Challenging SQL...

Question

Hi Everyone,
I'm have a table that has the following kind of structure.
Table X
cont_id Number
rid Number
cont_nm_flg Char(1)
bus_nm_flg Char(1)
bt_add_flg Char(1)
Upd_date Date
Now the table will have multiple cont_id and also duplicates.I have to roll up ( group by) by cont_id. But here is the hitch :
I have to select on based on bt_add_flg. If this field had a value 1 then we select that record. If that is not '1' then get the record having the most recent upd_date and if there is a tie (means thare are two recs having same max Upd_date) then get the record having max rid.
Could anyone help me with this type of SQL...
Thanks in Advance...

Answer 1

Given the following data:
SQL> SELECT rownum,cont_id,rid,cont_nm_flg,bus_nm_flg,bt_add_flg,upd_date
  2  FROM x;
    ROWNUM    CONT_ID        RID C B B UPD_DATE
         1          1          1 a a 1 12-nov-2002 13:12:59
         2          2          1 a a x 12-nov-2002 13:05:00
         3          2          2 a a y 12-nov-2002 13:05:00
         4          3          1 a a x 11-nov-2002 13:12:59
         5          3          2 a a x 12-nov-2002 13:12:59if I understand your requirement correctly, I would expect to get the following rows returned.
Rownum=1 by rule a) bt_add_flg = '1'
Rownum=3 by rule c) max(upd_date) and max(rid) to break the tie
Rownum 5 by rule b) max(upd_date) no tie to break
So,
SQL> SELECT *
  2  FROM x
  3  WHERE exists (SELECT 1
  4                FROM x sub_x
  5                WHERE x.cont_id = sub_x.cont_id and
  6                      bt_add_flg = '1')
  7  UNION ALL --Changed this to avoid a sort
  8  SELECT *
  9  FROM x
10  WHERE to_char(upd_date,'yyyymmddhh24miss')||
11        to_char(rid) = (SELECT MAX(to_char(sub_x.upd_date,'yyyymmddhh24miss')
||
12                                   to_char(sub_x.rid))
13                        FROM x sub_x
14                        WHERE x.cont_id = sub_x.cont_id) and
15        NOT EXISTS (SELECT 1
16                    FROM x sub_x2
17                    WHERE x.cont_id = sub_x2.cont_id and
18                          sub_x2.bt_add_flg = '1');
   CONT_ID        RID C B B UPD_DATE
         1          1 a a 1 12-nov-2002 13:12:59
         2          2 a a y 12-nov-2002 13:05:00
         3          2 a a x 12-nov-2002 13:12:59as expected, 3 rows, one for each cont_id despite having duplicate rows for cont_id 2 and cont_id 3.
If it is possible that there may be multiple rows for a cont_id one with bt_add_flg = 1 and one or more with bt_add_flg <> 1, then you need to add one more condition to the top select in the union:
INSERT INTO x VALUES (1,2,'a','a','z',sysdate + 3/24);
SELECT *
FROM x
WHERE EXISTS (SELECT 1
              FROM x sub_x
              WHERE x.cont_id = sub_x.cont_id and
                    sub_x.bt_add_flg = '1') and
bt_add_flg = '1'
UNION ALL 
SELECT *
FROM x
WHERE to_char(upd_date,'yyyymmddhh24miss')||
      to_char(rid) = (SELECT MAX(to_char(sub_x.upd_date,'yyyymmddhh24miss')||
                                 to_char(sub_x.rid))
                      FROM x sub_x
                      WHERE x.cont_id = sub_x.cont_id) and
      NOT EXISTS (SELECT 1
                  FROM x sub_x2
                  WHERE x.cont_id = sub_x2.cont_id and
                        sub_x2.bt_add_flg = '1')John