Home Previous Next

CSC240 :: Lecture Note :: Week 04
Assignments | Code | Handouts | Resources | Email Thurman {Twitter::@compufoo Facebook::CSzero}
{GDT::Bits:: Time  |  Weather  |  Populations  |  Special Dates}

Overview

Assignment(s): #stats0 (due 10/13/2017)

Code: itofi.cpp (continued) | funcs0.cpp | funcs1.cpp | funcs2.cpp | funcs3.cpp | refparams.cpp | funcs4.cpp
RandomInts.cpp | ArrayEG.cpp | Fibonacci.cpp | BingoCard.cpp | bingo.cpp | LotteryTickets.cpp
pointer0.cpp | pointer1.cpp | cmdline.cpp | temps.cpp


Reference Parameters

C++ passes arguments to functions "by value." This calling mechanism prohibits called functions from modifying the variables defined in the caller functions.

Reference parameters  allows arguments to be passed "by reference" instead of by value. If you pass an argument to a function by reference, then the called function can access that variable.

Passing an argument by reference causes the address of the variable argument to be passed rather than its value; consequently, the called function receives a "pointer" to the variable. Although the called function is dealing with a pointer, pointer notation is not required.

A parameter that is received as a reference becomes an alias for the variable that was passed.

Reference parameters are often used when passing large arguments to a function (e.g. a structure). It is more efficient to by reference than it is to pass by value (less data must be copied).

Suppose we have a function that converts inches to feet and inches. We have a problem: our function needs to return two pieces of information, but is allowed to return only one value. One solution to the problem is to use reference parameters:

   int convertInches(int totalInches, int& feet);
      // the & after the type  int  indicates that the function
      // receives a reference to an  int
   
   ...

   int feet;
   int inches = convertInches(80, feet);  
      // feet is passed by reference; no special notation is needed

   ...

   /**
      The function receives the total inches.  The return value
      of the function is the inches, and the feet are copied
      into the reference parameter supplied as the 2nd argument.
   **/

   int convertInches(int totalInches, int& feet) {
      const int INCHES_PER_FOOT = 12;
      feet = totalInches / INCHES_PER_FOOT;
      return totalInches % INCHES_PER_FOOT;
   }

{TopOfPage} {Resources}


Default Function Arguments

Default function arguments can be specified in either the function definition or prototype (but not both). Convention is to specify them in the function prototype.

Only the rightmost arguments can be defaulted. Once a default function argument is used, all remaining arguments must be default arguments.

Default arguments are useful when a function needs more arguments than are necessary to handle simple cases; in particular, functions that construct objects often provide several options for flexibility.

   Syntax:  data-type function_name(Type param = value);

   If  param  is not passed by the caller, then it will be set
       to  value  .

   Example:

      void printReport(char filename[], bool condensed = false);
      // one argument is required:  a filename, but the second is a
      //     default argument and it defaults to  false  if not passed

      printReport("report.out");           //1
      printReport("report.out", true);     //2
      printReport("report.out", false);    //3
          //statements 1 and 3 are equivalent

{TopOfPage} {Resources}


Function Overloading

Function overloading is the ability to give different functions the same name.

Stroustrup says:

Most often, it is a good idea to give different functions different names, but when some functions conceptually perform the same task on objects of different types, it can be more convenient to give them the same name.
Use descriptive overloaded function names to describe similar operations, not different behaviors.
   Good:
      int max(int, int);        //find max of two ints
      float max(float, float);  //find max of two floats

