Comments on: NOT EXISTS vs NOT IN

By: Gail

Gail — Wed, 30 Jun 2010 07:06:30 +0000

Interesting, thanks Otto.

By: Otto Astorga

Otto Astorga — Wed, 30 Jun 2010 00:24:33 +0000

Works the exact same way in Oracle. I was actually testing this issue and the difference between not exists and not in in the particular query I was running is 688 miliseconds with NOT EXISTS, 9 minutes 35 seconds with NOT IN.

That’s is a monumental difference. Column referenced is indexed.

By: E Mulwa

E Mulwa — Mon, 28 Jun 2010 19:57:38 +0000

One question: Is this a standard ANSI SQL behaviour or is it specific for SQL Server? For example, will this behaviour be the same in MySQL, Oracle, etc?

By: praneeth

praneeth — Mon, 10 May 2010 18:59:21 +0000

nice article, thanks gail

By: SQL Server Central

SQL Server Central — Sun, 28 Mar 2010 16:49:10 +0000

[...] Not in vs Not Exists [...]

By: Muthukkumaran

Muthukkumaran — Tue, 23 Mar 2010 12:22:00 +0000

Gail,Thanks for the great post.

By: Eric

Eric — Sun, 14 Mar 2010 22:00:10 +0000

Thank you for this. I hadn’t come across it in a lot of the things I’d read, and found it very useful.

By: Fabri

Fabri — Tue, 23 Feb 2010 09:45:52 +0000

Excellent post!

By: Gail

Gail — Fri, 19 Feb 2010 08:11:27 +0000

Yes, it is possible. The classic case is when the correlation condition is an inequality. However that’s not something likely in an IN/EXISTS condition, more likely when working with subqueries in the SELECT clause

By: Raj

Raj — Fri, 19 Feb 2010 03:13:26 +0000

Hi Gail,
Just some observation.
Is there any situation where a non correlated sub query outperforms corelated one?.
I have noticed it either to have identical plans when it comes to ‘in and exists’( as written by you already ) or correlated outperforming as in the case of Not exists vs not in when Nulls are involved.
To play it safe, in general, I suggest people to use correlated ones as even a structure change later by someone else doesn’t cause big a performance down. Not many will suspect a alter column to NULL as the cause of issue. Would like to know your take on the same.

Regards,
Raj