JAL Computing

Chapter 5 "C++ and Java Gumption Traps" 

It's time for another break from theory and on to some real world programming challenges. In this chapter, you take a detour into the world of gumption traps. A gumption trap is a situation that is so frustrating that it sucks the "gumption" or energy out of you. Surprisingly, a lot of gumption traps come from preconceived notions of what "should" be correct. C++ and Java programmers are prone to gumption traps due to their assumptions about the C# language. In other words, C++ and Java programmers have a bit of "language baggage" that needs to be left behind. So, if you have a language neurosis, this chapter is for you.

Top Ten C++ Gumption Traps

In no particular order.

1) Arrays in C# are Objects

In C#, arrays are objects derived from System.Array. The following code snippet creates an array object that can contain ten elements of MyClass:

MyClass[] arrayMC= new MyClass[10];

[STAThread]
static void Main(string[] args)
{
    //
    // TODO: Add code to start application here
    //
    Class1 c1= new Class1();
    for (int i=0; i<10; i++)
    {
        c1.arrayMC[i]= new MyClass();
    }
    Console.ReadLine();
}

2) C# Classes Use Reference Semantics, C++ Classes Use Value Semantics.

In C++, when you declare an object, an object is created (value semantics). Not so in C#. In C#, declaring an object in _local scope_ simply creates an un-initialized object variable. The following code simply declares an un-initialized reference variable of Type MyClass. The variable occupies memory on the stack to hold an address. No object is created on the heap. If you try to use c, the compiler will complain that the variable has not been initialized.

MyClass c;

Note: The compiler will initialize an instance variable to null under the following conditions:

The following code snippet, creates a variable and initializes the variable to null, so that the reference does not refer to any object. If you try to use the reference variable to touch a field or method of the class, the code will throw a null reference exception.

MyClass c= null;

So, if you want to declare and initialize a reference variable and create an object in C# you must use the new keyword like this:

MyClass c= new MyClass();

"c" is now a reference variable of Type MyClass. If this is a local variable, then "c" uses memory on the stack which contains the address of an object of Class MyClass on the heap.

As discussed earlier, the string class (strings are immutable and shared) is an exception to the rule. You can create a string like this:

string s= "Hello";

3) C# Uses Automatic Garbage Collection, Destructors are not Deterministic

The call to new may bother a C++ programmer who may now feel obligated to call delete appropriately. All I can say is lighten up! Objects are (mostly) reclaimed automatically in C#. In a nutshell, when available memory falls, the garbage collector is called, freeing up any objects that are unreachable. If you are done with an object, you can assist the garbage collector by releasing all references to the object. Even if two objects contain references to each other (circular references), the objects can still be collected if they become "unreachable."

 Alternatively, you can use C#'s try, catch, finally construct to insure that any allocated resources are reclaimed in the finally clause. Here is a twisted example of try, catch, finally in C#:

Note: Visual Studio C++ 2005 supports the concept of class instances on the stack. When a managed C++ object on the stack goes out of scope, Dispose is called.

4) The Assignment Operator Does Not Call the Copy Constructor

This one really confuses a lot of coders. In C# the assignment operator simply assigns a reference variable to an existing object, to a new object, or to null. After the assignment, the reference variable contains a reference to an object or to no object (is null). The assignment operator does not call the copy constructor to create a new object. It is quite legal in C# to have two reference variables that contain references to the same object. The two variables occupy different memory locations on the stack, but contain values that point to the same memory address on the heap. Both references would have to be released (e.g. the reference variables go out of scope, be reassigned, set to null) before the object can be reclaimed. As a result, the "destructor" will only be called once. The following code simply creates two references to the same object:

MyClass c1= new MyClass();
MyClass c2;
c2= c1; 
bool isSameReference= (c1 == c2);  // c1 and c2 contain references to the same object on the heap
Console.WriteLine(isSameReference.ToString()); // output--> true

5) Values and References to Objects are Passed By Value

Note: This topic has been a source of great confusion to C++ coders, so I will elaborate. If you want to write a swap routine then you need to add another degree of indirection to the method call. You can do this by use the keyword ref. Here is some sample code that demonstrates the concept. The swap routine only works if you pass a reference to a concrete Drawable object by reference.

You can mimic pass by reference in Java by "packing" the concrete Drawable object into a wrapper class. You can then pass a reference to the packing object to a swap routine by value. This added level of indirection allows you to implement a swap routine, so that you can swap the concrete Drawable object in each packing object.

