C# Keywords, Part I: Modifier Keywords

Introduction

I expected that learning C# would be easy considering my extensive C++ background. Interestingly, I haven�t found this to be particularly true. There are subtle differences between the languages which will trip up the C++ programmer, and there�s quite a few new things in C# (and features that C++ has but are missing from C#) to make life just a little more complicated than I expected. I realized that while I am �functional� in C#, I am far from being �proficient�.

I figured it was time to get back to basics and really explore the language in the way languages used to be taught in the old days: start with the keywords, then move into the syntax, and then start writing some basic (as in simple!) programs.

So, for all those interested, this article gives a synopsis of the various keywords in C#. The reader should be familiar with C++ or at least the concepts of classes and objects. In some ways, this is a beginner�s tutorial, but in other ways, it can be a very useful quick reference guide to the C++ veteran (especially the use of the override and new modifiers).

A really good place to begin is this site provided by Microsoft.

The Keywords

Following is a table of the keywords in C#. Unfortunately, they are organized alphabetically, instead of categorically.

Note that �value� is left out of this list, but it is syntax highlighted like a keyword in the IDE.

Let�s take this list and create some categories so that we can better organize and understand the keywords. Now, a disclaimer: I am not a compiler expert, nor have I cracked open a book on programming in a long time, so you�ll have to forgive me if you don�t like the categorization of some of your favorite keywords!

Keywords that modify a class, method, property, or field:

• abstract
• const
• extern
• internal
• new
• override
• private
• protected
• public
• readonly
• sealed
• static
• virtual
• volatile
• void

Keywords dealing with object type and type conversions:

• as
• explicit
• implicit
• is
• operator
• sizeof
• typeof

Keywords that are aliases to built in types, are specific types, or have to do with enumerations:

• bool
• byte
• char
• class
• decimal
• double
• enum
• float
• int
• interface
• long
• object
• sbyte
• short
• string
• struct
• uint
• ulong
• ushort

Keywords that control program flow:

• break
• case
• continue
• default
• do
• else
• for
• foreach, in
• goto
• if
• return
• switch
• while

Keywords that are used in exception handling:

• catch
• checked
• finally
• throw
• try
• unchecked

Keywords that are like C++ function pointers and related subjects:

• delegate
• event

Keywords that affect garbage collection:

• fixed

Keywords to help with critical sections of code:

• lock

Keywords that declare scope:

• namespace

Keywords that control object allocation/destruction:

• new
• stackalloc

Keywords that affect method parameter passing:

• out
• params
• ref

Keywords that are literals or refer to the current instance of an object:

• null
• false
• true
• this
• value

Keywords dealing with unmanaged code:

• unsafe

Keywords, miscellaneous:

• base
• void

Keywords That Modify Accessibility

In Part I, I'll look specifically at the keywords that modify the accessibility of classes, methods, fields and properties. MSDN provides excellent descriptions of these keywords, so I have two choices here�I can either plagiarize what they�ve already done, or I could write my own descriptions based on my own understanding, which may or may not be as complete, accurate or useful. I opted for the second option, as this way, I�ll end up writing little code snippets to try things out, and I�ll have learned something along the way (whether you will learn something, well, I don�t know about that).

abstract

Abstract class

When used as a modifier for a class, this keyword declares the class as being intended to be used only as a base class for other classes. You cannot directly instantiate an abstract class. For this reason, the sealed modifier can�t be used, because this prevents an abstract class from being inherited. An abstract class can have members that implement actual functionality, as compared to an interface.

Abstract classes are useful when providing generic functionality to a class structure. As the name implies, an abstract class should represent the abstraction of some object.

Abstract method

When used as a modifier for a method, the method does not have an implementation�the implementation must be provided by the derived class. Because abstract methods are implicitly virtual (meaning a derived class must provide the implementation), the derived class must use the keyword override as part of the declaration of the inherited class� abstract method.

Abstract methods are useful as a placeholder for a common process shared by specific implementations. Unlike virtual methods, in which the base class can provide some generic implementation, an abstract method forces the programmer to provide an implementation in all derived classes. This may be done because there is no generic implementation that is reasonable to provide.

Abstract attribute

An attribute can be abstract as well, which also requires that its implementation be provided by the derived class. Furthermore, abstract attributes cannot be static because they must be implemented in the derived class.

Abstract attributes are useful in the same situations as described for abstract methods.

