C++

Pocket Reference

Kyle Loudon

Beijing • Cambridge • Farnham • Köln • Paris • Sebastopol • Taipei • Tokyo

C++ Pocket Reference
by Kyle Loudon

Copyright © 2003 O’Reilly Media, Inc. All rights reserved.

Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North,
Sebastopol, CA 95472.
O’Reilly Media, Inc. books may be purchased for educational,
business, or sales promotional use. Online editions are also available
for most titles (safari.oreilly.com). For more information, contact our
corporate/institutional sales department: (800) 998-9938 or
[email protected].

Editor: Jonathan Gennick

Production Editor: Emily Quill
Cover Designer: Ellie Volckhausen
Interior Designer: David Futato

Printing History:
May 2003: First Edition.

Nutshell Handbook, the Nutshell Handbook logo, and the O’Reilly

logo are registered trademarks of O’Reilly Media, Inc. The Pocket
Reference series designations, C++ Pocket Reference, the image of a
chipmunk, and related trade dress are trademarks of O’Reilly Media,
Inc. Many of the designations used by manufacturers and sellers to
distinguish their products are claimed as trademarks. Where those
designations appear in this book, and O’Reilly Media, Inc. was aware of
a trademark claim, the designations have been printed in caps or initial
caps.
While every precaution has been taken in the preparation of this book,
the publisher and author assume no responsibility for errors or
omissions, or for damages resulting from the use of the information
contained herein.

0-596-00496-6
[C] [6/06]
 Contents

Introduction 1
Typographic Conventions 2
Acknowledgments 2
Compatibility with C 2
Program Structure 3
Startup 3
Termination 5
Header Files 5
Source Files 7
Preprocessor Directives 8
Preprocessor Macros 11
Fundamental Types 12
bool 12
char and wchar_t 13
short, int, long 15
float, double, long double 16
Compound Types 17
Enumerations 18
Arrays 19
Strings 22
Pointers 24
Pointers to Members 26

v
 References 27
Class Types 28
Type Conversions and Definitions 28
Type Conversions 28
Type Definitions 31
Lexical Elements 31
Comments 32
Identifiers 32
Reserved Words 33
Literals 34
Operators 34
Expressions 46
Scope 47
Local Scope 47
Class Scope 47
Namespace Scope 48
File Scope 48
Other Scopes 49
Enclosing Scopes 49
Declarations 50
Declaring Variables 51
Declaring Functions 52
Storage Classes 55
Qualifiers 57
Statements 59
Expression Statements 59
Null Statements 59
Compound Statements 59
Iteration Statements 60

vi | Contents
 Selection Statements 62
Jump Statements 64
Namespaces 66
using Declarations 67
using Directives 67
Unnamed Namespaces 68
Classes, Structs, and Unions 68
Declaring Objects 69
Accessing Members 69
Declaring Data Members 70
Declaring Member Functions 74
Access Levels for Members 78
Friends 79
Constructors 80
Destructors 83
Nested Declarations 85
Forward Declarations 86
Structs 86
Unions 86
Inheritance 88
Constructors and Inheritance 89
Destructors and Inheritance 90
Virtual Member Functions 91
Abstract Base Classes 94
Access Levels for Inheritance 94
Multiple Inheritance 95
Virtual Base Classes 97
Templates 98
Template Classes 98
Template Functions 101

Contents | vii
Overloading 104
Overloading Functions 104
Overloading Operators 105
Memory Management 108
Memory Allocation 108
Memory Reclamation 110
Casts and Runtime Type Information 112
C-Style Casts 112
Casts in C++ 112
Runtime Type Information 115
Exception Handling 117
try 117
throw 117
catch 118
Exception Specifications 119
The C++ Standard Library 120
The std Namespace 121
C Standard Library Support 121
C++ Standard Header Files 122
I/O Streams 122

viii | Contents
Chapter 1

C++ Pocket Reference

Introduction
The C++ Pocket Reference is a quick reference to the C++
programming language as defined by the international stan-
dard INCITS/ISO/IEC 14882–1998. It consists of a number
of short sections, each further divided into specific topics.
Many of the topics include pointed, canonical examples.
At the outset, it is important to recognize that C++ is a vast
language, admittedly difficult to describe in a pocket refer-
ence. As a result, this reference is devoted almost exclusively
to presenting the language itself. Other references are avail-
able from O’Reilly & Associates describing the C++ Stan-
dard Library, a vast subject on its own. The C++ Standard
Library includes all of the facilities of the C Standard Library
plus many new ones, such as the Standard Template Library
(STL) and I/O streams.
This book has been written for developers with a variety of
backgrounds and experience levels in C++. Those with expe-
rience using C++ will find this book to be a uniquely focused
reference to its most commonly used features. If you are new
to C++, you may wish to work with a tutorial first and return
to this reference later to research specific topics.

1
Typographic Conventions
This book uses the following typographic conventions:
Italic
This style is used for filenames and for items emphasized
in the text.
Constant width
This style is used for code, commands, keywords, and
names for types, variables, functions, and classes.
Constant width italic
This style is used for items that you need to replace.

Acknowledgments
I would like to thank Jonathan Gennick, my editor at
O’Reilly, for his support and direction with this book.
Thanks also to Uwe Schnitker, Danny Kalev, and Ron Passe-
rini for taking the time to read and comment on an early
draft of this book.

Compatibility with C
With some minor exceptions, C++ was developed as an
extension, or superset, of C. This means that well-written C
programs generally will compile and run as C++ programs.
(Most incompatibilities stem from the stricter type checking
that C++ provides.) So, C++ programs tend to look syntacti-
cally similar to C and use much of C’s original functionality.
This being said, don’t let the similarities between C and C++
fool you into thinking that C++ is merely a trivial derivation
of C. In fact, it is a rich language that extends C with some
grand additions. These include support for object-oriented
programming, generic programming using templates,
namespaces, inline functions, operator and function over-
loading, better facilities for memory management, refer-
ences, safer forms of casting, runtime type information,
exception handling, and an extended standard library.

2 | C++ Pocket Reference

Program Structure
At the highest level, a C++ program is composed of one or
more source files that contain C++ source code. Together,
these files define exactly one starting point, and perhaps vari-
ous points at which to end.
C++ source files frequently import, or include, additional
source code from header files. The C++ preprocessor is
responsible for including code from these files before each
file is compiled. At the same time, the preprocessor can also
perform various other operations through the use of prepro-
cessor directives. A source file after preprocessing has been
completed is called a translation unit.

Startup
The function main is the designated start of a C++ program,
which you as the developer must define. In its standard form,
this function accepts zero or two arguments supplied by the
operating system when the program starts, although many
C++ implementations allow additional parameters. Its return
type is int. For example:
int main( )
int main(int argc, char *argv[])

argc is the number of arguments specified on the command

line; argv is an array of null-terminated (C-style) strings con-
taining the arguments in the order they appear. The name of
the executable is stored in argv[0], and may or may not be
prefixed by its path. The value of argv[argc] is 0.
The following shows the main function for a simple C++ pro-
gram that prompts the user for actions to perform on an
account:
#include <iostream>
#include <cmath>
#include <cstdlib>
using namespace std;

Program Structure | 3
 #include "Account.h"

int main(int argc, char *argv[])

{
Account account(0.0);
char action;
double amount;

if (argc > 1)
account.deposit(atof(argv[1]));

while (true)
{
cout << "Balance is "
<< account.getBalance( )
<< endl;

cout << "Enter d, w, or q: ";

cin >> action;

switch (action)
{
case 'd':
cout << "Enter deposit: ";
cin >> amount;
account.deposit(amount);
break;

case 'w':
cout << "Enter withdrawal: ";
cin >> amount;
account.withdraw(amount);
break;

case 'q':
exit(0);

default:
cout << "Bad command" << endl;
}
}

return 0;
}

4 | C++ Pocket Reference

The class for the account is defined in a later example. An
initial deposit is made into the account using an amount
specified on the command line when the program is started.
The function atof (from the C++ Standard Library) is used
to convert the command-line argument from a string to a
double.

Termination
A C++ program terminates when you return from main. The
value you return is passed back to the operating system and
becomes the return value for the executable. If no return is
present in main, an implicit return of 0 takes places after fall-
ing through the body of main. You can also terminate a pro-
gram by calling the exit function (from the C++ Standard
Library), which accepts the return value for the executable as
an argument.

Header Files
Header files contain source code to be included in multiple
files. They usually have a .h extension. Anything to be
included in multiple places belongs in a header file. A header
file should never contain the following:
• Definitions for variables and static data members (see
“Declarations” for the difference between declarations
and definitions).
• Definitions for functions, except those defined as tem-
plate functions or inline functions.
• Namespaces that are unnamed.

NOTE
Header files in the C++ Standard Library do not use the .h
extension; they have no extension.

