# SQL 2008 calculate AVG on computed column

I have a computed column (MyColumn) defined as decimal (4,2) which calls a user-defined function. The function calculates a score (think of it as a credit score) based on transactional activity in two other tables.

The computed column is non deterministic and so the data is non persistent. I have 200,000+ rows (users who's credit score is calculated by the computed column and which changes daily). I want to calculate AVG(MyColumn) however the processing time would likely take several hours. What is the best way to calculate AVG(MyColumn) given that the data upon which it needs to calculate is not persistent?
Microsoft SQL Server 2008

Last Comment
drl1

8/22/2022 - Mon

Better would be to use actual tables while calculating the average...

Actual tables means tables  used in UDF..
Guy Hengel [angelIII / a3]

you could dump the computed columns' value into some temp table...  asnyc (for example, early in the morning), so you could use that table for your reporting ...
at least that is the way I work most often if such "computed columns" change daily once.
means also that I would not have the computed column in the table directly, but be computed once per day anyhow.
drl1

Thanks for the feedback.

@Saurv...Can you elaborate on how a table in UDF might work in this example? Currently I pass an ID (@id) in to the UDF and it conducts the following (simplified) process:

set @count1 = (select count(id) from table A where id=@id)
set @count2 = (select count(id) from table B where id=@id)
set @result = @count1/@count2
return @result

@angelIII... That's the line of thought I was heading down, so compute the value in the early hours of the morning so that I at least have persistent values from which to report in aggregate but it doesn't give me the real-time value per ID that I was looking for. I guess I could complete the early hours conputation and also retain the computed column for ad-hoc calls for each ID thus capturing the real-time value (which may have changed since that morning).
Experts Exchange has (a) saved my job multiple times, (b) saved me hours, days, and even weeks of work, and often (c) makes me look like a superhero! This place is MAGIC!
Walt Forbes

What i meant.... to put your logic inside the SP and do the calculation at once

I have written simple code just give you an idea... Wrap you code inside a Sp .
Since you have around 200,000 record you can also put looping with indexes on temp tables..

And you can call that Sp when ever you are running report..    let me know if you have any concern..

I am assuming the @id is  some column value of tablec    ........

`````` Create table #tempa
(
id_A int,
Cnt_A bigint
)

Create table #tempb
(
id_B int,
Cnt_b bigint
)

insert into #tempa
select A.id,count(A.id)
from tableA A  join tablec C on A.id=c.id
group by A.id

insert into #tempb
select b.id,count(b.id)
from tableb b  join tablec C on b.id=c.id
group by b.id

--Return Table

select  A.ID , cnt_a/cnt_b from
#tempa a join #tempb b on a.id_A=b.id_B
``````

Thanks,
Saurabh
Scott Pletcher

Create an in-line-table function that does the same computation.  In-line table functions will be much faster than row-by-row.  The computed column r-b-r makes sense for a few rows here and there, but to do a lot of them an in-line table function will be much more efficient.
drl1

Thanks for the guidance. So this is where I'm up to...
****************************************************************
CREATE FUNCTION [dbo].[TEST] (@id int)

RETURNS @TableA TABLE
(
id bigint,
FieldA bigint,
FieldB bigint,
Score (5,2)
)
AS
BEGIN
DECLARE @CountA int = (Select count(id) from tableA where id = @id)
DECLARE @CountB int = (Select count(id) from tableB where id = @id and status = 'Whatever')
DECLARE @CScore decimal(5,2) = (CAST((@CountB) as float))/(CAST((@CountA) as float))
INSERT @TableA
SELECT @id, @CountA, @CountB, @CScore
RETURN
END
****************************************************************

So I can pass the ID for a user in without problem and it returns a single row table containing the id, two counts and calculated score. How would I now convert this to work out the scores for every ID (from 1 - 200,000) and then conduct an AVG function on the 'Score' field?
Get an unlimited membership to EE for less than \$4 a week.
Unlimited question asking, solutions, articles and more.
drl1

Think I've got it now...
****************************************************************
CREATE FUNCTION [dbo].[TEST] (@id int)

RETURNS @TableA TABLE
(
id bigint,
FieldA bigint,
FieldB bigint,
Score (5,2)
)
AS
BEGIN
DECLARE @ProcID int = 1
DECLARE @MaxID int = (SELECT MAX(id) from tableA)

WHILE @ProcID <= @MaxID
BEGIN

DECLARE @CountA int = (Select count(id) from tableA where id = @ProcID)
DECLARE @CountB int = (Select count(id) from tableB where id = @ProcID and status = 'Whatever')
DECLARE @CScore decimal(5,2) = (CAST((@CountB) as float))/(CAST((@CountA) as float))
INSERT @TableA
SELECT @ProcID, @CountA, @CountB, @CScore
SELECT @ProcID = (Select top 1 ID from tableA where ID > @ProcID order by ID)
END
RETURN
END
****************************************************************

Takes 35 secs to return 1000 rows. @ID as an import value is now redundant as I start the loop from 1 (set @MaxID to 1000 for testing). Reason I'm incrementing the @ProcID using a select statement is because not all IDs exist in tableA due to some accounts being physically removed and IDs deleted. Any ideas on further streamlining this?
Scott Pletcher

That's not an inline table function, that's a multiline table function.

