Обсуждение: RULE: ON DELETE doesn't stack deletes
According to the manual you can stack multiple queries in a RULE:
CREATE RULE name AS ON event
TO object [ WHERE condition ]
DO [ INSTEAD ] action
where action can be:
NOTHING
|
query
|
( query ; query ... )
|
[ query ; query ... ]
This seems to work provided 'query' is not "DELETE"; if it is, only
the first one is executed.
My installation is Postgres 7.2.2, using the binary package shipped in
Redhat 8.0
[malcolm@localhost STRUCT]$ uname -sr
Linux 2.4.18-18.8.0
Here is a test case:
CREATE TABLE A
(a_data int);
CREATE TABLE B
(b_data int);
CREATE VIEW V_AB AS
SELECT a_data,b_data FROM A,B;
CREATE RULE R_DEL_AB AS
ON DELETE TO V_AB
DO INSTEAD
(
DELETE FROM A WHERE a_data=OLD.a_data;
DELETE FROM B WHERE b_data=OLD.b_data;
);
INSERT INTO A values (1);
INSERT INTO B values (1);
test=# select * from V_AB;
a_data | b_data
--------+--------
1 | 1
(1 row)
test=#delete from v_ab;
DELETE 0
test=# select * from a;
a_data
--------
(0 rows)
test=# select * from b;
b_data
--------
1
(1 row)
Confirmed this problem on cvs-tip. Replacing the DO INSTEAD ( DELETE....) with DO INSTEAD (INSERT...) allows multiple insert statements which function fine using OLD.a_data and OLD.b_data. So it must be something else. Could it be because once the DELETE FROM A has run the tuple no longer exists in the view? On Wed, 2002-11-20 at 10:49, Malcolm Hutty wrote: > According to the manual you can stack multiple queries in a RULE: > CREATE RULE R_DEL_AB AS > ON DELETE TO V_AB > DO INSTEAD > ( > DELETE FROM A WHERE a_data=OLD.a_data; > DELETE FROM B WHERE b_data=OLD.b_data; > ); -- Rod Taylor <rbt@rbt.ca>
On Wed, 20 Nov 2002, Malcolm Hutty wrote: > According to the manual you can stack multiple queries in a RULE: > > CREATE RULE name AS ON event > TO object [ WHERE condition ] > DO [ INSTEAD ] action > > where action can be: > > NOTHING > | > query > | > ( query ; query ... ) > | > [ query ; query ... ] > > > This seems to work provided 'query' is not "DELETE"; if it is, only > the first one is executed. I think it's something like: delete from v_ab turns into something equivalent to the two statements (I don't know what the actual form of the queries is however) delete from a where a_data=ANY(select a_data from v_ab); delete from b where b_data=ANY(select b_data from v_ab); So when the second runs there is no matching rows (having already had all the a rows deleted).
Rod Taylor <rbt@rbt.ca> writes:
> Could it be because once the DELETE FROM A has run the tuple no longer
> exists in the view?
Precisely. OLD is only a macro for the view, and the first delete
changes what's in the view ...
regards, tom lane