Program Structure | 5
Often you create one header file for each major class that you
define. For example, Account is defined in the header file
Account.h, shown below. Of course, header files are used for
other purposes, and not all class definitions need to be in
header files (e.g., helper classes are defined simply within the
source file in which they will be used).
#ifndef ACCOUNT_H
#define ACCOUNT_H

class Account
{
public:
Account(double b);

void deposit(double amt);

void withdraw(double amt);
double getBalance( ) const;

private:
double balance;
};

#endif

The implementation of this class is in Account.cpp. You use

the preprocessor directive #include to include a header file
within another file (see “Preprocessor Directives”).
Because header files are often included by other headers
themselves, care must be taken not to include the same file
more than once, which can lead to compilation errors. To
avoid this, it is conventional to wrap the contents of header
files with the preprocessor directives #ifndef, #define, and
#endif, as done in the previous example.
The tactic of wrapping a header file forces the preprocessor
to test an identifier. If that identifier is not defined, the pre-
processor defines it and processes the file’s contents. As an
example, the contents of Account.h are processed only when
ACCOUNT_H is undefined, and the first thing that processing
does is to define ACCOUNT_H to ensure the header is not

6 | C++ Pocket Reference

processed a second time. To ensure uniqueness, X_H is typi-
cally used as the identifier, where X is the name of the header
file without its extension.

Source Files
C++ source files contain C++ source code. They usually
have a .cpp extension. During compilation, the compiler typi-
cally translates source files into object files, which often have
a .obj or .o extension. Object files are joined by the linker to
produce a final executable or library.
Often you create one source file for each major class you
implement. For example, the implementation of Account is in
Account.cpp, shown below. Of course, there is no require-
ment about this; source files often contain more than just the
implementation of a single class.
#include "Account.h"

Account::Account(double b)
{
balance = b;
}

void Account::deposit(double amt)

{
balance += amt;
}

void Account::withdraw(double amt)

{
balance -= amt;
}

double Account::getBalance( ) const

{
return balance;
}

Program Structure | 7
Preprocessor Directives
The C++ preprocessor can be used to perform a number of
useful operations controlled via several directives. Each direc-
tive begins with a pound sign (#) as the first character that is
not whitespace on a line. Directives can span multiple lines
by including a backslash (\) at the end of intermediate lines.

#define
The #define directive replaces an identifier with the text that
follows it wherever the identifier occurs in a source file. For
example:
#define BUFFER_SIZE 80

char buffer[BUFFER_SIZE];

If you specify no text after the identifier, the preprocessor

simply defines the identifier so that any check for its defini-
tion tests true and it expands to nothing in the source code.
(You can see this in use earlier where ACCOUNT_H was defined.)

NOTE
In C++, it is preferable to use enumerations, and to a
lesser degree, variables and data members declared using
the keywords const or static const for constant data,
rather than the #define directive.

The #define directive can accept arguments for macro substi-

tution in the text. For example:
#define MIN(a, b) (((a) < (b)) ? (a):(b))

int x = 5, y = 10, z;

z = MIN(x, y); // This sets z to 5.

In order to avoid unexpected problems with operator prece-

dence, parameters should be fully parenthesized in the text,
as shown above.

8 | C++ Pocket Reference

 NOTE
In C++, it is preferable to use templates and inline func-
tions in place of macros. Templates and inline functions
eliminate unexpected results produced by macros, such
as MIN(x++, y) incrementing x twice when a is less than b.
(Macro substitution treats x++, not the result of x++, as
the first parameter.)

#undef
The #undef directive undefines an identifier so that a check
for its definition tests false. For example:
#undef LOGGING_ENABLED

#ifdef, #ifndef, #else, #endif

You use the #ifdef, #ifndef, #else, and #endif directives
together. The #ifdef directive causes the preprocessor to
include different code based on whether or not an identifier
is defined. For example:
#ifdef LOGGING_ENABLED
cout << "Logging is enabled" << endl;
#else
cout << "Logging is disabled" << endl;
#endif

Using #else is optional. #ifndef works similarly but includes

the code following the #ifndef directive only if the identifier
is not defined.

#if, #elif, #else, #endif

The #if, #elif, #else, and #endif directives, like the direc-
tives of #ifdef, are used together. These cause the preproces-
sor to include or exclude code based on whether an
expression is true. For example:
#if (LOGGING_LEVEL == LOGGING_MIN && \
LOGGING_FLAG)
cout << "Logging is minimal" << endl;

Program Structure | 9
 #elif (LOGGING_LEVEL == LOGGING_MAX && \
LOGGING_FLAG)
cout << "Logging is maximum" << endl;
#elif LOGGING_FLAG
cout << "Logging is standard" << endl;
#endif

The #elif (else-if) directive is used to chain a series of tests

together, as shown above.

#include
The #include directive causes the preprocessor to include
another file, usually a header file. You enclose standard
header files with angle brackets, and user-defined header files
with quotes. For example:
#include <iostream>
#include "Account.h"

The preprocessor searches different paths depending on the

form of enclosure. The paths searched depend on the system.

#error
The #error directive causes compilation to stop and a speci-
fied string to be displayed. For example:
#ifdef LOGGING_ENABLED
#error Logging should not be enabled
#endif

#line
The #line directive causes the preprocessor to change the
current line number stored internally by the compiler during
compilation in the macro _ _LINE_ _. For example:
#line 100

A filename optionally can be specified in double quotes after

the line number. This changes the name of the file stored
internally by the compiler in the macro _ _FILE_ _. For
example:
#line 100 "NewName.cpp"

10 | C++ Pocket Reference

#pragma
Some operations that the preprocessor can perform are
implementation-specific. The #pragma directive allows you to
control these operations by specifying the directive along
with any parameters in a form that the directive requires. For
example:
#ifdef LOGGING_ENABLED
#pragma message("Logging enabled")
#endif

Under Microsoft Visual C++ 6.0, the message directive

informs the preprocessor to display a message during compi-
lation at the point where this line is encountered. The direc-
tive requires one parameter: the message to display. This is
enclosed in parentheses and quoted.

Preprocessor Macros
The C++ preprocessor defines several macros for insert-
ing information into a source file during compilation.
Each macro begins and ends with two underscores, except
for _ _cplusplus, which has no terminating underscores.
_ _LINE_ _
Expands to the current line number of the source file
being compiled.
_ _FILE_ _
Expands to the name of the source file being compiled.
_ _DATE_ _
Expands to the date on which the compilation is taking
place.
_ _TIME_ _
Expands to the time at which the compilation is taking
place.
_ _TIMESTAMP_ _
Expands to the date and time at which the compilation is
taking place.

Program Structure | 11
_ _STDC_ _
Will be defined if the compiler is in full compliance with
the ANSI C standard.
_ _cplusplus
Will be defined if the program being compiled is a C++
program. How a compiler determines whether a given
program is a C++ program is compiler-specific. You may
need to set a compiler option, or your compiler may look
at the source file’s extension.

Fundamental Types
The type for an identifier determines what you are allowed to
do with it. You associate a type with an identifier when you
declare it. When declaring an identifier, you also may have
the opportunity to specify a storage class and one or more
qualifiers (see “Declarations”).
The fundamental types of C++ are its Boolean, character,
integer, floating-point, and void types. The Boolean, charac-
ter, and integer types of C++ are called integral types. Inte-
gral and floating-point types are collectively called arithmetic
types.

bool
Booleans are of type bool. The bool type is used for values of
truth. For example:
bool flag;
...
if (flag)
{
// Do something when the flag is true.
}

Boolean values
Booleans have only two possible values: true or false. The
typical size of a bool is one byte.

12 | C++ Pocket Reference

Boolean literals
The only Boolean literals are the C++ keywords true and
false. By convention, false is defined as 0; any other value is
considered true.

char and wchar_t

Characters are of type char or wchar_t. The char type is used
for integers that refer to characters in a character set (usually
ASCII). For example:
char c = 'a';

cout << "Letter a: " << c << endl;

The wchar_t type is a distinct type large enough to represent

the character sets of all locales supported by the implementa-
tion. To use facilities related to the wchar_t type, you include
the standard header file <cwchar>.
Character types may be specified either as signed or unsigned
and are sometimes used simply to store small integers. For
example:
signed char small = -128;
unsigned char flags = 0x7f;

A signed char represents both positive and negative values,

typically by sacrificing one bit to store a sign. An unsigned
char doesn’t have a sign and therefore can hold larger posi-
tive values, typically twice as large. If neither signed nor
unsigned is specified, characters are usually signed by default,
but this is left up to the compiler.

Character values
The range of values that characters may represent is found in
the standard header file <climits>. The size of a char is one
byte. The size of a byte technically is implementation-
defined, but it is rarely anything but eight bits. The size of

Fundamental Types | 13
the wchar_t type is also implementation-defined, but is typi-
cally two bytes.

Character literals
Character literals are enclosed by single quotes. For example:
char c = 'A';

To specify literals for wide characters, you use the prefix L.

For example:
wchar_t c = L'A';

To allow special characters, such as newlines and single

quotes, to be used within literals, C++ defines a number of
escape sequences, each of which begins with a backslash.
Table 1 presents these escape sequences. There is no limit to
the number of hexadecimal digits that can appear after \x in
a hexadecimal escape sequence. Octal escape sequences can
be at most three digits.