6) Const is Hard Coded Into the Caller and not Version-able

I find this one a bit weird. A const field value is hard coded into the caller for optimization so that the value is not updated until you recompile the caller. If you want to version-able read only field declare it readonly. Finally, you can provide a get only Property like this:

public string ModelName
{
    get
    {
        return modelName; // danger, return ModelName --> stack overflow!
    }
}

7) Supports Single Inheritance of Implementation and Multiple Inheritance of Interfaces

C# does not support multiple inheritance of implementation. It does support multiple inheritance of interfaces (pure virtual classes) and single inheritance of implementation.

8) Program to Return Single Values

Although C# does support an out parameter, in general, C# programs are designed to return single values. Consider redesigning your application or returning immutable objects.

Note: In .NET 2.0 the TryParse methods return false on error and return the single result using an out parameter. Resist the temptation to return multiple parameters using out.

9) Strings are Immutable

The string class is a very special class. First strings are immutable. Every time you concatenate a string, you may be creating a new string object. If you want a mutable class, use StringBuilder. Second, you can create a string object using the assignment operator. Third, strings are shared so that two strings that you create may occupy the same space in memory. Fourth, the equality operator is overloaded for string and checks for content equivalence, not "reference based equality."

Note: The overloaded string equality operator only works on reference variables of Type string. Be careful! Thanks to Jon S. for this insight.

object a= "Hello";
object b= "Hello";
bool isSameReference= (a == b); // test if a and b contain references to the same object or to no object (are both null)
bool isSameContent= ((string)a == (string)b); // test if string referenced by a and string referenced b have the same content/value or a and b are both null

Session["KEY"]="VALUE"; // Error! The left hand operand is of type object! This is a reference based comparison. Do this:
(String)Session["KEY"]="VALUE"; // content equivalence

10) bool is a Value Type

C# has a built in type for Boolean which can have a value true or false. The default value of bool is false. Enough said.

0) The Destructor is Called Even on Exceptional Construction

Be warned, the destructor will be called even if an exception is thrown in the constructor of an object. The runtime will allocate and initialize an object with default values and then call the user defined constructor. If an exception is thrown in the user defined constructor, the user defined destructor will still be called! Aaach.

Top Ten Java Gumption Traps

In no particular order.

1) No Checked Exceptions

The compiler in C# will not enforce catching of any checked exceptions. So there is no support for declaring a checked exception (no throws keyword).

2) All Methods are Not Virtual By Default. All Members are Private by Default.

In Java, all methods are virtual by default and can be overridden. Not so in C#. You must explicitly declare a method virtual if you want it to allow it to be overridden. You also cannot implicitly override or hide a virtual method, but must explicitly use the override or new keyword. In Java, members have package level access by default. In C# all members are private by default.

3) Const, ReadOnly and Get Only

I find this one a bit weird. A const field value is hard coded into the caller for optimization so that the value is not updated until you recompile the caller. If you want to version-able read only field declare it readonly. Finally, you can provide a get only Property like this:

public string ModelName
{
    get
    {
        return modelName; // danger, return ModelName --> stack overflow!
    }
}

4) All Types in the NET Framework are Polymorphic for Object

C# has a unified type system so that all reference types and boxed value types derive from Object. Since the CLR provides automatic boxing of value types when necessary, value types "indirectly appear" to derive from Object. More importantly, both value types and reference types exhibit polymorphic behavior for the interface of Object. 

For instance, int is just an alias for System.Int32 which is a value type that "indirectly" derives from object. As a result, you can call ToString() on any value or reference type. For reference types, the proper method is called using "subclass" polymorphism. For value types, the proper method will be called using "ad-hoc" polymorphism if necessary. For instance, if the value type does not provide an overridden Object method, the CLR may need to box the value type to access the associated reference type. This simplifies debugging in the Console mode. System.Int32 in turn provides useful properties such as MaxValue and MinValue. If you want to use an int with reference semantics, you can "box" the value (wrap a copy of the value in an object) like this:

Here is the graphical view of System.Int32.

What's this?

5) The Visual Studio IDE Startup Object is Empty By Default

If the startup object string is empty and you declare more than one "Main" method in a Visual Studio project, the project will not compile. This is very frustrating. Trying to set the startup object property in the Visual Studio IDE is non-trivial. Let me save you some pain. To set the startup object, select View --> Solution Explorer. In the Solution Explorer window select the name of the project. Now select View --> Property Pages. (Do not select the Property Window!) Alternatively, right click on the project name and select Properties. Now select the Common Properties folder in the left window. You can now select the appropriate startup object from the drop down list. Don't forget to click on "Apply."

6) Method Names Start with Upper Case

This one is pretty self explanatory.

