Solved

# SQL Server decimal calculations showing odd results

Posted on 2012-03-22
454 Views
Consider the script below ran in Management Studio on SQLServer 2008:

declare @N1 decimal(38,12)
declare @N2 decimal(38,12)
begin
set @N1=1111223333444455556666777.111222233
set @N2=1111223333444455556666777.111222233
select @N1/@N2  --expected result
select @N1*cast(-1 as decimal(38,12))/@N2
select @N1*(-1*(1.0))/@N2
select @N1*(-1.0/@N2)
select 1.0/@N2
end;

The results are:
1.000000
-0.999999
-0.999999
-0.888979
0.0000000000000000000000008

Can anyone explain these results? Our problem is our software has to parse user friendly representation of field and number summing into sql friendly script which in turn gives correct and consistent results. We do need to deal with large number and many decimal places, hence using decimal(38,12). We are aware of the need to make the nominator of a division into a decimal number otherwise the result is assumed to be an integer,
ie
select 2/5 results in 0, and select 2.0/5 results in 0.4

We have also noticed that SQLServer doesn't always apply BODMAS calculation ordering. So, is there a set of rules that can be followed to achieve reliable results?

Many thanks
Alex
0
Question by:stbdevelopment

LVL 18

Assisted Solution

deighton earned 50 total points
I think you are overflowing the maximum size of a float
0

LVL 42

Assisted Solution

dqmq earned 300 total points
Interesting...seems like the more precision you give, the less you get back.

declare @N1 decimal(38,12)
set @N1=1.111122223333
select @N1
select @N1/1
select @N1/1.0
select @N1/1.00
select @N1/1.000
select @N1/1.00000000000000000
select @N1/cast(1 as decimal(38,12))
select @N1/cast(1 as decimal(38,3))

-------------------

1.111122223333

1.111122223333

1.11112222333

1.1111222233

1.111122223

1.111122

1.111122

1.111122223
0

Author Comment

Thank you dqmq. That really highlights the problem we are having. However what is the solution?
Do we need to restrict the numbers used in calculations to so many decimal places and then cast them to a larger number of decimal places? But then the results would be the same. I can't see a way out of this.
0

LVL 42

Accepted Solution

dqmq earned 300 total points
I don't know.  Still trying to make sense of it:

For more weirdness, try this:

declare @N1 decimal(38,12)
set @N1=   1.111122223333
select @N1
select @N1/1.0
select @N1/1.000000000000
select 1.111122223333/1.000000000000
select 1.111122223333/1.0
0

LVL 2

Assisted Solution

winston33 earned 50 total points
I'm sorry if this does not completely answer your question, but I generally complete the equation using the native datatypes (assuming the two numbers are using the same datatype) and then convert the result to the desired data type. The only statement below where this does not apply is the third because the "1" needs to be converted prior to the division. I cannot explain WHY this works, but I generally get the proper results. For example, using your example:

declare @N1 decimal(38,12)
declare @N2 decimal(38,12)
begin
set @N1=1111223333444455556666777.111222233
set @N2=1111223333444455556666777.111222233
select CONVERT(DECIMAL(38,12), @N1/@N2)  --expected result
select CONVERT(DECIMAL(38,12), (@N1*-1)/@N2)
select CONVERT(DECIMAL(38,12), CONVERT(DECIMAL(38,12), 1)/@N2)
end;

1.000000000000

-1.000000000000

0.000000000000 --> This is actually the correct result, I think, given the precision and scale.

Anyway, hope this helps a little.
0

## Featured Post

Having an SQL database can be a big investment for a small company. Hardware, setup and of course, the price of software all add up to a big bill that some companies may not be able to absorb.  Luckily, there is a free version SQL Express, but does …
Load balancing is the method of dividing the total amount of work performed by one computer between two or more computers. Its aim is to get more work done in the same amount of time, ensuring that all the users get served faster.
Via a live example combined with referencing Books Online, show some of the information that can be extracted from the Catalog Views in SQL Server.
Via a live example, show how to setup several different housekeeping processes for a SQL Server.