Table 1. Character escape sequences

Escape sequence Description
\a Alert (system bell)
\b Backspace
\f Form feed
\n Newline
\r Carriage return
\t Horizontal tab
\v Vertical tab
\\ Backslash
\' Single quote
\" Double quote
\? Question mark
\ooo Octal number ooo
\xhhh... Hexadecimal number hhh...

14 | C++ Pocket Reference

short, int, long
Integers are of type short, int, or long. These types differ in
size and the range of values they can represent. For example:
short sval = 32767;
int ival = 2147483647;
long lval = 0x7fffffff;

Integers may be specified as either signed or unsigned. For

example:
signed short total;
unsigned short flags = 0xf0f0;

Signed integers represent both positive and negative values,

typically by sacrificing one bit to store a sign. Unsigned inte-
gers don’t have a sign and therefore can hold larger positive
values. If an integer is not specified as either signed or
unsigned, it is signed by default.

Integer values
The range of values that each of the integer types may repre-
sent is found in the standard header file <climits>. The exact
size of a short, int, or long is left up to the compiler, but is
typically two, four, or four bytes respectively. Although the
size of each type can vary, the compiler guarantees that the
size of a short is less than or equal to the size of an int, and
the size of an int is less than or equal to the size of a long.

Integer literals
Literals for integers have several forms, as shown in Table 2.
If U, u, L, or l is not used as a suffix, the compiler assigns a
type appropriate for the magnitude of the literal.

Fundamental Types | 15
Table 2. Integer literals
Examples Description
12 The most common form of integer literals.
-5
012 Literals that begin with 0 are octal values (e.g., 012 is the octal literal
0377 for the decimal number 10).
0x2a Literals that begin with 0x are hexadecimal values (e.g., 0x2a is the
0xffff hexadecimal literal for the decimal number 42).
256L Literals with L (or l) in the suffix are treated as long.
0x7fL
0x80U Literals with U (or u) in the suffix are treated as unsigned.
0xffffUL

float, double, long double

Floating points are of type float, double, or long double.
These types differ in size and in the range and precision of
values they can represent. For example:
float fval = 3.4e+38F;
double dval = 1.7e+308;

Floating-point values
The range and precision of values that each of the floating-
point types may represent is found in the standard header file
<cfloat>. The exact size, range, and precision of a float,
double, or long double is left up to the compiler, but is typi-
cally four, eight, or ten bytes respectively. Although the size
of each type can vary, the compiler guarantees that the size of
a float is less than or equal to the size of a double, and the
size of a double is less than or equal to the size of a long
double.

Floating-point literals
Literals for floating points can take on several forms, as
shown in Table 3. If F, f, L, or l is not used as a suffix, the
compiler assigns a type of double.

16 | C++ Pocket Reference

Table 3. Floating-point literals
Examples Description
1.2345 The most common form of literal floating points.
-57.0
0.4567
1.992e+2 Literals expressed in scientific notation.
1.71e-25
8.00275F Literals with the suffix F (or f) are given the type float; literals
3.4e+38L with the suffix L (or l) are given the type long double.

void
The void type indicates the absence of a value. One use is in
declaring functions that do not return a value. For example:
void sayHello( )
{
cout << "Hello" << endl;
}

Another use is in declaring a pointer that can point to any

type of data. For example:
int i = 200;
void *p = &i;

The variable p points to an int. Variables that are not point-

ers cannot be declared as void.

Compound Types
Arithmetic types are the building blocks for more complex
types, called compound types. These include enumerations,
arrays, strings, pointers, pointers to members, references,
and the various class types of C++, as well as functions.
Arithmetic types, enumerations, pointers, and pointers to
members are collectively called scalar types.

Compound Types | 17
Enumerations
An enumeration, specified by the keyword enum, is a set of
integer constants associated with identifiers, called
enumerators, that you define. In general, enumerations pro-
vide a way to use meaningful names where you might other-
wise use integer constants, perhaps defined using the
preprocessor directive #define. Enumerations are preferred
over the preprocessor for this in C++ because they obey the
language’s rules of scope. The following defines an enumera-
tion for the colors of the rainbow:
enum SpectrumColor
{
Red, Orange, Yellow,
Green, Blue, Indigo,
Violet
};

If you plan to instantiate variables to store values of an enu-

meration, you can give the enumeration a name (here,
SpectrumColor); however, a name is not required. With this
enumeration, you can write a loop to cycle through the col-
ors of the rainbow, for example:
for (SpectrumColor s = Red; s <= Violet; s++)
{
// Longest wavelength to shortest.
}

Following are some additional points to keep in mind about

enumerations:
• You can specify values for enumerators within an enu-
meration, which you can then use in place of integer con-
stants.
• When you let the compiler assign values to enumerators,
it assigns the next integer after the one assigned to the
preceding enumerator.
• Values start at 0 if you do not provide a value for the first
enumerator.

18 | C++ Pocket Reference

 • You can use enumerators anywhere that you would use
an int.
• You cannot assign arbitrary integers to a variable of an
enumeration type.
• The size of integers for enumerations is no larger than
the size of an int, unless a larger integer is needed for
explicit values.
The following example illustrates these points:
enum
{
ASCII_NUL, // 0
ASCII_SOH, // 1
ASCII_STX, // 2

ASCII_A = 65, // 65
ASCII_B, // 66

BufferSize = 8 // 8
};

char buffer[BufferSize];

Arrays
Arrays contain a specific number of elements of a particular
type. So that the compiler can reserve the required amount of
space when the program is compiled, you must specify the
type and number of elements that the array will contain
when it is defined. The compiler must be able to determine
this value when the program is compiled. For example:
enum
{
HandleCount = 100
};

int handles[HandleCount];

Once an array has been defined, you use the identifier for the
array along with an index to access specific elements within

Compound Types | 19
the array. The following sets each element in the previous
array to an initial value of –1:
for (int i = 0; i < HandleCount; i++)
{
handles[i] = -1;
}

Arrays are zero-indexed; that is, the first element is at index 0.

This indexing scheme is indicative of the close relationship in
C++ between pointers and arrays and the rules that the lan-
guage defines for pointer arithmetic. In short, the assignment
in the example above is equivalent to the following:
*(handles + i) = -1;

It is important to remember that no bounds-checking is per-

formed for arrays.

Multidimensional arrays
C++ supports multidimensional arrays, which are arrays
defined using more than one index, as follows:
enum
{
Size1 = 4,
Size2 = 4
};

double matrix[Size1][Size2];

Arrays can be defined with more than two indices in a simi-

lar manner. Once a multidimensional array is defined, you
use multiple indices to access a specific element, as follows:
for (int i = 0; i < Size1; i++)
for (int j = 0; j < Size2; j++)
matrix[i][j] = 0.0;

The relationship between pointers and arrays extends to

multidimensional arrays as well. In short, the assignment in
the example above is equivalent to the following:
*(*(matrix + i) + j) = 0.0;

20 | C++ Pocket Reference

Passing arrays to functions
When defining a function that has an array as a parameter,
all but the first dimension must be specified for the parame-
ter. This ensures that the proper pointer arithmetic can be
performed. In the case of an array with a single dimension,
this means that no dimension is required:
void f(int handles[])
{
handle[0] = 0;
}

In the case of an array with two dimensions, for example, the

second dimension must be specified:
void g(double matrix[][Size2])
{
matrix[0][Size2 – 1] = 1.0;
}

You can also define equivalent functions that use pointer

parameters:
void f(int *handles)
{
handles[0] = 0;
}

void g(double (*matrix)[Size2])

{
matrix[0][Size2 – 1] = 1.0;
}

The parentheses are needed in the second case so that the

array is a multidimensional array of double values, not a one-
dimensional array of double pointers.

Initializer lists for arrays

An initializer list for an array is a comma-delimited list of val-
ues by which to initialize the array’s elements. The list is
enclosed by braces ({}). Each value’s type must be accept-
able for the type of elements that the array has been declared
to contain. For example:

Compound Types | 21
 enum SwitchState
{
On, Off
};

SwitchState switches[] =
{
On, Off, On, Off
};

When you initialize an array with an initializer list, you may

omit the array size in the declaration; enough space will be
allocated for the array to accommodate the values specified.
If you provide a size but specify values for fewer elements
than the size indicates, the missing elements are default-ini-
tialized. The rules for default initialization are complicated;
you should not rely on them.
Initializer lists can also be used to initialize arrays that are
multidimensional. The rules are essentially the same as for
arrays of one dimension, except that an initializer list for a
multidimensional array uses nested braces to align its values
in a manner consistent with the size of each dimension.
char tictactoe[3][3] =
{
{'_', '_', '_',},
{'_', '_', '_',},
{'_', '_', '_',}
};

Strings
Character (C-style) strings are arrays of characters termi-
nated with a null character (\0). The characters of the string
are of type char, or type wchar_t for wide-character strings.
For example:
enum
{
NameLength = 81
};

22 | C++ Pocket Reference

 char name[NameLength];
wchar_t wide[NameLength];

You must allocate one extra character for the null terminator
in arrays of characters to be used for strings. Functions that
return a string’s length, such as strlen (from the C++ Stan-
dard Library), do not include a string’s null terminator in the
length returned. Wide-character versions of standard facili-
ties typically have the prefix w or use wcs instead of str (e.g.,
wostream, wcsncpy, etc.).

