POSTGRES: How to select rows with a certain value only if another value doesn't exist, and in that case select the other value?

Question

Here is an example of the table I am working with. What I would like to achieve is to select the most recent row where the type is not 'NONE' unless 'NONE' is the only type available for that id.

id	date	type
123	01-01-2021	NONE
123	12-31-2021	NONE
123	01-01-2021	METAL
123	12-31-2021	METAL

From the example table above I would expect the query to return this

id	date	type
123	12-31-2021	METAL

If the table were to only contain types of "NONE" such as this example...

id	date	type
123	01-01-2021	NONE
123	12-31-2021	NONE
123	01-01-2021	NONE
123	12-31-2021	NONE

Then I would expect the result set to be..

id	date	type
123	12-31-2021	NONE

I've tried a plethora of different ways to do this but my current attempt looked something like this. It works when there's only one ID in the table but not for when I try to select a row for every specific ID in the table.

SELECT DISTINCT ON (id),
       date,
       type
FROM 
    example_table
WHERE
    CASE
        WHEN 
            (   SELECT 
                    COUNT(*) 
                FROM 
                    example_table t 
                WHERE 
                    t.type <> 'NONE'
                AND t.id = example_table.id) 
            <> 0
        THEN type <> 'NONE'
        ELSE 1=1
    END
ORDER BY 
    id, date DESC

 

Answer 1

You can use Row_number() function together with a case statement to identify which row to pick.

with cte AS
(
  select id,
       date, 
       type,
       row_number() over(partition by id 
                          order by case when type <> 'NONE' THEN 1 ELSE 2 END, date desc
                          ) as RN
       
  from test
 )
 select *
 from cte
 where rn = 1

SQL Fiddle

Answer 2

Without using Analytical function: SQL:

with cte
as
(
 select id,type,flag, max(date) date from (select id , date , type, case when type='NONE' then 'flag_none' else 'flag_not_none' end as flag from test) x group by  id,type,flag)
select id,type,date from cte where flag='flag_not_none'  and  (id,date) in (select id,max(date) from cte group by id)
union
select id,type,date from cte where id not in (select id from cte where flag='flag_not_none'  and  (id,date) in (select id,max(date) from cte group by id)) and flag='flag_none';

Full Execution:

CREATE TABLE test 
(

  id int,
  date date,
  type text
);

insert into test
values
(123,  '01-01-2021', 'NONE'),
(123, '12-31-2021', 'NONE'),
(123, '01-01-2021', 'METAL'),
(123, '12-31-2021', 'PLASTIC'),
(124,  '01-01-2021', 'NONE'),
(124, '12-31-2021', 'NONE'),
(124, '01-01-2021', 'NONE'),
(124, '12-31-2021', 'NONE'),
(125, '12-25-2021', 'NONE'),
(125, '12-25-2021', 'RUBBER'),
(125, '12-31-2021', 'STEEL');

postgres=# with cte
postgres-# as
postgres-# (
postgres(#  select id,type,flag, max(date) date from (select id , date , type, case when type='NONE' then 'flag_none' else 'flag_not_none' end as flag from test) x group by  id,type,flag)
postgres-# select id,type,date from cte where flag='flag_not_none'  and  (id,date) in (select id,max(date) from cte group by id)
postgres-# union
postgres-# select id,type,date from cte where id not in (select id from cte where flag='flag_not_none'  and  (id,date) in (select id,max(date) from cte group by id)) and flag='flag_none';
 id  |  type   |    date
-----+---------+------------
 123 | PLASTIC | 2021-12-31
 124 | NONE    | 2021-12-31
 125 | STEEL   | 2021-12-31
(3 rows)