Inheritance

Syntax

  • class DerivedClass : BaseClass
  • class DerivedClass : BaseClass, IExampleInterface
  • class DerivedClass : BaseClass, IExampleInterface, IAnotherInterface

Remarks

Classes can inherit directly from only one class, but (instead or at the same time) can implement one or more interfaces.

Structs can implement interfaces but cannot explicitly inherit from any type. They implicitly inherit from System.ValueType, which in turn inherits directly from System.Object.

Static classes cannot implement interfaces.

Base class with recursive type specification

One time definition of a generic base class with recursive type specifier. Each node has one parent and multiple children.

/// <summary>
/// Generic base class for a tree structure
/// </summary>
/// <typeparam name="T">The node type of the tree</typeparam>
public abstract class Tree<T> where T : Tree<T>
{
    /// <summary>
    /// Constructor sets the parent node and adds this node to the parent's child nodes
    /// </summary>
    /// <param name="parent">The parent node or null if a root</param>
    protected Tree(T parent)
    {
        this.Parent=parent;
        this.Children=new List<T>();
        if(parent!=null)
        {
            parent.Children.Add(this as T);
        }
    }
    public T Parent { get; private set; }
    public List<T> Children { get; private set; }
    public bool IsRoot { get { return Parent==null; } }
    public bool IsLeaf { get { return Children.Count==0; } }
    /// <summary>
    /// Returns the number of hops to the root object
    /// </summary>
    public int Level { get { return IsRoot ? 0 : Parent.Level+1; } }
}

The above can be re-used every time a tree hierarchy of objects needs to be defined. The node object in the tree has to inherit from the base class with

public class MyNode : Tree<MyNode>
{
    // stuff
}

each node class knows where it is in the hierarchy, what the parent object is as well as what the children objects are. Several built in types use a tree structure, like Control or XmlElement and the above Tree<T> can be used as a base class of any type in your code.


For example, to create a hierarchy of parts where the total weight is calculated from the weight of all the children, do the following:

public class Part : Tree<Part>
{
    public static readonly Part Empty = new Part(null) { Weight=0 };
    public Part(Part parent) : base(parent) { }
    public Part Add(float weight)
    {
        return new Part(this) { Weight=weight };
    }
    public float Weight { get; set; }

    public float TotalWeight { get { return Weight+Children.Sum((part) => part.TotalWeight); } }
}

to be used as

// [Q:2.5] -- [P:4.2] -- [R:0.4]
//    \
//      - [Z:0.8]
var Q = Part.Empty.Add(2.5f);
var P = Q.Add(4.2f);
var R = P.Add(0.4f);
var Z = Q.Add(0.9f);

// 2.5+(4.2+0.4)+0.9 = 8.0
float weight = Q.TotalWeight;

Another example would in the definition of relative coordinate frames. In this case the true position of the coordinate frame depends on the positions of all the parent coordinate frames.

public class RelativeCoordinate : Tree<RelativeCoordinate>
{
    public static readonly RelativeCoordinate Start = new RelativeCoordinate(null, PointF.Empty) { };
    public RelativeCoordinate(RelativeCoordinate parent, PointF local_position)
        : base(parent)
    {
        this.LocalPosition=local_position;
    }
    public PointF LocalPosition { get; set; }
    public PointF GlobalPosition
    {
        get
        {
            if(IsRoot) return LocalPosition;
            var parent_pos = Parent.GlobalPosition;
            return new PointF(parent_pos.X+LocalPosition.X, parent_pos.Y+LocalPosition.Y);
        }
    }
    public float TotalDistance
    {
        get
        {
            float dist = (float)Math.Sqrt(LocalPosition.X*LocalPosition.X+LocalPosition.Y*LocalPosition.Y);
            return IsRoot ? dist : Parent.TotalDistance+dist;
        }
    }
    public RelativeCoordinate Add(PointF local_position)
    {
        return new RelativeCoordinate(this, local_position);
    }
    public RelativeCoordinate Add(float x, float y)
    {
        return Add(new PointF(x, y));
    }
}

to be used as

// Define the following coordinate system hierarchy
//
// o--> [A1] --+--> [B1] -----> [C1]
//             |     
//             +--> [B2] --+--> [C2]
//                         |
//                         +--> [C3]

var A1 = RelativeCoordinate.Start;
var B1 = A1.Add(100, 20);
var B2 = A1.Add(160, 10);

var C1 = B1.Add(120, -40);
var C2 = B2.Add(80, -20);
var C3 = B2.Add(60, -30);

double dist1 = C1.TotalDistance;

Constructors In A Subclass

When you make a subclass of a base class, you can construct the base class by using : base after the subclass constructor's parameters.

class Instrument
{
    string type;
    bool clean;

    public Instrument (string type, bool clean)
    {
        this.type = type;
        this.clean = clean;
    }
}

class Trumpet : Instrument
{
    bool oiled;

    public Trumpet(string type, bool clean, bool oiled) : base(type, clean)
    {
        this.oiled = oiled;
    }
}

Extending an abstract base class

Unlike interfaces, which can be described as contracts for implementation, abstract classes act as contracts for extension.

