simulating INSERT return values with default values of sequences

Thomas F.O'Connell <tfo@monsterlabs.com>

From: "Thomas F. O'Connell" <tfo@monsterlabs.com>
To: pgsql-general@postgresql.org
Date: 2002-07-22T19:18:54Z
Lists: pgsql-general
i didn't see this question asked (or answered in the mailing lists), but 
it seems like it would be an FAQ...

anyway, i'm just wondering if there's a slick way of any sort to be able 
to have a value generated from an INSERT available for immediate and 
secure re-use.

for example, if i have

CREATE SEQUENCE foo_id_seq;
CREATE TABLE foo (
	id int2 primary key default nextval( 'foo_id_seq' ),
	bar text
);

and then later have

INSERT INTO foo( bar ) VALUES( 'baz' );

is there any way safely to know what was inserted into the "id" field 
without encapsulating the INSERT statement in a transaction with 
whatever function needed that value?

i'm guessing a transaction probably makes the most sense, right?

e.g.,

BEGIN WORK;
INSERT INTO foo( bar ) VALUES( 'baz' );
UPDATE foo SET bar = 'ola' WHERE id = ( SELECT last_value FROM foo_id_seq );
END WORK;

if so, then i guess the next question is: is there a preference between 
a statement like

UPDATE foo SET bar = 'ola' WHERE id = ( SELECT last_value FROM foo_id_seq );

and

UPDATE foo SET bar = 'ola' WHERE id = ( SELECT MAX( id ) FROM foo );

thanks!

-tfo