How to exclude duplicate rows when joining a table with itself.

Question

Here's a sample table to help illustrate my problem:

mysql> select * from test;
+----+--------------+--------+
| id | type         | siteid |
+----+--------------+--------+
|  1 | First Visit  |    100 |
|  2 | Second Visit |    100 |
|  3 | First Visit  |    300 |
|  4 | First Visit  |    400 |
|  5 | Second Visit |    500 |
|  6 | Second Visit |    600 |
+----+--------------+--------+

I'm trying to join the table upon itself, to pull together rows that have the same siteid value. Here's my attempt:

mysql> select * from test T1
    -> LEFT OUTER JOIN test T2 on T1.siteid = T2.siteid and T1.id <> T2.id;
+----+--------------+--------+------+--------------+--------+
| id | type         | siteid | id   | type         | siteid |
+----+--------------+--------+------+--------------+--------+
|  1 | First Visit  |    100 |    2 | Second Visit |    100 |
|  2 | Second Visit |    100 |    1 | First Visit  |    100 |
|  3 | First Visit  |    300 | NULL | NULL         |   NULL |
|  4 | First Visit  |    400 | NULL | NULL         |   NULL |
|  5 | Second Visit |    500 | NULL | NULL         |   NULL |
|  6 | Second Visit |    600 | NULL | NULL         |   NULL |
+----+--------------+--------+------+--------------+--------+

This is basically the result I'm looking for, except for the 1st 2 rows. I'd like to eliminate one of those. So, I tried the following:

mysql> select * from test T1
    -> LEFT OUTER JOIN test T2 on T1.siteid = T2.siteid and T1.id <> T2.id
    -> GROUP BY T1.siteid;
+----+--------------+--------+------+--------------+--------+
| id | type         | siteid | id   | type         | siteid |
+----+--------------+--------+------+--------------+--------+
|  1 | First Visit  |    100 |    2 | Second Visit |    100 |
|  3 | First Visit  |    300 | NULL | NULL         |   NULL |
|  4 | First Visit  |    400 | NULL | NULL         |   NULL |
|  5 | Second Visit |    500 | NULL | NULL         |   NULL |
|  6 | Second Visit |    600 | NULL | NULL         |   NULL |
+----+--------------+--------+------+--------------+--------+

This is exactly the output I'm looking for. However, I come to learn this isn't the standard way to use GROUP BY, and the above statement fails on ORACLE, giving me a

General SQL error.
ORA-00979: not a GROUP BY expression

Can anyone offer some help on how to get the results like the last table and that works with ORACLE?

Answer 1

Just remove a single character from your query and it will get the job done... (replace "<>" with "<"):

select * from test T1
LEFT OUTER JOIN test T2 on T1.siteid = T2.siteid and T1.id < T2.id