An abstract class cannot be instantiated, it must be extended and the resulting class (or derived class) can then be instantiated.

Abstract classes are used to provide generic implementations

public abstract class Car
{
    public void HonkHorn() {
        // Implementation of horn being honked
    }
}

public class Mustang : Car
{
    // Simply by extending the abstract class Car, the Mustang can HonkHorn()
    // If Car were an interface, the HonkHorn method would need to be included
    // in every class that implemented it.
}

The above example shows how any class extending Car will automatically receive the HonkHorn method with the implementation. This means that any developer creating a new Car will not need to worry about how it will honk it's horn.

Inheritance Anti-patterns

Improper Inheritance

Lets say there are 2 classes class Foo and Bar. Foo has two features Do1 and Do2. Bar needs to use Do1 from Foo, but it doesn't need Do2 or needs feature that is equivalent to Do2 but does something completely different.

Bad way: make Do2() on Foo virtual then override it in Bar or just throw Exception in Bar for Do2()

public class Bar : Foo
{
    public override void Do2()
    {
        //Does something completely different that you would expect Foo to do
        //or simply throws new Exception 
    }
}

Good way

Take out Do1() from Foo and put it into new class Baz then inherit both Foo and Bar from Baz and implement Do2() separately

public class Baz
{
    public void Do1()
    {
        // magic
    }
}

public class Foo : Baz
{
    public void Do2()
    {
        // foo way
    }
}

public class Bar : Baz
{
    public void Do2()
    {
        // bar way or not have Do2 at all
    }
}

Now why first example is bad and second is good: When developer nr2 has to do a change in Foo, chances are he will break implementation of Bar because Bar is now inseparable from Foo. When doing it by latter example Foo and Bar commonalty has been moved to Baz and they do not affect each other (like the shouldn't).

Inheritance. Constructors' calls sequence

Consider we have a class Animal which has a child class Dog

class Animal
{
    public Animal()
    {
        Console.WriteLine("In Animal's constructor");
    }
}

class Dog : Animal
{
    public Dog()
    {
        Console.WriteLine("In Dog's constructor");
    }
}

By default every class implicitly inherits the Object class.

This is same as the above code.

class Animal : Object
{
    public Animal()
    {
        Console.WriteLine("In Animal's constructor");
    }
}

When creating an instance of Dog class, the base classes's default constructor (without parameters) will be called if there is no explicit call to another constructor in the parent class. In our case, first will be called Object's constructor, then Animal's and at the end Dog's constructor.

public class Program
{
    public static void Main()
    {
        Dog dog = new Dog();
    }
}

Output will be

In Animal's constructor
In Dog's constructor

View Demo

Call parent's constructor explicitly.

In the above examples, our Dog class constructor calls the default constructor of the Animal class. If you want, you can specify which constructor should be called: it is possible to call any constructor which is defined in the parent class.

Consider we have these two classes.

class Animal
{
    protected string name;

    public Animal()
    {
        Console.WriteLine("Animal's default constructor");
    }    

    public Animal(string name)
    {
        this.name = name;
        Console.WriteLine("Animal's constructor with 1 parameter");
        Console.WriteLine(this.name);
    }
}

class Dog : Animal
{
    public Dog() : base()
    {
        Console.WriteLine("Dog's default constructor");
    }  

    public Dog(string name) : base(name)
    {
        Console.WriteLine("Dog's constructor with 1 parameter");
        Console.WriteLine(this.name);
    }
}

What is going here?

We have 2 constructors in each class.

What does base mean?

base is a reference to the parent class. In our case, when we create an instance of Dog class like this

Dog dog = new Dog();

The runtime first calls the Dog(), which is the parameterless constructor. But its body doesn't work immediately. After the parentheses of the constructor we have a such call: base(), which means that when we call the default Dog constructor, it will in turn call the parent's default constructor. After the parent's constructor runs, it will return and then, finally, run the Dog() constructor body.

So output will be like this:

Animal's default constructor
Dog's default constructor

View Demo

Now what if we call the Dog's constructor with a parameter?

Dog dog = new Dog("Rex");

You know that members in the parent class which are not private are inherited by the child class, meaning that Dog will also have the name field.
In this case we passed an argument to our constructor. It in his turn passes the argument to the parent class' constructor with a parameter, which initializes the name field.

