Comments on: Performance Impact of Using SELECT COUNT(*) In Queries

By: Mark Freeman

Mark Freeman — Thu, 11 Apr 2013 15:05:50 +0000

This is the most efficient way that I have found, although it is obviously a little more time consuming to type.

SELECT i.rows
FROM sys.tables t
INNER JOIN sys.sysindexes i ON t.object_id = i.id AND i.indid < 2
WHERE t.schema_id = SCHEMA_ID('’) AND t.name = ”;

If you compare it to any flavor of SELECT COUNT() FROM ., it us the clear winner regardless of what is in the cache or not.

By: Mark Freeman

Mark Freeman — Thu, 11 Apr 2013 14:55:24 +0000

I use COUNT(1) as a holdover from my Oracle days, where it made a significant difference. I just tried SELECT(1), SELECT(*), and SELECT() on a table with 1.3M rows, and the explain plan showed 33% for each, so it may not make a difference to SQL Server.

By: jack

jack — Mon, 18 Feb 2013 20:14:38 +0000

You can use sp_spaceused to get a row count, too. It can report incorrect values for rows and space used, but a pretty lightweight way to find a row count.

USE AdventureWorks2008R2;
GO
EXEC sp_spaceused N’Purchasing.Vendor’;
GO

More info: http://msdn.microsoft.com/en-us/library/ms188776(SQL.105).aspx

By: Ville Vainio

Ville Vainio — Thu, 24 Jan 2013 08:20:38 +0000

Ed, I think that if you’re just counting all the rows in a table, you could just look it up at sys.sysindexes. Join it with sys.tables. Take a look here: http://geekswithblogs.net/TakeNote/archive/2007/09/22/115537.aspx

But in real world, i usually want to count customers by territory, or some other group. I don’t know of an alternative to that. Or if I’m actually interested in the total row count of a table, I can suffer the one-off impact of an index scan (as it would take more time to write a better query).

By: Sebastiaan Halsema

Sebastiaan Halsema — Wed, 23 Jan 2013 19:29:28 +0000

Whenever I need to know the number of records in a table or query I always use

Select Count(1) From …

Not sure though if it really makes any difference in the used plan, never tested it either. But you won’t have to know the name of any column that way and the result will be the same of course (each row will be counted)

By: Ed Miller

Ed Miller — Wed, 23 Jan 2013 14:15:21 +0000

Okay, good info. So what’s the solution? Would COUNT(Customer_ID) be better?