c# downcasting when binding to and interface

Question

Is there a better way of binding a list of base class to a UI other than downcasting e.g:

static void Main(string[] args) {
    List<Animal> list = new List<Animal>();  
    Pig p = new Pig(5);  
    Dog d = new Dog("/images/dog1.jpg");  
    list.Add(p);  
    list.Add(d);  
    foreach (Animal a in list)   
    {  
        DoPigStuff(a as Pig);  
        DoDogStuff(a as Dog);  
    }  

}  


static void DoPigStuff(Pig p)
{
    if (p != null) 
    {  
        label1.Text = String.Format("The pigs tail is {0}", p.TailLength);
    }  
}

static void DoDogStuff(Dog d) {
    if (d != null) 
    {
        Image1.src = d.Image;
    }
}

class Animal {
    public String Name { get; set; }
}

class Pig : Animal{
    public int TailLength { get; set; }

    public Pig(int tailLength) 
    {
        Name = "Mr Pig";
        TailLength = tailLength;
    }
}

class Dog : Animal {
    public String Image { get; set; }

    public Dog(String image) 
    {
        Name = "Mr Dog";
        Image = image;
    }
}

Answer 1

Why not make Animal include an abstract method that Pig and Dog are forced to implement

public class Animal
{
    public abstract void DoStuff();
}

public Dog : Animal
{
    public override void DoStuff()
    {
        // Do dog specific stuff here
    }
}

public Pig : Animal
{
    public override void DoStuff()
    {
        // Do pig specific stuff here
    }
}

This way each specific class takes responsibility for its actions, making your code simpler. You also won't need to cast inside your foreach loop.

Answer 2

You're not taking full advantage of your base class. If you had a virtual function in your Animal class that Dog & Pig override, you wouldn't need to cast anything.

Answer 3

Sorry I don't think I made this clear enough that I was asking about the specific situation of binding to a user interface. Obviously sticking UI code inside the class would be a bad idea.

Answer 4

Unless you have a more specific example, just override ToString().

Answer 5

Another way to do this is to perform a typecheck before calling the method:

if (animal is Pig) DoPigStuff();
if (animal is Dog) DoDogStuff();

What you are looking for is multiple-dispatch. NO - C# doesn't support multiple-dispatch. It only supports single-dispatch. C# can only dynamically invoke a method based on the type of the receiver (i.e. the object at the left hand side of the . in the method call)

This code uses double-dispatch. I'll let the code speak for itself:

class DoubleDispatchSample
{
    static void Main(string[]args)
    {
        List<Animal> list = new List<Animal>();
        Pig p = new Pig(5);
        Dog d = new Dog(@"/images/dog1.jpg");
        list.Add(p);
        list.Add(d);

        Binder binder = new Binder(); // the class that knows how databinding works

        foreach (Animal a in list)
        {
            a.BindoTo(binder); // initiate the binding
        }
    }
}

class Binder
{
    public void DoPigStuff(Pig p)
    {
        label1.Text = String.Format("The pigs tail is {0}", p.TailLength);
    }

    public void DoDogStuff(Dog d)
    {
        Image1.src = d.Image;
    }
}

internal abstract class Animal
{
    public String Name
    {
        get;
        set;
    }

    protected abstract void BindTo(Binder binder);
}

internal class Pig : Animal
{
    public int TailLength
    {
        get;
        set;
    }

    public Pig(int tailLength)
    {
        Name = "Mr Pig";
        TailLength = tailLength;
    }

    protected override void BindTo(Binder binder)
    {
        // Pig knows that it's a pig - so call the appropriate method.
        binder.DoPigStuff(this);
    }
}

internal class Dog : Animal
{
    public String Image
    {
        get;
        set;
    }

    public Dog(String image)
    {
        Name = "Mr Dog";
        Image = image;
    }

    protected override void BindTo(Binder binder)
    {
        // Pig knows that it's a pig - so call the appropriate method.
        binder.DoDogStuff(this);
    }
}

NOTE: Your sample code is much more simpler than this. I think of double-dispatch as one of the heavy artilleries in C# programming - I only take it out as a last resort. But if there are a lot of types of objects and a lot different types of bindings that you need to do (e.g. you need to bind it to an HTML page but you also need to bind it to a WinForms or a report or a CSV), I would eventually refactor my code to use double-dispatch.

Answer 6

When faced with this type of problem, I follow the visitor pattern.

interface IVisitor
{
  void DoPigStuff(Piggy p);
  void DoDogStuff(Doggy d);
}

class GuiVisitor : IVisitor
{
  void DoPigStuff(Piggy p)
  {
    label1.Text = String.Format("The pigs tail is {0}", p.TailLength);
  }

  void DoDogStuff(Doggy d)
  {
    Image1.src = d.Image;
  }
}

abstract class Animal
{
    public String Name { get; set; }
    public abstract void Visit(IVisitor visitor);
}

class Piggy : Animal
{
    public int TailLength { get; set; }

    public Piggy(int tailLength) 
    {
        Name = "Mr Pig";
        TailLength = tailLength;
    }

    public void Visit(IVisitor visitor)
    {
       visitor.DoPigStuff(this);
    }
}

class Doggy : Animal 
{
   public String Image { get; set; }

   public Doggy(String image) 
   {
     Name = "Mr Dog";
     Image = image;
   }

   public void Visit(IVisitor visitor)
   {
     visitor.DoDogStuff(this);
   }
}

public class AnimalProgram
{
  static void Main(string[] args) {
    List<Animal> list = new List<Animal>();  
    Pig p = new Pig(5);  
    Dog d = new Dog("/images/dog1.jpg");  
    list.Add(p);  
    list.Add(d);

    IVisitor visitor = new GuiVisitor();  
    foreach (Animal a in list)   
    {
      a.Visit(visitor);
    }  
  }
}

Thus the visitor pattern simulates double dispatch in a conventional single-dispatch object-oriented language such as Java, Smalltalk, C#, and C++.

The only advantage of this code over jop's is that the IVisitor interface can be implemented on a different class later when you need to add a new type of visitor (like a XmlSerializeVisitor or a FeedAnimalVisitor).

Answer 7

I think you want a view-class associated with a factory.

Dictionary<Func<Animal, bool>, Func<Animal, AnimalView>> factories;
factories.Add(item => item is Dog, item => new DogView(item as Dog));
factories.Add(item => item is Pig, item => new PigView(item as Pig));

Then your DogView and PigView will inherit AnimalView that looks something like:

class AnimalView {
  abstract void DoStuff();
}

You will end up doing something like:

foreach (animal in list)
  foreach (entry in factories)
    if (entry.Key(animal)) entry.Value(animal).DoStuff();

I guess you could also say that this is a implementation of the strategy pattern.