NOTE
Although many facilities in the C++ Standard Library
work with character (C-style) strings, the preferred way
to work with strings in C++ is to use the string class
from the C++ Standard Library. The wide-character ver-
sion is wstring.

String literals
String literals are enclosed in double quotes. For example:
char name[] = "Margot";

Long string literals can be broken into quoted strings sepa-

rated by whitespace for style, when needed. For example:
char s[] = "This string is "
"on two lines.";

To specify literals for wide-character strings, you use the pre-

fix L. For example:
wchar_t wide[] = L"Margot";

The compiler allocates enough space for a string, including

its null terminator. An empty string ("") actually has space
reserved for one character: the null terminator. The storage
for a string literal is guaranteed to exist for the life of the pro-
gram, even for a string literal defined locally within a block.
The type of a string literal is an array of const char or wchar_t
elements of static duration.

Compound Types | 23
Pointers
For any type T, there is a corresponding type pointer to T for
variables that contain addresses in memory of where data of
type T resides. T is the base type of a pointer to T. Pointers
are declared by placing an asterisk (*) before the variable
name in a declaration (see “Declaring Variables”). In the fol-
lowing example, i is an int while *iptr is a pointer to i:
int i = 20;
int *iptr = &i;

Normally you can set a pointer of a specific type only to the

address of data of that same type, as just shown. However, in
the case of a pointer to a class, the pointer can also be
assigned the address of an object of some type derived from
that class. This is essential for polymorphism (see “Virtual
Member Functions”). For example, if Circle were derived
from Shape (see “Inheritance”), we could do the following:
Circle c;
Shape *s = &c;

Pointer dereferencing
Dereferencing a pointer yields what the pointer points to. To
dereference a pointer, you precede it with an asterisk in an
expression, as shown in the commented lines below:
int i = 20;
int *iptr = &i;
int j;
int k = 50;

j = *iptr; // This sets j to i.

*iptr = k; // This sets i to k;

Pointer arithmetic
Pointers in expressions are evaluated using the rules of
pointer arithmetic. When an operator for addition, subtrac-
tion, increment, or decrement is applied to a pointer p of
type T, p is treated as an array of type T. As a result, p + n

24 | C++ Pocket Reference

points to the nth successive element in the array, and p – n
points to the nth previous element. If n is 0, p + n points to
the first element in the array. So, if T were a type with a size
of 24 bytes, p += 2 would actually increase the address
stored in p by 48 bytes.
Pointer arithmetic illustrates the close relationship between
pointers and arrays in C++. However, pointers and arrays do
have a fundamental difference: whereas a pointer can be mod-
ified to point to something else, an array cannot be changed
to point away from the data it was created to reference.

Void pointers
Pointers of type void are permitted to point to data of any
type. For example:
Circle c(2.0);
void *p;

p = &c; // c is a circle.

When assigning a void pointer to a pointer of some other

type, an explicit cast is required. For example:
Circle *c;

c = static_cast<Circle *>(p);

Void pointers cannot be dereferenced or used with pointer

arithmetic.

Null pointers
Pointers of any type can be assigned the value 0, which indi-
cates that the pointer points to nothing at all. A pointer with
the value 0 is called a null pointer. You should never derefer-
ence a null pointer.

Function pointers
A function pointer is a pointer that points to a function. Its
type is related to the signature of the function to which it

Compound Types | 25
points. For example, the following defines a function named
addOne, then defines inc as a pointer to a function that takes
a reference to an int as a parameter and returns void. inc is
then set to addOne, which has that same signature:
void addOne(int &x)
{
x += 1;
}

void (*inc)(int &x) = addOne;

The last line could also be written as shown below (using the
address-of operator, &, before addOne):
void (*inc)(int &x) = &addOne;

Parentheses are needed around inc so that the asterisk is

associated with the name of the pointer, not the type. Once a
function pointer points to a function, it can be used to
invoke the function, as follows:
int a = 10;

inc(a); // This adds 1 to a.

The last line could also be written as shown below (using the
indirection operator, *, before the pointer):
(*inc)(a);

Pointers to Members
Pointers to members are like alternative names for class
members (see “Classes, Structs, and Unions”). For example,
assume that class X has a member of type int called data:
int X::*p = &X::data;
X object;
X *objptr = new X;

int i = object.*p;
int j = objptr->*p;

26 | C++ Pocket Reference

This sets i to the value of data in object, and j to the value
of data in the object addressed by objptr.

References
References are used to provide alternative names for vari-
ables. They are declared by placing an ampersand (&) before
the variable name in a declaration. For example:
int i = 20;
int &r = i;

Because a reference always has to refer to something, refer-

ences must be initialized where they are defined. Therefore, a
reasonable way to think of a reference is as a constant
pointer. Once initialized, the reference itself cannot be made
to refer to anything else; however, the variable or object to
which it refers can be modified. Operations applied to the
reference affect the variable or object to which the reference
refers. For example:
int i = 20;
int &r = i;

r++; // This increments i.

Normally you can set a reference of a specific type to a vari-

able of that same type, as just shown. However, in the case of
a reference to a class, the reference can also refer to an object
of some type derived from that class. Therefore, like point-
ers, references support polymorphic behavior (see “Virtual
Member Functions”). For example, if Circle were derived
from Shape (see “Inheritance”), you could do the following:
Circle c;
Shape &s = c;

Reference parameters
A common use of references is with parameters for func-
tions. References allow changes to parameters to be reflected
in the caller’s environment. For example:

Compound Types | 27
 void xchg(int &x, int &y)
{
int t = x;
x = y;
y = t;
}

Using the definition above, you could swap two integers a

and b by doing the following:
xchg(a, b);

If x and y were not references in the definition of xchg, the

contents of x and y would be swapped within the function,
but the contents of a and b would be unchanged when the
function returned.

References as l-values
References are also often used in C++ as return values for
functions. This allows the return value of a function to be
used as an l-value, which is a value that can appear on the
left side of an assignment.

Class Types
The class types of C++ are classes, structs, and unions (see
“Classes, Structs, and Unions”).

Type Conversions and Definitions

In C++ you can convert a value of one type into a value of
another type. Such an action is called a type conversion. You
can also define your own type names using the typedef key-
word.

Type Conversions
Type conversions are performed when you use a cast explic-
itly (see “Casts and Runtime Type Information”), and at
times implicitly by the compiler. For example, the compiler

28 | C++ Pocket Reference

converts a type implicitly when the types in a binary opera-
tion are not the same. A compilation error occurs if no con-
version is possible.

Implicit conversions
Implicit conversions occur between C++’s arithmetic types,
between certain pointer types (see “Pointers”), and between
user-defined types and others. The implicit conversion of
arithmetic types and pointer types in binary operations is car-
ried out by converting the smaller or less precise type to the
larger or more precise one. Booleans, characters, and inte-
gers smaller than an int are first converted to an int using
integral promotion. When an integer and a floating point
appear in the same operation, the integer is converted to the
floating-point type.

Preservation of values
The implicit conversion of arithmetic types is performed in
such a way as to preserve the original values of the entities
being converted whenever possible. However, there are many
situations in which surprising results can occur. For exam-
ple, a compiler may not warn about conversions from wider
or more precise types to smaller or less precise ones (e.g.,
from long to short, or double to float), in which a wider
value may not be representable in the smaller type. In addi-
tion, the conversion from an unsigned type to a signed one
can result in a loss of information.

User-defined conversions
You can specify explicit conversions for user-defined types by
defining user-defined conversion operators (see “Overload-
ing Operators”). For example, the following user-defined
conversion operator, operator double, converts an Account
object to a double:
class Account
{
public:

Type Conversions and Definitions | 29

 Account(double b)
{
balance = b;
}

operator double( )
{
return balance;
}
...
private:
double balance;
};

This user-defined conversion operator allows you to use a

value of type Account where you might otherwise use a double:
Account account(100.0);
double balance = account;

When C++ sees the assignment of an Account value to a

double variable, it invokes operator double to perform the
conversion.

Converting constructors
A constructor that has a single parameter and is not declared
using explicit can be used by the compiler to perform
implicit conversions between the type of the parameter and
the class type. For example:
class Account
{
public:
Account(double b)
{
balance = b;
}
...
private:
double balance;
};

The constructor in this class allows you to do the following,

for example:
Account account = 100.0;

30 | C++ Pocket Reference

Type Definitions
Frequently it is useful to provide an alternative name for
types that have long or otherwise unwieldy names. This is
accomplished using typedef.
To define a new name for a type, you use the keyword
typedef followed by the old type, then the new type. The fol-
lowing example defines uint32 to mean unsigned long:
typedef unsigned long uint32;

uint32 value32bit;

This illustrates using typedef to define your own sized-

integer type (e.g., int8, int16, int32, etc.). Some compilers
define _ _int8, _ _int16, and so forth; typedef provides a
way to use types like these with any compiler. Another
common use of typedef is in providing alternative names
for parameterized types, which tend to be long, when work-
ing with the Standard Template Library. For example:
typedef map<int, string> IntStringMap;

IntStringMap m;

Lexical Elements
At the most fundamental level, a C++ program consists of
individual lexical elements called tokens. Tokens are delin-
eated by whitespace (spaces, newlines, tabs, etc.), or can be
formed once the start of another token is recognized, as
shown below:
ival+3

This stream actually consists of three tokens even though

there is no whitespace. The tokens are ival, +, and 3. In the
absence of whitespace, the compiler forms tokens by con-
suming the longest possible token as it scans from left to
right.

