How to select non empty value for columns from each group? - sql

Question

I have the following table.
| col1 | col2 | col3 | col4 |
| ---- | ---- | ---- | ---- |
| 123 | 345 | LA | 001 |
| 123 | 345 | AB | |
| 123 | 345 | LA | |
| 123 | 345 | | |
| abc | cde | LA | |
| abc | cde | | |
| abc | cde | AB | |
| abc | cde | LA | |
| ooo | zzz | LA | |
| ooo | zzz | LA | 001 |
| ooo | zzz | LA | |
| ooo | zzz | LA | 001 |
| 12 | 35 | LA | |
| 12 | 35 | LA | |
| 12 | 35 | LA | |
| 12 | 35 | LA | |
I would like to select for each group of col1 & col2, col3 must be AB if it is present in the group, else blank, and col4 should be the non-blank if non-blank value is present, else it should be blank. col4 for each group will either have a unique non-blank value or blank.
Desired output:
| col1 | col2 | col3 | col4 |
|------|------|------|------|
| 123 | 345 | AB | 001 |
| abc | cde | AB | |
| ooo | zzz | | 001 |
| 12 | 35 | | |
Desired code:
select col1, col2,
(AB if AB present else blank) as col3,
(non-blank value if non-blank is present else blank) as col4
from test
group by col1, col2;
What I have tried so far:
select col1, col2,
max(col4) as col4
from test
group by col1, col2;
I don't know how to do it for col3
SQL Fiddle: https://www.db-fiddle.com/f/dtEiDdu6Arn6VQTGwMXm4q/0
NOTE:
I require a vendor agnostic solution. I think an RDBMS specific solution might work in the platform I use, but I would have to test out that answer/solution and get back to you though.
I am not querying from the source, but from a grid, so I am not using an RDBMS at the backend.

Answer 1

For the column col3 you need a CASE statement and for the column col4 simply MAX() (or MIN()):
select
col1,
col2,
case when sum(case when col3 = 'AB' then 1 else 0 end) > 0 then 'AB' else null end col3,
max(col4) col4
from test
group by col1, col2;
See the demo.
Results:
| col1 | col2 | col3 | col4 |
| ---- | ---- | ---- | ---- |
| 12 | 35 | | |
| 123 | 345 | AB | 001 |
| abc | cde | AB | |
| ooo | zzz | | 001 |

Answer 2

This is a prioritization query. row_number() is often used:
select col1, col2, (case when col3 = 'AB' then col3 end) as col3,
col4
from (select t.*,
row_number() over (partition by col1, col2
order by (case when col3 = 'AB' then 1 else 2 end),
(case when col4 <> '' then 1 else 2 end)
) as seqnum
from t
) t
where seqnum = 1;

Answer 3

Check This-
SELECT B.col1,B.col2,B.N_Col3,MAX(B.col4)
FROM
(
SELECT your_table.*,A.col3 N_Col3
FROM your_table
LEFT JOIN
(
SELECT * FROM your_table
WHERE col3 = 'AB'
)A
ON A.col1 = CTE.col1 AND A.col2 = CTE.col2
)B
GROUP BY B.col1,B.col2,B.N_Col3
ORDER BY 3 DESC

Answer 4

select col1, col2, max(col3) col3, max(col4) col4
from
(
select col1, col2, case when col3 = 'AB' THEN 'AB' ELSE ' ' END col3, max(COL4) col4
from #table
group by col1, col2, case when col3 = 'AB' THEN 'AB' ELSE ' ' END
) A
group by col1, col2