SQL Postgres Invalidate Rows that reference invalid Id's - sql

Question

I am trying to create a stored procedure that will invalidate rows that contain id references of an id in another table. The catch is that the rows to be invalidated contain groupings of these id's stored as a comma separated string. Let's take a look at the tables:
table_a table_b
+----+------+ +---------+-------+
| id | name | | ids | valid |
+----+------+ +---------+-------+
| 1 | a | | 1,2,3 | T |
| 2 | b | | 4,3,8 | T |
| 3 | c | | 5,2,5,4 | T |
| 4 | d | | 7 | T |
| 5 | e | | 6,8 | T |
| 6 | f | | 9,7,2 | T |
| 7 | g | +---------+-------+
| 8 | h |
+----+------+
Above you can see that table_b contains groupings of ids from table_a and as you can imagine the table_a.id is an integer while table_b.ids is text. The goal is to look at each table_b.ids and if it contains an id that does not exist in table_a.id then set its validity to false.
I have not worked with any SQL in quite sometime and I have never worked with PostgreSQL which is why I am having such difficulty. The closest query I could come up with, is not working, but is along the lines of:
CREATE FUNCTION cleanup_records() AS $func$
BEGIN
UPDATE table_b
SET valid = FALSE
WHERE COUNT(
SELECT regexp_split_to_table(table_b.ids)
EXCEPT SELECT id FROM table_a
) > 0;
END;
$func$ LANGUAGE PLPGSQL;
The general idea is that I am trying to turn each row of table_b.ids into a table and then using the EXCEPT operator against table_a to see if it has any ids that are invalid. The error I receive is:
ERROR: syntax error at or near "SELECT"
LINE 1: ...able_b SET valid = FALSE WHERE COUNT(SELECT reg...
which is not very helpful as it just indicates that I do not have the correct syntax. Is this query viable? If so can you show me where I may have gone wrong - if not is there an easier or even more complicated way to achieve this?
Sample data:
CREATE TABLE table_b
(ids text, valid boolean);
INSERT INTO table_b
(ids, valid)
VALUES
('1,2,3' , 'T'),
('4,3,8' , 'T'),
('5,2,5,4' , 'T'),
('7' , 'T'),
('6,8' , 'T'),
('9,7,2' , 'T');
CREATE TABLE table_a
(id integer, name text);
INSERT INTO table_a
(id, name)
VALUES
(1,'a'),
(2,'b'),
(3,'c'),
(4,'d'),
(5,'e'),
(6,'f'),
(7,'g'),
(8,'h');

Answer 1

UPDATE table_b
SET valid = FALSE
WHERE EXISTS(
SELECT regexp_split_to_table(table_b.ids)
EXCEPT SELECT id FROM table_a
);
You can use 'exists' to check for the existence of a row. The previous syntax was incorrect as count can't be used that way.

Answer 2

groupings of these id's stored as a comma separated string
Don't do that. It's really bad database design, and is why you're having problems. See:
Is using multiple foreign keys separated by commas wrong, and if so, why?
PostgreSQL list of integers separated by comma or integer array for performance?
Also, there's a more efficient way to do your query than that shown by vkp. If you do it that way, you're splitting the string for every ID you're testing. There is no need to do that. Instead, join on a table of expanded ID lists.
Something like:
UPDATE table_b
SET valid = 'f'
FROM table_b b
CROSS JOIN regexp_split_to_table(b.ids, ',') b_ids(id)
LEFT JOIN table_a a ON (a.id = b_ids.id::integer)
WHERE table_b.ids = b.ids
AND a.id IS NULL
AND table_b.valid = 't';
You need to join on table_b even though it's the update target because you can't make a lateral function reference to the update target table directly.