Lexical Elements | 31
Tokens are passed to the parser, which determines if a
stream of tokens has the correct syntax. Tokens together
form more complex semantic constructs, such as declara-
tions, expressions, and statements that affect the flow of
execution.

Comments
Comments are notes written in the source code for develop-
ers; they are ignored completely by the compiler. The prepro-
cessor converts each comment to a single space before the
compiler ever gets the chance to see it.
A comment is any block of text enclosed between /* and */,
or following // on a single line. Comments of the first form
cannot be nested within one another. They usually span mul-
tiple lines. For example:
/* This comment has more than one line.
Here is another part of the comment.*/

Comments of the second form are useful for short explana-

tions that do not occupy more than a single line. For example:
z = MIN(x, y); // z is the smallest.

Once a single-line comment begins, it occupies the remain-

der of the line. There is no way to end the comment before
this.

Identifiers
Identifiers in C++ are sequences of characters that are used
for names of variables, functions, parameters, types, labels,
namespaces, and preprocessor macros. Identifiers may con-
sist of letters, digits, and underscores, but they must not
begin with a digit. For example, the following are all legal
C++ identifiers:
i addressBook Mgr item_count
ptr2 NAME_LENGTH class_ showWindow

32 | C++ Pocket Reference

The following rules apply to identifiers:
• Identifiers are case-sensitive, and they must not be one of
the C++ reserved words (see “Reserved Words”).
• Identifiers that begin with an underscore are reserved for
implementations of the language.
• Although C++ imposes no limit on the size of identifi-
ers, your compiler and linker will have size limits that
you should consider in practice.

NOTE
There is no one stylistic convention for identifiers upon
which everyone agrees. One common convention, howev-
er, is to use lowercase characters to begin names for local
variables, data members, and functions. Uppercase char-
acters are then used to begin the names of types,
namespaces, and global variables. Names processed by
the preprocessor are written entirely in uppercase. Names
of parameters in macros are written entirely in lowercase.

Reserved Words
C++ defines a number of keywords and alternative tokens,
which are sequences of characters that have special meaning
in the language. These are reserved words and cannot be
used for identifiers. The reserved words of C++ are listed
below:
and and_eq asm
auto bitand bitor
bool break case
catch char class
compl const const_cast
continue default delete
do double dynamic_cast
else enum explicit
export extern false
float for friend
goto if inline

Lexical Elements | 33
 int long mutable
namespace new not
not_eq operator or
or_eq private protected
public register reinterpret_cast
return short signed
sizeof static static_cast
struct switch template
this throw true
try typedef typeid
typename union unsigned
using virtual void
volatile wchar_t while
xor xor_eq

Literals
Literals are lexical elements that represent explicit values in a
program. C++ defines many types of literals. Each is
described under its respective type in “Fundamental Types.”

Operators
An operator is used to perform a specific operation on a set
of operands in an expression. Operators in C++ work with
anywhere from one to three operands, depending on the
operator.

Associativity
Operators may associate to the left or right. For example,
assignment operators (=, +=, <<=, etc.) associate to the right.
Therefore, the following:
i = j = k

actually implies:
i = (j = k)

On the other hand, the operator for addition (+) associates to

the left. Therefore, the following:
i + j + k

34 | C++ Pocket Reference

actually implies:
(i + j) + k

Precedence
Operators also have an order, or precedence, by which
expressions that contain them are evaluated. Expressions
containing operators with a higher precedence are evaluated
before those containing operators with a lower precedence.
You can use parentheses around an expression to force
grouping. Even when not essential, it’s best to use parenthe-
ses in expressions to document your intentions. The number
of operators in C++ often makes their precedence difficult to
remember.
Table 4 lists the operators of C++ from highest precedence
to lowest and describes how each operator associates. Each
section contains operators of equal precedence. The table
also describes the behavior of each operator when used with
the intrinsic types of C++. For most operators, C++ lets you
define additional behaviors for your own types (see “Over-
loading Operators”).

Table 4. Operators
Operator Description Associates
:: Scope resolution No
[] Array subscript Left
. Member selection Left
-> Member selection Left
() Function call Left
() Value construction No
++ Postfix increment No
-- Postfix decrement No
typeid Type information No
*_cast C++ cast No

Lexical Elements | 35
Table 4. Operators (continued)
Operator Description Associates
sizeof Size information No
++ Prefix increment No
-- Prefix decrement No
~ Bitwise NOT No
! Logical NOT No
- Unary minus No
+ Unary plus No
& Address-of No
* Indirection No
new Allocate No
new[] Allocate No
delete Deallocate No
delete[] Deallocate No
() C-style cast Right
.* Pointer-to-member selection Left
->* Pointer-to-member selection Left
* Multiply Left
/ Divide Left
% Modulo (remainder) Left
+ Add Left
- Subtract Left
<< Shift left Left
>> Shift right Left
< Less than Left
<= Less than or equal to Left
> Greater than Left
>= Greater than or equal to Left
== Equal to Left
!= Not equal to Left

36 | C++ Pocket Reference

Table 4. Operators (continued)
Operator Description Associates
& Bitwise AND Left
^ Bitwise XOR Left
| Bitwise OR Left
&& Logical AND Left
|| Logical OR Left
?: Conditional expression Right
= Simple assignment Right
*= Multiply and assign Right
/= Divide and assign Right
%= Modulo and assign Right
+= Add and assign Right
-= Subtract and assign Right
<<= Shift left and assign Right
>>= Shift right and assign Right
&= AND and assign Right
^= XOR and assign Right
|= OR and assign Right
throw Throw exception Right
, Sequence Left

Additional information about the behaviors of the operators

in C++ is summarized in the following sections.

Scope resolution (::)

The scope resolution operator is used to specify a scope (see
“Scope”). For example, the following invokes a static mem-
ber function of a class called Dialog:
dialog = Dialog::createDialog( );

Lexical Elements | 37
The scope operator can also be used without a scope name to
specify file (global) scope. For example:
::serialize(i);

This ensures that the global function serialize is invoked,

even if serialize has been declared within the local scope.

Array subscript ([ ])
The array subscript operator is used to access individual ele-
ments of arrays or memory referenced by pointers. For exam-
ple:
tmp = table[0];

This assigns the first element in an array called table to tmp.

The expression between the brackets indicates the element.

Member selection (. and ->)

Member selection operators are used to specify members of
objects (see “Classes, Structs, and Unions”). You use the dot
form with objects and the arrow form with pointers to
objects. For example:
object.f( );

This invokes member function f of an object called object.

The following illustrates the arrow form:
objptr->f( );

This invokes member function f for an object that is

addressed by the pointer objptr.

Function call (())

The function call operator, which is ( ), is used to invoke a
function. For example:
f(a, b);

This invokes a function called f with two arguments, a and b.

38 | C++ Pocket Reference

Value construction (())
The value construction operator, which is also ( ), is used to
create an instance of a type. For example:
g(Circle(5.0));

This constructs a temporary object that is an instance of the

Circle class, which is passed to g.

Postfix increment and decrement (++, ––)

The postfix increment and decrement operators increment or
decrement an operand, but the value of the operand within
its expression is the value prior to modification. For example:
void count( )
{
static int i = 0;

if (i++ == 0)
{
// This is the first time called.
}
}

The value of i prior to being incremented is tested for equal-

ity with 0. Because i is initialized to 0, the test is true during
the first invocation of the function.

typeid
The typeid operator gets runtime type information for an
operand. See “Casts and Runtime Type Information” for a
complete description of this operator.

C++ cast
Type cast operators specific to C++ are dynamic_cast,
static_cast, const_cast, and reinterpret_cast. See “Casts
and Runtime Type Information” for a complete description
of these operators.

Lexical Elements | 39
sizeof
The sizeof operator gets the size of its operand. For example:
size_t s = sizeof(c);

This initializes s to the size of c. The operand may be an

expression or type. The result is an integer of type size_t.

Prefix increment and decrement (++, ––)

The prefix increment and decrement operators increment or
decrement an operand. The value of the operand within its
expression is the value after modification. For example:
void count( )
{
static int i = 0;

if (++i == 1)
{
// This is the first time called.
}
}

The value of i after being incremented is tested for equality

with 1. Because i is initialized to 0, the test is true during the
first invocation of the function.

Bitwise NOT (~)

The bitwise NOT operator computes the bitwise comple-
ment of its operand. For example:
unsigned char bits = 0x0;
bits = ~bits;

This assigns 0xFF back into bits, assuming a character is

eight bits. The operand must be one of the Boolean, charac-
ter, or integer types of C++.

Logical NOT (!)

