How to Choose between IDisposable, IComponent, Component, MarshalByValueComponent and Control while Implementing a Class

In this post, I'll discuss about the criteria behind choosing IDisposable, IComponent, Component, MarshalByValueComponent and base Control classes (System.Windows.Form.Control and System.Web.UI.Control) while implementing a Class.

Prior to discussing further, I'd like to share the reason behind this post. I ran code analysis in a project code and got a warning that Dispose was not called for a DataSet in the project’s code. I then called the Dispose method for that particular DataSet and this time, the code analysis executed with zero errors/warnings. Out of curiosity, I thought of verifying the use of calling Dispose on DataSet as even though when Dispose was not being called everything was fine. I found that calling Dispose method did nothing as DataSet is a managed object and does not contain unmanaged resources (if I'm wrong, please correct me on this). I could verify this as I was able to access/update the DataSet even after Dispose had been called without any “Object Disposed Exception”. FeatureSchema is a typed DataSet as displayed in the code snippet below:

FeatureSchema fs = Proxy.GetFeatureSchema();
fs.Dispose(); 
 
foreach (FeatureSchema.FeatureRow featureRow in fs.Feature.Rows)
{
//Do processing of dataset
}

I went one step further to find the reason behind having Dispose method in DataSet/DataTable and the reason was inheritance as DataSet/DataTable inherits from System.ComponentModel.MarshalByValueComponent class. The reason DataSet/DataTable inherits MarshalByValueComponent class so that it can be designable, i.e., used on a design surface. Thus DataSet/DataTable has dispose method even though it doesn’t clean up the resources. DataSet/DataTable call GC.SuppressFinalize in the constructor to prevent the Garbage Collector from calling Object.Finalize on an object that does not require it.

This can be proved by the code snippet displayed below.

I have a typed dataset FeatureSchema and I have added a destructor to it. I have excluded the code that was auto generated as I am focusing on finalizers only:

public partial class FeatureSchema : global::System.Data.DataSet
{
//Added only a destructor to this types dataset. I have excluded all other code
// as I am focusing on finalizers
~FeatureSchema ()
{
}
}

As displayed below, I am loading data to this typed dataset from a .xml file:

//I am just loading the data into this typed dataset
FeatureSchema schema = new FeatureSchema();
schema.ReadXml(featureSchemaFile, XmlReadMode.Auto);

As DataSet calls GC.SuppressFinalize in the constructor, Object.Finalize will not be called by the Garbage Collector. This can be verified by placing a break point on the destructor and you will find that it is not getting called.

Now I will call GC.ReRegisterForFinalize as displayed in the code snippet below:

//I am just loading the data into this typed dataset
//This time calling GC.GC.ReRegisterForFinalize()
FeatureSchema schema = new FeatureSchema();
GC.ReRegisterForFinalize(schema);
schema.ReadXml(featureSchemaFile, XmlReadMode.Auto);

Calling GC.ReRegisterForFinalize will register the object for finalization. I am doing this as I know that DataSet calls GC.SuppressFinalize in the constructor. Now the code will break on the break point which I placed on the constructor at some random time, i.e., when GC will execute the “GC Finalizer Thread”.

This verifies the need of calling GC.SuppressFinalize from the constructor of DataSet/DataTable to prevent the garbage collector from requesting finalization.

Even though Dispose in DataSet/DataTable does nothing, we should not ignore it and stick to the .NET coding practices, otherwise in future versions of Framework, things may break. Please do correct me if I am wrong. This concludes the reason behind this post.

I’ll continue with the main theme of the post. The guidelines for implementers are:

• IDisposable

  If a class uses external resources and will not be used on a design surface IDisposable interface has to be implemented directly or indirectly e.g. System.Drawing.Font class, etc.

• IComponent

  If a class will be used on a design surface IComponent interface has be implemented directly or indirectly. IComponent implements IDisposable e.g. System.Web.UI.Control class, etc.

• Component

  If a class will be used on a design surface and is Marshalled by reference, then it has to derive from Component class e.g. System.Timers.Timer class, etc.

• MarshalByValueComponent

  If a class will be used on a design surface and is Marshalled by value, then it has to derive from MarshalByValueComponent class, e.g. DataSet, DataTable, etc.

• Control

  If a class will be used on a design surface and provides a user interface, then this class is a control and has to derive from System.Windows.Form.Control or System.Web.UI.Control, e.g. System.Windows.Forms.Button, etc.

The above discussion does not apply to WPF. The WPF Designer architecture is significantly different from the Windows Forms Designer architecture, which is characterized by the IComponent interface and the System.ComponentModel namespace. The WPF Designer architecture retains the TypeConverter and TypeDescriptor classes from the Windows Forms Designer object model. Most other aspects of the WPF Designer architecture are different. For more information, please read Comparing the Windows Forms Designer Framework to the WPF Designer Framework.

Reference

• http://msdn.microsoft.com/en-us/library/0b1dk63b.aspx