How to get number of rows using SqlDataReader in C#

Question

My question is how to get the number of rows returned by a query using SqlDataReader in C#. I\'ve seen some answers about this but none were clearly defined exc

Answer 1

I also face a situation when I needed to return a top result but also wanted to get the total rows that where matching the query. i finaly get to this solution:

   public string Format(SelectQuery selectQuery)
    {
      string result;

      if (string.IsNullOrWhiteSpace(selectQuery.WherePart))
      {
        result = string.Format(
@"
declare @maxResult  int;
set @maxResult = {0};

WITH Total AS
(
SELECT count(*) as [Count] FROM {2}
)
SELECT top (@maxResult) Total.[Count], {1} FROM Total, {2}", m_limit.To, selectQuery.SelectPart, selectQuery.FromPart);
      }
      else
      {
        result = string.Format(
@"
declare @maxResult  int;
set @maxResult = {0};

WITH Total AS
(
SELECT count(*) as [Count] FROM {2} WHERE {3}
)
SELECT top (@maxResult) Total.[Count], {1} FROM Total, {2} WHERE {3}", m_limit.To, selectQuery.SelectPart, selectQuery.FromPart, selectQuery.WherePart);
      }

      if (!string.IsNullOrWhiteSpace(selectQuery.OrderPart))
        result = string.Format("{0} ORDER BY {1}", result, selectQuery.OrderPart);

      return result;
    }

Answer 2

You can't get a count of rows directly from a data reader because it's what is known as a firehose cursor - which means that the data is read on a row by row basis based on the read being performed. I'd advise against doing 2 reads on the data because there's the potential that the data has changed between doing the 2 reads, and thus you'd get different results.

What you could do is read the data into a temporary structure, and use that in place of the second read. Alternatively, you'll need to change the mechanism by which you retrieve the data and use something like a DataTable instead.

Answer 3

Per above, a dataset or typed dataset might be a good temorary structure which you could use to do your filtering. A SqlDataReader is meant to read the data very quickly. While you are in the while() loop you are still connected to the DB and it is waiting for you to do whatever you are doing in order to read/process the next result before it moves on. In this case you might get better performance if you pull in all of the data, close the connection to the DB and process the results "offline".

People seem to hate datasets, so the above could be done wiht a collection of strongly typed objects as well.

Answer 4

If you do not need to retrieve all the row and want to avoid to make a double query, you can probably try something like that:

using (var sqlCon = new SqlConnection("Server=127.0.0.1;Database=MyDb;User Id=Me;Password=glop;"))
      {
        sqlCon.Open();

        var com = sqlCon.CreateCommand();
        com.CommandText = "select * from BigTable";
        using (var reader = com.ExecuteReader())
        {
            //here you retrieve what you need
        }

        com.CommandText = "select @@ROWCOUNT";
        var totalRow = com.ExecuteScalar();

        sqlCon.Close();
      }

You may have to add a transaction not sure if reusing the same command will automatically add a transaction on it...

Answer 5

to complete of Pit answer and for better perfromance : get all in one query and use NextResult method.

using (var sqlCon = new SqlConnection("Server=127.0.0.1;Database=MyDb;User Id=Me;Password=glop;"))
{
    sqlCon.Open();
    var com = sqlCon.CreateCommand();
    com.CommandText = "select * from BigTable;select @@ROWCOUNT;";
    using (var reader = com.ExecuteReader())
    {
        while(reader.read()){
            //iterate code
        }
        int totalRow = 0 ;
        reader.NextResult(); // 
        if(reader.read()){
            totalRow = (int)reader[0];
        }
    }
    sqlCon.Close();
}

Answer 6

There are only two options:

• Find out by reading all rows (and then you might as well store them)

• run a specialized SELECT COUNT(*) query beforehand.

Going twice through the DataReader loop is really expensive, you would have to re-execute the query.

And (thanks to Pete OHanlon) the second option is only concurrency-safe when you use a transaction with a Snapshot isolation level.

Since you want to end up storing all rows in memory anyway the only sensible option is to read all rows in a flexible storage (List<> or DataTable) and then copy the data to any format you want. The in-memory operation will always be much more efficient.