On Tue, 29 Jan 2008, David Owen wrote:
CREATE TABLE simple1(a varchar, b varchar, c timestamp DEFAULT now(),
d timestamp, PRIMARY KEY(a, c));
CREATE VIEW simple2 AS SELECT a, b FROM simple1 WHERE d IS NULL;
INSERT INTO simple1(a, b) VALUES('a', '1');
CREATE RULE simple2_upd AS ON UPDATE TO simple2 DO INSTEAD
(UPDATE simple1 SET d = now() WHERE a = NEW.a AND d IS NULL;
INSERT INTO simple1(a, b) VALUES(NEW.a, NEW.b));
Now, if I do
UPDATE simple2 SET b='2';
... I expect the the original row in simple1 to now have d=(some time), and a
new row with (a, b, c, d) = ('a', '2', some time, NULL).
However, the new row isn't inserted; only the previous row is changed.
I discovered that if I change the rule to only do the insert, I still have
troubles. The first update to simple2 will insert an row, but a second
update will give a duplicate primary key error.
The timestamp being inserted doesn't correspond to the creation of the
rule. Is the rule's query tree maybe absorbing the default value into
itself, but not doing so until the rules first use?
Thanks,
David Owen
---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?
http://www.postgresql.org/docs/faq
