct per spec.  Whether it's more useful than what we do is
> highly debatable --- it forces all sessions to use the same definition
> of any given temp table name, which is a bit silly for something that's
> supposed to support session-local data.
>

hmm. ALTER OR DROP is really strange. By contrast others op can be
relative simple - maybe only change in heap_insert and
relationisvisible.

Thenk you

Pavel Stehule



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-02 16:12  Jim Nasby <[email protected]>
  parent: Tom Lane <[email protected]>
  1 sibling, 1 reply; 288+ messages in thread

From: Jim Nasby @ 2007-07-02 16:12 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Pavel Stehule <[email protected]>; pgsql-hackers

On Jul 1, 2007, at 4:46 PM, Tom Lane wrote:
>> I have question. Is correct implementation of global temp in  
>> Oracle or
>> Firebird, where content of glob.temp table is session visible and
>> metadata of g.t.t is persistent?
>
> It's correct per spec.  Whether it's more useful than what we do is
> highly debatable --- it forces all sessions to use the same definition
> of any given temp table name, which is a bit silly for something  
> that's
> supposed to support session-local data.

Would it be possible to support both global and local?

I've often thought that having global temp tables would be a really  
good idea, since it would drastically reduce the need to vacuum  
catalog tables, but I've never looked into what would be required to  
do so.
--
Jim Nasby                                            [email protected]
EnterpriseDB      http://enterprisedb.com      512.569.9461 (cell)





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-02 16:43  Tom Lane <[email protected]>
  parent: Jim Nasby <[email protected]>
  0 siblings, 2 replies; 288+ messages in thread

From: Tom Lane @ 2007-07-02 16:43 UTC (permalink / raw)
  To: Jim Nasby <[email protected]>; +Cc: Pavel Stehule <[email protected]>; pgsql-hackers

Jim Nasby <[email protected]> writes:
> I've often thought that having global temp tables would be a really  
> good idea, since it would drastically reduce the need to vacuum  
> catalog tables,

I rather doubt that.  The most likely implementation would involve
cloning a "template" entry into pg_class.

			regards, tom lane



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-02 18:17  Pavel Stehule <[email protected]>
  parent: Tom Lane <[email protected]>
  1 sibling, 0 replies; 288+ messages in thread

From: Pavel Stehule @ 2007-07-02 18:17 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Jim Nasby <[email protected]>; pgsql-hackers

I

2007/7/2, Tom Lane <[email protected]>:
> Jim Nasby <[email protected]> writes:
> > I've often thought that having global temp tables would be a really
> > good idea, since it would drastically reduce the need to vacuum
> > catalog tables,
>
> I rather doubt that.  The most likely implementation would involve
> cloning a "template" entry into pg_class.
>

I am working on prototype, and cloning of template entry is propably
one possible solution. Every session's clon needs own statistic and
then needs own table oid.

Nice a day
Pavel Stehule



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 00:46  Alvaro Herrera <[email protected]>
  parent: Tom Lane <[email protected]>
  1 sibling, 2 replies; 288+ messages in thread

From: Alvaro Herrera @ 2007-07-03 00:46 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Jim Nasby <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers

Tom Lane escribió:
> Jim Nasby <[email protected]> writes:
> > I've often thought that having global temp tables would be a really  
> > good idea, since it would drastically reduce the need to vacuum  
> > catalog tables,
> 
> I rather doubt that.  The most likely implementation would involve
> cloning a "template" entry into pg_class.

How about a new relkind which causes the table to be located in
PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>

So each backend can have its own copy of the table with the same
relfilenode; there's no need for extra catalog entries.

-- 
Alvaro Herrera       Valdivia, Chile   ICBM: S 39º 49' 18.1", W 73º 13' 56.4"
"La victoria es para quien se atreve a estar solo"



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 01:46  Jaime Casanova <[email protected]>
  parent: Alvaro Herrera <[email protected]>
  1 sibling, 1 reply; 288+ messages in thread

From: Jaime Casanova @ 2007-07-03 01:46 UTC (permalink / raw)
  To: Alvaro Herrera <[email protected]>; +Cc: Tom Lane <[email protected]>; Jim Nasby <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers

On 7/3/07, Alvaro Herrera <[email protected]> wrote:
> Tom Lane escribió:
> > Jim Nasby <[email protected]> writes:
> > > I've often thought that having global temp tables would be a really
> > > good idea, since it would drastically reduce the need to vacuum
> > > catalog tables,
> >
> > I rather doubt that.  The most likely implementation would involve
> > cloning a "template" entry into pg_class.
>
> How about a new relkind which causes the table to be located in
> PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>
>
> So each backend can have its own copy of the table with the same
> relfilenode; there's no need for extra catalog entries.
>

we recently make the path for temp files to be just base/pgsql_tmp or
pg_tblspc/<tblspc_oid>/pgsql_tmp. do we want to complicate things
again?

while not just a new rekind indicating this is a template and not and
actual table. and using that template for creating the actual tables?

