C++ for Java Programmers

Barbara Staudt Lerner
September 1998

C++ was developed in the early 1980's. Its goal was to introduce object-orientation to C while maintaining backwards compatibility so that existing C programs would continue to work without change. Java, developed in the early 1990's inherited none of that baggage and instead was intent on developing a pure object-oriented language with syntactic similarities to C.

Besides the desire for backwards compatibility, C++ also maintained the philosophy that performance was critical. Many design decisions of C++ were done to allow programmers to get maximum performance out of their programs at the cost of making the code more difficult to understand and more error-prone.

Java was originally developed as a programming language for programs embedded in electronic devices, such as microwave ovens, CD players, telephones, etc. Since this was a new marketplace, there was not a lot of legacy code that the Java designers needed to maintain compatibility with. Also, since the performance of architectures has improved substantially and the computing demands of these devices were not great, performance of the language was much less of a concern, but productivity of the programmers was considered quite important.

The end result is that Java is much more programmer-friendly than C++. James Gosling, the creator of Java, quips that Java is C++ without the guns, knives, and clubs. Fortunately for us, Nachos uses a restricted subset of C++ that excludes many (but not all) of the hazardous features. This document should give you a basic understanding of the subset of C++ that Nachos uses. This should help you read the existing Nachos code. You will certainly need to supplement this with a C++ text to get all the details. This document assumes that you already know Java.

Differences among Features Common to C++ and Java

Low-level Syntax

At the level of statements and declarations, C++ and Java are quite similar. The syntax that you have learned in Java will carry over (mostly, anyway) to C++. C++ includes more constructs than Java, for which you will need to learn both the syntax and semantics.

The boolean data type consisting of the values true and false is called bool in C++.

In C++, the data type char is an 8 bit value capable of representing an ASCII character. An instance of the char type can also be treated as an 8 bit integer. Therefore, you can do arithmetic on variables declared as char.

In C++, you can declare any of the numeric data types to be unsigned. This results in the lower bound for the type being 0. For example, int normally ranges from -231 to 231. An unsigned int ranges from 0 to 232.

In C++, = is the assignment operator just as in Java. In C++, = can also be used as an expression that returns the value being assigned. This allows the following convenient way of initializing two variables to the same value:

    a = b = 1;
1 is assigned to b. The assignment expression returns the value assigned and assigns this to a.

C++ allows the condition controlling an if-statement or while-statement to be an integer expression. If the integer expression evaluates to 0, this is treated the same as the boolean value false. A non-zero value is treated the same as true. This is bad programming style and should be avoided, but you might run into this in the existing Nachos code. It would be better to say:

    if (someInt != 0)
then to just say
    if (someInt)
although they are semantically equivalent.

Combining these last two paragraphs shows one of the most common syntactic errors in C++ programs:

    int a = 0, b = 1;
    if (a = b) {
      ...
    }
Assume that the programmer intended to say a == b as the condition, which is almost certainly the case. The programmer therefore intended that the body of the if-statement would be executed only if a and b had the same value. In Java, the code above would give a compilation error, but it does not in C++ since = is an expression. Instead the value of b is assigned to a, so a now becomes 1. This value (1) is returned as the result of the assignment expression. Since integer expressions are allowed for conditions, this is ok. 1 is treated as true and the body of the if-statement is executed, which is not what the programmer intended. Be on the lookout for this simple error in your code!

In C++, you must declare the size of an array when you declare the array. Memory is allocated for the array when the array is declared:

    int intArray[10];
Beware! C++ does not check array bounds like Java does. If you pass in a negative number for an array bound or an array bound that is greater than the size of the array, C++ will happily access some (seemingly random) piece of memory. If this appears on the left side of an assignment statement, it will happily change some (seemingly random) piece of memory. Always check array bounds yourself if you are not absolutely certain that the value is in the correct range!!!

