SELECT
    [Shelf Location], date /*, ...*/
FROM (
    SELECT *, DENSE_RANK() OVER(PARTITION BY [Shelf Location] ORDER BY date DESC) AS rank_num
 ...

I see no way to accurately parse that data will all the variations presented (and any number even worse than that that could happen).

If ID is (nearly) always an identity column, you can use ORDER BY $IDENTITY, that way you don't need to know the exact column name.

I'm not sure which specific "problem" you're referring...

It actually does explain it.  If SQL has to do a sort to get the rows in the correct order, then the cursor would be required to do a sort...

• This reply was modified 4 years, 6 months ago by ScottPletcher.

Sorry to have bothered you then.  Good luck with this.  (I believe if you changed your cursor to LOCAL FAST_FORWARD (as in standard in SQL Server unless you explicitly need...

You should change the index so that it has the proper sort order already, which will (should) avoid the sort in the query:

CREATE NONCLUSTERED INDEX [IX_ImageProduct_MID] ON...

The one...

SQL can use an index for...

fake_pointer ???? NOT NULL);

The reason you are using big integers is because they look like pointers. But we don’t use pointers in RDBMS. You’re using SQL to write assembly...

I'm not sure which specific "problem" you're referring to.

But when Jonathon added both columns the WHERE clause, it might have taken care of it.

c.CustomerNumber = t.CustomerNumber AND c.Customer = t.Customer

Jonathon, your second query brought the spool back. ...

Removing NOLOCK would only slow down the process, never speed it up.  I'm not saying you should or shouldn't use it here...

I'm not...

What indexes do you have on that table?

Looks like an index on ( MessageType, NewValue, KeyValue, RecordDate ) INCLUDE ( CreateDate, OldValue )

might be useful for this query.