-- 
regards,
Jaime Casanova

"Programming today is a race between software engineers striving to
build bigger and better idiot-proof programs and the universe trying
to produce bigger and better idiots.
So far, the universe is winning."
                                       Richard Cook



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 01:59  Gregory Stark <[email protected]>
  parent: Jaime Casanova <[email protected]>
  0 siblings, 1 reply; 288+ messages in thread

From: Gregory Stark @ 2007-07-03 01:59 UTC (permalink / raw)
  To: Jaime Casanova <[email protected]>; +Cc: Alvaro Herrera <[email protected]>; Tom Lane <[email protected]>; Jim Nasby <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers


"Jaime Casanova" <[email protected]> writes:

> while not just a new rekind indicating this is a template and not and
> actual table. and using that template for creating the actual tables?

For precisely the reason stated upthread. That would mean creating and
deleting catalog entries for every transaction. Imagine a busy OLTP system
running hundreds of transactions per second trying to use a temporary table
for intermediate results. Mixing DDL and DML is just as bad an idea behind the
scenes as it is for users.

-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com




^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 03:00  Tom Lane <[email protected]>
  parent: Alvaro Herrera <[email protected]>
  1 sibling, 3 replies; 288+ messages in thread

From: Tom Lane @ 2007-07-03 03:00 UTC (permalink / raw)
  To: Alvaro Herrera <[email protected]>; +Cc: Jim Nasby <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers

Alvaro Herrera <[email protected]> writes:
> Tom Lane escribi�:
>> I rather doubt that.  The most likely implementation would involve
>> cloning a "template" entry into pg_class.

> How about a new relkind which causes the table to be located in
> PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>
> So each backend can have its own copy of the table with the same
> relfilenode; there's no need for extra catalog entries.

Uh-huh.  And what do you do with relpages, reltuples, relfrozenxid, and
pg_statistic entries?  What if one backend wants to TRUNCATE or CLUSTER
its copy (requiring a new relfilenode)?  Where does ALTER TABLE fit into
this?

			regards, tom lane



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 05:08  Pavel Stehule <[email protected]>
  parent: Gregory Stark <[email protected]>
  0 siblings, 1 reply; 288+ messages in thread

From: Pavel Stehule @ 2007-07-03 05:08 UTC (permalink / raw)
  To: Gregory Stark <[email protected]>; +Cc: Jaime Casanova <[email protected]>; Alvaro Herrera <[email protected]>; Tom Lane <[email protected]>; Jim Nasby <[email protected]>; pgsql-hackers

2007/7/3, Gregory Stark <[email protected]>:
>
> "Jaime Casanova" <[email protected]> writes:
>
> > while not just a new rekind indicating this is a template and not and
> > actual table. and using that template for creating the actual tables?
>
> For precisely the reason stated upthread. That would mean creating and
> deleting catalog entries for every transaction. Imagine a busy OLTP system
> running hundreds of transactions per second trying to use a temporary table
> for intermediate results. Mixing DDL and DML is just as bad an idea behind the
> scenes as it is for users.
>

Global temp table can be created from template only when is used. It's
has not negative efect on app which doesn't use it. The benefit of
g.t.t. is simplifycation of stored procedures.

regards
Pavel Stehule



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 05:14  Pavel Stehule <[email protected]>
  parent: Tom Lane <[email protected]>
  2 siblings, 0 replies; 288+ messages in thread

From: Pavel Stehule @ 2007-07-03 05:14 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Alvaro Herrera <[email protected]>; Jim Nasby <[email protected]>; pgsql-hackers

>
> > How about a new relkind which causes the table to be located in
> > PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>
> > So each backend can have its own copy of the table with the same
> > relfilenode; there's no need for extra catalog entries.
>
> Uh-huh.  And what do you do with relpages, reltuples, relfrozenxid, and
> pg_statistic entries?  What if one backend wants to TRUNCATE or CLUSTER
> its copy (requiring a new relfilenode)?  Where does ALTER TABLE fit into
> this?
>

This entries can be teoreticly virtual (in memory). If we have some
memory storage we can use it for it.

nice a day
Pavel Stehule



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 09:24  Gregory Stark <[email protected]>
  parent: Pavel Stehule <[email protected]>
  0 siblings, 1 reply; 288+ messages in thread

From: Gregory Stark @ 2007-07-03 09:24 UTC (permalink / raw)
  To: Pavel Stehule <[email protected]>; +Cc: Jaime Casanova <[email protected]>; Alvaro Herrera <[email protected]>; Tom Lane <[email protected]>; Jim Nasby <[email protected]>; pgsql-hackers

"Pavel Stehule" <[email protected]> writes:

> Global temp table can be created from template only when is used. It's
> has not negative efect on app which doesn't use it. The benefit of
> g.t.t. is simplifycation of stored procedures.

