multiple cursor using Npgsql?

Question

I'm used to (from Ada using libpq) to

• open a cursor

• get some key values

• use those key values as bind parameters for other statements.

But I get a Npgsql.NpgsqlOperationInProgressException instead.

I once got this working with sql-server, but solved it via adding 'MARS' to the connect string (Multiple Active Record Sets)

Can I do something like that here to?

Here what I'm trying to do:

            conn.Open();

        // Define a query
        NpgsqlCommand cmdGetSelectionIds = new NpgsqlCommand("select distinct(R.SELECTIONID) from ARUNNERS R, AMARKETS M " +
                                              "where true " +
                                              "and M.MARKETID = R.MARKETID " +
                                              "and M.MARKETTYPE = 'WIN' " +
                                              "and R.STATUS <> 'REMOVED'", conn);


        NpgsqlCommand cmdNumWins = new NpgsqlCommand("select count('a') from ARUNNERS R, AMARKETS M " +
                                              "where true " +
                                              "and M.MARKETID = R.MARKETID " +
                                              "and R.SELECTIONID = @selid " +
                                              "and M.MARKETTYPE = 'WIN' " +
                                              "and R.STATUS = 'WINNER'", conn);

        NpgsqlCommand cmdNumPlcs = new NpgsqlCommand("select count('a') from ARUNNERS R, AMARKETS M " +
                                              "where true " +
                                              "and M.MARKETID = R.MARKETID " +
                                              "and R.SELECTIONID = @selid " +
                                              "and M.MARKETTYPE = 'PLACE' " +
                                              "and R.STATUS = 'WINNER'", conn);

        NpgsqlCommand cmdNumLosses = new NpgsqlCommand("select count('a') from ARUNNERS R, AMARKETS M " +
                                              "where true " +
                                              "and M.MARKETID = R.MARKETID " +
                                              "and R.SELECTIONID = @selid " +
                                              "and M.MARKETTYPE = 'WIN' " +
                                              "and R.STATUS = 'LOSER'", conn);

        NpgsqlDataReader drGetSelectionIds = cmdGetSelectionIds.ExecuteReader();

        while (drGetSelectionIds.Read())
        {
            selid = drGetSelectionIds.GetInt32(0);


            cmdNumWins.Parameters.AddWithValue("selid", selid);
            using (NpgsqlDataReader drNumWins = cmdNumWins.ExecuteReader())
            { 
                if (drNumWins.Read()) numWins = drNumWins.GetInt32(0);
            }

            using (NpgsqlDataReader drNumPlcs = cmdNumPlcs.ExecuteReader())
            {
                if (drNumPlcs.Read()) numPlcs = drNumPlcs.GetInt32(0);
            }

            using (NpgsqlDataReader drNumLosses = cmdNumLosses.ExecuteReader()) {
                if (drNumLosses.Read()) numPlcs = drNumLosses.GetInt32(0);
            }

            Console.Write("selid : {0} \t num W {1} \t num P {2} num L {3} \t points {4}\n", selid, numWins, numPlcs, numLosses, (3.0*numWins + numPlcs )/(numWins+numLosses));


        }
        // Close connection
        conn.Close();

Yes, I could read the first statement into a list and loop over that, but I do have got quite some legacy code that is build like the above.

/Björn

Answer 1

The problem is NpgSql only allows a single data reader to be open at a time. If you have a situation where you need more than one data reader open against the same connection at the same time, I would argue there may be a better way of reaching your end goal.

So, disclaimer... don't do this. BUT, if you wanted to do what you described above, this would be a way to accomplish it:

cmdNumWins.Parameters.Add(new NpgsqlParameter("@selid", NpgsqlDbType.Integer));
cmdNumPlcs.Parameters.Add(new NpgsqlParameter("@selid", NpgsqlDbType.Integer));
cmdNumLosses.Parameters.Add(new NpgsqlParameter("@selid", NpgsqlDbType.Integer));

List<int> idList = new List<int>();

using (NpgsqlDataReader drGetSelectionIds = cmdGetSelectionIds.ExecuteReader())
{
    while (drGetSelectionIds.Read())
        idList.Add(drGetSelectionIds.GetInt32(0));

    drGetSelectionIds.Close();
}

foreach (int selid in idList)
{
    cmdNumWins.Parameters[0].Value = selid;
    cmdNumPlcs.Parameters[0].Value = selid;
    cmdNumLosses.Parameters[0].Value = selid;

    numWins = Convert.ToInt32(cmdNumWins.ExecuteScalar());
    numPlcs = Convert.ToInt32(cmdNumPlcs.ExecuteScalar());
    numLosses = Convert.ToInt32(cmdNumLosses.ExecuteScalar());

    Console.Write("selid : {0} \t num W {1} \t num P {2} num L {3} \t points {4}\n", selid, 
        numWins, numPlcs, numLosses, (3.0 * numWins + numPlcs) / (numWins + numLosses));
}

The better way would be to do this all as a single query:

select
  r.selectionid,
  count (1) filter (where m.markettype = 'WIN' and r.status = 'WINNER') as win,
  count (1) fitler (where m.markettype = 'PLACE' and r.status = 'WINNER') as place,
  count (1) filter (where m.markettype = 'WIN' and r.status = 'LOSER') as lose
from
  arunners r
  join amarkets m on m.marketid = r.marketid
where
  r.status != 'REMOVED'
group by
  r.selectionid

I may have missed some nuances about your logic, but hopefully you get the idea. This should do in one fell swoop, quickly and efficiently, what your C# logic is doing. It will be much friendlier to the database and for large datasets should be quite a bit quicker.