3. More about Data Structures and ADTs:
C++ data types
C++ types an be classified as:
• Fundamental (or simple or scalar):
A data object of one of these types is a single object.
- int, double, char, bool, complex,
and the related types (unsigned, short, etc.) - enumerations
• Structured:
Collections of data
- arrays, structs, unions, classes, valarrays, bitsets, the containers and adapters in STL
Read §3.1
& §3.2
Structs vs. Classes
Similarities
1. Essentially the same syntax
2. Both are used to model objects with different attributes (characteristics) represented as data members
(also called fields or instance or attribute variables).
Thus, both are used to process non-homogeneous data sets.
Structs vs. Classes
Differences
1. C does not provide classes;
C++ provides both structs and classes.
2. Members of a struct by default are public (can be accessed outside the struct by using the dot operator.
In C++ they can be declared to be private (cannot be accessed outside the struct.
3. Members of a class by default are private (cannot be
accessed outside the class) but can be explicitly
C Structs vs. C++ Classes (& Structs) ( "traditional" vs "OOP")
C++'s structs and classes model objects that have:
Attributes (characteristics) represented as and
Attributes Data Members Operations
Function Members
This leads to a whole new style of programming:
object-oriented. Objects are self-contained, possessing their own operations — commonly called the I can do it myself principle —
rather than being passed as a parameter to an
external function that operates on them and sends them back.
Operations (behaviors) represented as (also called methods).
data members
function members
Declaring a Class
Common Forms:
class ClassName {
// private by default
Declarations of private members public:
Declarations of public members };
class ClassName {
public: //"interface" of class given first Declarations of public members
private:
Declarations of private members
Some programmers prefer to put the private section first:
- Using the default access for classes - Can omit the private: specifier
Other programmers put the public interface of the class first and the hidden private details last. (In the text, the latter approach is used.) Although not commonly done, a class may have several private and public sections; the keywords private: and public: mark the beginning of each.
Attributes Data Members Operations
Function Members public:
private:
Notes:
Data members are normally placed in the private section of a class;
function members in the public section.
Access to Class Members
A particular instance of a class is called an object : ClassName object_name;
Private members can be accessed only within the class (except by friend functions to be described later).
Public members can be accessed within or outside
the class; to access them outside the class, one must use the dot operator:
object_name.public_member_name
Class declarations are usually placed in a header file whose
name is ClassName.h . The library is then called a class
library.
Header file for class Time — Version 1
/** Time.h --- This header file defines the data type Time for processing time.
Basic operations are:
Set: To set the time
Display: To display the time
---*/
#include <iostream>
using namespace std;
class Time
{/******** Member functions ********/
public:
/* Set sets the data members of a Time object to specified values.
* * Receive: hours, the number of hours in standard time
* minutes, the number of minutes in standard time * AMPM ('A' if AM, 'P' if PM
* Return: The Time object containing this function with its
* myHours, myMinutes, and myAMorPM members set to hours, * minutes, and am_pm, respectively, and myMilTime to
* the equivalent military time
********************************************************************/
/* Display displays time in standard and military format using * output stream out.
* * Receive: ostream out
* Output: The time represented by the Time object containing * this function
* Passes back: The ostream out with time inserted into it
*****************************************************************/
Notes:
1. "my" in data members names remind of internal ("I can do it myself") perspective.
2. const at end of Display()'s prototype makes it a const function whieh means it cannot modify any of the data members.
void Display(ostream & out) const;
/********** Data Members **********/
private:
unsigned myHours, myMinutes;
char myAMorPM; // 'A' or 'P'
unsigned myMilTime; // military time equivalent }; // end of class declaration
3. Why make data members private?
"Hidden" data members:
Cannot be accessed outside class
Application programs must interact with an object through its interface
Public member functions control interaction between programs and class.
Application program need not know about implementation!
Implementation may change (improve storage, simpler algorithms. etc.)
If interface is constant, programs using an object need not be changed.
What's wrong with tying application code to implementation details?
A change in implementation forces a change in the application code Increased upgrade time.
Increased programmer cost
Decreased programmer productivity Reduced profits due to
— Delayed releases/upgrades
— Loss of customer confidence in software reliability
Always define data members of a class as private.
Implementation of a Class
Class declaration contains:
• Declarations of data members
• Prototypes (declarations) of function members
Definitions of function members are not usually placed in class declaration Avoid cluttering up the interface
Definitions placed outside the class declaration must tell compiler where the corresponding declaration/prototype is :
Use the scope operator :: which has the form ClassName::ItemName
(the qualified or full name of ItemName.)
Example
:class Something {
public:
static const int CAPACITY = 100;
typedef double ArrayType[CAPACITY];
void Print(ArrayType a, int itsSize);
. . . }
. . .