The logical NOT operator reverses the truth of its operand; it
yields false if its operand is true (nonzero) and true if its
operand is false. For example:

40 | C++ Pocket Reference

 bool done = false;

while (!done)
{
// Set done to true when finished.
}

This loop is repeated until something in the loop sets done to

true. The result of the logical NOT operator is a bool.

Unary minus and plus (–, +)

The unary minus and plus operators compute the negative
and positive values their operands. For example:
i = -125;
j = +273;

Because the unary plus operator simply returns the value of

its operand (promoted to an int), it is seldom used.

Address-of (&)
The address-of operator gets the address at which its oper-
and resides in memory. For example:
Circle c;
Circle *p = &c;

This assigns the address of c to the Circle pointer p. The

address is a pointer derived from the type of the operand.

Indirection (*)
The indirection operator dereferences a pointer and gets the
value that it addresses. For example:
int i;
int *p = new int;

*p = 5;
i = *p;

This assigns the value 5 to i. The type of the result is the type
from which the pointer is derived. The operand must be a
pointer.

Lexical Elements | 41
Allocate and deallocate
The C++ memory management operators are new, new[],
delete, and delete[]. They allocate and reclaim memory on
the heap. See “Memory Management” for a complete
description of these operators.

C-style cast (())

The C-style cast operator converts the type of its operand to
a new type (see “C-Style Casts”). For example:
void *p = new int;
*p = 10;
int *q = (int *)p;

This casts p from a void pointer to an int pointer. No run-

time checking is performed to ensure that the cast is legal.

Pointer-to-member selection (.* and –>*)

The .* and ->* operators access a class member via a pointer
to the member. For example:
int X::*p = &X::data;
X object;
X *objptr = new X;

int i = object.*p;
int j = objptr->*p;

This sets i to the value of data in object, and j to the value of

data in the object addressed by objptr. You use the dot form
with objects and the arrow form with pointers to objects.

Arithmetic (*, /, %, +, –)
Arithmetic operators perform multiplication (*), division (/),
modulus (%), addition (+), and subtraction (–) using two
operands. For example:
if (x % 2 == 0)
{
// The integer x is an even number.
}

42 | C++ Pocket Reference

The modulo operator computes the remainder of dividing
the first operand by the second. For example, if dividing an
integer by 2 has no remainder, the integer is even.

Shift left and right (<<, >>)

The shift operators shift bits to the left (<<) or right (>>). For
example:
unsigned char bits = 0x1;
bits = bits << 2;

This assigns 0x4 back into bits. The first operand is the one
shifted; the second is the number of bits to shift. Both oper-
ands must be one of the Boolean, character, or integer types
of C++. These operators are commonly used for insertion
and extraction with I/O streams as well (see “I/O Streams”).

Relational (<, <=, >, >=, ==, !=)

Relational operators compare two operands, yielding true if
the comparison is true and false if the comparison is false.
For example:
for (int i = 0; i < 100; i++)
{
// Do something for each iteration.
}

This loop uses the less-than operator to determine when to

stop looping. The result of using a relational operator is a
bool.

Bitwise AND, XOR, and OR (&, ^, |)

The bitwise AND (&), XOR (^), and OR (|) operators per-
form bitwise operations. Each bit in the first operand is com-
pared with the same bit in the second. For the bitwise AND
operator, if both bits are 1, the corresponding result bit is 1;
otherwise, the bit is 0. For example, the following sets a to
0x0F:
unsigned char a, b = 0x0f, c = 0xff;
a = b & c;

Lexical Elements | 43
For the bitwise XOR operator, if one bit is 0 and the other is
1, the corresponding result bit is 1; otherwise, the bit is 0.
For example, the following sets a to 0xA0:
unsigned char a, b = 0xaa, c = 0x0a;
a = b ^ c;

For the bitwise OR operator, if both bits are 0, the corre-

sponding result bit is 0; otherwise, the bit is 1. For example,
the following sets a to 0x0F.
unsigned char a, b = 0x0f, c = 0x0a;
a = b | c;

The operands of these bitwise operators must both be one of

the Boolean, character, or integer types of C++.

Logical AND and OR (&&, ||)

The logical AND (&&) and OR (||) operators combine two
operands to evaluate their truth. The logical AND operator
yields true only if both operands are true (nonzero); other-
wise, it yields false. Both conditions in the following exam-
ple must be true in order to execute the block containing the
comment:
int i = 10;
int *p = &i;

if (p != NULL && *p < 100)

{
// Do something if both are true.
}

The logical OR operator yields true if either operand is true;

otherwise, it yields false. For example:
bool doneWithTask1 = false;
bool doneWithTask2 = false;

while (!doneWithTask1 || !doneWithTask2)

{
// One of the tasks has not finished.
}

44 | C++ Pocket Reference

Either condition can be true for the block containing the
comment to be executed.
For both operators, if a result can be determined from the
first operand alone, the second operand is not evaluated.
When these operators are overloaded, both operands are
always evaluated.

Conditional expression (?:)

The conditional expression operator uses the value of one
operand to determine whether to evaluate the second or
third operand. For example:
i = (p != NULL) ? *p : -1;

If the first operand is true, the result is the second operand;

otherwise, the result is the third. The first operand appears
before the question mark (?); the second and third operands
are separated by a colon (:).

Simple and compound assignments

(=, *=, /=, %=, +=, –=, <<=, >>=, &=, |=, ^=)
Assignment operators assign the value of one operand to
another. For example:
i = (j + 10) * 5;

This is the simplest form of assignment; the second operand

is simply evaluated and stored into the first. The other
assignment operators perform compound assignments. For
example:
i += 5;

This adds 5 to i and assigns the result back to i. Therefore, it

has the same effect as the following but avoids the need for i
to be evaluated twice:
i = i + 5;

After any assignment, the value of the expression is the value

that was assigned. This allows assignments to be chained
together, as follows:

Lexical Elements | 45
 a = b = c;

Exception (throw)
The throw operator is used to throw an exception. See
“Exception Handling” for a complete description of this
operator.

Sequence (,)
The sequence operator, which is a comma, evaluates two
operands from left to right. The value of the expression
becomes the value of the last operand. For example:
for (i = 0, j = 10; i < 10; i++, j--)
{
// Increase i while making j smaller.
}

In this case, the result of i=0, j=10 is 10. However, both

assignments are performed; both variables are initialized.

Expressions
An expression is something that yields a value. Nearly every
type of statement uses an expression in some way. For exam-
ple, the declaration below uses an expression for its initial-
izer:
int t = (100 + 50) / 2;

The simplest expressions in C++ are just literals or variables

by themselves. For example:
1.23 false "string" total
More interesting expressions are formed by combining liter-
als, variables, and the return values of functions with various
operators to produce new values. These can then be used in
expressions themselves. For example, the following are all
C++ expressions:
i->getValue( ) + 10
p * pow(1.0 + rate,(double)mos))
new char[20]
sizeof(int) + sizeof(double) + 1

46 | C++ Pocket Reference

Scope
A name can be used only within certain regions of a pro-
gram. These regions define its scope. The scope of a name is
based on where, and to some extent how, you declare it.
Most names have one of four scopes. Labels and prototype
parameters have their own special scopes.

Local Scope
A name has local scope when it is declared inside of a block.
A block is a compound statement that begins with a left
brace ({) and ends with a right brace (}). For example:
void f( )
{
int i = 10;
...
}

In this example, i has local scope. A name with local scope is

visible only within its block.

Class Scope
A name has class scope when it is declared within the con-
fines of a class and does not have local scope. For example:
class Event
{
public:
enum Type
{
keyDown,
...
};
...
Type getType( ) const
{
return type;
}
...
private:

Scope | 47
 Type type;
...
};

Here, Type, keyDown, getType, and type all have class scope. A
name with class scope is visible inside of the class in which it
is declared, and outside of the class using a selection or scope
operator, depending on what the name represents. To use a
name with class scope outside of its declaring class, the
access level for the name must also allow access to it (see
“Access Levels for Members”).

NOTE
A name declared within a block that is within a class has
a scope local to that block, and does not have class scope.

Namespace Scope
A name declared inside of a namespace has namespace scope
(see “Namespaces”). For example:
namespace Aviation
{
const double NMPerSM = 0.826201;
}

This places NMPerSM, used for converting nautical to statute

miles, in a namespace called Aviation. A name with
namespace scope is visible inside of the namespace in which
it is declared, and outside of the namespace with the scope
operator.

File Scope
A name that is not declared in a block, class, or namespace
has file scope. A name with file scope can be used anywhere
within a file after the point where the name is declared. A
name for a variable, object, or function with file scope that
has not been declared using the keyword static is called a

48 | C++ Pocket Reference

global. This is because it can be used potentially anywhere in
a program.

Other Scopes
Labels (see “Jump Statements”) and the parameters in proto-
types have their own scopes. The scope of a label is the func-
tion in which the label is used, even if the label is defined
inside of a block. This allows jumping into or out of a block.
The scope of a prototype parameter goes to the end of the
prototype; however, a parameter cannot be used to define
default values for other parameters.

