RE: Re: Unanswered questions about Postgre

No, SELECT FOR UPDATE locks the rows matching the constraint and if no rows
match, none are locked.

as in (somewhat oracle syntax)

BEGIN
  SELECT 'x' INTO a FROM test WHERE y = 10 FOR UPDATE;
  UPDATE test SET z = 12;
EXCEPTION
  WHEN no_data_found THEN
    INSERT INTO test VALUES (10, 12);
END;


-----Original Message-----
From: Edward Q. Bridges [mailto:ed.bridges@buzznik.com]
Sent: den 12 december 2000 17:10
To: PostgreSQL general mailing list; Roger Wernersson
Subject: RE: [GENERAL] Re: Unanswered questions about Postgre


On Tue, 12 Dec 2000 12:04:46 +0100, Roger Wernersson wrote:

>
> I can't BEGIN - SELECT FOR UPDATE - INSERT or UPDATE - COMMIT as someone
> might insert after my SELECT and before my INSERT.
>

correct me if i'm wrong, but a select for update locks the table for
an insert or an update until the end of the transaction.

--e--



> Is there a right solution?
>
> /Roger
>
> -----Original Message-----
> From: Tim Kientzle [mailto:kientzle@acm.org]
> Sent: den 12 december 2000 02:10
> To: PostgreSQL general mailing list
> Subject: [GENERAL] Re: Unanswered questions about Postgre
>
>
> > Of course, people really shouldn't be inserting
> > objects which already exist, ...
>
> On the contrary, the best way to test if
> something already exists is to just try the
> INSERT and let the database tell you if
> it's already there.  Both faster and more
> reliable than doing SELECT then INSERT.
>
>             - Tim
>