Why is my SQL Server CE code failing?

Question 1

For fun I'd try two things. First, replace the '?' parameter with a named parameter like '@tablename' and see if that changes things. Yes, I know '?' should work, but it's a confusing, ugly precedent and maybe since it's a system table it's wonky. Yes, it's a stretch, but worth a try just to know.

The second thing I'd do is something like this method from the SQLCE implementation of the OpenNETCF ORM:

    public override bool TableExists(string tableName)
    {
        var connection = GetConnection(true);
        try
        {
            using (var command = GetNewCommandObject())
            {
                command.Transaction = CurrentTransaction as SqlCeTransaction;
                command.Connection = connection;
                var sql = string.Format("SELECT COUNT(*) FROM information_schema.tables WHERE table_name = '{0}'", tableName);
                command.CommandText = sql;
                var count = Convert.ToInt32(command.ExecuteScalar());

                return (count > 0);
            }
        }
        finally
        {
            DoneWithConnection(connection, true);
        }
    }

Note that I didn't even bother parameterizing, largely because I doubt it will provide any perf benefit (queue the hordes whining about SQL injection). This way definitely works - we've got it deployed and in use in many live solutions.

EDIT

For completeness (though I'm not sure it adds to clarity).

    protected virtual IDbConnection GetConnection(bool maintenance)
    {
        switch (ConnectionBehavior)
        {
            case ConnectionBehavior.AlwaysNew:
                var connection = GetNewConnectionObject();
                connection.Open();
                return connection;
            case ConnectionBehavior.HoldMaintenance:
                if (m_connection == null)
                {
                    m_connection = GetNewConnectionObject();
                    m_connection.Open();
                }
                if (maintenance) return m_connection;
                var connection2 = GetNewConnectionObject();
                connection2.Open();
                return connection2;
            case ConnectionBehavior.Persistent:
                if (m_connection == null)
                {
                    m_connection = GetNewConnectionObject();
                    m_connection.Open();
                }
                return m_connection;
            default:
                throw new NotSupportedException();
        }
    }

    protected virtual void DoneWithConnection(IDbConnection connection, bool maintenance)
    {
        switch (ConnectionBehavior)
        {
            case ConnectionBehavior.AlwaysNew:
                connection.Close();
                connection.Dispose();
                break;
            case ConnectionBehavior.HoldMaintenance:
                if (maintenance) return;
                connection.Close();
                connection.Dispose();
                break;
            case ConnectionBehavior.Persistent:
                return;
            default:
                throw new NotSupportedException();
        }
    }

Question 2

wow... still struggling... I did too when I first got started on a handheld device SQL-CE. My current project is running with C#.Net 3.5 but I think the principles you are running into are the same. Here is what is working for my system in it's close parallels to yours.

First, the connection string to the handheld. It is just

string myConnString  = @"Data Source=\MyFolder\MyData.sdf";

no reference to the sql driver

Next, the TableExists

SqlCeCommand oCmd = new SqlCeCommand( "select * from INFORMATION_SCHEME.TABLES "
   + " where TABLE_NAME = @pTableName" );
oCmd.Parameters.Add( new SqlCeParameter( "pTableName", YourTableParameterToFunction ));

The "@pTableName" is to differentiate between the "TABLE_NAME" column and to absolutely prevent any issues about ambiguity. The Parameter does NOT get the extra "@". In SQL, the @ indicates to look for a variable... The SqlCeParameter of "pTableName" must match as it is in the SQL Command (but without the leading "@").

Instead of issuing a call to ExecuteScalar, I am actually pulling the data down into a DataTable via

DataTable oTmpTbl = new DataTable();
SqlCeDataAdapter da = new SqlCeDataAdapter( oCmd );
da.Fill( oTmpTbl );
bool tblExists = oTbl.Rows.Count > 0;

This way, I either get records back or I dont... if I do, the number of records should be > 0. Since I'm not doing a "LIKE", it should only return the one in question.

When you get into your insert, updates and deletes, I have always tried to prefix my parameters with something like "@pWhateverColumn" and make sure the SqlCeParameter is by the same name but without the "@". I haven't had any issues and this project has been running for years. Yes it's a .net 3.5 app, but the fundamental basics of connecting and querying SHOULD be the same.

If it IS all within your application, I would try something like creating a single global static "Connection" object. Then, a single static method to handle it. Then, instead of doing a NEW connection during every "using" attempt, change it to something like...

public static class ConnectionHandler
{
   static SqlCeConnection myGlobalConnection;

   public static SqlCeConnection GetConnection()
   {
      if( myGlobalConnection == null )
         myGlobalConnection = new SqlCeConnection();

      return myGlobalConnection;
   }

   public static bool SqlConnect()
   {
      GetConnection();   // just to ensure object is created

      if( myGlobalConnection.State != System.Data.ConnectionState.Open)
      {
         try
         {
            myGlobalConnection.ConnectionString = @"Data Source=\MyFolder\MyDatabase.sdf";
            myGlobalConnection.Open();
         }
         catch( Exception ex)
         {
            // optionally messagebox, or preserve the connection error to the user
         }
      }

      if( myGlobalConnection.State != System.Data.ConnectionState.Open )
         MessageBox.Show( "notify user");

      // return if it IS successful at opening the connection (or was already open)
      return myGlobalConnection.State == System.Data.ConnectionState.Open;
   }

   public static void SqlDisconnect()
   {
      if (myGlobalConnection!= null)
      {
         if (myGlobalConnection.State == ConnectionState.Open)
            myGlobalConnection.Close();

         // In case some "other" state, always try to force CLOSE
         // such as Connecting, Broken, Fetching, etc...
         try
         { myGlobalConnection.Close(); }
         catch
         { // notify user if issue}
      }
   }
}

... in your other class / function...

   if( ConnectionHandler.SqlConnect() )
      Using( SqlCeConnection conn = ConnectionHandler.GetConnection )
      {
         // do your stuff
      }

... finally, when your app is finished, or any other time you need to...

ConnectionHandler.SqlDisconnect();

This keeps things centralized, and you don't have to worry about open/close, what the connection string is buried all over the place, etc... If you can't connect, you can't run a query, don't try to run the query if it can't even get that far.

Question 3

I think it may be a permission issue on INFORMATION_SCHEMA system views. Try the following.

GRANT VIEW DEFINITION TO your_user;

See here for more details

Why is my SQL Server CE code failing?

UPDATE

UPDATE 2

UPDATE 3

UPDATE 4

UPDATE 5

UPDATE 6