Something::ArrayType x = {0};
for (int i = 0; i < Something::CAPACITY; i++) . . .
void Something::Print(Something::ArrayType a, int itsSize)
{ . . . }
Traditionally, definitions of member functions have been put in an implementation file ClassName.cpp corresponding to the class' header file. This is done to enforce data abstraction — separating the interface of the ADT from its implementation details.
(Unfortunately, the class data members, which store data and are therefore part of the implementation, must be in the .h file.)
With the increasing use of templates, however, this practice is becoming less common because current compiler technology
doesn't permit this split for templates — everything has to be in the
same file. Thus the reason for dropping the ".h" from standard
class libraries. They're really class-template libraries, and there are
therefore no corresponding ".cpp" files.
Implementation of class Time — Version 1
// Time.cpp -- implements the Time member functions
#include "Time.h"
/*** Utility functions ***/
/* ToMilitary converts standard time to military time.
Receive: hours, minutes, am_pm
Return: The military time equivalent
Could implement this as a private class method
*/
int ToMilitary (unsigned hours, unsigned minutes, char am_pm) {
if (hours == 12) hours = 0;
return hours * 100 + minutes + (am_pm == 'P' ? 1200 : 0);
}
Implementation of class Time — cont.
//--- Function to implement the Set operation
void Time::Set(unsigned hours, unsigned minutes, char am_pm) {
// Check class invariant
if (hours >= 1 && hours <= 12 &&
minutes >= 0 && minutes <= 59 &&
(am_pm == 'A' || am_pm == 'P')) {
myHours = hours;
myMinutes = minutes;
myAMorPM = am_pm;
myMilTime = ToMilitary(hours, minutes, am_pm);
} else
cerr << "*** Can't set time with these values ***\n";
// Object's data members remain unchanged }
//--- Function to implement the Display operation void Time::Display(ostream & out) const
{
out << myHours << ':'
<< (myMinutes < 10 ? "0" : "") << myMinutes << ' ' << myAMorPM << ".M. ("
<< myMilTime << " mil. time)";
}
Implementation of class Time — cont.
Test driver for class Time
#include "Time.h"
#include <iostream>
using namespace std;
int main() {
Time mealTime;
mealTime.Set(5, 30, 'P');
cout << "We'll be eating at ";
mealTime.Display(cout);
cout << endl;
}
Execution:
We'll be eating at 5:30 P.M. (1730 mil. time)
Object-Oriented Perspective
Procedural: Send object off to some function for processing OOP: Send a message to the object to operate on itself.
To set my digital watch to 5:30 P.M., I don't wrap it up and mail it off to Casio. I push a button!
To display the time on my watch, I don't wrap it up and mail it
off to Casio and have them tell me what time it is. I have it
display the time itself, perhaps pushing a button to turn on the
backlight so I can see it .
Notes:
1. Member functions: "Inside" an object, so don't pass object to them as a parameter. (They receive the object to be operated implicitly, rather than explicitly via a parameter.)
Non-member functions: "Outside" an object, so to operate on an object, they must receive it via a parameter.
2. Public items must be qualified when referred to outside the class declaration:
ClassName::ItemName
Public constants are usually declared static so they are global class
properties that can be accessed by all objects of that class type rather than each object having its own copy.
3. Simple member functions: Usually specified as inline functions.
This suggests to compiler to replace a function call with actual code of the function with parameters replaced by arguments
— saves overhead of function call.
Two ways to inline a class function member:
1. Prototype in class declaration; inline definition below the class declaration in the header file,
qualifying the name as usual.
In ClassName.h:
class ClassName {
public:
RetType SimpleFun(param_list); //prototype . . .
};
inline RetType ClassName::SimpleFun(param_list) { . . . } // definition
2.Put function’s definition (instead of prototype) inside
Class Invariant: A condition (boolean expression) that ensures that the data members always contain valid values.
Example: 1 < myHours < 12 && 0 < myMinutes < 59 &&
myAMorPM == 'A' or 'P' && 0 < myMilTime < 2359 Operations that modify data members
must check the class invariant.
Then other operations can be sure data members have valid values.
2. Use assert(class_invar) (must #include <cassert> )
true false
How?
1. Use an if statement — see Set()
continue execution halt execution and display error message
3. Throw an exception that the calling function can catch and take appropriate action:
//--- Function to implement the Set operation ---
void Time::Set(unsigned hours, unsigned minutes, char am_pm) {
// Check class invariant
if (hours >= 1 && hours <= 12 &&
minutes >= 0 && minutes <= 59 &&
(am_pm == 'A' ||am_pm == 'P')) {
. . . }
else {
char error[] = "*** Illegal initializer values ***\n";
throw error;
} }
To catch this exception, a calling function might contain:
try {
mealTime.Set(13, 30, 'P');
cout << "This is a valid time\n";
}
catch (char badTime[]) {
cerr << "ERROR: " << badTime << endl;
exit(-1);
}
cout << "Proceeding. . .\n";
When executed, the output produced will be
ERROR: *** Illegal initializer values ***
24
Class Constructors
Constructing an object consists of:
(1) allocating memory for the object, and (2) initializing the object.
