Comments on: composite WHERE condition in SQL Server

By: marcw

marcw — Fri, 11 Dec 2009 08:59:11 +0000

Okay… I don’t have any control over the table, this is an MIS application pulling from a database someone else has created. What we’re pulling from is actually a snapshot: would it be better to ask them to create a field in the snapshot that’s pre-formatted?

By: mrdenny

mrdenny — Fri, 11 Dec 2009 03:16:38 +0000

While you can do this keep in mind that any indexes that you have on the table won’t be used. You’ll get better performance if you add a computed column and index that column then use that column in your WHERE clause.

By: carlosdl

carlosdl — Thu, 10 Dec 2009 17:27:29 +0000

Since you are using concatenations and conversions in your where clause, the optimizer is not using any indexes and is already doing a full table scan to get your data, so, I think you could avoid the double conversion by doing something like this:

SELECT code, ver
FROM ( 
SELECT code, right(’0′ + convert(VARCHAR,codeversion),2) AS ver
FROM table) AS temp
where code+ver IN (’00102′,’00202′,’00302′);

By: marcw

marcw — Thu, 10 Dec 2009 16:39:10 +0000

Thanks- that works.

In this instance, the version field is a derived field, because the field in the database is an 11-char int for some reason.

So it looks like this:

SELECT code, right(’0′ + convert(VARCHAR,codeversion),2) AS ver
FROM table
WHERE
code + right(’0′ + convert(VARCHAR,codeversion),2) IN (’00102′,’00202′,’00302′) ;

If I use WHERE code + ver IN (’00102′,’00202′,’00302′) as the predicate, the SQL Server ODBC driver complains that “ver” is an invalid field name.

I’m concerned that it’s inefficient doing the conversion multiple times. Am I missing a trick here?