   Not so Good:
      void draw(Image*);        //draw an image
      void draw(Card*);         //draw a card 
The following is an example of an overloaded function:
   void print(int), print(double), print(long), print(char), 
        print(int, int), print(double, double);

   char c;
   short s;
   int i;
   float f;

   print(c);            // invoke print(char)
   print(i);            // invoke print(int)
   print(s);            // invoke print(int); s promoted to int
   print(f);            // invoke print(double); f promoted to double
   print('A');          // invoke print(char); 'A' is a char
   print(200);          // invoke print(int); 200 is an int
   print(200L);         // invoke print(long); 200L is a long
   print(99.9);         // invoke print(double); 99.9 is a double
   print(i, i);         // invoke print(int,int)
   print(i, 'a');       // invoke print(int,int); 'a' promoted to int
   print(s, 'A');       // invoke print(int,int); s - 'A' promoted to int
   //! print(1.1L);     // ambiguous -- compiler can't decide
   print(200L, i);      // invoke print(int,int);
   //! print(i, 3.14);  // ambiguous -- compiler can't decide
When  print()  is called, the compiler must figure out which of the functions with the name "print" is to be invoked. This is done by comparing the types of the actual arguments with the types of the parameters of all functions called "print". The function with the best match is called; if none exist, then a compile-time error.

Definition:

The  signature  of a function is defined to consist of the name of a function and its ordered set of parameter data types.

Criteria used to determine a match (partial):
  1. exact match of function call arguments with an overloaded function signature
  2. trivial conversions (exact match after applying promotions to argument data types -- char ==> int, short ==> int, float ==> double)
  3. exact match after applying promotions and standard conversions to argument data types (float ==> int, double ==> float)
  4. exact match after programmer defined conversions (e.g. typecast)
Note:  const  arguments can be distinguished from non-const arguments.
	void foo(const char*);
	void foo(char*);

	These two functions have different signatures.
Return types are not considered in overload resolution.

Functions declared in different scopes do not overload.

If not careful, you can be surprised as to which function is called. It is poor programming practice to redefine library functions (your code may end up calling the correct function, but existing code in the library may end up invoking the wrong function).

Internally, the compiler accomplishes function overloading by  mangling  function names. Function name mangling makes each translated function name unique. A function name is mangled by adding prefixes and suffixes to the the name. The prefixes and suffixes are determined in part by the ordered lists of the function's parameter data types.

	print(200);             // mangled name: print_Fi
	print(float(3.14));     // mangled name: print_Ff

Function Overloading
C: no   C++: yes   Java: yes

{TopOfPage} {Resources}


Lifetime and Visibility (scope)

Lifetime is the period, during execution of a program, in which a variable or function exists (all functions in a program exist at all times during its execution).

Visibility is the portions of the program in which a variable or function can be referenced by name (also referred to as scope (scope units: file, function, block or function prototype).

Local and Global Variables

Local variables are declared and/or defined within a block (either a function or a compound statement).

Global variables are declared and/or defined outside of any function.

Storage Classes
auto

The auto storage class indicates that a variable is a local variable and that memory is automatically allocated/de-allocated. By default, local variables are defined to be auto; therefore, this storage class is rarely specified.

Using auto on a global variable is illegal.

Automatic variables are not initialized to any known value.

extern

The extern storage class allows a global variable to be visible across multiple source files.

A global variable defined in file A can be accessed in file B if and only if file B contains the following declaration.

	extern type variable_name;

	extern int errno;  /* errno is defined in some other file */

Use of extern results in a declaration -- not a definition. In other words, an extern declaration does not result in the allocation of memory.

It is not legal to initialize an external variable at the time it is declared (i.e. extern float rate = 5.3; is illegal).

static

When static is applied to a global variable, then it limits the scope of an object to the rest of the file (i.e. it cannot be externally declared in other source files).

When static is applied to a local variable, then it causes the local variable to remain in existence across function calls providing private, permanent storage within a single function.

When static is applied to a function, then the function is visible only to the source file in which is defined (in a way, the function is made "private").

register

A register is a high speed memory location located on the CPU.

The register storage class is used on local variables that are going to be accessed many times (for example, loop control variables).

{TopOfPage} {Resources}


Arrays

An array is a collection of individual data values with two characteristics: it is ordered and homogeneous.

The following are some terms commonly used when talking about arrays:

An array is homogeneous because each element must be of the same type. For example, an array of int's, an array of float's, an array of char's.

An array is ordered -- it has a 1st element, a 2nd element, a 3rd element, and so on. Array elements are stored in contiguous memory locations.

Just like any other variable, an array must be defined before it is used.

