Avatar of HRISTeam
HRISTeamFlag for United States of America asked on

INDEXING with Cast / Convert function in SQL Server (Enhancing Query Performance)

I have several SQL Server tables which are dropped and reloaded through text files on a regular basis. The result is dozens of tables which all have the data type NVARCHAR(255). In the past this has worked fine since I can cast the data type to whatever I want later without having to worry about type conversion issue on the daily job. This practice has now created an issue since I would like to create an index on a datetime derived field that is stored as a multiple text fields. Is it possible to create an index on a value that is derived from the table? In my case I would like something similar to:

CREATE NONCLUSTERED INDEX NDX_prks_Position_Dta_01
ON [dbo].[Position_Dta]
([Position Number - Position Dta],
DATEADD(Minute, cast([Effective Date Sequence # - Position Dta] as int), [Effective Date - Position Dta]),
cast(tbl_Secondary.[Action Date - Position Dta] as date))

Obviously the above syntax does not work but hopefully that helps explain what I am looking for. Essentially the query I have in the example is one of dozens of subqueries with the same issue and the result is queries that take 4 hours to run. I suppose I would create a temp table with the correct data types and index the temp table but I was really hoping for a solution that would not cause me to have to update a lot of functioning code (even though it is super slow it works). Additionally I am not the only person who accesses these tables so I am really trying to avoid modifying the source tables.

Attached is a copy of a typical example where I am trying to create a more effective index.  I am open to any ideas that will increase performance and the less modification to the existing code the better! I appreciate the help experts!
EE-IndexPositionQuery-01.png
DatabasesMicrosoft SQL ServerSQL

Avatar of undefined
Last Comment
Jim Horn

8/22/2022 - Mon
TheAvenger

You can add computed columns to the tables and then make the index on those.

Here is a topic about computed columns: http://technet.microsoft.com/en-us/library/ms191250(v=sql.105).aspx
And here about indexes on them: http://technet.microsoft.com/en-us/library/ms189292.aspx
Pavel Celba

To create an index on TWO tables as you showed in the question is not good solution probably...

You should try to create index on  [Position Number - Position Dta]  column on both tables and see what happens. Then you may add indexes on [Effective Date Sequence # - Position Dta] and [Effective Date - Position Dta] which will optimize your subquery.

The "less than comparison part" of the WHERE condition isn't easily optimizable and the index on computed column could be a good start BUT I am skeptic...
ASKER
HRISTeam

Sorry, the screen shot is not 100% accurate. All the data is coming from the same table. So the table alias tbl_Primary and tbl_Secondary are actually the exact same table.
This is the best money I have ever spent. I cannot not tell you how many times these folks have saved my bacon. I learn so much from the contributors.
rwheeler23
Pavel Celba

In such case I would rather create a temporary table which will be used in place of the first subselect. This subselect must slow down the whole query because it calculates minimum from unoptimized data.

So create a table where each  [Position Number - Position Dta]  value has one corresponding value calculated as MIN(DateAdd(MINUTE, ...  WHERE ... etc.
and join this table by LEFT JOIN to your Position_dta table. This should optimize the subselect part.

If the result will be still slow then you may optimize the correlated subquery in the WHERE part.
TheAvenger

What about the computed columns I wrote about?
ASKER
HRISTeam

The computed columns looks promising but it would seemingly require me to rewrite much of my code. I am going to try a small test with the concept and test the performance gains.
Get an unlimited membership to EE for less than $4 a week.
Unlimited question asking, solutions, articles and more.
ASKER CERTIFIED SOLUTION
Jim Horn

Log in or sign up to see answer
Become an EE member today7-DAY FREE TRIAL
Members can start a 7-Day Free trial then enjoy unlimited access to the platform
Sign up - Free for 7 days
or
Learn why we charge membership fees
We get it - no one likes a content blocker. Take one extra minute and find out why we block content.
See how we're fighting big data
Not exactly the question you had in mind?
Sign up for an EE membership and get your own personalized solution. With an EE membership, you can ask unlimited troubleshooting, research, or opinion questions.
ask a question