In our example, after the declaration Time mealTime;
memory has been allocated for object mealTime and it's data members have some default (perhaps "garbage") initial values.
Need to:
Specify initial values for mealTime
Provide default values to be used if no initial values are specified.
This is the role of a class' constructor.
Class Constructors - Properties
1. Names are always the same as the class name.
2. Initialize the data members of an object with default values or with values provided as arguments.
3. Do not return a value; have no return type (not even void).
4. Often simple and can be inlined.
5. Called whenever an object is declared.
6. If no constructor is given in the class, compiler supplies a default constructor which allocates memory
and initializes it with some default (possibly garbage) value.
A default constructor is one that is used when the declaration of an object contains no initial values:
ClassName object_name;
7. If we supply a class constructor, we must also provide a default
26
Constructors for Time class
In Time.h class Time { public:
/* --- Construct a class object (default).
* Precondition: A Time object has been declared.
* Postcondition: Data members initialized to 12, 0, 'A', and 0.
****************************************************************/
Time();
/* --- Construct a class object (explicit values).
* Precondition: A Time object has been declared.
* Receive: Initial values initHours, initMinutes,and * initAMPM
* Postcondition: Data members initialized to initHours,initMinutes, * initAMPM, & correspoding military time
**********************************************************/
Time(unsigned initHours, unsigned initMinutes, char initAMPM);
. . .
// other member function prototypes private:
/********** Data Members **********/
. . .
Constructors for Time class - Implementation
In Time.h, after class declaration:
inline Time::Time() {
myHours = 12;
myMinutes = 0;
myAMorPM = 'A';
myMilTime = 0;
}
In Time.cpp:
Time::Time(unsigned initHours, unsigned initMinutes, char initAMPM) {
// Check class invariant
assert(initHours >= 1 && initHours <= 12 &&
initMinutes >= 0 && initMinutes <= 59 &&
(initAMPM == 'A' || initAMPM == 'P'));
myHours = initHours;
myMinutes = initMinutes;
myAMorPM = initAMPM;
Testing #1:
Time mealTime, //default constructor
bedTime(11,30,'P'); //explicit-value constructor
Create and initialize 2 Time objects:
mealTime.Display(cout);
cout << endl;
bedTime.Display(cout);
cout << endl;
Execution:
12:00 A.M. (0 mil. time) 11:30 P.M. (2330 mil. time) mealTime
myHours myMinutes
myAMorPM myMilTime
Function members Default Constructor
bedTime myHours
myMinutes myAMorPM myMilTime
Function members
Explicit Value Constructor
120 A0
1130 P 2330
Constructors for Time class — Default Arguments
Can combine both constructors into a single constructor function by using default arguments:
Replace constructors in Time.h with:
/*--- Construct a class object.
Precondition: A Time object has been declared.
Receive: Initial values initHours, initMinutes, and initAMPM (defaults 12, 0, 'A')
Postcondition: Data members initialized to initHours, initMinutes, initAMPM, & correspoding military time.
*/
Time(unsigned initHours = 12, unsigned initMinutes = 0, char initAMPM = 'A');
Note: Any parameter with default argument must appear after all
Testing:
Time mealTime, t1(5), t2(5, 30), t3(5, 30, 'P');
Creates 4 Time objects:
120 A0
05 500A
35 5300A
35 17300P
mealTime.Display(cout);
cout << endl;
t1.Display(cout); cout << endl;
t2.Display(cout); cout << endl;
t3.Display(cout); cout << endl;
Execution:
12:00 A.M. (0 mil. time) 5:00 A.M. (500 mil. time) 5:30 A.M. (530 mil. time) 5:30 P.M. (1730 mil. time)
Copy Operations
Two default copy operations are provided:
1. Copy in initialization (via ) 2. Copy in assignment (via )
copy constructor assignment operator
Each makes a raw (bitwise) copy of the memory allocated to the
data members of the object.
Examples:
Time t = bedTime;
Time t(bedTime);
Both:
1. Allocate memory for t
2. Copy data members of bedTime so t is a copy of bedTime
In contrast :
Time t = Time(11, 30, 'P');
calls the explicit-value constructor to construct a (temporary) Time object and then copies it into t .
Note: These are not assignments; a default
11 30 P 2330 11
30 P 2330
What about
Time t = 3;
There is a default copy operation for assignment.
Example :
t = mealTime;
copies the members of mealTime into t , replacing any previous values:
120 A 0 120
A 0
It returns this Time object as the value of this expression .
