views:

978

answers:

4

I've just come across this in a WHERE clause:

AND NOT (t.id = @id)

How does this compare with:

AND t.id != @id

Or with:

AND t.id <> @id

I'd always write the latter myself, but clearly someone else thinks differently. Is one going to perform any better than the other? I know that using <> or != is going to bust any hopes for using an index that I might have had, but surely the first approach above will suffer the same problem?

+3  A: 

There will be no performance hit, both statements are perfectly equal.

HTH

Tim Sullivan
+14  A: 

Note that the != operator is not standard SQL. If you want your code to be portable (that is, if you care), use <> instead.

DannySmurf
+15  A: 

These 3 will get the same exact execution plan

declare @id varchar(40)
select @id = '172-32-1176'

select * from authors
where au_id <> @id

select * from authors
where au_id != @id

select * from authors
where not (au_id = @id)

It will also depend on the selectivity of the index itself of course. I always use au_id <> @id myself

SQLMenace
A: 

Just a little adjustement fors those who come later:

The equality operator generate a unknow value when there is a null and the unknown value is treated a false. Not (unknown) is unknown

In the example below I'll try to say if a couple (a1, b1) is equal to (a2, b2). Note that each columns has 3 values 0, 1 and NULL.

DECLARE @t table (a1 bit, a2 bit, b1 bit, b2 bit)

Insert into @t (a1 , a2, b1, b2) 
values( 0 , 0 , 0 , NULL )

select 
a1,a2,b1,b2,
case when (
    (a1=a2 or (a1 is null and a2 is null))
and (b1=b2 or (b1 is null and b2 is null))
)
then 
'Equal'
end,
case when not (
    (a1=a2 or (a1 is null and a2 is null))
and (b1=b2 or (b1 is null and b2 is null))
)
then 
'not Equal'
end,
case when (
    (a1<>a2 or (a1 is null and a2 is not null) or (a1 is not null and a2 is null))
or (b1<>b2 or (b1 is null and b2 is not null) or (b1 is not null and b2 is null))
)
then 
'Different'
end
from @t

Note that here we expect results :

  • Equal to be null
  • not equal to be not equal
  • different to be difFerent

but we get another result

  • Equal is null OK
  • Not Equal is null ???
  • Different is different
LAGARRIGUE Mattiheu