Question

There are some duplicate (appearing twice or more) combinations of (id, name, dept_name). Note: the dept_name is nullable so there are often some nulls in that column.

I am trying to get a result of the duplicate (id, name, dept_name) combos where one record has got a null mgr_id or sal but another record of that same duplicated combo has a non-null mgr_id or sal.

This is what I have so far:

 select t1.id, t1.name, t1.dept_name, t1.mgr_id, t2.mgr_id, t1.sal, t2.sal from employee t1
    inner join employee t2 on t2.id = t1.id and t2.name = t1.name
    where t1.id in
    (select id from 
    (select id, name, dept_name, count(*) from employee
    group by id, name, dept_name
    having count(*) > 1))
    and (nvl(t1.mgr_id,0) <> nvl(t2.mgr_id,0) 
    or nvl(t1.sal,0) <> nvl(t2.sal,0)) 
    and t1.mgr_id is not null ;

If I don’t leave the last line “and t1.mgr_id is not null” in the query I effectively receive the same resultset twice (which I don’t want). However I need some way of incorporating the (t1.sal is not null) clause to not get the same resultset twice.

Also, note there is no PK or sequence number column where I can just go:
t1.seq_no <> t2.seq_no.

Answer 1

Can’t you just add the filter to the OR?:

 select t1.id, t1.name, t1.dept_name, t1.mgr_id, t2.mgr_id, t1.sal, t2.sal from employee t1
    inner join employee t2 on t2.id = t1.id and t2.name = t1.name
    where t1.id in
    (select id from 
    (select id, name, dept_name, count(*) from employee
    group by id, name, dept_name
    having count(*) > 1))
    and ((nvl(t1.mgr_id,0) <> nvl(t2.mgr_id,0) AND t1.mgr_id is not null) 
    or (nvl(t1.sal,0) <> nvl(t2.sal,0) AND t1.sal is not null));