[Serializable] does not inherit (aka RTFM)

Of course, everybody knows that if you want to serialize an object, you need to decorate the class with the [Serializable] attribute. Right? Well, yes, and so do I, but I ran into a snag today: it is an attribute, not a property so it does not inherit. If you consider the following code

using System;

namespace demo
{
    [Serializable]
    public class BaseClass
    {
        public string AProperty { get; set; }
    }

    public class ChildClass : BaseClass
    {
        public string AnotherProperty { get; set; }
    }
}

you will find that Childclass will not serialize with BinaryFormatter, although its base class will. This is of course quite evident, but a more interesting gotcha occurs when you add a third class:

    [Serializable]
    public class ChildClass2 : ChildClass
    {
        public string FinalProperty { get; set; }
    }

This won't serialize either, because halfway up it's class hierarchy there's a non-serializable class. In plain vanilla .NET this will generate an "Failed to serialize. Reason: Type 'demo.ChildClass' in Assembly 'bla bla bla' is not marked as serializable", precisely pinpointing the problem. If it happens (as in my case) deep down inside a CSLA dataportal, all you get is an error message like

System.ServiceModel.CommunicationException: The underlying connection was closed: The connection was closed unexpectedly. ---> System.Net.WebException: The underlying connection was closed: The connection was closed unexpectedly.

And you might spend quite some time finding out it is caused by a non-serializable class somewhere in your class hierarchy. So, check your attributes, folkes ;-) Or run FxCop (and read the warnings)

Calling Oracle stored functions from ADO.NET

Die-hard Microsofties may be flabbergasted, but there are people out there using Oracle as a database in stead of SQLServer and - oh horror - are even daft enough to try to retrieve data from it using ADO.NET. But while Microsoft cranks out code samples for using SQLServer by the the truckload, those that are living in a hybrid environment may have a hard time finding actual working data access sample code. Like calling a stored function in Oracle and retrieving the value. Suppose whe have the following trivial stored function, created by my collegue Louis Ywema for unit testing purposes. It takes a numerical value and a string value and concatenates them with a hyphen between. Not quite the functionality you've been waiting for all your life but it proves the point

create function get_test  (p_parameter1 in number
                   ,p_parameter2 in varchar2)
return varchar2
is
   l_parameter1 number;
   l_parameter2 varchar2(1024);
begin
   l_parameter1 := p_parameter1;
   l_parameter2 := p_parameter2;
   return to_char(l_parameter1)||' - '||l_parameter2;
exception
   when others then
       return(sqlerrm);       
end get_test

You run this code in a database user that has CREATE PROCEDURE privileges Next, you define an Oracle connection in your connectionStrings section of your config file:

<connectionStrings>
  <add name="MYCONNECTION" 
    connectionString="Password=TEST;User ID=TEST;Data Source=JOOST;"
    providerName="System.Data.OracleClient" />
</connectionStrings>

I use a database JOOST with username TEST and password TEST Using raw ADO.NET code you can run the stored function like this

[TestFixture]
[Test(Description = "Stored function - Raw ADO.NET")]
public void TestStoredFunctionRaw()
{
  var myConnection = 
      ConfigurationManager.ConnectionStrings["MYCONNECTION"];
  using (var connection = 
      new OracleConnection(myConnection.ConnectionString))
  {
     connection.Open();
     using(var command = connection.CreateCommand() )
     {
       var p1 = new OracleParameter("p_parameter1", OracleType.Number);
       p1.Value = 1;
       command.Parameters.Add(p1);

       var p2 = new OracleParameter("p_parameter2", OracleType.VarChar);
       p2.Value = "Hello";
       command.Parameters.Add(p2);

       var retVal = new OracleParameter("returnvalue", OracleType.VarChar);
       retVal.Size = 1024;
       retVal.Direction = ParameterDirection.ReturnValue;
       command.Parameters.Add(retVal);

       command.CommandText = "get_test";
       command.CommandType = CommandType.StoredProcedure;
       command.ExecuteNonQuery();

       Assert.IsTrue(((string)retVal.Value) == "1 - Hello");
     }
  }
}

I would not recommend this to anyone but it proves the point. For real life situations, use the Enterprise Library and write your code like this

[Test(Description = "Stored function -  EL")]
public void TestStoredFunctionEntLib()
{
    Database Db = DatabaseFactory.CreateDatabase("MYCONNECTION");
    using (var cmd = Db.GetSqlStringCommand("get_test" ) )
    {
      Db.AddInParameter(cmd, "p_parameter1", DbType.VarNumeric, 1);
      Db.AddInParameter(cmd, "p_parameter2", DbType.String, "Hello");
      Db.AddParameter(cmd, "returnvalue", DbType.String,1024, 
        ParameterDirection.ReturnValue, 
        false, 0, 0, null, DataRowVersion.Default, null);

      cmd.CommandType = CommandType.StoredProcedure;
      Db.ExecuteNonQuery(cmd);
      Assert.IsTrue(((string) cmd.Parameters[2].Value) == "1 - Hello");
    }
}

Notice the "returnvalue" parameter - it has a ParameterDirection.ReturnValue and more importantly - it has a size. This is important: the default size of a parameter is 0, and if Oracle tries to write back te parameter value into the parameter you will get an error indicating so. In this example I use 1024, but you can set this to any size you need. In my production code I set it to 32K (32768) but I am not sure of the absolute maximum value. The crazy thing about this is, of course, there are situations where you cannot know the size of the return value. Your only option is to set is as high pas possible and hope for the best ;-) Complete test project downloadable here. This example still uses the May 2007 EntLib 3.1, by the way.

Converting colors from RGB to HTML and back

A very short one this time: it turns out to be pretty easy to convert a RGB color to a HTML color and back. This may come handy when (like me) you are using configuration files in which colors for map features are specified in RGB. Add a reference to System.Drawing.dll and a "using System.Drawing" to your class and you can use the following code:

Color rgbColor = Color.FromArgb(255, 0, 0);
string HTMLColor = ColorTranslator.ToHtml(rgbColor);
Color rgbColor2 = ColorTranslator.FromHtml(HTMLColor);
Color rgbColor3 = ColorTranslator.FromHtml("Fuchsia");

 Console.WriteLine(HTMLColor);
 Console.WriteLine(string.Format("{0},{1},{2}",
     rgbColor2.R, rgbColor2.G, rgbColor2.B));
 Console.WriteLine(string.Format("{0},{1},{2}",
     rgbColor3.R, rgbColor3.G, rgbColor3.B));

this wil output the following: #FF0000 255,0,0 255,0,255 I translated RGB red to HTML red, and back. For an encore, I demonstrated the fact that named colors are translated into RGB as well.