An inline table function has just a RETURN statement.
drl1

Thanks. Would you be able to re-code the above as an inline function, just so I'm clear on how to achieve that implementation? When I tried declaring the temp tables within the function I got some compile errors so must have been doing it incorrectly.
Experts Exchange is like having an extremely knowledgeable team sitting and waiting for your call. Couldn't do my job half as well as I do without it!
James Murphy
Scott Pletcher

Become an EE member today7-DAY FREE TRIAL
Members can start a 7-Day Free trial then enjoy unlimited access to the platform
or
Learn why we charge membership fees
We get it - no one likes a content blocker. Take one extra minute and find out why we block content.
Not exactly the question you had in mind?
Sign up for an EE membership and get your own personalized solution. With an EE membership, you can ask unlimited troubleshooting, research, or opinion questions.
drl1

Wow...now that is pretty slick. 211,379 rows calculated and returned in 85 secs. Thanks a million ;o)
drl1

Perfect, thanks!
Scott Pletcher

Inline functions are literally "compiled" right into the statement being run, so they are much more efficient.  Multiline functions are called like a regular scalar function, and so have much more overhead.
Get an unlimited membership to EE for less than \$4 a week.
Unlimited question asking, solutions, articles and more.
drl1

Yeah, it makes sense now I can see good working examples of each method in front of me. Could I pester you for another solution? I could start a new Q but it is related to the above....

I need to calculate a similar process to the one above however this time I'm pulling values from one table only instead of two tables.

Essentially...

select count(id) as Count1 from tableA where status = 'X' or status = 'Y'
select count(id) as Count2 from tableA
NewScore = Count1/Count2

Each row in the result set would show: id, Count1, Count2, NewScore

Again, I need to calculate this on-the-fly for each ID and be able to run AVG(NewScore) without waiting an inordinate amount of time. Maybe i've been looking at the screen too long to see the obvious answer to this one, which is definitely more straight forward.
Scott Pletcher

SELECT
id,
SUM(CASE WHEN status = 'X' OR status = 'Y' THEN 1 ELSE 0 END) AS Count1,
SUM(1) AS Count2,
(SUM(CASE WHEN status = 'X' OR status = 'Y' THEN 1 ELSE 0 END) * 1.0) /
SUM(1) AS NewScore
FROM dbo.tableA
GROUP BY
id
ORDER BY
id

If you need to see NewScore as a %, change the "* 1.0" to "* 100.0".
drl1

That's great for getting all of the rows, thanks. Calculating AVG(NewScore) is my final requirement. Would I simply select all of the above to a temp table then run AVG(NewScore) on the temp table to get the most effective solution to that requirement?
This is the best money I have ever spent. I cannot not tell you how many times these folks have saved my bacon. I learn so much from the contributors.
rwheeler23
Scott Pletcher

Do you need to see the detail at all?  If not, we can compute the avg in a single SELECT with just that as the result, we don't need a temp table.

If you need the detail and the final avg, we can add a WITH ROLLUP to the group by and calc it from that :-) .
drl1

Having the detail would be good ;o)
Scott Pletcher

Try this and see if it gives you what you need:

SELECT
CASE WHEN id IS NULL THEN 'Summary' ELSE CAST(id AS varchar(10)) END AS id,
Count1, Count2,
CASE WHEN id IS NULL THEN (Count1 * 1.0 / Count2) ELSE NewScore END AS NewScore
FROM (
SELECT
id,
SUM(CASE WHEN status = 'X' OR status = 'Y' THEN 1 ELSE 0 END) AS Count1,
SUM(1) AS Count2,
(SUM(CASE WHEN status = 'X' OR status = 'Y' THEN 1 ELSE 0 END) * 1.0) /
SUM(1) AS NewScore
FROM dbo.tableA
GROUP BY
id WITH ROLLUP
) AS derived
ORDER BY
id

If you want to leave id as numeric, you can use 0 or 2000000000 for the summary row, depending on whether you want it to be first or last.
Get an unlimited membership to EE for less than \$4 a week.
Unlimited question asking, solutions, articles and more.
drl1