I was wondering if it's possible to drop a column identity (not the column itself) while keeping the attached sequence. This would avoid recreating an identical sequence (especially with a correct start value and owner).
Why doesn't this work?
BEGIN;
DROP SEQUENCE t_id;
CREATE SEQUENCE new_t_id_seq AS INTEGER OWNED BY t.id;
ALTER SEQUENCE new_t_id_seq OWNER TO new_owner;
SELECT setval('new_t_id', (SELECT MAX(id) FROM t));
SELECT nextval('new_t_id');
COMMIT;
Changing the sequence owner to NONE before dropping identity is not allowed. Also changing pg_class.relowner to some role did not help. The sequence is still dropped together with the column identity.
Manually diigging around the system catalog is never recommended.
--
Born in Arizona, moved to Babylonia.
Born in Arizona, moved to Babylonia.