   Syntax:  element-data-type array-name [ length ];

      int intArray[10]; 
         name:  intArray
         type:  int
         length:  10
         size:  10 * sizeof(int)

      char someString[32];
         name:  someString
         type: char
         length: 32
         size:  32 * sizeof(char)

Each element of the array is identified by a numeric value called its index (index numbers always start at 0).

When defining an array, its length must be specified at compile-time. In addition, the specification of the length must be a constant integral EXPR.

   
   int i = 10;
   /* double salaries[i];   // illegal (not a constant) */
   /* char name[15.5];      // illegal (not an integral type) */
   const int MAX_ELEMENTS = 10;
   short scores[MAX_ELEMENTS];  /* okay in C++, not C */
   #define LENGTH 5
   float someArrayOfFloats[LENGTH];

Typically, array lengths are defined to be manifest constants.

Array Initialization

Arrays can be initialized at the time they are defined.

   int evenNumbers[5] = { 2, 4, 6, 8, 10 };

   note:  it is a syntax error if the # of initializers is
          greater than the array length (i.e. # of elements)

   The elements of the initializer list must be constant
   EXPRs.  If the number of initializers is less than the
   array length, then remaining elements are set to zero.

If an array is initialized when defined, then the length is not needed -- the compiler will set the length depending on the number of initializers. If the length of the array needs to be greater than the number of initializers specified, then the length must be specified.

   float radioStations[] = { 103.1, 93.3, 100.7 };

   radioStations  has a length of 3; to figure out the length
   of the array using code:

      sizeof(radioStations) / sizeof(radioStations[0])

   sizeof(radiostations)  evaluates to the size of the array
   and  sizeof(radioStations[0])  evaluates to the size of a
   single element of the array (recall, an array size is equal 
   to the length of the array times the size of the array type)

   the following EXPR also works to determine the length of
   an array    sizeof(radioStations) / sizeof(float)    but could 
   result in the defect if the type of the array is changed and
   the EXPR is not

There is not a convenient mechanism for initializing all elements of an array to a single value (exception: it is easy to set all elements of an array to zero -- int a[10] = { 0 };).

For efficiency, locally declared arrays that are initialized at definition may be declared to be static.

   static short tvStations[99] = { 3, 10, 12, 61 };

   tvStations  is an array of length 99; elements 0, 1, 2, 3
   have non-zero values, whereas elements 4 through 98 equal 0;
   the array is initialized once -- at program load time

Once an array has been defined, its length (i.e. number of elements) cannot be altered.

Elements of the array are accessed using the unary array operator [].

   const int LEN = 10;
   int a[LEN];

   a[3] = 150;  /* set element #4 to 150 */
   a[1] = 200;  /* assign the value 200 to element #2 */

   if (a[1] == a[3]) /* compare the value of element #2 with element #4*/

Array indicies must be integral values, but they are not restricted to being constants.

   int i, j, k;

   a[0] = 100;
   a[i] = 150;
   a[i * j - k] = 210;
   a[a[a[i]]] = 250;
   a[rand() % LEN] = 178;

The language does not protect against indexing beyond the ends of an array. Typically, if this happens, then a run-time error is encountered.

   a[-1] = 100;
   a[sizeof(a)/sizeof(a[0]) + 2] = 200;

Array 'a' cannot be copied to array 'b' using simple assignment.

   #define LEN 5 
   int a[LEN] = { 100, 200, 210, 120, 240 }, b[LEN];

   /* b = a;   //syntax error -- array name not a lvalue */
   for (int i = 0; i < LEN; i++)
      b[i] = a[i];
Arrays and Functions

Arrays are passed to functions "by reference." (Think about the overhead if arrays were passed by value.)

A function that receives an array as a parameter, obtains a constant pointer to the first element of the array.

Unless specifically qualified to be const, the function can modify the content of the array.