Output will be

Animal's constructor with 1 parameter
Rex
Dog's constructor with 1 parameter
Rex

Summary:

Every object creation starts from the base class. In the inheritance, the classes which are in the hierarchy are chained. As all classes derive from Object, the first constructor to be called when any object is created is the Object class constructor; Then the next constructor in the chain is called and only after all of them are called the object is created

base keyword

  1. The base keyword is used to access members of the base class from within a derived class:
  2. Call a method on the base class that has been overridden by another method. Specify which base-class constructor should be called when creating instances of the derived class.

Inheriting from a base class

To avoid duplicating code, define common methods and attributes in a general class as a base:

public class Animal 
{
    public string Name { get; set; }
    // Methods and attributes common to all animals
    public void Eat(Object dinner)
    {
        // ...
    }
    public void Stare()
    {
        // ...
    }
    public void Roll()
    {
        // ...
    }
}

Now that you have a class that represents Animal in general, you can define a class that describes the peculiarities of specific animals:

public class Cat : Animal
{
    public Cat() 
    {
        Name = "Cat";
    }
    // Methods for scratching furniture and ignoring owner
    public void Scratch(Object furniture)
    {
        // ...
    }
}

The Cat class gets access to not only the methods described in its definition explicitly, but also all the methods defined in the general Animal base class. Any Animal (whether or not it was a Cat) could Eat, Stare, or Roll. An Animal would not be able to Scratch, however, unless it was also a Cat. You could then define other classes describing other animals. (Such as Gopher with a method for destroying flower gardens and Sloth with no extra methods at all.)

Inheriting from a class and implementing an interface

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

public interface INoiseMaker
{
    string MakeNoise();
}

//Note that in C#, the base class name must come before the interface names
public class Cat : Animal, INoiseMaker
{
    public Cat() 
    {
        Name = "Cat";
    }

    public string MakeNoise()
    {
        return "Nyan";
    }
}

Inheriting from a class and implementing multiple interfaces

public class LivingBeing
{
    string Name { get; set; }
}

public interface IAnimal 
{
    bool HasHair { get; set; }
}

public interface INoiseMaker
{
    string MakeNoise();
}

//Note that in C#, the base class name must come before the interface names
public class Cat : LivingBeing, IAnimal, INoiseMaker
{
    public Cat() 
    {
        Name = "Cat";
        HasHair = true;
    }

    public bool HasHair { get; set; }

    public string Name { get; set; }

    public string MakeNoise()
    {
        return "Nyan";
    }
}

Inheriting methods

There are several ways methods can be inherited

public abstract class Car
{
    public void HonkHorn() {
        // Implementation of horn being honked
    }

    // virtual methods CAN be overridden in derived classes
    public virtual void ChangeGear() {
        // Implementation of gears being changed
    }

    // abstract methods MUST be overridden in derived classes
    public abstract void Accelerate();
}

public class Mustang : Car
{
    // Before any code is added to the Mustang class, it already contains 
    // implementations of HonkHorn and ChangeGear.

    // In order to compile, it must be given an implementation of Accelerate,
    // this is done using the override keyword
    public override void Accelerate() {
        // Implementation of Mustang accelerating
    }

    // If the Mustang changes gears differently to the implementation in Car
    // this can be overridden using the same override keyword as above
    public override void ChangeGear() {
        // Implementation of Mustang changing gears
    }
}

Testing and navigating inheritance

interface BaseInterface {}
class BaseClass : BaseInterface {}

interface DerivedInterface {}
class DerivedClass : BaseClass, DerivedInterface {}

var baseInterfaceType = typeof(BaseInterface);
var derivedInterfaceType = typeof(DerivedInterface);
var baseType = typeof(BaseClass);
var derivedType = typeof(DerivedClass);

var baseInstance = new BaseClass();
var derivedInstance = new DerivedClass();  

Console.WriteLine(derivedInstance is DerivedClass); //True
Console.WriteLine(derivedInstance is DerivedInterface); //True
Console.WriteLine(derivedInstance is BaseClass); //True
Console.WriteLine(derivedInstance is BaseInterface); //True
Console.WriteLine(derivedInstance is object); //True

Console.WriteLine(derivedType.BaseType.Name);  //BaseClass
Console.WriteLine(baseType.BaseType.Name);  //Object
Console.WriteLine(typeof(object).BaseType);  //null

Console.WriteLine(baseType.IsInstanceOfType(derivedInstance));  //True
Console.WriteLine(derivedType.IsInstanceOfType(baseInstance));  //False

Console.WriteLine(
    string.Join(",", 
    derivedType.GetInterfaces().Select(t => t.Name).ToArray()));
//BaseInterface,DerivedInterface
    
Console.WriteLine(baseInterfaceType.IsAssignableFrom(derivedType)); //True
Console.WriteLine(derivedInterfaceType.IsAssignableFrom(derivedType)); //True
Console.WriteLine(derivedInterfaceType.IsAssignableFrom(baseType)); //False