Computed column formula for row_number function - sql

Question

I am trying to find a formula for a computed column that will return same results as the ROW_NUMBER() function. since I cannot use the ROW_NUMBER() function and save results on the table.
I have a table as below;
ID YEAR
1 2018
2 2018
3 2019
4 2019
5 2020
6 2018
7 2019
I would like a formulae that will compute and assign numbers the rows depending with year as shown below;
ID YEAR COMPUTED COLUMN
1 2018 1
2 2018 2
3 2019 1
4 2019 2
5 2020 1
6 2018 3
7 2019 3

Answer 1

I am trying to find a formula for a computed column that will return same results as the ROW_NUMBER() function.
You just can't use window functions in a computed column; window functions operate on a range of rows (called the window frame), while a computed column has visibility to the row it belongs to only.
I cannot use the ROW_NUMBER() function and save results on the table
While this is technically possible, I would not recommend it. This is derived information, that can be computed on the fly whenever needed. You could use a view instead:
create view myview
as
select
id,
year,
row_number() over (partition by year order by id) rn
from mytable

Answer 2

There is always a way around it.
It's not a good way though...
select
id,
year,
row_number() over(partition by year order by id) rn,
(select count(*) from Table1 t where t.YEAR=d.YEAR and t.ID<=d.ID) PoorRN
from Table1 d
order by id
• This is going to have a very bad performance.
If your table has over a million rows - just go ahead and through it on production and tell us what happened ;)

Answer 3

Update: Based on the comments by OP, I have a new suggestion -
From the original table - create a view and then give numbers to each year
create view original_table_view
as
select *, ROW_NUMBER() over ( partition by [year] order by ID) as numbering
from original_table
Create a new table and apply a Merge on the new table from view
create table New_table (id int, [year] int, numbering int)
Create Merge syntax -
merge new_table t using original_table_view v
on (t.id = v.id) --merge_condition
when matched then update set
t.[year]=v.[year],
t.numbering=v.numbering
when not matched by target
then insert ( id, [year], numbering)
values (v.id, v.[year],numbering)
when not matched by source
then delete;
SO now if in future, if new values are inserted in the original table, you just need to run the merge query to update the new table.
view data -
select * from original_table order by ID
select * from original_table_view order by ID
select * from New_table order by ID
Old Answer:
I confused your query with calculated column instead of the computed column (sorry about that)
You cannot store function as a computed column in the SQL server. However, you can create a view on the table and then call the same.
example -
create table as below
create table year_computed2 (ID int, Year int )
go
insert into year_computed2 values (1,2018 )
insert into year_computed2 values (2,2018 )
insert into year_computed2 values (3,2019 )
insert into year_computed2 values (4,2019 )
insert into year_computed2 values (5,2020 )
insert into year_computed2 values (6,2018 )
insert into year_computed2 values (7,2019 )
go
Now Create a view using Row_number() function (you can also use Rank() and Ntile())
alter view year_computed_view as
select *, ROW_NUMBER() over ( partition by [year] order by ID) as Rn from year_computed2
Now Query this view like below
select * from year_computed_view order by ID
enter image description here
In-case you were talking about the calculated column - below answer is
available
Dense_Rank()
select DISTINCT id, [year], DENSE_RANK() over ( order by [year]) as [Dense_rank]
from year_computed
ORDER BY ID
Output is like:
If Dense_rank does't fulfill then try below two functions -
Rank()
select DISTINCT id, [year], RANK() over ( partition by [year] order by ID) as [Rank]
from year_computed
ORDER BY ID
enter image description here
NTile()
declare #ntile_count int = (select count (distinct [year])from year_computed)
select id, [year], ntile(#ntile_count) over (partition by [year] order by id) as rn from year_computed ORDER BY ID
enter image description here