When a function receives an array as a parameter, it does not know, nor can it figure out, the length or size of the array. In many cases, the caller passes the array length (size_t) to the function.

   void func1(int[]);         //do *not* specify the array length
   void func2(const int[]);   //func2() not allowed to modify the array
   void func3(int [], int);
   #define A_LEN 10
   int a[A_LEN];
   
   func1(a);  //using array name w/o [] operator "decays" into a
              //constant pointer to the 1st element of the array
   func2(a);
   func3(a, A_LEN);  //pass the array length to the function
   ...
   
   void func1(int i[]) {  //again, array length not specified
      int len = sizeof(i) / sizeof(i[0]);
         // does not work -- sizeof(i) evaluates to the size
         // allocated for pointer variables, which is typically
         // the sizeof(int)...
      ...
   }
   
   void func2(const int i[]) {
      //! i[0] = 100;    //use of  const  implies the content of the
                         //array cannot be changed...
   }
   
   void func3(int i[], int len) {
      for (int i = 0; i < len; i++)
         //loop through each element of the array
       ...
    }

Pointer notation can be used when prototyping and defining functions that receive arrays as parameters.

   void func1(int*);  //this syntax indicates that the function receives
                      //a pointer to an array
   void func2(const int*);  //array content cannot be modified

A Tutorial on Pointers and Arrays in C

{TopOfPage} {Resources}


Pointers

A pointer is a variable whose value is an address of another variable.

Defining and Initializing Pointer Variables
When defining a variable, prefixing the variable's name with an asterik * causes the variable to be a pointer.

   int* iptr;   /*define variable iptr that will point to an int variable*/
   float * fptr; /*define variable fptr that will point to a float variable*/
   char *cptr;  /*define variable cptr that will point to a character*/
   int i, j;
   float f;
   char c;

   iptr = &i;  /*assign the address-of variable 'i' to iptr*/
   fptr = &f;  /*assign the address-of variable 'f' to fptr*/
   cptr = &c;  /*assign the address-of variable 'c' to cptr*/
   int* iptr2 = &j; /*define and initialize an int pointer*/
   iptr = iptr2;  /*now iptr points to the variable 'j'*/

   note:  Placement of the  *  when defining a pointer variable is
          a matter of style -- I like to place the  *  next to the
          data type, but others prefer to place it next to the 
          variable name.  Placing it next to the data type does
          require caution when multiple variables are defined on
          the same declaration statement.  Example:

              int* ip1, ip2;  
                   // ip1  is a pointer to an  int  , but  ip2  is a
                      regular  int

              int* ip1, *ip2;  
                   // ip1  and  ip2  are both pointers to an  int

The address-of operator only applies to objects in memory: variables and array elements.

Locally defined non-static pointers, unless explicitly initialized, are garbage and using them without initialization can cause a program to execute incorrectly (or abort).

Global and statically defined pointers are initialized to the NULL pointer.

Accessing Data Using Pointers

The unary operator * is the indirection or dereferencing operator; when applied to a pointer, it accesses the object the pointer points to.

   int i = 200;
   int* iptr = &i;

   cout << *iptr;    /*prints 200 -- the value of 'i' which is the object
                       iptr points to*/
Pointers and Function Arguments

Since C passes arguments to functions by value, there is no way for the called function to alter a variable in the calling function. A way to obtain the desired effect is for the calling program to pass pointers to the values to be changed.

   void swap(int*, int*);

   int i, j;

   swap(&i, &j);

   void swap(int* a, int* b) {
      int tmp = *a;
      *a = *b;
      *b = tmp;
   }

Using pointers is similar to using reference variables; however, reference variables are part of C++ and they are not part of C.

Advantage of using reference variables over pointers.

Introduction to Pointers and Arrays

There is a strong relationship between pointers and arrays.

When an array name is used by itself, it evaluates to a constant pointer to the first element of the array.

Everywhere you use arrayName[index] you can use *(ptr + index) (assuming ptr points to some part of the array).

   int scores[10];

   scores[0] = 2;      // or  *scores = 2
   scores[1] = 3;      // or  *(scores + 1) = 3
   *(scores + 2) = 5;  // or  scores[2] = 5

{TopOfPage} {Resources}


Home Previous Next