Following from a comment on IsDbNull performance, I thought I would measure how the performance of IsDBNull.

http://weblogs.asp.net/justin_rogers/archive/2004/04/11/111151.aspx

Original test

This was the optimised access code from test 3 of the previous entry

http://www.noelwatson.com/blog/PermaLink,guid,e93c0b50-233e-47d9-b3cf-bfacf754af05.aspx

while (reader.Read()) 

{
     
     logDetails = (String)reader[0];
     logTime =(DateTime)reader[1];

}

taking around 8.3 seconds for 100 iterations

1. IsDbNull()

while (reader.Read()) 

{
     if (!reader.IsDBNull(0))
     {
      logDetails = (String)reader[0];
     }
     if (!reader.IsDBNull(1))
     {
      logTime =(DateTime)reader[1];
     }

}

The duration of the experiment has increased dramtically to over 13 seconds.

2. Comparing to System.DBNull.Value

while (reader.Read())
{
        if (reader[0]!=System.DBNull.Value)
      {
         logDetails = (String)reader[0];
     }
     if (reader[1]!=System.DBNull.Value)
     {
         logTime =(DateTime)reader[1];
     }
}

 

This test is quicker (approx 20%) and I think this is the 20% the article was referring to. However, as can be seen in the screendump, the number of calls to get_item has increased. Storing the value in a local variable will solve this

3. Storing result in local variable

while (reader.Read()) 

{
     detailsNullCheck = reader[0];
     if (detailsNullCheck!=System.DBNull.Value)
     {
         logDetails = (String)detailsNullCheck;
     }
     timeNullCheck = reader[1];
     if (timeNullCheck!=System.DBNull.Value)
     {
         logTime =(DateTime)timeNullCheck;
     }

}

This value is approximately the same as the original test in the previous entry - checking for null values should not be an overhead.