'Oracle - Rank and transform on columns

I got a sample data as below in this cte_orders. Each row is a order and contains the user_id, zone_code and zone_name.

I need to rank the top 3 zone_code for each user_id, and I need the output be on columns. Like this

|user_id |  top1_zone  |  top2_zone  |  top3_zone  |
|--------+-------------+-------------+-------------+
|1000    |  5555-ABCD  |  4567-ZMNY  |  7888-IXPO  |
|--------+-------------+-------------+-------------+
|9999    |  3456-JJKL  |  7688-HBGT  |  5555-ABCD  |
|--------+-------------+-------------+-------------+

Here's the SQL with the test data and the query that I'm trying. The problem with this query is the result is given a row for each position of the rank.

(
    SELECT 1 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
    SELECT 2 as order_id, '1000' as user_id, '4567' as zone_code, 'ZMNY' as zone_name from dual union all
    SELECT 3 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
    SELECT 4 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
    SELECT 5 as order_id, '1000' as user_id, '5599' as zone_code, 'HZTR' as zone_name from dual union all
    SELECT 6 as order_id, '1000' as user_id, '4567' as zone_code, 'ZMNY' as zone_name from dual union all
    SELECT 7 as order_id, '1000' as user_id, '4567' as zone_code, 'ZMNY' as zone_name from dual union all
    SELECT 8 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
    SELECT 9 as order_id, '1000' as user_id, '1999' as zone_code, 'LNKJ' as zone_name from dual union all
    SELECT 10 as order_id, '1000' as user_id, '5677' as zone_code, 'OPLH' as zone_name from dual union all
    SELECT 11 as order_id, '1000' as user_id, '7888' as zone_code, 'IXPO' as zone_name from dual union all
    SELECT 12 as order_id, '9999' as user_id, '5599' as zone_code, 'HZTR' as zone_name from dual union all
    SELECT 13 as order_id, '9999' as user_id, '3456' as zone_code, 'JJKL' as zone_name from dual union all
    SELECT 14 as order_id, '9999' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
    SELECT 15 as order_id, '9999' as user_id, '3456' as zone_code, 'JJKL' as zone_name from dual union all
    SELECT 16 as order_id, '9999' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
    SELECT 17 as order_id, '9999' as user_id, '7688' as zone_code, 'HBGT' as zone_name from dual union all
    SELECT 17 as order_id, '9999' as user_id, '7688' as zone_code, 'HBGT' as zone_name from dual union all
    SELECT 18 as order_id, '9999' as user_id, '1566' as zone_code, 'LNOI' as zone_name from dual union all
    SELECT 19 as order_id, '9999' as user_id, '3456' as zone_code, 'JJKL' as zone_name from dual union all
    SELECT 20 as order_id, '9999' as user_id, '7654' as zone_code, 'NNJJ' as zone_name from dual union all
    SELECT 21 as order_id, '9999' as user_id, '4433' as zone_code, 'NHJE' as zone_name from dual union all
    SELECT 22 as order_id, '9999' as user_id, '4111' as zone_code, 'ABHJ' as zone_name from dual

)

select 
    user_id,
    CASE WHEN rank_zones = 1 then zone_concat end as top1_zone,
    CASE WHEN rank_zones = 2 then zone_concat end as top2_zone,
    CASE WHEN rank_zones = 3 then zone_concat end as top3_zone
    from

( 
    select  
        user_id, 
        zone_code||'-'||zone_name as zone_concat,
        row_number() over (partition by user_id order by count(*) desc) rank_zones 
    from cte_orders
    group by
        user_id,
        zone_code||'-'||zone_name

)
where rank_zones <= 3

group by 
    user_id,
    CASE WHEN rank_zones = 1 then zone_concat end,
    CASE WHEN rank_zones = 2 then zone_concat end,
    CASE WHEN rank_zones = 3 then zone_concat end
order by user_id;

The output that i getting

|user_id |  top1_zone  |  top2_zone  |  top3_zone  |
|--------+-------------+-------------+-------------+
|1000    |  5555-ABCD  |    (null)   |    (null)   |
|--------+-------------+-------------+-------------+
|1000    |    (null)   |  4567-ZMNY  |    (null)   |
|--------+-------------+-------------+-------------+
|1000    |    (null)   |    (null)   |  7888-IXPO  |
|--------+-------------+-------------+-------------+
|9999    |  3456-JJKL  |    (null)   |    (null)   |
|--------+-------------+-------------+-------------+
|9999    |    (null)   |  7688-HBGT  |    (null)   |
|--------+-------------+-------------+-------------+
|9999    |    (null)   |    (null)   |  5555-ABCD  |
|--------+-------------+-------------+-------------+

How can i fix my query to get one row for each user_id without the nulls?



