Query to find the maximum and second maximum

Question

i have a table temptab of 3 columns
col1 col2 col3
1 a 08-JAN-08
1 a 09-JAN-08
1 a 10-JAN-08
1 b 10-JAN-08
1 c 11-mar-08
1 c 10-mar-08
i want to select 1st maxm and 2nd maxm of col3 group by (col1,col2)
o/p will be like
1 a 10-jan-08 08-jan-08
1 b 10-jan-08 null
1 c 11-mar-08 10-mar-08
select first.col1, first.col2, first.MAX_DATE, second.SEC_DATE from (select a.col1, a.col2, max(a.col3) as MAX_DATE
from tab1 a group by a.col1, a.col2) first,
(select b.col1, b.col2, max(b.col3) as SEC_DATE from tab1 b,
(select a.col1, a.col2, max(a.col3) as MAX_DATE from tab1 a
group by a.col1, a.col2) c
WHERE c.col1 = b.col1
AND c.col1=b.col1
AND c.MAX_DATE > b.col3
group by b.col1, b.col2) second
where first.col1 = second.col1
and first.col2 = second.col2;
this is not working..
please give a query which will do this.

Answer 1

1. Your query have 3 subqueries where mine only one. Feel free to make your choice regarding the performance.
2. You cannot have a column named date, unless you use double-quotes
3. Your query doesn't work as well when you have not any second highest value
4. Your query return a cartesian product of your table.
Please check the results of your query (after applying the required modications) with the mine above, and make your choice :
SQL> with tbl as
  2  (select 1 col1, 'a' col2, to_date('08-JAN-08','DD-MON-YY') col3 from dual union all
  3   select 1 col1, 'a' col2, to_date('09-JAN-08','DD-MON-YY') col3 from dual union all
  4   select 1 col1, 'a' col2, to_date('10-JAN-08','DD-MON-YY') col3 from dual union all
  5   select 1 col1, 'b' col2, to_date('10-JAN-08','DD-MON-YY') col3 from dual union all
  6   select 1 col1, 'c' col2, to_date('11-mar-08','DD-MON-YY') col3 from dual union all
  7   select 1 col1, 'c' col2, to_date('10-mar-08','DD-MON-YY') col3 from dual )
  8  --end of data sample
  9  select t1.col1 ,t1.col2 ,t1.col3,t2.col3
10  from (select col1,col2,max(col3) col3
11        from tbl
12        group by col1,col2 ) t1,
13       (select col1, col2, col3
14        from (select col1, col2, col3,
15                     row_number() over (partition by col1, col2 order by col3 desc) rn
16              from tbl)
17       where rn <= 2) t2
18  where t1.col3 !=t2.col3;
      COL1 C COL3            COL3
         1 c 11-MAR-08       10-JAN-08
         1 a 10-JAN-08       09-JAN-08
         1 b 10-JAN-08       09-JAN-08[i] <-- obviously wrong, there is no 09-JAN-08 for 1b
         1 c 11-MAR-08       09-JAN-08[i] <-- obviously wrong, there is no 09-JAN-08 for 1c
         1 c 11-MAR-08       10-JAN-08[i] <-- obviously wrong, there is no 10-JAN-08 for 1c
         1 a 10-JAN-08       11-MAR-08[i] <-- obviously wrong, there is no 11-MAR-08 for 1a
         1 b 10-JAN-08       11-MAR-08[i] <-- obviously wrong, there is no 11-MAR-08 for 1b
         1 a 10-JAN-08       10-MAR-08[i] <-- obviously wrong, there is no 10-MAR-08 for 1a
         1 b 10-JAN-08       10-MAR-08[i] <-- obviously wrong, there is no 10-MAR-08 for 1b
         1 c 11-MAR-08       10-MAR-08[i] <-- obviously wrong, duplicate row
10 rows selected.Nicolas.