SQL 改造-DB2

Question

I have a dataset in DB2. The PK column is PK_Ind and there is a Logic_Ind column as well. The transformation I am looking for is for the same PK column if Logic_Ind is NULL then try to find a match where END_DATE of that specific row is BETWEEN START_DATE and END_DATE of any other row for same PK_Ind and Logic_Ind is NOT NULL (ie Logic_Ind='A'). If match found then the values for Col1, Col2 and Col3 should be taken and populated for the row where Logic_Ind is NULL.

In the below example 3rd row should find a match with 2nd row and 7th row should match with 4th row and Col1, COl2 and Col3 should reflect the modified values 3rd and 7th row.

But if there is no match found then still the same record should be presented as it is in the source dataset.

Source dataset:

Target dataset:

Answer 1

Try this:

SELECT A.PK_IND, A.LOGIC_IND
, NVL2(A.LOGIC_IND, A.COL1, B.COL1) AS COL1
, NVL2(A.LOGIC_IND, A.COL2, B.COL2) AS COL2
, NVL2(A.LOGIC_IND, A.COL3, B.COL3) AS COL3
, A.START_DATE, A.END_DATE
FROM MYTAB A
LEFT JOIN TABLE
(
SELECT B.COL1, B.COL2, B.COL3
FROM MYTAB B
WHERE B.PK_IND = A.PK_IND 
    AND B.LOGIC_IND IS NOT NULL 
    AND A.END_DATE BETWEEN B.START_DATE AND B.END_DATE
FETCH FIRST 1 ROW ONLY  
) B ON A.LOGIC_IND IS NULL;

db<>fiddle example

Answer 2

You can perform an left join with all possible matches (when Logic_Ind is null) and then pick one of them, using ROW_NUMBER() .

For example, you can do:

select *
from (
  select
    a.pk_ind,
    a.logic_ind,
    case when a.logic_ind is null then b.col1 else a.col1 end as col1,
    case when a.logic_ind is null then b.col2 else a.col2 end as col2,
    case when a.logic_ind is null then b.col3 else a.col3 end as col3,
    a.start_date,
    a.end_date,
    row_number() over(partition by a.pk_ind, a.start_date 
                      order by b.start_date) as rn
  from t a
  left join t b on b.pk_ind = a.pk_ind
    and a.logic_ind is null
    and b.end_date between a.start_date and a.end_date
) x
where rn = 1

Note : As a side note, for the ROW_NUMBER() function to work as needed, you need to provide a PK for the table. Since PK_Ind is not a primary key (your data example shows duplicate values), I guessed the tuple ( a.pk_ind , a.start_date ) could act as one. Change as needed.