Solution 1:[1]

How can i fix my query to get one row for each user_id without the nulls?

With a little bit of aggregation (and comments, as its consequence).

Sample data:

SQL> with cte_orders as (
  2      SELECT 1 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
  3      SELECT 2 as order_id, '1000' as user_id, '4567' as zone_code, 'ZMNY' as zone_name from dual union all
  4      SELECT 3 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
  5      SELECT 4 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
  6      SELECT 5 as order_id, '1000' as user_id, '5599' as zone_code, 'HZTR' as zone_name from dual union all
  7      SELECT 6 as order_id, '1000' as user_id, '4567' as zone_code, 'ZMNY' as zone_name from dual union all
  8      SELECT 7 as order_id, '1000' as user_id, '4567' as zone_code, 'ZMNY' as zone_name from dual union all
  9      SELECT 8 as order_id, '1000' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
 10      SELECT 9 as order_id, '1000' as user_id, '1999' as zone_code, 'LNKJ' as zone_name from dual union all
 11      SELECT 10 as order_id, '1000' as user_id, '5677' as zone_code, 'OPLH' as zone_name from dual union all
 12      SELECT 11 as order_id, '1000' as user_id, '7888' as zone_code, 'IXPO' as zone_name from dual union all
 13      SELECT 12 as order_id, '9999' as user_id, '5599' as zone_code, 'HZTR' as zone_name from dual union all
 14      SELECT 13 as order_id, '9999' as user_id, '3456' as zone_code, 'JJKL' as zone_name from dual union all
 15      SELECT 14 as order_id, '9999' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
 16      SELECT 15 as order_id, '9999' as user_id, '3456' as zone_code, 'JJKL' as zone_name from dual union all
 17      SELECT 16 as order_id, '9999' as user_id, '5555' as zone_code, 'ABCD' as zone_name from dual union all
 18      SELECT 17 as order_id, '9999' as user_id, '7688' as zone_code, 'HBGT' as zone_name from dual union all
 19      SELECT 17 as order_id, '9999' as user_id, '7688' as zone_code, 'HBGT' as zone_name from dual union all
 20      SELECT 18 as order_id, '9999' as user_id, '1566' as zone_code, 'LNOI' as zone_name from dual union all
 21      SELECT 19 as order_id, '9999' as user_id, '3456' as zone_code, 'JJKL' as zone_name from dual union all
 22      SELECT 20 as order_id, '9999' as user_id, '7654' as zone_code, 'NNJJ' as zone_name from dual union all
 23      SELECT 21 as order_id, '9999' as user_id, '4433' as zone_code, 'NHJE' as zone_name from dual union all
 24      SELECT 22 as order_id, '9999' as user_id, '4111' as zone_code, 'ABHJ' as zone_name from dual
 25  )

This is what makes what you ask for (the MAX aggregate function; lines #28 - 30):

 26  select
 27      user_id,
 28      max(CASE WHEN rank_zones = 1 then zone_concat end) as top1_zone,
 29      max(CASE WHEN rank_zones = 2 then zone_concat end) as top2_zone,
 30      max(CASE WHEN rank_zones = 3 then zone_concat end) as top3_zone
 31      from
 32  (
 33      select
 34          user_id,
 35          zone_code||'-'||zone_name as zone_concat,
 36          row_number() over (partition by user_id order by count(*) desc) rank_zones
 37      from cte_orders
 38      group by
 39          user_id,
 40          zone_code||'-'||zone_name
 41  )
 42  where rank_zones <= 5

Comment CASE expressions (lines #45 - 47):

 43  group by
 44      user_id
 45  --    CASE WHEN rank_zones = 1 then zone_concat end,
 46  --    CASE WHEN rank_zones = 2 then zone_concat end,
 47  --    CASE WHEN rank_zones = 3 then zone_concat end
 48  order by user_id;

And the result is then:

USER TOP1_ZONE TOP2_ZONE TOP3_ZONE
---- --------- --------- ---------
1000 5555-ABCD 4567-ZMNY 5599-HZTR
9999 3456-JJKL 5555-ABCD 7688-HBGT

SQL>

Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution Source
Solution 1 Littlefoot