Question

I have a procedure, where I have to check a certain view for some specified entries and delete them accordingly. I have used the following approach for this purpose –

SELECT      id_1,
            id_2, 
            id_3, 
            id_4
INTO        v_id_1,
            v_id_2, 
            v_id_3,
            v_id_4
FROM        v_doc
WHERE       parent_id_1 = p_id_1    -- 'p_' suffix stands for function parameters
            AND parent_id_2 = p_id_2
            AND parent_id_3 = p_id_3
LIMIT       1
;

WHILE v_id_1 IS NOT NULL
LOOP
    -- Code for child document line deletion goes here


    SELECT      id_1,
                id_2, 
                id_3, 
                id_4
    INTO        v_id_1,
                v_id_2, 
                v_id_3,
                v_id_4
    FROM        v_doc
    WHERE       parent_id_1 = p_id_1
                AND parent_id_2 = p_id_2
                AND parent_id_3 = p_id_3
    LIMIT       1
    ;
END LOOP;

Is this is the efficient way, or there is a more efficient way to do this type of query? I am talking about the way I am selecting the records, of course.

Answer 1

I think you’re wondering how you can delete each matching item, if your query returns many rows. A quicker and correcter way is to run the query once, and loop over its rows:

DECLARE
    r RECORD;
BEGIN
    FOR r IN SELECT id_1, id_2, id_3, id_4 
               FROM v_doc 
              WHERE id_1 = p_id_1 
                AND id_2 = p_id_2 
                AND id_3 = p_id_3 LOOP
        -- delete item for r.id_1, r.id_2, etc.
    END LOOP;
END;

See http://www.postgresql.org/docs/8.4/static/plpgsql-control-structures.html#PLPGSQL-RECORDS-ITERATING

An even better way might be to simply use a DELETE FROM x WHERE ... statement, if possible. It depends how simple the deletion is.