Jeremy Massel's Blog » database

Select DataSet and number of total rows with one stored procedure

jeremy.massel — Wed, 11 Aug 2010 17:16:18 +0000

when you want to write a search using .net and MSSQL, it’s a pain. This is because you’re forced to select every row in the table and then only display a small subset of it. This works okay for tables that have a few hundred rows, as query caching can make this faster. But what happens when you’re searching a table with half a million rows?

Unless you’re a complete masochist, you’re doing to want to split this into a more manageable data set, otherwise you’re gonna eat all the memory on your server. But this means that you can no longer use the DataSet.Tables[0].Rows.Count property to figure out how many rows you have. You can write a second stored procedure that’ll count the rows. But who wants to clog up their database with tons of stored procedures for no reason? Let’s consolidate it into one.

So what does this look like?

First: the stored procedure.

We’ll use output parameters to pass the row count back to our code

create procedure [dbo].[Search] @searchText varchar(512), @recordsToReturn INT, @pageNumber INT, @numberofrows INT OUTPUT AS


-- get the page we want to view

select * from

(

     select *,  ROW_NUMBER() OVER (ORDER BY creation_timestamp DESC) AS row from [table] where [table].columnName like '%' + @searchText + '%'

)

AS results WHERE row between (@pageNumber - 1) * @recordsToReturn + 1 and @pageNumber*@recordsToReturn;
-- get the total number of rows, not just the subset we want

set @numberofrows = (select count(*)  from [table] where [table].columnName like '%' + @searchText + '%')
END

Now the C# (this’ll work in VB too, but feel free to convert it yourself)

SqlConnection conn = new SqlConnection(); conn.ConnectionString = ".....your connection string here....."; conn.Open();


      DataSet returnData = new DataSet();
      SqlDataAdapter da = new SqlDataAdapter( "SearchMessages", conn);

      da.SelectCommand.CommandType = CommandType.StoredProcedure;
      da.SelectCommand.Parameters.Add("@searchText", SqlDbType.VarChar).Value = "bob";

      da.SelectCommand.Parameters.Add("@recordsToReturn", SqlDbType.Int).Value = 10;

      da.SelectCommand.Parameters.Add("@pageNumber", SqlDbType.Int).Value = 1;
      //number of rows

      SqlParameter outputParameter = new SqlParameter("@numberofrows", SqlDbType.Int, 2);

      outputParameter.Direction = ParameterDirection.Output;  
      da.SelectCommand.Parameters.Add(outputParameter);
      da.Fill(returnData, "theData");
      int numberOfRowsInDataSet = (int)outputParameter.Value;

da.Dispose(); conn.Close();

Best of luck! As always, leave a message in the comments if you have questions

Batch Delete Performance SQL Server

jeremy.massel — Thu, 17 Jun 2010 17:38:32 +0000

Deleting old records from a table with > 3 000 00 rows. What’s the best way to do this?

It seems the fastest way to do this is simply to:
delete from [table] where creation_timestamp < dateadd (mm, -6, getdate())
(deleting anything older than 6 months)

It took 3 hours (10916 seconds to be exact) to delete 1.6 million (1,619,433) records this way. (148.35 / second).

We needed to do a second batch the next day, but wanted to split it into batches to try to get better performance.

Running:

delete from [table] where pk_id in( select pk_id from ( SELECT ROW_NUMBER() OVER (ORDER BY creation_timestamp desc) AS RowNumber, pk_id FROM [table] where creation_timestamp < '2009-12-16 13:52:08.673') _objectsToDelete WHERE RowNumber between 1 and 100000)

takes 12 minutes. (732 seconds) (136.61 / second).

Strangely, using the TOP command with a subquery takes the longest:

delete from [table] where pk_id in (select TOP 100000 pk_id from [table] where creation_timestamp < '2009-12-16 13:52:08.673')

15 minutes (904 seconds) (110.61 / second)

Have a better way? Let me know in the comments!

mysql command doesn’t work in OS X snow leopard

jeremy.massel — Wed, 20 Jan 2010 20:09:01 +0000

So you install mySql on your mac, and it’s working great, but then you go into terminal to run a mysqldump or something, and it doesn’t work!

you get a message something like:
-bash: mysqldump: command not found

so how do you fix this? well, copy and paste the following into your terminal window and press enter:
echo export PATH="/usr/local/bin:/usr/local/sbin:/usr/local/mysql/bin:$PATH" >> ~/.profile

source ~/.profile

whoa, whoa, you say. what is this doing? It makes it so that when you type any command into the terminal, it does a check of these folders before giving you the “command not found” error.

If you want to undo it, you can always edit your .profile, or just delete it.