Example:

abstract class AbstractClass
{
    public int i=5;
    public void Method()
    {
        Console.WriteLine("AbstractClass::Method()  i="+i.ToString());
    }

    // can't have an implementation:

    public abstract void AbstractMethod();

    // the property cannot have an implementation either:

    public abstract int I
    {
        get;
    }
}

class ConcreteClass : AbstractClass
{
    // because there's an abstract method in the inherited class,

    // we have to provide the implementation here.  Note that we must

    // also provide the "override" keyword.

    public override void AbstractMethod()
    {
        Console.WriteLine("ConcreteClass:AbstractMethod");
    }

    // similarly, for the property:

    public override int I
    {
        get {return i;}
    }
}

// can't do: AbstractClass abstractClass=new AbstractClass();

ConcreteClass concreteClass=new ConcreteClass();
concreteClass.Method();
concreteClass.AbstractMethod();
Console.WriteLine("I="+concreteClass.I.ToString());

const and readonly

The const keyword only applies to variables, which MSDN calls �fields�. A field with the const keyword indicates that the field cannot be changed. The value of the field must be initialized in the declaration. Interestingly, attributes cannot be declared as const, and furthermore, attributes referring to const fields can only have getters, not setters.

In C++, the const keyword was also used to indicate that the method did not modify any variables:

void Foo() const;

or that the method returned a reference that could not be modified:

const int& Foo();

or that the parameter being passed as a reference could not be modified by the method:

void Foo(const int& i);

which could lead to some ridiculous but valid statements:

const int& Foo(const int& i) const;

Happily, C# treats the keyword const with only one meaning.

Declaring fields as const protects both you and other programmers from accidentally changing the value of the field. Also note that with const fields, the compiler performs some optimization by not declaring any stack space for the field. For example:

const int foo=10;
int bar=16;

results in stack space being used for bar:

00000020  mov         dword ptr [ebp-8],10h 

but not for foo, which is loaded into a register as an immediate value:

Console.WriteLine("foo="+foo.ToString()+", bar="+bar.ToString());

00000096  mov         dword ptr [ebp-18h],0Ah     ; foo

0000009d  mov         edi,dword ptr ds:[01C40070h] 
000000a3  lea         ecx,[ebp-18h] 
000000a6  call        dword ptr ds:[02EA2758h] 
000000ac  mov         dword ptr [ebp-1Ch],eax 
000000af  mov         eax,dword ptr ds:[01C40074h] 
000000b5  mov         dword ptr [ebp-20h],eax 
000000b8  lea         ecx,[ebp-8]             ; bar

Furthermore, const can only be applied to intrinsic types. You can�t say:

const HasConstField hcf2=new HasConstField();

Example:

class HasConstField
{
    public HasConstField(); // can't do: {i=3;}

    public const int i=1;

    public int I
    {
        get {return i;}
        // can't do: set {i=5;}

    }
}
HasConstField hcf=new HasConstField();
// can't do: hcf.i=3;

The readonly keyword is similar to const, with two exceptions. First, the storage of a readonly field is the same as a regular (read-write) field, and thus there is no performance benefit. Secondly, readonly fields can be initialized in the constructor of the containing class.

For example (from MSDN):

public class ReadOnlyTest 
{
   class MyClass 
   {
      public int x;
      public readonly int y = 25; // Initialize a readonly field

      public readonly int z;

      public MyClass() 
      {
         z = 24;   // Initialize a readonly instance field

      }

      public MyClass(int p1, int p2, int p3) 
      {
         x = p1; 
         y = p2; 
         z = p3;
      }
   }
}

extern

The extern keyword is basically used just for non-C# function calls, which also requires:

using System.Runtime.InteropServices;

and is used in conjunction with the DllImport attribute, so that your C# program can call functions in other DLLs. This is a very useful capability of the C# language, because there are cases where you need direct access to the underlying Windows API.

Example:

[DllImport("User32.dll")]
public static extern int MessageBox(int h, string m, string c, int type);

MessageBox(0, "Test", "My Message Box", 0);

internal

The internal keyword restricts the visibility of the member to just the specific files in the assembly in which the internal class is declared.

For classes, an internal class and its members (public or not) are not visible to any programs that reference the assembly.

For methods, fields and properties, using the internal keyword limits the visibility of that member to just the assembly. Thus, you can have a class visible to the outside world, but the class can contain methods, fields and properties that are visible only within the assembly. Compare this to the protected and private keywords, which restrict access of members to derived classes and the class methods themselves, respectively.