In C++, it is possible to declare that a variable should be kept in a register. This is typically done to improve performance but is really unnecessary with modern compilers. Compilers are smart about recognizing which variables are used most often and keeping those values in registers. If not used extremely carefully, register declarations actually degrade performance. You should avoid using them, but the existing Nachos code does use them so you need to recognize what they are.

    register int i;

Classes

Both Java and C++ use classes as an abstraction mechanism. Classes encapsulate data and methods that operate on that data. In C++, a class definition is broken into a declaration and a separate definition of each member. It is not possible to attach the keywords public, private, or protected to the class definition. If the class is public (in the Java sense), you put its declaration in a separate file whose name ends in .h, while you put the definitions in a file with the same name but ending in .cc. A class declaration can be broken into three sections: a public section, a protected section, and a private section. Instead of attaching these keywords to each class member as in Java, you put the member in the appropriate section of the declaration as follows:
    class Pair {
      public:
        Pair (int x, int y);  // The constructor
        int getX();
        int getY();
        void setX (int newValue);
        void setY (int newValue);
    
      private:
        int x;
        int y;
    };

Also, note that C++ requires a semicolon at the end of a class declaration.

To make this class visible to another file, it must be included in that file:

    #include <pair.h>
There is no notion of a package in C++.

The member definitions appear in the .cc file as mentioned earlier. Since they appear outside the class declaration, each definition needs to declare which class it is in using the class_name:: syntax:

    int Pair::getX() {
      return x;
    }
Any members that are fully defined in the class declaration (usually just the variables) should not be defined in the .cc file.

If you want to specify a particular member from a specific class when doing a function call, for example, you use the :: operator, as in

    SomeClass::someMethod()

The syntax for declaring a subclass is different in C++:

    class SubClass : public SuperClass {
      ...
    };
This is a declaration of SubClass as a subclass of SuperClass. Note that you must include the keyword public before the superclass name. If you want to allow a method to be overridden in a subclass, the superclass must include the keyword virtual in its declaration of the method. Any class that contains a virtual function and no definition of that function is implicitly abstract. It is not possible to declare a class or function to be abstract. There is no equivalent to Java's super keyword. If you want to refer to a superclass function that is overridden in a subclass, you must explicitly qualify the function name with the class from which it is inherited using the :: syntax. Nachos does not use subclasses, so I will not go further into their details.

Input and Output

There are several ways to do input and output in C++. The first way is a hold-over from C. Here, you use the function printf to produce output. printf takes a variable number of arguments. The first argument is a string. The string may have embedded within it zero or more control sequences. For each control sequence there must be an additional parameter to printf defining the value to use for that control sequence. The control sequences used in Nachos are the following:
%c A character.
%s String
%d An integer to be represented as a decimal string.
%x An integer to be represented as a hexadecimal string.
%f A floating point number.
An important variant of printf is fprintf. fprintf is like printf, except that it takes an additional argument before the string which is a file pointer (for an already-open file). The output is written to the file instead of to standard output. Here is an example use of printf:
    char month [4];
    int year;
    strcpy (month, "Sep");  // Assigns a value to a string variable.
    year = 1998;
    printf ("The month is %s.  The year is %d.\n", month, year);

scanf is the input function from C. Its format is similar to printf. The first argument is a string often exclusively consisting of control sequences and whitespace. For each control sequence, there must be an argument that is the address of a variable of the appropriate type. The memory must be allocated already. sscanf is a variant of scanf with an additional first argument representing a string to parse instead of reading from standard input. sscanf is typically used to convert a string to an integer. For example, suppose s contains "1998",

    sscanf(s, "%d", &year);
will set year equal to the integer 1998.

The second (and preferred) way of doing input and output in C++ is using streams. Nachos uses printf rather than streams so I will not discuss them here.

(There is also an sprintf function that places the formatted output in a string and a fscanf function that reads input from a file, but these are not used in Nachos.)

Features Unique to C++ (and used in Nachos)

Constants and Macros

To declare a constant in C++, you use #define (not final) as in:
    #define MAX_SIZE 10
#define is actually a much more powerful macro mechanism. Everywhere that the defined name appears within the scope, it is replaced by the definition, which could be a complex expression. This is often done to make a statement look like a function call but without having the runtime overhead of making a procedure call. For example, here is a macro defining minimum:
    #define min(a,b)  (((a) < (b)) ? (a) : (b))
Later in the code, the programmer can say:
    min (i, 4)
and it is expanded by the preprocessor to the compiler to:
    (((i) < (4)) ? (i) : (4))
Since this is done by the preprocessor, the expression is inlined rather than being executed as a function call.

