Comments on: The most optimal join type

By: Thomas LeBlanc

Thomas LeBlanc — Wed, 10 Nov 2010 22:21:01 +0000

Great article, again…keep them coming

By: Gail

Gail — Mon, 01 Nov 2010 19:48:18 +0000

Pretty much, yes in most cases.

There are times that you will want to overule the optimiser, the trick is to know when those times are and to have full understanding of why you’re overruling it and what the effects will be.

By: Pradeep

Pradeep — Mon, 01 Nov 2010 11:23:15 +0000

Great post!

One thing I understand from here is it is better to let the query optimizer decide what method should be used, rather than forcing a join type using any hints.

If my understanding is correct, I wonder why there should be so much complicated literature elsewhere (including the MSDN) about explicit hints.

By: Tony

Tony — Tue, 01 Dec 2009 15:41:57 +0000

Wow, I did not know that you would respond so soon.

Thanks and I will check out the rest of your index posts soon.

By: Gail

Gail — Mon, 30 Nov 2009 15:22:37 +0000

Depends. Both can work. In this specific case, the where clause refers to joinTable1, which becomes the outer table of the join, so an index on the join column won’t be used. If it was the inner table, that index would be very useful.

In a real-life situation, I’d probably test both options (ID, someString) and (Somestring, ID) and see which one produced the more optimal performance.

Go have a look at some of my indexing posts (if you haven’t), specifically the two on selectivity and column order

By: Tony

Tony — Mon, 30 Nov 2009 15:16:58 +0000

Hi:

I have a question on indexing one of your examples.

For instance if we take this example from up above…..

1. SELECT SomeString
2. FROM jointable1 j1
3. INNER JOIN jointable2 j2 ON j1.somestring = j2.joinstring
4. WHERE ID BETWEEN 0 AND 5000
5. OPTION (LOOP JOIN)

Where would you start the index in the where clause or the join clause?