Search code examples
sqloracle-databaseplsqloracle10gcascading-deletes

Deleting rows from multiple related tables


There are three tables related one-to-one on the identifier. I need to delete all the records from the three tables that match the criteria A.ID = B.ID = C.ID

Now I do it in the following way:

DECLARE
    CURSOR CUR IS 
        SELECT C.ID FROM A 
        INNER JOIN B ON A."ID" = B."ID" 
        INNER JOIN C ON B."ID" = C."ID"
        WHERE A.STATUS = 'ERROR';
    IDX NUMBER;
BEGIN
    FOR REC IN CUR LOOP 
        IDX := REC.ID;
        DELETE FROM C WHERE C."ID" = IDX;
        DELETE FROM B WHERE B."ID" = IDX;
        DELETE FROM A WHERE BP."ID" = IDX;
    END LOOP;
    COMMIT;
END;

A lot of data and this way for very long runs. Is there any way to delete faster?


Solution

  • You could create a PL/SQL type to store the IDs.

    CREATE TYPE t_ids AS TABLE OF NUMBER;
    

    Delete all records from table a that match the criterias, and return the IDs into a variable of that type. Then delete all records from b and c with these IDs.

    DECLARE
      ids_to_delete   t_ids;
    BEGIN
      DELETE FROM a
       WHERE a.status = 'ERROR'
         AND EXISTS ( SELECT 1 FROM b WHERE b.id = a.id )
         AND EXISTS ( SELECT 1 FROM c WHERE c.id = a.id )
      RETURNING a.id
      BULK COLLECT INTO ids_to_delete;
    
      DELETE FROM b
       WHERE id IN ( SELECT COLUMN_VALUE FROM TABLE( ids_to_delete ) );
    
      DELETE FROM c
       WHERE id IN ( SELECT COLUMN_VALUE FROM TABLE( ids_to_delete ) );
    END;
    

    This should perform a lot better, since it requires no loop and does everything in three SQL statements, instead of three statements per ID.