7) Getters and Setters are Implemented Using Properties, Which are Upper Case

C# formally supports the concept of getters and setters with Properties. Note that properties are upper case by convention. Here are the setter and getter methods as properties:

          // Properties
         public int SecondsToToast 
        {
            get {return secondsToToast;}  // danger, return SecondsToToast --> stack overflow!
            set  
            {
                 if (value > 0) 
                {
                    secondsToToast= value;
                }
                else  
                {
                    secondsToToast= 0;
                }
            }
        }

The reserved word value is used to represent the caller’s input. Note the syntax used to call a property. To set a property use:

t.SecondsToToast= 12;

8) Use "is" instead of "instanceof", ArrayList instead of ArrayList/Vector, StringBuilder instead of StringBuffer, Hashtable instead of HashMap/Hashtable, System.Environment.NewLine instead of line.separator. There are no wrapper classes for value types, but value types can be boxed (see #4 above).

Just trying to save you some headaches.

9) There is an "enum"

Java evolved before the concept of type safe enum but support enums now. C# has built in support for type safe enums like this:

public enum ColorType {Black, Red, Yellow, White};
private static ColorType DEFAULT_COLOR= ColorType.Black;

Here is a graphical representation of a enum.

You can still create a custom enum in C# like this:

sealed class MyEnum
{
    private String name;
    private static int nextOrdinal= 1;
    private int ordinal= nextOrdinal++;
    private MyEnum(String name)
    {
        this.name= name;
    }
    public override String ToString()
    {
        return name;
    }
    public int ToOrdinal()
    {
        return ordinal;
    }
    public static MyEnum INVALID= new MyEnum("Invalid"); // ordinal 1
    public static MyEnum OPENED= new MyEnum("Opened"); // ordinal 2
    public static MyEnum CLOSED=new MyEnum("Closed"); // ordinal 3
    /// <summary>
    /// The main entry point for the application.
    /// </summary>
    [STAThread]
    static void Main(string[] args)
    {
        //
        // TODO: Add code to start application here
        //
        Console.WriteLine(MyEnum.OPENED.ToString());
        Console.WriteLine(MyEnum.OPENED.ToOrdinal().ToString());
        Console.WriteLine(MyEnum.INVALID.ToString());
        Console.WriteLine(MyEnum.INVALID.ToOrdinal().ToString());
        Console.WriteLine(MyEnum.CLOSED.ToString());
        Console.WriteLine(MyEnum.CLOSED.ToOrdinal().ToString());
        Console.ReadLine();
    }
}

10) String Equality Operator is Overridden for Content Comparison

Thankfully, the string equality operator in C# has been overloaded and checks for content/value equivalence. One less gotcha for the C# novice.

Note: The overloaded string equality operator only works on reference variables of Type string. Be careful! Thanks to Jon S. for this insight.

object a= "Hello";
object b= "Hello";
bool isSameReference= (a == b); // test if a and b contain references to the same object or to no object (are both null)
bool isSameContent= ((string)a == (string)b); // test if string referenced by a and string referenced b have the same content/value or a and b are both null

Session["KEY"]="VALUE"; // Error! The left hand operand is of type object! This is a reference based comparison. Do this:
(String)Session["KEY"]="VALUE"; // content equivalence

OK, boys and girls. It's not your father's language. Get over it<g>.

The Equality Operator Revisited

Thanks to Jon S. I have revisited the concept of equality on operands of the reference type. The equality operator, in general, compares operands by reference. The string equality operator is overloaded to compare operands by value. In plain English, the equality operator normally checks to see if two object variables refer to the same object on the heap. The string equality operator is overloaded to check for content or value equivalence.

In not so plain English, the equality operator normally checks to see if two reference variables contain references to the same object on the heap. In C# local reference variables go on the stack and contain the address of the object on the heap. So local reference based equality in C# checks to see if two reference variables, stored on the stack, contain addresses which are equal. In other words, it checks for address equality. This is referred to as "reference based" equality or "compare by reference" or "referential equality". Whew!

As previously stated, the string equality operator is overloaded to check for value or content equivalence. In not so plain English, the string equality operator compares the content or value of string objects on the heap. It does this by obtaining the address of the objects from the values of the reference variables stored on the stack.

Now I would like to simplify this discussion to that of "object" equality vs. "content" equality, but I cannot! By convention, object equality refers to content equality! I would also like to reduce this discussion to referential "equality" vs. content "equivalence", but I cannot. Unfortunately, by convention, the Equals() method is used to determine content equivalence!

All Rights Reserved Jeff Louie 2003

Send mail to [email protected] with questions or comments about this web site. Copyright © 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 © 
Last modified: 08/04/09