Also, all internal members are treated as public within the scope of the assembly. The only exception to this is the �protected internal� member which limits access to just derived classes. This allows the programmer to control the access to the member within the assembly (ensuring non-derived objects don�t tamper with the value or access the function), while also making the method or field inaccessible outside of the assembly.

Example:

// This class is only visible to this assembly.

internal class InternalClass
{
}

class InternalTestClass
{
    // this field is visible only within this assembly.

    internal int i=2;
    // this method is visible only within this assembly.

    internal void InternalMethod()
    {
        Console.WriteLine("InternalMethod");
    }
    // this protected method can only be invoked by derived classes.

    internal protected void InternalProtectedMethod()
    {
        Console.WriteLine("InternalProtectedMethod");
    }
    // this attribute is visible only within this assembly.

    internal int I
    {
        get {return i;}
        set {i=value;}
    }
}

InternalTestClass itc=new InternalTestClass();
itc.I+=5;
Console.WriteLine("I="+itc.i.ToString());
itc.InternalMethod();

new, override, and virtual

Well, just when I thought C# tried to give only one meaning to keywords, I discover this gem. The keyword new, in addition to being used as an operator, can be used as a modifier to explicitly hide a member inherited from a base class (MSDN). Furthermore, MSDN states: Using new and virtual together guarantees a new point of specialization. Compare this with the MSDN definition for virtual: An override method provides a new implementation of a member inherited from a base class. Let�s explore what all this means. Also note that it is impossible to discuss the new modifier without also discussing the override modifier, therefore these two have been lumped together.

To explicitly hide a member inherited from a base class

First off, the new modifier is used exclusively with members of a derived class. A basic use of the new modifier is to declare a method in the derived class that just so happens to have the same name as a method in the base class. Without the new modifier, one gets the well known error �foo hides inherited member blah�, which is a common problem in C++. The new modifier allows a method in the derived class to have the same name as a method in the inherited class (assuming all else is the same�return type and parameter lists).

Example 1:

class NewBaseClass
{
    public void NonVirtualMethod()
    {
        Console.WriteLine("NewBaseClass:NonVirtualMethod");
    }
}

class NewClassA : NewBaseClass
{
    public new void NonVirtualMethod()
    {
        Console.WriteLine("NewClassA:NonVirtualMethod");
        base.NonVirtualMethod();
    }
}

In this example, the NonVirtualMethod method of the base class is hidden from the derived class, so it can have its own implementation but use the same name.

Using new and virtual together guarantees a new point of specialization

Now, things get more interesting when the inherited method has been declared as virtual. First, take a look at the following C# classes:

Example 2A:

class VirtualClass
{
    public virtual void VirtualMethod()
    {
        Console.WriteLine("VirtualClass:VirtualMethod");
    }
}

class OverrideClassA : VirtualClass
{
    public override void VirtualMethod()
    {
        Console.WriteLine("OverrideClassA:VirtualMethod");
        base.VirtualMethod();
    }
}

class OverrideClassB : OverrideClassA
{
    public override void VirtualMethod()
    {
        Console.WriteLine("OverrideClassB:VirtualMethod");
        base.VirtualMethod();
    }
}

and the test code:

Example 2B:

OverrideClassB ocb=new OverrideClassB();
ocb.VirtualMethod();
((OverrideClassA)ocb).VirtualMethod();
((VirtualClass)ocb).VirtualMethod();

resulting in:

OverrideClassB:VirtualMethod
OverrideClassA:VirtualMethod
VirtualClass:VirtualMethod
OverrideClassB:VirtualMethod
OverrideClassA:VirtualMethod
VirtualClass:VirtualMethod
OverrideClassB:VirtualMethod
OverrideClassA:VirtualMethod
VirtualClass:VirtualMethod

Now compare the above class definition with the following:

Example 2C:

class NewBaseClass
{
    public virtual void VirtualMethod()
    {
        Console.WriteLine("NewBaseClass:VirtualMethod");
    }
}

class NewClassA : NewBaseClass
{
    public new void VirtualMethod()
    {
        Console.WriteLine("NewClassA:VirtualMethod");
        base.VirtualMethod();
    }
}

class NewClassB : NewClassA
{
    public new void VirtualMethod()
    {
        Console.WriteLine("NewClassB:VirtualMethod");
        base.VirtualMethod();
    }
}

And note the use of the new modifier. When the following code is run:

Example 2D:

NewClassB ncb=new NewClassB();
ncb.VirtualMethod();
((NewClassA)ncb).VirtualMethod();
((NewBaseClass)ncb).VirtualMethod();

we get:

OverrideClassB:VirtualMethod
OverrideClassA:VirtualMethod
VirtualClass:VirtualMethod
OverrideClassA:VirtualMethod
VirtualClass:VirtualMethod
VirtualClass:VirtualMethod

Now, the cast allows us to access the specific base class member, because the derived class provides a new specialization, meaning that it doesn�t simply provide an implementation of the base class method, it provides a specialization. This is a critical distinction.

