I have the following example:
Table A
-some_id
Table B
-another_id
Table C
-some_id_fk
-another_id_fk
I want to cascade a row on
I suggest two foreign key constraints with ON DELETE SET NULL and a trigger that takes care of the rest
Tables:
CREATE TABLE a (a_id serial PRIMARY KEY, a text NOT NULL);
CREATE TABLE b (b_id serial PRIMARY KEY, b text NOT NULL);
CREATE TABLE ab (
ab_id serial PRIMARY KEY
, a_id int REFERENCES a ON DELETE SET NULL
, b_id int REFERENCES b ON DELETE SET NULL
, UNIQUE (a_id, b_id)
);
Trigger:
CREATE OR REPLACE FUNCTION trg_ab_upbef_nulldel()
RETURNS trigger AS
$func$
BEGIN
DELETE FROM ab WHERE ab_id = NEW.ab_id;
RETURN NULL;
END
$func$ LANGUAGE plpgsql;
CREATE TRIGGER upbef_nulldel
BEFORE UPDATE OF a_id, b_id ON ab
FOR EACH ROW
WHEN (NEW.a_id IS NULL AND
NEW.b_id IS NULL)
EXECUTE PROCEDURE trg_ab_upbef_nulldel();
SQL Fiddle.
Be sure to have a surrogate PK column for the connecting table. (a_id, b_id)
cannot be the PK anyway, because that would disallow NULL in both. Add a UNIQUE constraint instead, which allows NULL values.
The trigger is optimized for performance and only kicks in when one of the two FK columns is updated, and only when that results in both being NULL
.
The trigger function is trivial: deletes the row and returns NULL to cancel the now void cascaded UPDATE
.