Reputation: 1875
Given that I have following table table1
field1 field2
A B
A C
D B
D C
E F
G F
I am trying to get a query that partitions my data and ranks members within each partition. I am expecting that my query will look something like the one below but I am not sure how I can achieve the field1 OR field2
condition:
SELECT
field1, field2, ???? as partition,
RANK() OVER ( PARTITION BY field1 OR field2) as rank
FROM table1
field1 field2 partition rank
A B 1 1
A C 1 2
D B 1 3
D C 1 4
E F 2 1
G F 2 2
Upvotes: 2
Views: 334
Reputation: 168361
It's really not a simple/nice solution but this will do it:
Oracle Setup:
CREATE TABLE table_name (
field1 VARCHAR2(10),
field2 VARCHAR2(10)
);
INSERT INTO table_name
SELECT 'A', 'B' FROM DUAL UNION ALL
SELECT 'A', 'C' FROM DUAL UNION ALL
SELECT 'D', 'B' FROM DUAL UNION ALL
SELECT 'D', 'C' FROM DUAL UNION ALL
SELECT 'E', 'F' FROM DUAL UNION ALL
SELECT 'G', 'F' FROM DUAL;
CREATE OR REPLACE TYPE StringList AS TABLE OF VARCHAR2(50);
/
Query:
SELECT field1,
field2,
"partition",
ROW_NUMBER() OVER ( PARTITION BY "partition" ORDER BY field1, field2 )
AS "rank"
FROM (
SELECT field1,
f.COLUMN_VALUE AS field2,
DENSE_RANK() OVER ( ORDER BY partition_string ) AS "partition"
FROM (
SELECT field1,
CAST( COLLECT( field2 ORDER BY field2 ) AS StringList ) AS field2s,
LISTAGG( field2, ',' ) WITHIN GROUP ( ORDER BY field2 )
AS partition_string
FROM table_name
GROUP BY field1
) t,
TABLE( field2s ) f
);
Output:
FIELD1 FIELD2 partition rank
------ ------ ---------- ----
A B 1 1
A C 1 2
D B 1 3
D C 1 4
E F 2 1
G F 2 2
Upvotes: 2