Access Member Functions
Data members are private: they cannot be accessed outside the class. To
make the values stored in some or all of these members accessible, provide access (or extractor) member functions
Example:
To provide access to myHours of class Time.
(Access for the remaining members is analogous.)
simply retrieves and returns the value stored in a data member
inline, because it's simple
prototype (and define) as a const function Add in Time class declaration
/* Hour Accessor
* Return: value stored in myHours data member of * Time object containing this function */
unsigned Hour() const;
Add below Time class declaration
inline unsigned Time::Hour() const { return myHours; }
Testing:
Time mealTime;
. . .
cout << "Hour: " << mealTime.Hour() << endl;
Execution:
Hour: 12
Output and Input
Add output operation(s) to a class early so it can be used for debugging other operations.
Example: overload operator<<() for a Time object Instead of:
cout << "We'll be eating at " ; mealTime.Display(cout);
cout << endl;
we can write:
cout << "We'll be eating at "
<< mealTime << endl;
Overloading operators
In C++, operator can be implemented with the function operator() .
If a member function of a class C, and a is of type C, the compiler treats a b as
a.operator(b)
If not a member function of a class C, the compiler treats a b as
operator(a, b)
Overloading Output Operator <<
Can operator<<() be a member function?
No, because the compiler will treat
cout << t
as
cout.operator<<(t)
which means that operator<<(const Time &) would have to be a member of class ostream (or cout of type Time )
and we can't (or don't want to) modify standard C++ classes.
Putting the prototype
ostream& operator<<(ostream & out, const Time& t);
inside our class declaration producs a compiler error like:
'Time::operator <<(ostream &, const Time &)'
must take exactly one argument
Why is out a reference parameter?
So corresponding actual ostream argument gets modified when out does.
Why is t a const reference parameter?
Avoid overhead of copying a class object.
Why is return type ostream & (a reference to an ostream)?
Else a copy of out is returned.
Why return out?
So we can chain output operators.
Since << is - associative:
operator<<(cout, t1) << endl << t2 << endl;
first function must return cout
cout << endl << t2 << endl;
operator<<(cout, endl) << t2 << endl;
left ( )
( )
( ) cout << t1 << endl << t2 << endl;
Method 1: Put definition in .h file, after class declaration, and have it call an output function member.
Inline it, because it's simple.
. . .
} // end of class declaration . . .
/* operator<< displays time in standard and military format.
Receive: ostream out and Time object t
Output: time represented by Time object t Pass back: ostream out with t inserted into it Return: out
*/
inline ostream & operator<<(ostream & out, const Time & t) {
t.Display(out);
return out;
}
Overloading << for a Class
Method 2: Define operator<<() outside the class declaration as a non-member function and:
(i) use access functions to display data members, or (ii) declare it to be a friend in the class declaration.
In class declaration
/* doc. as before */
friend ostream & operator<<(ostream & out, const Time & t);
Outside class declaration (in .cpp file)
ostream & operator<<(ostream & out, const Time & t)
{
out << t.myHours << ':'
<< (t.myMinutes < 10 ? "0" : "") << t.myMinutes
<< ' ' << t.myAMorPM << ".M. ("
<< t.myMilTime << " mil. time)";
return out;
Friend Functions
A function that a class declares as a friend is a
non-member function to which the class has granted permission to access members in its private sections.
Note: Because a friend function is not a function member:
Don't qualify its definition with class name and scope operator (::).
Don't put friend in definition.
It receives the object on which it operates as a parameter.
It uses the dot operator to access the data members.
To add an input operator to our Time class, we proceed in much the same way as for output. We could either:
1. Add a member function ReadTime() that reads values and stores them in the data members of a Time object; then call it from non- member function operator>>()
2. Declare operator>>() to be a friend function so that it can
access the data members of a Time object and store input values in them.
Is one of two methods for input/output preferred?
We'll see later when we study inheritance and polymorphism that the
first method is preferred (or perhaps required).
Relational Operators (<)
Specification:
Receives: Two Time objects
Returns: True if the first object is less than the second; false otherwise.
Should operator<() be a member function?
Internal perspective: I compare myself with another Time object and determine if I am less than that other object
External perspective: Two Time objects are compared by an external function to determine if the first is less than the second.
OOP: "I-can-do-it-myself" principle ( objects self-contained):
Use member functions whenever possible.
Rephrased specification:
Receives: A Time object (and the current object implicitly)
Returns: True if I (the Time object containing this function) am
less than the Time object received; false otherwise.
// Internal perspective : Add to Time.h
class Time {
public: // member functions . . .
/***** Relational operators *****/
/* operator< determines if one Time is less than another Time * Receive: A Time t (and the current object implicitly)
* Return: True if time represented by current object is < t.
*/
bool operator<(const Time & t) const;
. . .
}; // end of class declaration
inline bool Time::operator<(const Time & t) const { return myMilTime < t.myMilTime; }