Compiler Directives

#define is one example of a compiler preprocessor command. This is a command that is executed by a preprocessor that scans the code prior to compilation. The preprocessor is run automatically when you run the compiler. Two other common directives in C++ are #ifdef and #ifndef. #ifdef takes a variable name for its condition. If that variable name is defined, it evaluates to true and its body is included in the source code that is compiled. #ifndef is similar but includes its body if the variable is not defined. Both may have #else clauses. They both end with the delimiter #endif.
    #ifdef HOST_SPARC
    #include 
    #endif
This is how C++ programmers typically port programs between architectures. Architecture-dependent code is placed inside #ifdef statements. When the code is compiled, the appropriate variable is set for the architecture allowing the correct code to be compiled in. Unlike normal if-statements, these if-statements are evaluated at compilation time. The branch that is true at compilation time is compiled into the program. Branches that are false are not compiled in. The condition is not tested at runtime.

Life Outside of a Class

In Java, everything is declared inside of a class. Since C++ needed to maintain backwards compatibility with C, this is not true for C++. Data types, variables, and functions can all be declared outside of classes. These are referred to by simply using their names. There is no . syntax required to dereference them.

If a data declaration is to be global and shared between multiple files, it is declared in one file and declared to be an extern in the other files. We all know that global variables are bad, so we shouldn't do this, however, there are some instances of this in Nachos. A function can also be declared this way, but it is better style to put the function declaration in a .h file and #include the .h file rather than extern the function declaration. Occasionally you will see something like:


    extern "C" {
      <a list of declarations>
    }
This indicates that the list of declarations has been declared as pure C code, not within classes.

The main program for a C++ program is called main, but it is declared externally to any class. Its signature is:

void main (int argc, char **argv);

The first parameter is the number of arguments. The second parameter is an array of strings, each string containing one command-line argument.

Struct Types

Data type declarations outside of classes are encapsulated inside a struct:
struct Date {
  char *month;
  int date;
  int year;
};

struct Date someDate;

Typically, when declaring a type one gives the type a name. Oddly enough, creates a type named struct Date. To give it a simple type name, a slightly different syntax is required:

typedef struct {
  char *month;
  int date;
  int year;
} Date;

Date someDate;

You will find quite a few struct declarations in existing Nachos code. You should not create any new struct declarations. Instead you should create a class whose data members are like the struct fields.

Enumerated Types

Enumerated types is one feature of C++ that I really wish Java had. With an enumerated type, you can define your own type with its own set of discrete values. For example,
enum PrimaryColor {
  red,
  yellow,
  blue
};

PrimaryColor c;
c = red;

Enumerated types are simulated in Java in the following way. The programmer declares a class (equivalent to the enumerated type) that provides a number of public constants (representing the enumerated values).

Union Types

A union type is a type that allows a particular piece of memory to store a value of different types at different types (a primitive precursor to subtyping). A union declaration looks a lot like a struct declaration:
union String_or_int {
  char *someString;
  int someInt;
};
The union itself does not keep track of which type is in it, so a union is typically used inside a struct where a second field of the struct remembers the type currently in the union field:
struct S_or_i {
  bool containsInt;
  union String_or_int x;
};

Pointers

In Java, all references to objects are pointers to objects. All references to primitive types, like int are values. In C++, using a type name always means that the variable will have a value of that type. It is possible to introduce pointers to values explicitly and also to create types whose values are pointers to other values. Suppose we have a Date type, here is how we would declare a variable that is to contain a pointer to a date and also a type to represent a pointer to a date:
    Date *someDate;         // Variable containing a pointer to a Date

    typedef Date *DatePtr;  // Type defining a pointer to a Date
    DatePtr date2;	    // Variable containing a pointer to a Date

    date2 = someDate;
someDate and date2 both contain pointers to dates. The assignment statement results in both variables pointing to the same memory location and therefore sharing the same value as happens in Java.

Contrast the above with the following similar code that does not use pointers:

    Date someDate;
    Date date2;

    date2 = someDate;
Assuming that Date is simply a struct type, not a pointer type, the assignment statement above copies the value from someDate to date2. If the value referenced in either variable is changed, it has no effect on the other value. In Java, you would need to explicitly clone the value to have this effect. Unless you know the definition of the type involved in an assignment, you cannot tell whether the assignment results in value-sharing or value-copying.

