views:

3034

answers:

5

In SQL Server 2005, is there a way of deleting rows and being told how many were actually deleted?

I could do a select count(*) with the same conditions, but I need this to be utterly trustworthy.

My first guess was to use the @@ROWCOUNT variables - but that isn't set, e.g.

delete 
from mytable 
where datefield = '5-Oct-2008' 

select @@ROWCOUNT

always returns a 0.

MSDN suggests the OUTPUT construction, e.g.

delete from mytable 
where datefield = '5-Oct-2008' 
output datefield into #doomed

select count(*) 
from #doomed

this actually fails with a syntax error.

Any ideas?

+1  A: 

Create temp table with one column, id.

Insert into temp table selecting the ids you want to delete. That gives you your count.

Delete from your table where id in (select id from temp table)

Corey Trager
Although something that would work, the overhead of something like this is a waste.
Mitchel Sellers
+12  A: 

Did you try SET NOCOUNT OFF?

wcm
I had thought that I had tried this - but apparently I hadn't because it worked like a charm - thanks.
Unsliced
I'm glad it helped. And thank you! This was my first accepted answer. I couldn't be more proud.
wcm
A: 

In your example @@ROWCOUNT should work - it's a proper way to find out a number of deleted rows. If you're trying to delete something from your application then you'll need to use SET NOCOUNT ON

According to MSDN @@ROWCOUNT function is updated even when SET NOCOUNT is ON as SET NOCOUNT only affects the message you get after the the execution.

So if you're trying to work with the results of @@ROWCOUNT from, for example, ADO.NET then SET NOCOUNT ON should definitely help.

Ilya Kochetov
+2  A: 

I use @@ROWCOUNT for this exact purpose in SQL2000 with no issues. Make sure that you're not inadvertantly resetting this count before checking it though (BOL: 'This variable is set to 0 by any statement that does not return rows, such as an IF statement').

Simon
A: 

Out of curiosity, how are you calling the procedure? (I'm assuming it is a stored procedure?). The reason I ask is that there is a difference between a stored procedure's return value (which would be 0 in this case), and a rowset result -- which in this case would be a single row with a single column. In ADO.Net, the former would be accessed by a parameter and the latter with a SqlDataReader. Are you, perhaps, mistaking the procedure's return value as the rowcount?

Decker
This was code all in one place (but yes, in a stored procedure) - I was deleting some data then writing that fact to a status table, so it wasn't really a case of being confused over return values.
Unsliced