Can I use analytical function in this problem?

Question

Hi,
I want to use query only for the following . I don't want to wright any function or procedure for this.
create temp table test_3 (user_id number, auth_id number);
insert into test_3 values (133,609);
insert into test_3 values (133,610);
insert into test_3 values (133,611);
insert into test_3 values (133,612);
insert into test_3 values (133,613);
insert into test_3 values (133,614);
insert into test_3 values (144,1);
insert into test_3 values (134,610);
insert into test_3 values (135,610);
insert into test_3 values (135,610);
insert into test_3 values (135,610);
insert into test_3 values (136,610);
insert into test_3 values (136,610);
insert into test_3 values (137,610);
insert into test_3 values (137,610);
insert into test_3 values (137,609);
insert into test_3 values (137,11);
I want to count:
1. for each auth_id, how many users are there who is assigned to this aut_id only
example
user_id 134 and 135 is assigned to auth_id 610 only and the count is 3 and 2 respectively .
user_id 144 is assigned to auth_id 1 only and the count is 1.
2.how many user_id is common between auth_id 609 and 610
how many user_id is common between auth_id 609 and 611
how many user_id is common between auth_id 609 and 612
and so on.
I have re-written the problem bellow
Regards,
Edited by: user576726 on Feb 13, 2011 3:54 AM

Answer 1

Hi,
user576726 wrote:
Hi,
Thanks for the response.
drop table test_3;
create table test_3 (user_id number, auth_id number);
insert into test_3 values (133,609);     --row 1
...Thanks. That makes the problem a lot clearer.
My desired output is:
auth_id_1 auth_id_2 count1 count2
1 12 1 --(user_id 144) 2 --(row 15, row 16)
1 610 1 --(user_id 144) 1 --(row 19)
11 609 1 --(user_id 137) 1 --(row 13)
11 610 1 --(user_id 137) 2 --(row 11, row 12)
12 1 1 --(user_id 144) 1 --(row 4)
12 610 1 --(user_id 144) 1 --(row 19)
609 11 1 --(user_id 137) 1 --(row 14)
609 610 2 --(user_id 133 & 137) 3      --(row 2, row 11 and row 12)
609 611 1 --(user_id 133) 1 --(row 3)
610 1 1 --(user_id 144) 1 --(row 4)
610 11 1 --(user_id 137) 1 --(row 14)
610 12 1 --(user_id 144) 2 --(row 15, row 16)
610 609 2 --(user_id 133 & 137) 4 --(row 1, row 13, row 17 and row 18)
610 611 1 --(user_id 133) 1 --(row 3)
611 609 1 --(user_id 133) 3 --(row 1, row 17 and row 18)
611 610 1 --(user_id 133) 1 --(row 2)               1 --(user_id 133)               1 --(row 2)
Count1 is the number of common different user id between auth_id_1 and auth_id_2
example
for the first row in the output:-
common user ids between 609 and 610 are 133 and 137. so the count1 should be 2
Count2 is how many rows are there for auth_id_2 where user id is common for auth_id_1 and auth_id_2
example
for the first row in the output:-
the common user_id for 609 and 610 are 133 & 137
the rows in the test_3 table that has auth_id 610 and user_id 133 & 137 are
row 2, row 11 and row 12 so the count is 3.
What I have done is
I have writtent the following query to get the first two columns of the output:
select tab1.auth_id auth_id_1, tab2.auth_id auth_id_2
from
(select user_id, auth_id
from test_3
group by user_id, auth_id
) tab1,
(select user_id, auth_id
from test_3
group by user_id, auth_id
) tab2
where tab1.user_id = tab2.user_id
and tab1.auth_id <> tab2.auth_id
group by tab1.auth_id, tab2.auth_id
order by 1,2;You're on the right track. You're doing a self-join and getting the right combinations of auth_id_1 and auth_id_2.
Why are you doing the GROUP BY in sub-queries tab1 and tab2? Eventually, you'll need to count identical rows, like these:
insert into test_3 values (137,610);     --row 11
insert into test_3 values (137,610);     --row 12If you do a GROUP BY in the sub-queries, all you'll know is that user_id=137 was related to auth_id=610. You won't know how many times, which is what count2 is based on. So don't do a GROUP BY in the sub-queries; just do the GROUP BY in the main query. That means you won't need to do sub-queries; you might as well just join two copies of the original test_3 table.
Count1 is the number of common different user id between auth_id_1 and auth_id_2Great; that's very clear. In SQL, how do you count the number of different user_ids in such a group? (Hint "different" means the same thing as "distinct".)
Count2 is how many rows are there for auth_id_2 where user id is common for auth_id_1 and auth_id_2
example
for the first row in the output:-The first row in the output you posted was
1 12 1 --(user_id 144) 2 --(row 15, row 16)Isn't this one that you're explaining here the 8th row of output?
the common user_id for 609 and 610 are 133 & 137
the rows in the test_3 table that has auth_id 610 and user_id 133 & 137 are
row 2, row 11 and row 12 so the count is 3.So, for count2, you want to know how many distinct rows from tab2 are in each group. If you had a primary key in the table, or anything that uniquely identified the rows, you could count the distinct occurrences of that, but you're not storing anything unique on each row (at least you haven't mentioned it in your sample data). If that's really the case, then this is one place where the ROWID pseudocolumn is handy; it uniquely identifies any row in any table, so you can just count how many different values of tab2.ROWID are in each group.