Enclosing Scopes
A name is visible in any scope that its declaring scope
encloses. For example:
const double NMPerSM = 0.826201;

double convertToSM(double nm)

{
return nm * NMPerSM;
}

Since the local scope of convertToSM is enclosed by the file

scope in which NMPerSM is declared, NMPerSM is visible within
convertToSM.
A name declared in one scope hides a declaration of the same
name in the scope that encloses the first scope. For example:
const double NMPerSM = 0.826201;

double convertToSM(double nm)

{
const double NMPerSM = 0.826;

return nm * NMPerSM;
}

Here, the local NMPerSM hides the declaration of NMPerSM with

file scope. As a result, the value used in the computation is
0.826, not 0.826201.

Scope | 49
 NOTE
In general, hiding names leads to errors that are difficult
to discover. Therefore, this sort of thing should be avoid-
ed as much as possible.

Don’t confuse hiding function names with overriding and

overloading functions (see “Virtual Member Functions” and
“Overloading Functions”).

Declarations
A name must be declared within the necessary scope before it
can be used. A declaration is often a definition as well. A
name may be declared in multiple places throughout a pro-
gram; however, it must be defined only once. If multiple dec-
larations for a name exist, all must be identical.
The declaration of a function is a definition when you pro-
vide a body for the function; the declaration of a variable is a
definition whenever storage is allocated. In short, a declara-
tion is a definition except in the following situations:
• A variable is declared using the keyword extern (see
“Storage Classes”), and no initializer is provided.
• The declaration is for a static data member; static data
members are defined outside of their class.
• The declaration introduces a class name with no defini-
tion, in other words a forward declaration.
• The declaration is a prototype for a function; prototypes
have no body.
• The declaration is a typedef statement, which declares a
synonym for an existing type.

50 | C++ Pocket Reference

Declaring Variables
Declarations for variables introduce names that refer to data.
They contain the following, in order: an optional storage
class, optional qualifiers, a type, and a comma-delimited list
of one or more names to declare. For example:
int i, j, k;
char buffer[80];
static int counter, a;
volatile float x;

Data members of classes are declared in a similar manner

(see “Classes, Structs, and Unions”); however, they can only
have the storage classes static (see “Static data members”)
and mutable (see “Mutable data members”).
Declarations for variables may appear anywhere within a
block, not just at the start. This makes code like the follow-
ing common in C++:
void spin(int n)
{
cout << "Spinning" << endl;

for (int i = 0; i < n; i++)

;
}

In this example, the variable i is declared within the for

loop, as opposed to at the start of the function in which the
loop appears.

Pointer variables
Declarations for pointers follow the same rules as for other
types of variables, except you must be sure to precede each
name with an asterisk (*). For example:
int *p, *q, *r;

Special situations arise when the qualifier const is used in the

declaration of pointer variables (see “Qualifiers”).

Declarations | 51
Initialization
Optionally, you can initialize variables using an initializer
where they are defined. Variables declared using the key-
word const must be initialized. When an initializer appears,
it must evaluate to the correct type. For example:
bool done = false;
static const int max = 100;
int timers[] = {5, 5, 5};
int *p, *q = 0, *r;

Arrays are initialized using initializer lists, as shown here for

the timers[] array, and in “Initializer lists for arrays.” Vari-
ables that are instances of class types are initialized via con-
structors (see “Constructors”). Variables that are instances of
classes that do not explicitly define a constructor may be ini-
tialized using initializer lists similar in syntax to those used
for arrays. For example:
Rectangle r = {0.0, 0.0, 3.0, 4.0};

If the class defines any constructor explicitly, this form of ini-

tialization cannot be used.

Declaring Functions
At their simplest, declarations for functions consist of the fol-
lowing: a return type, a name, and a comma-delimited list of
zero or more parameters enclosed by parentheses. Names are
not required for parameters, but they serve as good docu-
mentation. For example:
void xchg(int &x, int &y);

Functions may be declared using the keyword static. Types

for parameters and return values may be qualified using
const and volatile. For example:
static char *format(const char *s);

You can use references for parameters to have changes to the

parameters reflected in the caller’s environment, as shown
for xchg (see “References”).

52 | C++ Pocket Reference

Function definitions
Declarations for functions are called prototypes. They do not
define a function; they simply inform the compiler of your
intention to define and use it. To define a function, you spec-
ify a body for it, as follows:
void xchg(int &x, int &y)
{
int t = x;
x = y;
y = t;
}

This function has a return type of void. If the return type is

anything other than void, the function must use a return
statement (see “Jump Statements”) to return a value suitable
for the function’s return type. Functions that return void can
use a return statement without a value.

Default arguments
Default arguments can be specified for the parameters of
functions. You do this by setting a parameter equal to its
default value in the function declaration, as shown below:
void isTempOK(const int t,
const int low = 20,
const int high = 50);

A default argument is used for a parameter when nothing is

specified for it in an invocation of the function. For example,
assuming a temperature declared as temp, the following uses
the default arguments 20 for low and 50 for high:
if (!isTempOK(temp))
{
// Do something if too low or high.
}

If a function is declared with default arguments, the parame-

ters with defaults must appear last in the parameter list. Val-
ues are assigned to parameters from left to right. Remaining
parameters are then assigned their default values. Therefore,

Declarations | 53
there is no way to provide an argument for a parameter after
any that have assumed default values. So, for example, the
following invocation passes 30 to low:
if (!isTempOK(temp, 30))
{
// Do something if too low or high.
}

There is no way to pass 30 for high without also using an

explicit argument for low.

Inline functions
An inline function is a function whose body is substituted
directly at every point in a program where the function is
called, as opposed to generating a call using the stack and a
single copy of the function. To make a function inline, you
precede its definition with the keyword inline, as follows:
inline void xchg(int &x, int &y)
{
int t = x;
x = y;
y = t;
}

By inlining a function, you avoid the overhead required to set

up each function call. However, since inlining requires that
separate copies of the same function be inserted in poten-
tially numerous places throughout your program, inlining is
typically used only for very small functions. Because an inline
function must be defined within every file in which it is used,
inline functions are usually defined in header files.

NOTE
The inline keyword is just a request to the compiler to
inline a function. The compiler makes the ultimate
decision.

54 | C++ Pocket Reference

Storage Classes
The storage classes of C++ are static, extern, mutable, auto,
and register. The storage class of a variable or data member
determines its lifetime and linkage; the storage class of a
function determines its linkage and other treatment. If no
storage class is specified, the following rules apply:
Local variables
Have the storage class auto by default.
Global variables
Have file scope, unless declared using the keyword
extern elsewhere.
Data members
There is a separate instance of data members for each
instance of their class.
Nonmember functions
Have the storage class extern by default.
Member functions
Must be invoked through an instance of their class.

static
The following list summarizes the meanings of the storage
class static in various contexts:
Local variables
Persist between executions of their enclosing block. They
are constructed only once, or never if their declaration is
never reached.
Global variables
Have file scope. Unnamed namespaces are a better way
to achieve this.
Data members
A single instance of a static data member is shared by all
instances of its class. Static data members are initialized
before main is called.

Declarations | 55
Nonmember functions
Have file scope. Unnamed namespaces offer a better way
to achieve this.
Member functions
Can be called without an instance of their class, but can-
not themselves access members of their class that are not
also declared static.

extern
When applied to a global variable or to a nonmember func-
tion, the extern storage class specifies that the variable or
function is defined in another source file. Nonmember func-
tions have the storage class extern by default.

mutable
The mutable storage class can be applied only to a class data
member. It specifies that the member can be modified even
though its containing object has been declared using the key-
word const.

auto
The auto storage class instructs the compiler to allocate stor-
age automatically for a variable on the stack each time a
block is entered. Local variables are automatic by default;
therefore, auto is rarely used.

register
The register storage class requests that the address of a vari-
able be stored in a machine register for better performance.
The register storage class is only a request; the compiler
decides whether or not to use a register.

56 | C++ Pocket Reference

Qualifiers
The qualifiers of C++ are const and volatile. In some situa-
tions they are used together.

const
The const qualifier keeps the entity it qualifies from being
modified, except for data members declared using keyword
mutable (see “Storage Classes”). The following rules apply:
Local variables
Once initialized, cannot be modified. They must be ini-
tialized.
Global variables
Once initialized, cannot be modified. They must be ini-
tialized.
Data members
Once initialized, cannot be modified. They must be ini-
tialized when an object is constructed.
Nonstatic member functions (where const appears at the end
of the signature)
Cannot modify nonstatic data members of their class
except those declared using mutable. The member func-
tions are allowed to be invoked through const instances
of their class.
Function parameter
Cannot be changed by the function.
Return value of a function
Can be used only where a const value is permitted.
Pointer declarations use a syntax with const that depends on
what you want to protect from modification. If you want to
protect what a pointer addresses from being modified, you
use a declaration like the following:
int i = 100, j = 200;
const int *p = &i;

Declarations | 57
 *p = j; // This is an error.
p = &j; // This is OK.

This is the most common use. If you want to protect the

pointer itself from being modified, you use a declaration in
which const is written after the asterisk (*), as follows:
int i = 100, j = 200;
int *const p = &i;

*p = j; // This is OK.
p = &j; // This is an error.