And if it's used in 200 txns/s? Imagine the earlier poster who was looking for
a way to display the count of records matching a search followed by the ten
records on the page without re-executing the search.

-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com




^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 09:59  Gregory Stark <[email protected]>
  parent: Tom Lane <[email protected]>
  2 siblings, 0 replies; 288+ messages in thread

From: Gregory Stark @ 2007-07-03 09:59 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Alvaro Herrera <[email protected]>; Jim Nasby <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers


"Tom Lane" <[email protected]> writes:

> Alvaro Herrera <[email protected]> writes:
>> Tom Lane escribió:
>>> I rather doubt that.  The most likely implementation would involve
>>> cloning a "template" entry into pg_class.
>
>> How about a new relkind which causes the table to be located in
>> PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>
>> So each backend can have its own copy of the table with the same
>> relfilenode; there's no need for extra catalog entries.
>
> Uh-huh.  And what do you do with relpages, reltuples, relfrozenxid, and
> pg_statistic entries?  What if one backend wants to TRUNCATE or CLUSTER
> its copy (requiring a new relfilenode)?  Where does ALTER TABLE fit into
> this?

I would have suggested that when we construct the relcache entry for the table
we substitute a local version of refilenode for the global one. 

None of those sound like hard problems. Certainly it's more invasive this way
but the other way is just a hack for complying with the letter of the spec
without actually making it work right. It would be silly and in many use
cases useless to have regular DML operating on data which has no business
being anything but backend-local generate garbage in on-disk catalog tables.

I had a strange thought though. The ideal data structure for local
pg_statistic data in the unlikely case that users analyze their local tables
would in fact be a global temporary table as well. I wonder if we could
bootstrap something similar for pg_class as well.

Incidentally, for what would imagine relfozenxid would be useful for these
tables anyways?

-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com




^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 10:01  Pavel Stehule <[email protected]>
  parent: Gregory Stark <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Pavel Stehule @ 2007-07-03 10:01 UTC (permalink / raw)
  To: Gregory Stark <[email protected]>; +Cc: Jaime Casanova <[email protected]>; Alvaro Herrera <[email protected]>; Tom Lane <[email protected]>; Jim Nasby <[email protected]>; pgsql-hackers

> > Global temp table can be created from template only when is used. It's
> > has not negative efect on app which doesn't use it. The benefit of
> > g.t.t. is simplifycation of stored procedures.
>
> And if it's used in 200 txns/s? Imagine the earlier poster who was looking for
> a way to display the count of records matching a search followed by the ten
> records on the page without re-executing the search.
>

I wrote about comparation global temp tables and current temp tables.

Counting of result's records is problem. I know. It's incompleteness
of current cursor's implementation. Every cursor can be materialised
and then can be counted. We need operation OPEN which matarialise
cursor and returns real row_count.

Regards
Pavel



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 15:49  Bruce Momjian <[email protected]>
  parent: Tom Lane <[email protected]>
  2 siblings, 1 reply; 288+ messages in thread

From: Bruce Momjian @ 2007-07-03 15:49 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Alvaro Herrera <[email protected]>; Jim Nasby <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers

Tom Lane wrote:
> Alvaro Herrera <[email protected]> writes:
> > Tom Lane escribió:
> >> I rather doubt that.  The most likely implementation would involve
> >> cloning a "template" entry into pg_class.
> 
> > How about a new relkind which causes the table to be located in
> > PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>
> > So each backend can have its own copy of the table with the same
> > relfilenode; there's no need for extra catalog entries.
> 
> Uh-huh.  And what do you do with relpages, reltuples, relfrozenxid, and
> pg_statistic entries?  What if one backend wants to TRUNCATE or CLUSTER
> its copy (requiring a new relfilenode)?  Where does ALTER TABLE fit into
> this?

And what is the use-case for this functionality?  What does it give us
that we don't already have?

