Обсуждение: Problem With A Rule (clear version)

Many thanks for the patience! Tom,

I am afraid that this too-complete example makes
readers headache.

Thank you all!

CN
===============
CREATE TABLE table1 (
--the invoice table.
PRIMARY KEY(column10,column1,column2),
column10 VARCHAR(255), --id#='1'
column1  VARCHAR(255), --invoice#
column2  SMALLINT, --serial number
column6  NUMERIC, --amount
column7  "char"  --indicator: either 'r' or 'E'
);

CREATE TABLE table2 (
--the voucher detail table.
CONSTRAINT c1 FOREIGN KEY (column10,column1)
REFERENCES table3 (column10,column1) ON UPDATE
CASCADE ON DELETE CASCADE,
PRIMARY KEY (column10,column1,column2),
column10 VARCHAR(255), --id#='1'
column1  VARCHAR(255), --voucher#
column6  "char",  --indicator: either '1' or '2'
column7  NUMERIC, --amount
column8  VARCHAR(255), --matched column for
table1.column7
column9  VARCHAR(255) --matched column for
table1.column2
);

CREATE TABLE table3 (
--the voucher master table.
PRIMARY KEY (column10,column1),
column10 VARCHAR(255), --id#='1'
column1  VARCHAR(255), --voucher#
column4  "char",  --status='N'
column5  VARCHAR(255) --matched column for
table1.column1
);

CREATE RULE rule1 AS ON UPDATE TO table1
DO UPDATE table2 SET column7=NEW.column6 WHERE
column10=NEW.column10
AND column1=(SELECT column1 FROM table3 WHEREcolumn10=NEW.column10 AND column4='N'AND column5=NEW.column1)
AND column6='2'
AND column8=OLD.column7
AND column9=OLD.column2;


db1=# SELECT column10,column1,column2,column6,column7
FROM table1;
column10 | column1  | column2 | column6 | column7
----------+----------+---------+---------+---------1        | Collect1 |       2 |      11 | r1        | Collect1 |
 1 |     130 | r 
(2 rows)

db1=# SELECT column1,column5 FROM table3 WHERE
column10='1' AND column4='N' AND column5='Collect1';  column1   | column5
-------------+----------AC200108083 | Collect1
(1 row)

db1=# SELECT
column10,column1,column6,column7,column8,column9 FROM
table2 WHERE column10='1' AND column1=(SELECT column1
FROM table3 WHERE column10='1' AND column4='N' AND
column5='Collect1') AND column6='2' AND column8='r'
AND column9='2';
column10 |   column1   | column6 |  column7  |
column8 | column9
----------+-------------+---------+-----------+------
---+---------1        | AC200108083 | 2       | 11.000000 | r     | 2
(1 row)

db1=# UPDATE table1 SET column6=13 WHERE column2=1;
<postgresql waits forever here. So, ctrl-c to
interrupt psql.>
Cancel request sent
ERROR:  Query was cancelled.
db1=# update table2 set column7=13 where column10='1'
and column1=(SELECT column1 FROM table3 WHERE
column10='1' AND column4='N' AND column5='Collect1')
and column6='2' and column8='r' and column9=1;
UPDATE 1

IF rule1 is changed to the following version, then
the update to table1 works:

CREATE RULE rule1 AS ON UPDATE TO table1
DO UPDATE table2 SET column7=NEW.column6 WHERE
column10=NEW.column10
AND column1='AC200108083'
AND column6='2'
AND column8=OLD.column7
AND column9=OLD.column2;

--------------------------------------------------------
You too can have your own email address from Eurosport.
http://www.eurosport.com

<cnliou@eurosport.com> writes:
> I am afraid that this too-complete example makes
> readers headache.

I'm afraid you omitted some detail that's necessary to show the 
problem ... because it works fine for me.

Please try to make a pg_dump script that can be loaded into a fresh
database to produce the problem.  I'm not willing to hand-edit SELECT
output into INSERT statements again.
        regards, tom lane

Thank you again! Tom,

You have reminded me to narrow down the problem. I
tested it in a brand new simplified database where
only related tables and rule are created. The rule I
had problem works fine, and fast, in this simplified
database.

I wonder my working database is corrupted by some
orphan pg_class records, constraints, triggers, etc.
So, please ignore this rule problem as it is actually
ok.

Regards,

cn

--------------------------------------------------------
You too can have your own email address from Eurosport.
http://www.eurosport.com