When implementing a class, the programmer must consider whether the class methods provide specialization vs. always override the base class functionality. One rule of thumb is to always assume specialization, except in the case of abstract classes, methods or interfaces, all of which must be overridden by the derived class.

Note that if the inherited class� method has been declared private, it is already hidden from the derived class and therefore the new modifier is not necessary. Also note that a virtual member cannot be private, because it is impossible to override a private member�it is private to the class that implements it.

One final point�inspecting the resulting assembly code shows that there is no difference in performance between invoking members that are modified with the new keyword vs. the override keyword.

private, protected, and public

These three modifiers affect the accessibility of methods, fields and attributes. The private modifier ensures that the method, field or attribute is visible only within the class it is defined. The protected modifier extends this visibility to derived classes, and the public modifier declares the method, field or attribute to be visible to everyone.

These three modifiers are fundamental to a good object oriented design. Much progress can be made towards reducing potential bugs by restricting visibility and access to methods, fields and attributes. However, inappropriate use of these modifiers can result in classes that are not extensible because visibility/access is incorrectly configured.

For example:

class PPP
{
    private int a=0;
    public int A
    {
        get {return a;}
        set {a=value;}
    }
    protected int A2
    {
        get {return a*2;}
        set {a=value/2;}
    }
    public int GetA() {return a;}
    protected int GetA2() {return a*2;}
    private void SetA2(int a2) {a=a2/2;}
}

sealed

The sealed modifier ensures that a class cannot be inherited. Therefore, sealed classes cannot be abstract, nor can they have abstract or virtual methods. There have been several complaints that some classes in the .NET environment shouldn�t have been sealed by the designers at Microsoft, so be careful when considering whether to seal your own classes. Similar to using the public, protected and private modifiers, using the sealed modifier on a class can make your object model difficult, if not impossible, to extend.

static

A method, field or property that has been modified with the static keyword is accessible without requiring an instance of the containing class. A static field is shared between all instances of an object. Also, methods and fields modified by the static keyword is a mechanism for providing �global� functions and variables.

For example:

class StaticClass
{
    public static int a=0;

    public int A
    {
        get {return a;}
        set {a=value;}
    }
}

StaticClass sc1=new StaticClass();
StaticClass sc2=new StaticClass();
sc1.A=10;
Console.WriteLine("sc2.A="+sc2.A.ToString());
StaticClass.a=20;
Console.WriteLine("sc1.A="+sc1.A.ToString());
Console.WriteLine("sc2.A="+sc2.A.ToString());

Outputs:

Sc2.A=10
Sc1.A=20
Sc2.A=20

Illustrating how even though a was set using instance sc1, it affected instance s1, and similarly how field a can be set without even needing an instance of the class, and its effects on instances.

volatile

This keyword modifies how the compiler generates code that reads and writes the contents of a field that uses this keyword. Using the volatile modifier ensures that the compiler generates code that always reads from the field�s memory location (as opposed to optimizing by perhaps re-using a register that was previously loaded with the field�s value), and immediately writes a value to the field (as opposed to postponing the write because it may be more efficient to do it later).

This keyword is useful when an external process (like the operating system or a thread in your application) can potentially modify the value of a field, and you need to ensure that you are reading the most current value.

Unfortunately, I cannot seem to get the C# compiler to generate code to demonstrate the differences between volatile and non-volatile memory reads, even with the optimization flag turned on. This is probably because of poor optimization in C#.

void

The void keyword modifies a method, indicating that it does not have a return value. Unlike C++, you cannot use the void keyword in the argument list, which was an allowable (if not acceptable) practice for indicating that a method did not have any parameters.

Conclusion

By thoroughly reviewing the keywords in C#, especially the modifier keywords, the transition from C++ to C# can be much less confusing and much simpler for the programmer.