-- 
  Bruce Momjian  <[email protected]>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-03 20:57  Jim C. Nasby <[email protected]>
  parent: Bruce Momjian <[email protected]>
  0 siblings, 1 reply; 288+ messages in thread

From: Jim C. Nasby @ 2007-07-03 20:57 UTC (permalink / raw)
  To: Bruce Momjian <[email protected]>; +Cc: Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers

On Tue, Jul 03, 2007 at 11:49:05AM -0400, Bruce Momjian wrote:
> Tom Lane wrote:
> > Alvaro Herrera <[email protected]> writes:
> > > Tom Lane escribió:
> > >> I rather doubt that.  The most likely implementation would involve
> > >> cloning a "template" entry into pg_class.
> > 
> > > How about a new relkind which causes the table to be located in
> > > PGDATA/base/<dboid>/pg_temp_<backendid>/<relfilenode>
> > > So each backend can have its own copy of the table with the same
> > > relfilenode; there's no need for extra catalog entries.
> > 
> > Uh-huh.  And what do you do with relpages, reltuples, relfrozenxid, and
> > pg_statistic entries?  What if one backend wants to TRUNCATE or CLUSTER
> > its copy (requiring a new relfilenode)?  Where does ALTER TABLE fit into
> > this?
> 
> And what is the use-case for this functionality?  What does it give us
> that we don't already have?

The use case is any system that uses temp tables in an OLTP setting,
which certainly isn't uncommon. The problem is that today (and as well
with a global temp table that is still writing to the catalogs) is that
every OLTP operation that creates or drops a temp table is doing DDL.
At best, that leads to a lot of catalog bloat. Right now, it appears to
also expose some race conditions (we've got a customer that's been bit
by this and we've been able to reproduce some odd behavior in the lab).
-- 
Jim Nasby                                      [email protected]
EnterpriseDB      http://enterprisedb.com      512.569.9461 (cell)


^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-04 03:20  Bruce Momjian <[email protected]>
  parent: Jim C. Nasby <[email protected]>
  0 siblings, 1 reply; 288+ messages in thread

From: Bruce Momjian @ 2007-07-04 03:20 UTC (permalink / raw)
  To: Jim C. Nasby <[email protected]>; +Cc: Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; Pavel Stehule <[email protected]>; pgsql-hackers

> The use case is any system that uses temp tables in an OLTP setting,
> which certainly isn't uncommon. The problem is that today (and as well
> with a global temp table that is still writing to the catalogs) is that
> every OLTP operation that creates or drops a temp table is doing DDL.
> At best, that leads to a lot of catalog bloat. Right now, it appears to
> also expose some race conditions (we've got a customer that's been bit
> by this and we've been able to reproduce some odd behavior in the lab).

The solution is to fix the bloat, not add a work-around.

-- 
  Bruce Momjian  <[email protected]>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-04 07:22  Pavel Stehule <[email protected]>
  parent: Bruce Momjian <[email protected]>
  0 siblings, 2 replies; 288+ messages in thread

From: Pavel Stehule @ 2007-07-04 07:22 UTC (permalink / raw)
  To: Bruce Momjian <[email protected]>; +Cc: Jim C. Nasby <[email protected]>; Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers

2007/7/4, Bruce Momjian <[email protected]>:
> > The use case is any system that uses temp tables in an OLTP setting,
> > which certainly isn't uncommon. The problem is that today (and as well
> > with a global temp table that is still writing to the catalogs) is that
> > every OLTP operation that creates or drops a temp table is doing DDL.
> > At best, that leads to a lot of catalog bloat. Right now, it appears to
> > also expose some race conditions (we've got a customer that's been bit
> > by this and we've been able to reproduce some odd behavior in the lab).
>
> The solution is to fix the bloat, not add a work-around.
>

Catalog bloat is one unwanted effect. Second is different behave of
temp tables  than other mayor rdbms, and uncomfortable work with temp
tables in stored procedures. Third argument for implementation of
global temp tables is full support of ANSI SQL,

Regards
Pavel Stehule



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-04 17:18  Bruce Momjian <[email protected]>
  parent: Pavel Stehule <[email protected]>
  1 sibling, 0 replies; 288+ messages in thread

From: Bruce Momjian @ 2007-07-04 17:18 UTC (permalink / raw)
  To: Pavel Stehule <[email protected]>; +Cc: Jim C. Nasby <[email protected]>; Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers

Pavel Stehule wrote:
> 2007/7/4, Bruce Momjian <[email protected]>:
> > > The use case is any system that uses temp tables in an OLTP setting,
> > > which certainly isn't uncommon. The problem is that today (and as well
> > > with a global temp table that is still writing to the catalogs) is that
> > > every OLTP operation that creates or drops a temp table is doing DDL.
> > > At best, that leads to a lot of catalog bloat. Right now, it appears to
> > > also expose some race conditions (we've got a customer that's been bit
> > > by this and we've been able to reproduce some odd behavior in the lab).
> >
> > The solution is to fix the bloat, not add a work-around.
> >
> 
> Catalog bloat is one unwanted effect. Second is different behave of
> temp tables  than other mayor rdbms, and uncomfortable work with temp
> tables in stored procedures. Third argument for implementation of
> global temp tables is full support of ANSI SQL,

OK, so the idea of global temp tables is actually implemented in other
dbmss.  OK.

