Update postgres table to squash duplicate values in second table - database

Question

I have a postgresql schema with two tables:
tableA: tableB:
| id | username | | fk_id | resource |
| 1 | user1 | | 2 | item1 |
| 2 | user1 | | 1 | item3 |
| 3 | user1 | | 1 | item2 |
| 4 | user2 | | 4 | item5 |
| 5 | user2 | | 5 | item8 |
| 6 | user3 | | 3 | item9 |
The foreign key fk_id in tableB references id in tableA.
How can I update all of the foreign key id's of tableB to point to the lowest entry for a unique username in tableA?

Answer 1

update table_b b
set fk_id = d.id
from table_a a
join (
select distinct on (username) username, id
from table_a
order by 1, 2
) d using(username)
where a.id = b.fk_id;
Test it here.
The query used inside the update gives actual_id, username, desired_id:
select a.id actual_id, username, d.id desired_id
from table_a a
join (
select distinct on (username) username, id
from table_a
order by 1, 2
) d using(username)
actual_id | username | desired_id
-----------+----------+------------
1 | user1 | 1
2 | user1 | 1
3 | user1 | 1
4 | user2 | 4
5 | user2 | 4
6 | user3 | 6
(6 rows)

Answer 2

We define your tables:
CREATE TABLE tableA (id, username) AS
SELECT * FROM
(
VALUES
(1, 'user1'),
(2, 'user1'),
(3, 'user1'),
(4, 'user2'),
(5, 'user2'),
(6, 'user2')
) AS x ;
CREATE TABLE tableB (fk_id, resource) AS
SELECT * FROM
(
VALUES
(2, 'item1'),
(1, 'item3'),
(1, 'item2'),
(4, 'item5'),
(5, 'item8'),
(3, 'item9')
) AS x ;
With that info, you can create a (virtual) conversion table, and use it to update your data:
-- Using tableA, make a new table with the
-- minimum id for every username
WITH username_to_min_id AS
(
SELECT
min(id) AS min_id, username
FROM
tableA
GROUP BY
username
)
-- Convert the previous table to a id -> min_id
-- conversion table
, id_to_min_id AS
(
SELECT
id, min_id
FROM
tableA
JOIN username_to_min_id USING(username)
)
-- Use this conversion table to update tableB
UPDATE
tableB
SET
fk_id = min_id
FROM
id_to_min_id
WHERE
-- JOIN condition with table to update
id_to_min_id.id = tableB.fk_id
-- Take out the ones that won't change
AND (fk_id <> min_id)
RETURNING
* ;
The result you would get is:
+-------+----------+----+--------+
| fk_id | resource | id | min_id |
+-------+----------+----+--------+
| 1 | item1 | 2 | 1 |
| 1 | item9 | 3 | 1 |
| 4 | item8 | 5 | 4 |
+-------+----------+----+--------+
Shows you that three rows have been updated, that had fk_id = (2, 3, 5), and have now (1, 1, 4). (The id is the "old" fk_id value).
You can check it at http://rextester.com/EQPH47434
You can "squeeze everything" [change every virtual table name by its definition, and do a couple of SELECT optimizations] and get this equivalent query (probably less clear, yet totally equivalent):
UPDATE
tableB
SET
fk_id = min_id
FROM
tableA
JOIN
(
SELECT
min(id) AS min_id, username
FROM
tableA
GROUP BY
username
) AS username_to_min_id
USING (username)
WHERE
tableA.id = tableB.fk_id
AND (fk_id <> min_id)
RETURNING
* ;