How to fill NULL value using min-max in SQL Server? - sql

Question

I want to get front value and back value using min-max comparing.
Here is my sample table.
https://dbfiddle.uk/?rdbms=sqlserver_2017&fiddle=322aafeb7970e25f9a85d0cd2f6c00d6
For example, this is temp01 table.
I want to get start&end NULL value in temp01.
So, I fill
price = [NULL, NULL, 13000]
to
price = [12000, 12000, 13000]
Because the 12000 is minimum value in [230, 1]. And I fill end NULL is filled maximum value in [cat01, cat02] group
| SEQ | cat01 | cat02 | dt_day | price |
+-----+-------+-------+------------+-------+
| 1 | 230 | 1 | 2019-01-01 | NULL |
| 2 | 230 | 1 | 2019-01-02 | NULL |
| 3 | 230 | 1 | 2019-01-03 | 13000 |
...
| 11 | 230 | 1 | 2019-01-11 | NULL |
| 12 | 230 | 1 | 2019-01-12 | NULL |
| 1 | 230 | 2 | 2019-01-01 | NULL |
| 2 | 230 | 2 | 2019-01-02 | NULL |
| 3 | 230 | 2 | 2019-01-03 | 12000 |
...
| 12 | 230 | 2 | 2019-01-11 | NULL |
| 13 | 230 | 2 | 2019-01-12 | NULL |
[result]
| SEQ | cat01 | cat02 | dt_day | price |
+-----+-------+-------+------------+-------+
| 1 | 230 | 1 | 2019-01-01 | 12000 | --START
| 2 | 230 | 1 | 2019-01-02 | 12000 |
| 3 | 230 | 1 | 2019-01-03 | 13000 |
| 4 | 230 | 1 | 2019-01-04 | 12000 |
| 5 | 230 | 1 | 2019-01-05 | NULL |
| 6 | 230 | 1 | 2019-01-06 | NULL |
| 7 | 230 | 1 | 2019-01-07 | 19000 |
| 8 | 230 | 1 | 2019-01-08 | 20000 |
| 9 | 230 | 1 | 2019-01-09 | 21500 |
| 10 | 230 | 1 | 2019-01-10 | 21500 |
| 11 | 230 | 1 | 2019-01-11 | 21500 |
| 12 | 230 | 1 | 2019-01-12 | 21500 |
| 13 | 230 | 1 | 2019-01-13 | 21500 | --END
| 1 | 230 | 2 | 2019-01-01 | 12000 | --START
| 2 | 230 | 2 | 2019-01-02 | 12000 |
| 3 | 230 | 2 | 2019-01-03 | 12000 |
| 4 | 230 | 2 | 2019-01-04 | 17000 |
| 5 | 230 | 2 | 2019-01-05 | 22000 |
| 6 | 230 | 2 | 2019-01-06 | NULL |
| 7 | 230 | 2 | 2019-01-07 | 23000 |
| 8 | 230 | 2 | 2019-01-08 | 23200 |
| 9 | 230 | 2 | 2019-01-09 | NULL |
| 10 | 230 | 2 | 2019-01-10 | 24000 |
| 11 | 230 | 2 | 2019-01-11 | 24000 |
| 12 | 230 | 2 | 2019-01-12 | 24000 |
| 13 | 230 | 2 | 2019-01-13 | 24000 | --END
Please, let me know what a good way to fill NULL using linear relationships.

Answer 1

find the min() and max() for the price GROUP BY cat01, cat02.
Also find the min and max seq for the row where price is not null
after that it is just simply inner join to your table and update where price is null
with val as
(
select cat01, cat02,
min_price = min(price),
max_price = max(price),
min_seq = min(case when price is not null then seq end),
max_seq = max(case when price is not null then seq end)
from temp01
group by cat01, cat02
)
update t
set price = case when t.seq < v.min_seq then min_price
when t.seq > v.max_seq then max_price
end
FROM temp01 t
inner join val v on t.cat01 = v.cat01
and t.cat02 = v.cat02
where t.price is null
dbfiddle
EDIT : returning the price as a new column in SELECT query
with val as
(
select cat01, cat02, min_price = min(price), max_price = max(price),
min_seq = min(case when price is not null then seq end),
max_seq = max(case when price is not null then seq end)
from temp01
group by cat01, cat02
)
select t.*,
new_price = coalesce(t.price,
case when t.seq < v.min_seq then min_price
when t.seq > v.max_seq then max_price
end)
FROM temp01 t
left join val v on t.cat01 = v.cat01
and t.cat02 = v.cat02
Updated dbfiddle