A sql query needed

Question

Hi All,
I have belwo two record values.
select 1 as seq_no,'test1' as data_set from dual
union
select 1,'12test' from dual
union
select 2,'abcd' from dual
union
select 2,'ilm' from dual
union
select 2,'12test444' from dual
here i need to extract the below o/p from from the above dataset
1,12test
1,test1
2,12test444
select 1 as seq_no,'test1' as data_set from dual
union
select 1,'12test' from dual
union
select 2,'abcd' from dual
union
select 2,'ilm' from dual
here i need to extract the below o/p from from the above dataset
1,12test
1,test1
2,abcd
2,ilm
The logic behind the above two types of o/p is
if the search string 'test' is present in any data set (data set is defined by same seq_no values) then print the whole row containing the search string. As in the first record value data set with seq_val 1 and 2 both have test string 'test' , o/p is all the rows contating the search string 'test'
If the search string 'test' is not present in any data set (data set is defined by same seq_no values) then print all other rows not containing the search string. As in the second record value data set with seq_val 2 both doesn't have test string 'test' , o/p is all the rows not contating the search string 'test' in data test with seq_val 1as well as all the rows contating the search string 'test' in data set with seq_val 2
Hope i could explain my o/p logic and i need to perform this in a sql query.
Waiting for your responses.
Thanks in advance.

Answer 1

Hi,
Here's one way:
WITH     got_rnk          AS
     SELECT     seq_no, data_set
     ,     DENSE_RANK () OVER ( PARTITION BY  seq_no
                               ORDER BY          CASE
                                   WHEN  INSTR (data_set, 'test') > 0
                                   THEN  1
                                   ELSE  2
                                 END
                       )     AS rnk
     FROM    table_x
SELECT     seq_no, data_set
FROM     got_rnk
WHERE     rnk     = 1
;The CASE expression returns 1 if data_set contains 'test', and it returns 2 if it does not. Let's call this number x.
You're not really interested in x itself. Rather, you want all the rows that have the lowest value of x among rows with the same seq_no. That is, if any row with a given_seq_num has x=1, then you want all rows with that seq_no and x=1, but it the lowest x for the a seq_no is 2, then you want the rows with x=2 instead.
DENSE_RANK returns the the number 1 if the given riow is the lowest (or tied for the lowest) in its partition, so if the are any rows with x=1 in the group, rnk=1 will correspond to x=1. But if all the rows with the same seq_no have x=2, then rnk=1 will correspond to x=2.