Assistance understanding query structure

Question

Hi,
I was wondering if any of you know the answer to this question.
There are 3 tables.
A
A1 & A2
"A: table has rows with quasi composite keys -- meaning they're not primary keys
but both their combinations make them unique ie. (id_1, id_2). The combination of these two keys make the rows unique.
Now, table A1 and A2 have these keys as well. The relationship between A to A1 & A2 is one-to-many. My question is: When I count how many entries there are for a particular key-combination in table A and A1 do I have to do use the SQl keyword "Distinct"?
I have found that using distinct and not using distinct give me different results. Conceptually I'm having a problem understanding. If I combine the last table "A2" it makes it more interesting.
-Jim

Answer 1

Actually, your problem has less to do with composite primary keys than it does with the one to many realtionship between the tables.
A join will return one row for each matching record in the joined table, so if there are two matching rows, you should exect to get back 2 rows. For example:
SQL> CREATE TABLE t (col1 number, col2 number, col3 varchar2(10));
Table created.
SQL> ALTER TABLE t ADD constraint t_pk PRIMARY KEY (col1, col2);
Table altered.
SQL> CREATE TABLE t1 (t1_pk number, col1 number, col2 number, descr varchar2(10));
Table created.
SQL> ALTER TABLE t1 ADD constraint t1_pk PRIMARY KEY (t1_pk);
Table altered.
SQL> ALTER TABLE t1 ADD constraint t1_fk
  2  FOREIGN KEY (col1, col2) REFERENCES t(col1, col2);
Table altered.
SQL> INSERT INTO t VALUES (1, 1, 'OneOne');
1 row created.
SQL> INSERT INTO t VALUES (1, 2, 'OneTwo');
1 row created.
SQL> INSERT INTO t1 VALUES (1, 1, 1, 'T1One');
1 row created.
SQL> INSERT INTO t1 VALUES (2, 1, 1, 'T1Two');
1 row created.
SQL> INSERT INTO t1 VALUES (3, 1, 2, 'T1Three');
1 row created.
SQL> COMMIT;
SQL> SELECT t.col1, t.col2, t.col3, t1.t1_pk, t1.descr
  2  FROM t
  3     JOIN t1 ON t.col1 = t1.col1 and
  4                t.col2 = t1.col2;
      COL1       COL2 COL3            T1_PK DESCR
         1          1 OneOne              1 T1One
         1          1 OneOne              2 T1Two
         1          2 OneTwo              3 T1ThreeWhen you add a third table, each row from the first join gets joined to the third table. So, if we do:
SQL> CREATE TABLE t2 (t2_pk number, col1 number, col2 number,
  2  descr varchar2(10));
Table created.
SQL> ALTER TABLE t2 ADD CONSTRAINT t2_pk PRIMARY KEY (t2_pk);
Table altered.
SQL> ALTER TABLE t2 ADD constraint t2_fk
  2  FOREIGN KEY (col1, col2) REFERENCES t(col1, col2);
Table altered.
SQL> INSERT INTO t2 VALUES (1, 1, 1, 'T2One');
1 row created.
SQL> INSERT INTO t2 VALUES (2, 1, 1, 'T2Two');
1 row created.
SQL> INSERT INTO t2 VALUES (3, 1, 2, 'T2Three');
1 row created.
SQL> COMMIT;
Commit complete.and then query all three tables, we get:
SQL> SELECT t.col1, t.col2, t.col3, t1.t1_pk, t1.descr, t2.t2_pk, t2.descr
  2  FROM t
  3     JOIN t1 ON t.col1 = t1.col1 and
  4                t.col2 = t1.col2
  5     JOIN t2 ON t.col1 = t2.col1 and
  6                t.col2 = t2.col2;
      COL1       COL2 COL3            T1_PK DESCR             T2_PK DESCR
         1          1 OneOne              1 T1One                 1 T2One
         1          1 OneOne              2 T1Two                 1 T2One
         1          1 OneOne              1 T1One                 2 T2Two
         1          1 OneOne              2 T1Two                 2 T2Two
         1          2 OneTwo              3 T1Three               3 T2ThreeThe results would be exactly the same if table t had a single column PK.
HTH
John