-- 
  Bruce Momjian  <[email protected]>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-04 21:27  Gregory Stark <[email protected]>
  parent: Pavel Stehule <[email protected]>
  1 sibling, 3 replies; 288+ messages in thread

From: Gregory Stark @ 2007-07-04 21:27 UTC (permalink / raw)
  To: Pavel Stehule <[email protected]>; +Cc: Bruce Momjian <[email protected]>; Jim C. Nasby <[email protected]>; Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers

"Pavel Stehule" <[email protected]> writes:

> 2007/7/4, Bruce Momjian <[email protected]>:
>> > The use case is any system that uses temp tables in an OLTP setting,
>> > which certainly isn't uncommon. The problem is that today (and as well
>> > with a global temp table that is still writing to the catalogs) is that
>> > every OLTP operation that creates or drops a temp table is doing DDL.
>> > At best, that leads to a lot of catalog bloat. Right now, it appears to
>> > also expose some race conditions (we've got a customer that's been bit
>> > by this and we've been able to reproduce some odd behavior in the lab).
>>
>> The solution is to fix the bloat, not add a work-around.

The bloat is a direct consequence of performing DDL in the midst of an OLTP
transaction. And it's not the only consequence either. Off the top of my head
trying to do DDL in an OLTP environment will cause OID inflation, locking
issues, catcache problems, unnecessary prepared query replans, and the list
goes on, what happens to views defined on the temporary tables? Foreign key
references to the temporary tables?

You've got it backwards: addressing the artificially imposed requirement to do
DDL to create new tables for what should be purely DML operations is fixing
the root problem, not a work-around. What would be a work-around is trying to
deal with the consequences as they come up.

> Catalog bloat is one unwanted effect. Second is different behave of
> temp tables  than other mayor rdbms, and uncomfortable work with temp
> tables in stored procedures. Third argument for implementation of
> global temp tables is full support of ANSI SQL,

I think the ANSI concept of temporary tables which are defined once but give
you a fresh empty work-space for each transaction only makes sense if you're
thinking in terms of an OLTP environment. Otherwise you would just go ahead
and do the DDL to create new tables for each query and not worry about the
down-sides.

The advantages of the ANSI temporary tables are all things you would worry
about in an OLTP environment but not a data warehousing environment:

1) Overhead to perform DDL

2) Replanning overhead

3) Security issues of doing DDL at run-time

4) Difficulty structuring code when multiple procedures need the same
   temporary tables but the procedures may be called in different orders for
   different jobs and need different sets of tables.

-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com




^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-05 04:01  Tom Lane <[email protected]>
  parent: Gregory Stark <[email protected]>
  2 siblings, 1 reply; 288+ messages in thread

From: Tom Lane @ 2007-07-05 04:01 UTC (permalink / raw)
  To: Gregory Stark <[email protected]>; +Cc: Pavel Stehule <[email protected]>; Bruce Momjian <[email protected]>; Jim C. Nasby <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers

Gregory Stark <[email protected]> writes:
>> 2007/7/4, Bruce Momjian <[email protected]>:
> The solution is to fix the bloat, not add a work-around.

> The bloat is a direct consequence of performing DDL in the midst of an OLTP
> transaction.

Hardly.  It's a consequence of our current implementation of temp
tables; that does not necessarily imply that we cannot fix it without
an API change.

			regards, tom lane



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-05 06:26  Gregory Stark <[email protected]>
  parent: Tom Lane <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Gregory Stark @ 2007-07-05 06:26 UTC (permalink / raw)
  To: Tom Lane <[email protected]>; +Cc: Pavel Stehule <[email protected]>; Bruce Momjian <[email protected]>; Jim C. Nasby <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers


"Tom Lane" <[email protected]> writes:

> Gregory Stark <[email protected]> writes:
>>> 2007/7/4, Bruce Momjian <[email protected]>:
>> The solution is to fix the bloat, not add a work-around.
>
>> The bloat is a direct consequence of performing DDL in the midst of an OLTP
>> transaction.
>
> Hardly.  It's a consequence of our current implementation of temp
> tables; that does not necessarily imply that we cannot fix it without
> an API change.

Sure, we could change our regular temporary tables to not create new records
in pg_class at all, but I don't think it would make a big difference to DSS
users. And I think for OLTP you would still want all the other advantages the
standard api gives you.

