How to use a variable within my SELECT (T-SQL) - sql

Question

I've written a crazy INSERT/SELECT statement that works pretty well, except that I think it can be tuned a wee bit more if I can avoid using the LEN([stats].[dbo].[Domain].[DomainName]) twice. Right now it takes 20 seconds to run, and if I replace these LEN sections with actual numbers for a test, it literally takes a second, hence my hope!
I've been racking my brain on how to get that into a variable so I can use it twice but only have one performance hit. I'm on SQL Server 2008 R2 for what it is worth.
Thanks much!
INSERT INTO [stats].[dbo].[5MinStats] (Qty, MsgRequest, MsgRecType, MsgDateTime, DomainID)
SELECT
COUNT([stats].[dbo].[RawMsgLog].[MsgRequest]) AS Qty,
[stats].[dbo].[RawMsgLog].[MsgRequest],
[stats].[dbo].[RawMsgLog].[MsgRecType],
DATEADD(minute, DATEDIFF(minute, 0, [stats].[dbo].[RawMsgLog].[MsgDateTime])/ 5 * 5, 0) AS MsgDateTime,
[stats].[dbo].[Domain].[DomainID]
FROM
[stats].[dbo].[RawMsgLog], [stats].[dbo].[Domain]
WHERE
RIGHT([stats].[dbo].[Domain].[DomainName], LEN([stats].[dbo].[Domain].[DomainName])) = RIGHT([stats].[dbo].[RawMsgLog].[MsgRequest], LEN([stats].[dbo].[Domain].[DomainName]))
AND [stats].[dbo].[RawMsgLog].[switch] = 1
GROUP BY
[stats].[dbo].[RawMsgLog].[MsgRequest],
[stats].[dbo].[RawMsgLog].[MsgRecType],
DATEADD(minute, DATEDIFF(minute, 0, [stats].[dbo].[RawMsgLog].[MsgDateTime]) / 5 * 5, 0),
[stats].[dbo].[Domain].[DomainID]
ORDER BY
MsgDateTime ASC

Answer 1

Change your WHERE to the following:
WHERE stats.dbo.Domain.DomainName LIKE '%' + stats.dbo.RawMsgLog.MsgRequest
This avoids the RIGHT() and LEN() functions, and allows your query to use any available indexes... applying functions to your indexed columns will obfuscate them and cause a scan to be used instead.
I would also make a few other changes...
Alias your tables.
Don't use [] if you don't need to...
Use explicit JOIN.
Which would give you the following query:
INSERT INTO stats.dbo.5MinStats (
Qty,
MsgRequest,
MsgRecType,
MsgDateTime,
DomainID
)
SELECT
COUNT(rml.MsgRequest) as Qty,
rml.MsgRequest,
rml.MsgRecType,
DATEADD(minute,
DATEDIFF(minute, 0, rml.MsgDateTime)/ 5 * 5, 0) as MsgDateTime,
d.DomainID
FROM
stats.dbo.RawMsgLog rml
JOIN stats.dbo.Domain d
ON d.DomainName LIKE '%' + rml.MsgRequest
WHERE rml.switch=1
GROUP BY
rml.MsgRequest,
rml.MsgRecType,
dateadd(minute, datediff(minute, 0, rml.MsgDateTime)/ 5 * 5, 0),
d.DomainID
ORDER BY MsgDateTime ASC

Answer 2

You could use CHARINDEX to speed this up:
WHERE CHARINDEX([stats].[dbo].[Domain].[DomainName], [stats].[dbo].[RawMsgLog].[MsgRequest]) > 0
This will return true if the value in DomainName is found anywhere in the MsgRequest column.

Answer 3

In the first half of your condition, you do a "right" with the length of the substring being the same as length of the field itself (i.e. right(field, len(field)). You can remove this and just use the field itself. Then, your comparison can be wildcard text:
WHERE stats.dbo.RawMsgLog.MsgRequest like '%' + stats.dbo.Domain.DomainName
you can also use this substring comparison, but I imagine it would be slower.
right(stats.dbo.RawMsgLog.MsgRequest, LEN(stats.dbo.Domain.DomainName)) = stats.dbo.Domain.DomainName