Remove reverse duplicates from an SQL query

Question

Let's say that a query result is supposed to return a list of string pairs (x, y). I am trying to eliminate the reverse duplicates. What I mean is, if (x, y) was one of the results, (y, x) should not appear later.

example:

column 1 (foreign key)    column 2 (int)     column 3 (string)
4                         50                 Bob
2                         70                 Steve 
3                         50                 Joe

The people represented in this table can appear multiple times with a different column 2 value.

My query needs to print every pair of names that have the same column 2 value :

select e.column3, f.column3 from example as e, example as f where e.column2 = f.column2 

(Bob, Bob)
(Bob, Joe)
(Joe, Bob)
(Joe, Joe)

I upgraded the query so that it removes the doubles:

select e.column3, f.column3 from example as e, example as f where e.column2 = f.column2
       and e.column3 <> f.column3

(Bob, Joe)
(Joe, Bob)

Now I want it to only return:

(Bob, Joe). 

(Joe, Bob) is a reverse duplicate, so I don't want it in the result. Is there anyway to handle that in one query?

Answer 1

First of all, welcome to 2012. We have migrated away from relating tables using commas. It was introdued in ANSI 89 but is severely lacking. Nowaways, the correct way is to write queries using the ANSI 92/99/2003 JOIN syntax.

The solution to your problem is to turn your bidirectional inequality <> into a unidirectional inequality, either < or > whichever you prefer.

select e.column3, f.column3
from example as e
join example as f on e.column2 = f.column2 and e.column3 < f.column3

Answer 2

select e.column3, f.column3 from example as e, example as f where e.column2 = f.column2
       and e.column3 <> f.column3 where e.id < f.id

adding a simple where clause should do it.