-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com




^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMPTABLES in PostgreSQL
@ 2007-07-05 06:34  Simon Riggs <[email protected]>
  parent: Gregory Stark <[email protected]>
  2 siblings, 0 replies; 288+ messages in thread

From: Simon Riggs @ 2007-07-05 06:34 UTC (permalink / raw)
  To: Gregory Stark <[email protected]>; +Cc: Pavel Stehule <[email protected]>; Bruce Momjian <[email protected]>; Jim C. Nasby <[email protected]>; Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers

On Wed, 2007-07-04 at 22:27 +0100, Gregory Stark wrote:
> "Pavel Stehule" <[email protected]> writes:

> > Catalog bloat is one unwanted effect. Second is different behave of
> > temp tables  than other mayor rdbms, and uncomfortable work with temp
> > tables in stored procedures. Third argument for implementation of
> > global temp tables is full support of ANSI SQL,
> 
> I think the ANSI concept of temporary tables which are defined once but give
> you a fresh empty work-space for each transaction only makes sense if you're
> thinking in terms of an OLTP environment. Otherwise you would just go ahead
> and do the DDL to create new tables for each query and not worry about the
> down-sides.
> 
> The advantages of the ANSI temporary tables are all things you would worry
> about in an OLTP environment but not a data warehousing environment:

IIRC there were similar problems with temp table usage at many DW sites
using Teradata. The issue was about locking, specifically the access
rights required. We might have that problem, or not, but the issues
related to significant numbers of temp tables effect many types of
application., not just OLTP.
-- 
  Simon Riggs
  EnterpriseDB  http://www.enterprisedb.com




^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* Re: what is difference between LOCAL and GLOBAL TEMP TABLES in PostgreSQL
@ 2007-07-18 00:16  Bruce Momjian <[email protected]>
  parent: Gregory Stark <[email protected]>
  2 siblings, 0 replies; 288+ messages in thread

From: Bruce Momjian @ 2007-07-18 00:16 UTC (permalink / raw)
  To: Gregory Stark <[email protected]>; +Cc: Pavel Stehule <[email protected]>; Jim C. Nasby <[email protected]>; Tom Lane <[email protected]>; Alvaro Herrera <[email protected]>; pgsql-hackers


Added to TODO:

        o Allow GLOBAL temporary tables to exist as empty by default in
          all sessions

          http://archives.postgresql.org/pgsql-hackers/2007-07/msg00006.php


---------------------------------------------------------------------------

Gregory Stark wrote:
> "Pavel Stehule" <[email protected]> writes:
> 
> > 2007/7/4, Bruce Momjian <[email protected]>:
> >> > The use case is any system that uses temp tables in an OLTP setting,
> >> > which certainly isn't uncommon. The problem is that today (and as well
> >> > with a global temp table that is still writing to the catalogs) is that
> >> > every OLTP operation that creates or drops a temp table is doing DDL.
> >> > At best, that leads to a lot of catalog bloat. Right now, it appears to
> >> > also expose some race conditions (we've got a customer that's been bit
> >> > by this and we've been able to reproduce some odd behavior in the lab).
> >>
> >> The solution is to fix the bloat, not add a work-around.
> 
> The bloat is a direct consequence of performing DDL in the midst of an OLTP
> transaction. And it's not the only consequence either. Off the top of my head
> trying to do DDL in an OLTP environment will cause OID inflation, locking
> issues, catcache problems, unnecessary prepared query replans, and the list
> goes on, what happens to views defined on the temporary tables? Foreign key
> references to the temporary tables?
> 
> You've got it backwards: addressing the artificially imposed requirement to do
> DDL to create new tables for what should be purely DML operations is fixing
> the root problem, not a work-around. What would be a work-around is trying to
> deal with the consequences as they come up.
> 
> > Catalog bloat is one unwanted effect. Second is different behave of
> > temp tables  than other mayor rdbms, and uncomfortable work with temp
> > tables in stored procedures. Third argument for implementation of
> > global temp tables is full support of ANSI SQL,
> 
> I think the ANSI concept of temporary tables which are defined once but give
> you a fresh empty work-space for each transaction only makes sense if you're
> thinking in terms of an OLTP environment. Otherwise you would just go ahead
> and do the DDL to create new tables for each query and not worry about the
> down-sides.
> 
> The advantages of the ANSI temporary tables are all things you would worry
> about in an OLTP environment but not a data warehousing environment:
> 
> 1) Overhead to perform DDL
> 
> 2) Replanning overhead
> 
> 3) Security issues of doing DDL at run-time
> 
> 4) Difficulty structuring code when multiple procedures need the same
>    temporary tables but the procedures may be called in different orders for
>    different jobs and need different sets of tables.
> 
> -- 
>   Gregory Stark
>   EnterpriseDB          http://www.enterprisedb.com
> 
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 2: Don't 'kill -9' the postmaster