A pointer is dereferenced using the -> syntax:

    some_pointer->some_field
In C++, this is a pointer, not an object. To dereference it, you must say
    this->member

To get a pointer to an object, you use the & operator:

    int *IntPtr;
    int anInt;

    anInt = 1;
    intPtr = &anInt;

To get the value pointed to by a pointer, you use the * operator:

    int *intPtr;
    int anInt, int2;

    anInt = 1;
    intPtr = &anInt;
    int2 = *anInt;

In C++, all parameters are passed by value. If you want to be able to change the value of a parameter as a side effect, you must declare the parameter type to be a pointer and you must pass in the address of the variable that you want to change:

    void increment (int * anInt) {
        (*anInt)++;
    }

    int i = 0;
    increment (&i);

If you want to pass a pointer or an array to a function, but you do not want the object to be changed, you can say that the parameter type is const:

    void doSomething (const int * anInt) {...}

    int *intArray;
    // Assume the array has been allocated and given memory.
    doSomething (intArray);

Memory Management

Java is a garbage-collected language. C++ is not. In Java, memory is allocated for an object when that object is constructed. The memory is deallocated when there are no more references to that object.

In C++, objects (and structs) can either be automatic or manually allocated. Variables whose types are not pointers (such as classes or structs) are automatically allocated and deallocated. They are allocated when they are declared and deallocated at the end of the block in which they are declared. You do not use new to allocate a variable whose type is a class.

With pointer types, the programmer must explicitly allocate and deallocate memory. You must allocate memory before assigning a value to the object. Allocation for classes is done with new as in Java. You should deallocate an object when you believe there are no more references to that object. The syntax for deallocating an object is:

    delete list1;
where list1 is the name of an object. If you have an array of objects, and you want to delete all the objects in the array, say
    delete [] objArray;
For every object allocated with new there should be a deallocation with delete.

If you want to do anything special when deleting an object, you must define a deconstructor for the object's class. A typical thing to do is to delete the objects referenced by the object being deleted (if you are sure they are the last reference to that object!). The syntax for declaring a deconstructor is:

    ~MyClass();
where MyClass is the name of the class containing the deconstructor.

You must also allocate memory for pointers to other types (typically structs), as in C. To do this you need to know how big the structure is. You can find this out using the sizeof function:

    sizeof (some_type)
The syntax for allocating memory is:
    some_struct = (some_struct_type *) malloc (sizeof (some_struct_type));
To free such memory, you use the free function:
free (some_struct);
For every variable allocated with malloc there should be a deallocation with free.

Similarity between Arrays and Pointers

Suppose you want to have an array variable, but you do not know how big the array should be. Since C++ requires you to declare the size of the array when you declare the array variable, you cannot declare it to be an array. Instead you must declare a pointer to the desired element type and later allocate the appropriate amount of memory yourself:
    int *intArray;
    intArray = (int *) calloc (10, sizeof (int));
Even though you declared the variable to be a pointer, you can still dereference it as an array!

There is no string data type in C++. Strings are simply arrays of characters. Since we typically want to allow variable length strings, string variables are typically declared to be pointers to characters:

    char *someString;
All strings must have the special null character '\0' as their last character to identify the end of the string since there is no length recorded with the string. String constants are enclosed in "" and implicitly end in the null character. Since strings are pointers, assigning one string variable to another results in the two variables pointing to the same piece of memory. Changing one string changes the other. If you do not want this effect, you must use the strcpy function:
    char *month;	    // Declare the string
    month = malloc (4);     // Allocate memory for string ending in null.
    strcpy (month, "Sep");  // Assigns a value to a string variable.
Remember to free the string when it is no longer being used.

In a few places in Nachos, you will syntax like the following:

    char **stringArray;
This is a pointer to a pointer of a character. Keeping in mind that pointer declarations often mean variably-sized arrays, this syntax represents a variably-sized array of strings (since char * means string).

Features Unique to Java

There is no equivalent to the synchronized keyword. Threads are also not built into the language. When we discuss threads and synchronization in class, we will discuss how this is done in C++.

There is no equivalent of JavaDoc for C++.

C++ does not have an instanceof operator.

C++ does not have interfaces.

C++ does not come with a large standardized class library as Java does. If you look at sections 2 and 3 of the Unix manual (using xman), however, you will see a large collection of C functions that can be called from C++. Some of the common functions, like strcpy, are the same across all operating systems, but, in general, the routines in these libraries are not standardized (particularly the system calls in section 2). You need to explore the man pages to see what is on the particular operating system you are using. This lack of standardization is one reason that C++ programs are not as portable as Java ones.


Last modified by Barbara Lerner on September 11, 1998.