Efficient way to pull counts for all permutations of a field - sql

Question

I have an oracle DB w/ a table that contains records associated to a person (based on an ID). The records are categorized as category = 1, 2, or 3.
I would like to pull as follows:
- # of people with only a category 1 record (no category=2 or 3)
- # of people with only a category 2 record (no category=1 or 3)
- # of people with only a category 3 record (no category=1 or 2)
- # of people with both category 1 & 2 records (no category=3)
- # of people with both category 1 & 3 records (no category=2)
- # of people with all category records 1,2, & 3
- # of people with both a category 2 & 3 records (no category=1)
I could only think of the following solution (modified for each case):
select count(*) from table1
where id in (select id from table1 where category=1)
and id not in (select id from table1 where category=2)
and id not in (select id from table1 where category=3)
But, I believe this is a highly inefficient way of doing this, was wondering if anyone had quicker/better way of getting this info.
Thanks!

Answer 1

One way to do this is to bring the categories together, using listagg() and then reaggregate:
select categories, count(*)
from (select listagg(t1.category, ',') within group (order by t1.category) as categories, personid
from table1 t1
group by personid
) x
group by categories;
EDIT:
If you need distinct values:
select categories, count(*)
from (select listagg(t1.category, ',') within group (order by t1.category) as categories, personid
from (select distinct t1.category, t1.personid from table1 t1) t1
group by personid
) x
group by categories;

Answer 2

Here is a query that, for each ID, shows the count of distinct categories and the MIN and MAX category. This query can be used as a sub-query in further processing (you didn't explain exactly HOW you want the results to be presented). When the COUNT is 1, then the single category is that in the MIN_CAT column; when the COUNT is 3, then all three categories are present for that ID; and when the COUNT is 2, then the two categories that are present are in the MIN and the MAX columns. Whatever else you need to do from here should be very simple; for example you can now GROUP BY CT, MIN_CAT, MAX_CT and count ID's.
I do a count(distinct category) to allow the possibility of non-unique (id, category) - as illustrated in the sample data I include in a WITH clause (which is NOT part of the SQL query!)
with
test_data ( id, category ) as (
select 101, 3 from dual union all
select 101, 1 from dual union all
select 101, 3 from dual union all
select 104, 2 from dual union all
select 105, 2 from dual union all
select 105, 2 from dual union all
select 105, 1 from dual union all
select 106, 1 from dual union all
select 106, 2 from dual union all
select 106, 3 from dual union all
select 106, 3 from dual
)
select id,
count(distinct category) as ct,
min(category) as min_cat,
max(category) as max_cat
from test_data
group by id
;
ID CT MIN_CAT MAX_CAT
--- -- ------- -------
101 2 1 3
105 2 1 2
104 1 2 2
106 3 1 3

Answer 3

Oracle Setup:
CREATE TABLE test_data ( id, category ) as
select 101, 3 from dual union all
select 101, 1 from dual union all
select 101, 3 from dual union all
select 104, 2 from dual union all
select 105, 2 from dual union all
select 105, 2 from dual union all
select 105, 1 from dual union all
select 106, 1 from dual union all
select 106, 2 from dual union all
select 106, 3 from dual union all
select 106, 3 from dual union all
select 107, 1 from dual union all
select 107, 3 from dual;
Query:
SELECT c1,
c2,
c3,
LTRIM(
DECODE( c1, 1, ',1' ) || DECODE( c2, 1, ',2' ) || DECODE( c3, 1, ',3' ),
','
) AS categories,
COUNT(1) AS num_people,
LISTAGG( id, ',' ) WITHIN GROUP ( ORDER BY id ) AS people
FROM ( SELECT DISTINCT * FROM test_data )
PIVOT ( COUNT(1) FOR category IN ( 1 AS c1, 2 AS c2, 3 AS c3 ) )
GROUP BY c1, c2, c3;
Output:
C1 C2 C3 CATEGORIES NUM_PEOPLE PEOPLE
-- -- -- ---------- ---------- ----------
0 1 0 2 1 104
1 0 1 1,3 2 101,107
1 1 0 1,2 1 105
1 1 1 1,2,3 1 106