-- 
  Bruce Momjian  <[email protected]>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +



^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread

* [PATCH v4] Restructure CreateStatsStmt parse-analysis processing
@ 2025-10-17 11:55  Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 288+ messages in thread

From: Álvaro Herrera @ 2025-10-17 11:55 UTC (permalink / raw)

Previously, a lot of code would be run in the ProcessUtility pipeline,
where it is not welcome (that's supposed to be just a giant dispatcher
switch).  Move the parse analysis code to CreateStatistics() instead,
which then means we don't need to open the relation twice; it also
allows us to give a better error message when something other than a
relation name is given in the FROM clause.  It also means we no longer
need to do the transform step in ATPostAlterTypeParse(), which saves a
bit of code.

Author: Jian He <[email protected]>
Discussion: https://postgr.es/m/CACJufxFcvo=fM6J9RC88n1U=eS9+5QdftwUNQMu5h7OgjmqDwQ@mail.gmail.com
---
 src/backend/commands/statscmds.c   | 13 +++++++++++--
 src/backend/commands/tablecmds.c   | 16 +---------------
 src/backend/parser/parse_utilcmd.c | 13 +++----------
 src/backend/tcop/utility.c         | 29 +----------------------------
 src/include/commands/defrem.h      |  2 +-
 src/include/parser/parse_utilcmd.h |  2 +-
 6 files changed, 18 insertions(+), 57 deletions(-)

diff --git a/src/backend/commands/statscmds.c b/src/backend/commands/statscmds.c
index 27bf67e7c4b..632228c13ed 100644
--- a/src/backend/commands/statscmds.c
+++ b/src/backend/commands/statscmds.c
@@ -30,6 +30,7 @@
 #include "miscadmin.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/optimizer.h"
+#include "parser/parse_utilcmd.h"
 #include "statistics/statistics.h"
 #include "utils/acl.h"
 #include "utils/builtins.h"
@@ -58,9 +59,14 @@ compare_int16(const void *a, const void *b)
 
 /*
  *		CREATE STATISTICS
+ *
+ * queryString is the source text of the CREATE STATISTICS command, if any.
+ *
+ * Normally stmt has not already undergone transformation, but in some cases
+ * it might have, such as CREATE TABLE (LIKE).
  */
 ObjectAddress
-CreateStatistics(CreateStatsStmt *stmt)
+CreateStatistics(CreateStatsStmt *stmt, const char *queryString)
 {
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			nattnums = 0;
@@ -113,7 +119,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (!IsA(rln, RangeVar))
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only a single relation is allowed in CREATE STATISTICS")));
+					 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
 
 		/*
 		 * CREATE STATISTICS will influence future execution plans but does
@@ -151,6 +157,9 @@ CreateStatistics(CreateStatsStmt *stmt)
 	Assert(rel);
 	relid = RelationGetRelid(rel);
 
+	/* Run parse analysis */
+	stmt = transformStatsStmt(rel, stmt, queryString);
+
 	/*
 	 * If the node has a name, split it up and determine creation namespace.
 	 * If not, put the object in the same namespace as the relation, and cons
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 5fd8b51312c..a1732b71e55 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -9650,16 +9650,7 @@ static ObjectAddress
 ATExecAddStatistics(AlteredTableInfo *tab, Relation rel,
 					CreateStatsStmt *stmt, bool is_rebuild, LOCKMODE lockmode)
 {
-	ObjectAddress address;
-
-	Assert(IsA(stmt, CreateStatsStmt));
-
-	/* The CreateStatsStmt has already been through transformStatsStmt */
-	Assert(stmt->transformed);
-
-	address = CreateStatistics(stmt);
-
-	return address;
+	return CreateStatistics(stmt, NULL);
 }
 
 /*
@@ -15632,11 +15623,6 @@ ATPostAlterTypeParse(Oid oldId, Oid oldRelId, Oid refRelId, char *cmd,
 			querytree_list = lappend(querytree_list, stmt);
 			querytree_list = list_concat(querytree_list, afterStmts);
 		}
-		else if (IsA(stmt, CreateStatsStmt))
-			querytree_list = lappend(querytree_list,
-									 transformStatsStmt(oldRelId,
-														(CreateStatsStmt *) stmt,
-														cmd));
 		else
 			querytree_list = lappend(querytree_list, stmt);
 	}
diff --git a/src/backend/parser/parse_utilcmd.c b/src/backend/parser/parse_utilcmd.c
index e96b38a59d5..89c8315117b 100644
--- a/src/backend/parser/parse_utilcmd.c
+++ b/src/backend/parser/parse_utilcmd.c
@@ -3137,16 +3137,14 @@ transformIndexStmt(Oid relid, IndexStmt *stmt, const char *queryString)
  * transformStatsStmt - parse analysis for CREATE STATISTICS
  *
  * To avoid race conditions, it's important that this function relies only on
- * the passed-in relid (and not on stmt->relation) to determine the target
- * relation.
+ * the passed-in rel (and not on stmt->relation) as the target relation.
  */
 CreateStatsStmt *
-transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
+transformStatsStmt(Relation rel, CreateStatsStmt *stmt, const char *queryString)
 {
 	ParseState *pstate;
 	ParseNamespaceItem *nsitem;
 	ListCell   *l;
-	Relation	rel;
 
