当 SQL 中有多个继续 Null 值时，如何根据前几个月的值填充 Null 值

Question

Below is the input table:

Month	Value
1	200
2	-
3	-
4	300
5	-

Expected Output:

Month	Value
1	200
2	200
3	200
4	300
5	300

I did try using LAG function in SQL, as a result I was able to populate value for the immediate NULL values which is Month 2 in above case but next month which is Month 3 was still having Null values

Answer 1

You can build a query using the LEAD function to convert your data into ranges as the following:

from_month	to_month	Value
1	3	200
4	5	300

Then use the update join statement to update your table as the following:

Update table_name Set Value = D.Value
From table_name T Join
(
  Select Month As from_month, 
       Lead(Month, 1, (select max(month) from table_name) +1)
       Over(Order By Month) -1 As to_month,
       Value
  From table_name 
  Where Value Is Not Null
) D
On T.month Between D.from_month And D.to_month
Where T.Value Is Null;

See demo

Answer 2

You can emulate the LAST_VALUE function to ignore nulls by creating partitions with:

• one non-null value
• following null values (excluding next non-null value)

then updating null values to the max for each partition.

WITH cte AS (
    SELECT *, COUNT(Value) OVER(ORDER BY Month) AS partitions
    FROM tab
), cte2 AS (
    SELECT Month, MAX(Value) OVER(PARTITION BY partitions) AS Value 
    FROM cte
)
UPDATE tab
SET tab.Value = cte2.Value
FROM cte2
WHERE tab.Month = cte2.Month;

Check the demo here .

Answer 3

You can do this with a simple correlation and an updatable table expression:

update u set value = v
from (
    select *, 
        (select top(1) value 
        from t t2 where t2.month < t.month 
            and t2.value is not null 
            order by month desc
        )v
    from t
    where t.value is null
)u;

Answer 4

If you don't like the old hat solutions you can do it in one pass too:

with testdata as (
    select *
    from    (
        values (1, 200), (2, 50), (3,NULL), (4,300), (5, NULL), (6, NULL), (7, 10)
        ) x (mon, val)
    )
select mon, val AS orig
,   cast(substring(maxvalue2, 31, 30) as int) AS new
FROM (
        SELECT  MAX(CAST(mon AS BINARY(30)) + CAST(val AS BINARY(30))) OVER(ORDER BY mon) AS maxValue2
        ,   *
        FROM    testdata
    ) x