From: Tom Lane <tgl@sss.pgh.pa.us>
To: Daniel Gustafsson <daniel@yesql.se>
cc: jochen.bandhauer@gmx.net,
        PostgreSQL-documentation <pgsql-docs@lists.postgresql.org>
Subject: Re: remark regarding 4.2.13. Row Constructors
In-reply-to: <8B3166A3-E46D-4A21-B755-37301A6102AF@yesql.se>
References: <178090456543.1285959.7355583344220664853@wrigleys.postgresql.org>
 <8B3166A3-E46D-4A21-B755-37301A6102AF@yesql.se>
Comments: In-reply-to Daniel Gustafsson <daniel@yesql.se>
	message dated "Tue, 09 Jun 2026 10:06:20 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <891847.1781023598.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Tue, 09 Jun 2026 12:46:38 -0400
Message-ID: <891848.1781023598@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/891848.1781023598%40sss.pgh.pa.us>
Precedence: bulk

Daniel Gustafsson <daniel@yesql.se> writes:
>> On 8 Jun 2026, at 09:42, PG Doc comments form <noreply@postgresql.org> =
wrote:
>> in
>> https://www.postgresql.org/docs/current/sql-expressions.html#SQL-SYNTAX=
-ROW-CONSTRUCTORS
>> there is:
>> "The key word ROW is optional when there is more than one expression in=
 the
>> list."
>> I think it should be:
>> "The key word ROW is optional when there are more than zero expressions=
 in
>> the list."

> "more than zero" sounds a bit odd, my suggestion would be "The key word =
ROW is
> optional when there is one, or more, expressions in the list."

More to the point, the statement is correct as written and either of
these changes would make it wrong.  For example, both of these things
produce a two-column composite value:

postgres=3D# select row(1,2), (1,2);
  row  |  row  =

-------+-------
 (1,2) | (1,2)
(1 row)

postgres=3D# select pg_typeof(row(1,2)), pg_typeof((1,2));
 pg_typeof | pg_typeof =

-----------+-----------
 record    | record
(1 row)

But adding more parentheses around a scalar value does not make
it a composite value:

postgres=3D# select row(1), (1), ((((1))));
 row | ?column? | ?column? =

-----+----------+----------
 (1) |        1 |        1
(1 row)

postgres=3D# select pg_typeof(row(1)), pg_typeof((1)), pg_typeof(((((1))))=
);
 pg_typeof | pg_typeof | pg_typeof =

-----------+-----------+-----------
 record    | integer   | integer
(1 row)

Maybe there is something we can do to make this clearer, but
the above isn't it.

			regards, tom lane