 	/* Nothing to do if statement already transformed. */
 	if (stmt->transformed)
@@ -3158,10 +3156,8 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	/*
 	 * Put the parent table into the rtable so that the expressions can refer
-	 * to its fields without qualification.  Caller is responsible for locking
-	 * relation, but we still need to open it.
+	 * to its fields without qualification.
 	 */
-	rel = relation_open(relid, NoLock);
 	nsitem = addRangeTableEntryForRelation(pstate, rel,
 										   AccessShareLock,
 										   NULL, false, true);
@@ -3196,9 +3192,6 @@ transformStatsStmt(Oid relid, CreateStatsStmt *stmt, const char *queryString)
 
 	free_parsestate(pstate);
 
-	/* Close relation */
-	table_close(rel, NoLock);
-
 	/* Mark statement as successfully transformed */
 	stmt->transformed = true;
 
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
index 918db53dd5e..027356f6e77 100644
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -1866,34 +1866,7 @@ ProcessUtilitySlow(ParseState *pstate,
 				break;
 
 			case T_CreateStatsStmt:
-				{
-					Oid			relid;
-					CreateStatsStmt *stmt = (CreateStatsStmt *) parsetree;
-					RangeVar   *rel = (RangeVar *) linitial(stmt->relations);
-
-					if (!IsA(rel, RangeVar))
-						ereport(ERROR,
-								(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-								 errmsg("CREATE STATISTICS only supports relation names in the FROM clause")));
-
-					/*
-					 * CREATE STATISTICS will influence future execution plans
-					 * but does not interfere with currently executing plans.
-					 * So it should be enough to take ShareUpdateExclusiveLock
-					 * on relation, conflicting with ANALYZE and other DDL
-					 * that sets statistical information, but not with normal
-					 * queries.
-					 *
-					 * XXX RangeVarCallbackOwnsRelation not needed here, to
-					 * keep the same behavior as before.
-					 */
-					relid = RangeVarGetRelid(rel, ShareUpdateExclusiveLock, false);
-
-					/* Run parse analysis ... */
-					stmt = transformStatsStmt(relid, stmt, queryString);
-
-					address = CreateStatistics(stmt);
-				}
+				address = CreateStatistics((CreateStatsStmt *) parsetree, queryString);
 				break;
 
 			case T_AlterStatsStmt:
diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h
index dd22b5efdfd..684f62109fd 100644
--- a/src/include/commands/defrem.h
+++ b/src/include/commands/defrem.h
@@ -85,7 +85,7 @@ extern void RemoveOperatorById(Oid operOid);
 extern ObjectAddress AlterOperator(AlterOperatorStmt *stmt);
 
 /* commands/statscmds.c */
-extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt);
+extern ObjectAddress CreateStatistics(CreateStatsStmt *stmt, const char *queryString);
 extern ObjectAddress AlterStatistics(AlterStatsStmt *stmt);
 extern void RemoveStatisticsById(Oid statsOid);
 extern void RemoveStatisticsDataById(Oid statsOid, bool inh);
diff --git a/src/include/parser/parse_utilcmd.h b/src/include/parser/parse_utilcmd.h
index 4965fac4495..67255799f0a 100644
--- a/src/include/parser/parse_utilcmd.h
+++ b/src/include/parser/parse_utilcmd.h
@@ -26,7 +26,7 @@ extern AlterTableStmt *transformAlterTableStmt(Oid relid, AlterTableStmt *stmt,
 											   List **afterStmts);
 extern IndexStmt *transformIndexStmt(Oid relid, IndexStmt *stmt,
 									 const char *queryString);
-extern CreateStatsStmt *transformStatsStmt(Oid relid, CreateStatsStmt *stmt,
+extern CreateStatsStmt *transformStatsStmt(Relation rel, CreateStatsStmt *stmt,
 										   const char *queryString);
 extern void transformRuleStmt(RuleStmt *stmt, const char *queryString,
 							  List **actions, Node **whereClause);
-- 
2.47.3


--xzcma6eafn4gmiua--





^ permalink  raw  reply  [nested|flat] 288+ messages in thread