Here's a tricky one:
Versions: I'm using PostgreSQL 9.0.7 and PHP 5.3.8.
First, I have a table "bob" with two columns: "bobid" (integer) and "bobtext"
(text). The "bobid" field defaults to a sequence called "bob_seq" which is just
a simple counter (1,2,3 etc.)
I can do an INSERT query into "bob":
INSERT INTO bob (bobtext) VALUES ('Hello Bob') RETURNING bobid;
and get the value the sequence has given me for "bobid": this works fine using
both the psql tool and pg_query in PHP.
I can also do an UPDATE query in the same vein:
UPDATE bob SET bobtext='Hello Bob revisited' WHERE bobid=<x> RETURNING bobid;
where <x> is a valid "bobid" value for an existing record in the table. This
returns the value of "bobid" for the updated record which should be the same as
<x> and again it works in the psql tool and in pg_query.
Next, I have a VIEW based on "bob", called "fred". It is simply defined as
CREATE VIEW fred AS (
SELECT * FROM bob
);
I have two rules on "fred" which allow me to alter the view.
The first is used for INSERT into "fred", which translates them into INSERT into
"bob":
CREATE OR REPLACE RULE fred_insert AS
ON INSERT TO fred
DO INSTEAD
INSERT INTO bob (bobtext) VALUES (NEW.bobtext) RETURNING *
;
The second is used for UPDATE on "fred", which translates to creating a new
record in "bob" and deleting the old record:
CREATE OR REPLACE RULE fred_update AS
ON UPDATE TO fred
DO INSTEAD (
INSERT INTO bob (bobtext) VALUES (NEW.bobtext) RETURNING *;
DELETE FROM bob WHERE bobid=OLD.bobid;
)
;
Note that these rules both return all of the fields of the newly inserted "bob"
records back up the chain to be accessible to the original query.
Now, I can insert into "fred" knowing that the rule will handle it:
INSERT INTO fred (bobtext) VALUES ('Hello Fred') RETURNING bobid;
and that should return the new "bobid" value of the inserted record (from the
"bob_seq" sequence). It still works in the psql tool and in pg_query.
I should also be able to exercise the update rule on "fred":
UPDATE fred SET bobtext='Hello Fred revisited' WHERE bobid=<x> RETURNING bobid
where <x> is a valid "bobid" value for an existing record in the table.
This should insert a new row with a new "bobid" from the sequence and the new
"bobtext" value, delete the row with "bobid"=<x>, and return the "bobid" for the
new row.
Here's the punchline (at last):
It works in the psql tool.
It doesn't work in pg_query.
<?php
$result = pg_query("UPDATE fred SET bobtext='Hello Fred revisited' WHERE
bobid=42 RETURNING bobid"); // Assume there is a record with bobid==42
if ($result && pg_numrows($result))
{
echo pg_numrows($result).PHP_EOL;
echo pg_fetch_result($result, 0, 0).PHP_EOL;
)
?>
Tells me there that although the $result resource is valid, there are no rows
and therefore no result to fetch (PHP Warning: pg_fetch_result(): Unable to
jump to row 0 on PostgreSQL result ...)
So: why is the return from the UPDATE rule different to the return from the
INSERT rule in PHP pgsql?
Thanks for taking the time...
Cheers
Pete
--
Peter Ford, Developer phone: 01580 893333 fax: 01580 893399
Justcroft International Ltd. www.justcroft.com
Justcroft House, High Street, Staplehurst, Kent TN12 0AH United Kingdom
Registered in England and Wales: 2297906
Registered office: Stag Gates House, 63/64 The Avenue, Southampton SO17 1XS
--
PHP General Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php
