Counting Combinations - sql

Question

I have an Oracle table with multiple columns some populated with a variable, there a large number of possible variables the example below is not exhaustive.
ID Col1 Col2 Col3 Col4 Col5 Col6
-------------------------------------
1 X2 B2
2 C3 D1 R4
3 B2 X2
4 E4 T1 W2
5 X2 B2
6 R4 D1
7 D1 R4 C3
I need to identify the number of distinct combinations where row 1, row 3 and row 5 in the above example are considered the same combination and rows 2 and 7 are also considered the same. So the desired result would look like:
Col1 Col2 Col3 Col4 Col5 Col6 Count(*)
------------------------------------------------
B2 X2 3
C3 D1 R4 2
E4 T1 W2 1
D1 R4 1
But if I use this:
SELECT Col1, Col2, Col3, Col4, Col5, Col6, Count(*)
FROM MyTable
GROUP BY Col1, Col2, Col3, Col4, Col5, Col6
ORDER BY Count(*) DESC
Then row 3 in my data is considered unique. However, it has the same combination as rows 1 and row 5. Also row 2 and 7 are not considered the same and the result is:
Col1 Col2 Col3 Col4 Col5 Col6 Count(*)
------------------------------------------------
X2 B2 2
C3 D1 R4 1
B2 X2 1
E4 T1 W2 1
R4 D1 1
D1 R4 C3 1
It looks like I need to sort the col variables before comparing them. But is there an elegant solution to doing this for large record sets (3 million+ records) with up to 20 columns of data in Oracle?

Answer 1

There are two ways that come to my mind. First you can write a function accepting six or more strings and concatenating them in order. Then:
select colstring, count(*)
from
(
select id, concat_sorted(col1, col2, col3, col4, col5, col6) as colstring
from MyTable
)
group by colstring;
Another way would be to make each column a separate record and use listagg on them, provided you have Oracle 11g or higher available:
select colstring, count(*)
from
(
select id, listagg (colx, ',') within group (order by colx) as colstring
from
(
select id, col1 as colx from MyTable
union all
select id, col2 from MyTable
union all
select id, col3 from MyTable
union all
select id, col4 from MyTable
union all
select id, col5 from MyTable
union all
select id, col6 from MyTable
)
group by id
)
group by colstring

Answer 2

Try like this,
WITH t AS (
SELECT 1 ID, 'X2' col1, 'B2' col2, NULL col3, NULL col4, NULL col5, NULL col6 FROM dual
UNION
SELECT 2, 'C3', 'D1', 'R4', NULL, NULL, NULL FROM dual
UNION
SELECT 3, 'B2', 'X2', NULL, NULL, NULL, NULL FROM dual
UNION
SELECT 4, 'E4', 'T1', 'W2', NULL, NULL, NULL FROM dual
UNION
SELECT 5, 'X2', 'B2', NULL, NULL, NULL, NULL FROM dual
UNION
SELECT 6, 'R4', 'T1', NULL, NULL, NULL, NULL FROM dual
UNION
SELECT 7, 'D1', 'R4', 'C3', NULL, NULL, NULL FROM dual
)
SELECT col1, col2, col3, col4, col5, col6, tot_count
FROM (
SELECT col1, col2, col3, col4, col5, col6, cnt,
MAX(cnt) OVER (PARTITION BY val) AS tot_count,
row_number() OVER (PARTITION BY val ORDER BY cnt DESC) AS rn
FROM (
SELECT col1, col2, col3, col4, col5, col6, val, count(*) OVER (PARTITION BY val) cnt
FROM (
SELECT A.ID, col1, col2, col3, col4, col5, col6, val
FROM (SELECT ID, col1, col2, col3, col4, col5, col6
FROM t
) A,
(SELECT ID, listagg( val,',') WITHIN GROUP(ORDER BY val DESC) AS val
FROM (
SELECT ID, val
FROM t
unpivot ( val FOR origin IN (col1, col2, col3, col4, col5, col6))
)
GROUP BY ID
)b
WHERE A.ID = b.ID
)
ORDER BY val
)t1
)t2
WHERE tot_count = cnt
AND rn = 1
ORDER BY tot_count DESC;