Deleting cross referenced data - mysql

Question

I have the following MySQL table:
id rid
----- ------
1 2
2 1
2 3
3 2
1 3
3 1
I want to change this so only one row per relation exists.
e.g:
id rid
----- ------
1 2
2 3
1 3

Answer 1

If you always have pairs (as in your example):
delete from table
where id > rid;
This keeps the record where id is smaller.
If there is the possibility that no all pairs exist, then:
delete t
from table t left outer join
(select least(id, rid) as lid, greatest(id, rid) as gid, count(*) as cnt
from table t2
group by least(id, rid), greatest(id, rid)
) t2
on least(t.id, t.rid) = t2.lid and greatest(t.id, t.rid) = gid
where id < rid or t2.cnt = 1;
EDIT (explanation):
How does the second query work? Let me be honest, what I want to write is this:
delete t from table t
where id < rid or
(id > rid and
not exists (select 1 from table t2 where t2.id = t.rid and t2.rid = t.id
);
That is, I want to keep all records where id < rid. But then, I also want to keep all singleton records where rid > id. I don't think MySQL allows the syntax with the where clause.
Instead, the query in the answer counts the number of times that a pair exists, by looking at the smallest value and the largest value. For the data in the question, the result of the subquery is:
id rid cnt
1 2 2
2 3 2
1 3 2
So, all of these would use the id < rid to select the row. If you had one more row, say 4, 1. It would look like:
lid gid cnt
1 2 2
2 3 2
1 3 2
1 4 1
In this case, the first three would take the row with id < rid. But the new row would also be selected because the cnt is 1.
If you had duplicates in the table and a primary key, there would be a slight variation on the query that would do the same thing.