In the event that you want to protect both the data and the
pointer from being modified, you use a declaration that is a
combination of the two presented previously, as in this
example:
int i = 100, j = 200;
const int *const p = &i;

p = &j; // This is an error.

*p = j; // This is an error.

volatile
The volatile qualifier informs the compiler that a variable,
data member, or parameter may be modified unexpectedly,
be it by another process, the hardware, or something else.
As a result, the compiler avoids optimizations that could
conflict with changes happening asynchronously to the
entity (e.g., normally a compiler will skip accessing memory
again if it needs a value that was just obtained in the previ-
ous instruction).
When applied to member functions, volatile has a slightly
different shade of meaning. When you create a volatile object
of a class, the compiler will only allow you to call member
functions that have also been declared volatile. Thus, you
should apply the volatile qualifier to member functions in
cases when you expect to create volatile objects of a class,
and even then only apply volatile to those functions you
intend to invoke on those volatile objects.

58 | C++ Pocket Reference

Statements
There are many types of statements in C++. Some simply
evaluate expressions, while others change the order in which
statements are executed in the future.

Expression Statements
An expression statement is an expression followed by a single
semicolon (;). Expression statements cause an expression to
be evaluated. Side effects, such as an assignment to a vari-
able, are completed before the next statement is executed.
For example:
a = 10;

Null Statements
A null statement is written as a semicolon (;). Null state-
ments are useful when the syntax of C++ requires a state-
ment but you don’t need anything performed. For example:
void spin(int n)
{
for (int i = 0; i < n; i++)
;
}

This loop simply counts to a specified value, as might be

required to insert a delay in a real-time system (assuming the
compiler doesn’t optimize the loop away altogether).

Compound Statements
A compound statement is a group of zero or more statements
beginning with a left brace ({) and ending with a right brace
(}). For example:
while (true)
{
// Start of a compound statement.
...

Statements | 59
 if (!done)
{
// Another compound statement.
}
else
{
// Another compound statement.
}
}

Compound statements are often called blocks. A block

defines a region that has its own local scope.

Iteration Statements
Iteration statements cause a statement or block to be exe-
cuted repeatedly. There are three types of iteration state-
ments in C++: while, do, and for.

while
A while loop repeats a statement or block as long as an
expression (which can be a declaration) evaluated at the top
of the loop is true. For example:
char ch = 'y';

while (ch == 'y')

{
// Do something to be repeated.
...

cout << "Do it again (y or n)? ";

cin >> ch;
}

This repeats a block as long as ch is 'y'. The body of the loop

is never executed if the expression at the top of the loop is
false when the loop is first encountered.

do
A do loop repeats a statement or block as long as an expres-
sion evaluated at the bottom of the loop is true. For example:

60 | C++ Pocket Reference

 char ch;

do
{
// Do something to be repeated.
...

cout << "Do it again (y or n)? ";

cin >> ch;
} while (ch == 'y');

This repeats a block as long as ch is 'y'; however, the body

of the loop is executed at least once because the condition for
looping is evaluated at the end of each iteration.

for
A for loop is similar to a while loop, but additional mecha-
nisms are provided for initializing the loop and making
adjustments after each of its iterations. For example:
// Prevent warnings in Visual C++.
#pragma warning(disable:4786)

typedef map<int, string> IntStringMap;

IntStringMap m;
char s[4];

for (int i = 0; i < 10; i++)

{
s[0] = 'a' + i; s[1] = 'b' + i;
s[2] = 'c' + i; s[3] = '\0';

m.insert(IntStringMap::value_type(i,
string(s)));
}

The key to for loops is understanding the statements that go

within the parentheses following the for keyword. The state-
ment before the first semicolon is the statement executed to
initialize the loop. Before each iteration, including the first,
the expression between the two semicolons is evaluated. If the
expression is true, the loop body is executed; otherwise, the

Statements | 61
loop terminates. After each iteration, the rightmost expres-
sion is evaluated, and the cycle is repeated. for loops can con-
tain more complicated expressions as well. For example:
void upperString(char *t, const char *s)
{
for (; *s != '\0'; *(t++) = toupper(*(s++)))
;

*(t++) = '\0';
}

This function uses a for loop to translate string s to upper-

case and copy it to t. The function assumes that storage has
already been allocated for t. A null statement is used for ini-
tialization since s and t are already initialized when the func-
tion starts. A null statement is also used for the loop body.

NOTE
A name declared in a for initialization statement is visi-
ble until the end of the for loop.

Selection Statements
Selection statements execute a different statement or block
based on the result produced by an expression. There are
two types of selection statements in C++: if and switch.

if
An if statement evaluates an expression (which may be a
declaration) and uses the result to determine which of up to
two statements or blocks to execute next. For example:
if (i > 0 && i < 100)
{
// Do something when within range.
}
else
{
// Do something when not in range.
}

62 | C++ Pocket Reference

If the expression is true, the statement or block immediately
after the terminating parentheses of the if section is exe-
cuted; otherwise, the statement or block after else is exe-
cuted. If the else clause is omitted, nothing is performed
when the expression is not true. When if statements are
nested within one another, else clauses associate with the
nearest if.

switch
A switch statement selects one of several sections of code to
execute based on the value of a controlling expression. For
example:
switch (type)
{
case keyDown:
// Do something for a key down.
...
break;

case keyUp:
// Do something for a key up.
...
break;
...
default:
// Handle anything not handled.
...
}

Each section is identified by the keyword case followed by an

expression that must evaluate to a distinct, constant, integral
value at compile time. At runtime, execution branches to the
section identified by the value matching that of the control-
ling expression, and continues from that point onward. A
break statement (see “Jump Statements”) is used at the end
of each case to prevent the code associated with all subse-
quent cases from being executed. An optional default case
can be provided to handle the situation when the value of the
controlling expression does not match any of the cases.

Statements | 63
Jump Statements
Jump statements jump unconditionally to a different state-
ment. There are four types of jump statements in C++:
break, continue, goto, and return.

break
A break statement is used to jump out of an innermost loop
or switch statement. For example:
for (;;)
{
if (done)
break;

// When we're done, set done to true

// so we break on the next iteration.
...
}

This is a for loop that has no condition specified for stop-

ping it. A break statement is used to break out of the loop
when done is true.

continue
A continue statement is used to jump to the start of an inner-
most enclosing loop. For example:
while (!done)
{
// If we need to skip the end portion
// of the function, set skip to true.
...

if (skip)
continue;

// This gets skipped when continuing.

...
}

This is a while loop that contains statements to be skipped

when skip has been set to true. A continue statement is used

64 | C++ Pocket Reference

to jump to the start of the loop in order to skip these
statements.

goto
A goto statement jumps to a label that you specify. For exam-
ple:
if (GetLastError( ) != ERROR_SUCCESS)
goto handleError;

// This code gets skipped on an error.

...

handleError:
// Do something to handle error state.

Because goto statements lead to unstructured code, they are

seldom used.

return
A return statement jumps out of a function, and if needed
sets a return value. For example:
double convertToSM(double nm)
{
return nm * NMPerSM;
}

A return statement can be called from anywhere within a

function. The type of the value returned must be, or be con-
vertible to, the return type of the function. Functions that
return void do not require return statements. For example:
void sayHello( )
{
cout << "Hello" << endl;
}

A return is performed automatically when this function com-

pletes. However, a return statement with no value can be
used to jump out of the function before the end of the func-
tion block is reached.

Statements | 65
Namespaces
A namespace defines a named scope. Namespaces are used
to group related names and to avoid clashes between similar
names in different parts of large programs. You declare a
namespace as follows:
namespace Aviation
{
const double NMPerSM = 0.826201;

double convertToSM(double nm);

}

Namespaces can contain anything that you would otherwise

declare outside of a namespace, even other namespaces. In
addition, namespaces are open, which means you can extend
a namespace using multiple declarations. For example, the
following extends the previous declaration of Aviation:
namespace Aviation
{
const double SMPerNM = 1.21036;

double convertToNM(double sm);

}

You use a name outside of its declaring namespace by quali-

fying it with the name of the namespace using the scope
operator (::). For example:
Aviation::convertToSM(20.0);

To define this function outside of the namespace, you do the

following:
double Aviation::convertToSM(double nm)
{
return nm * NMPerSM;
}

Names not declared within a namespace belong to a special

namespace: the global namespace. To avoid qualifying names
everywhere they are used, you use a using declaration or
directive.

66 | C++ Pocket Reference

 NOTE
The C++ Standard Library uses the namespace std to dis-
tinguish between new and pre-standard versions of its fa-
cilities (see “The C++ Standard Library”).

using Declarations
A using declaration lets you use a name from a namespace
without having to qualify it. For example:
namespace Aviation
{
double convertToSM(double nm);
double convertToNM(double sm);
}
...
using Aviation::convertToSM;
...
double sm = convertToSM(20.0);

A using declaration adds a name to the scope in which the

using declaration exists. This means the following:
• A compilation error occurs if the name is declared else-
where in the same scope.
• If the same name is declared in an enclosing scope, the
name in the namespace hides it.

using Directives
A using directive lets you use all names from a namespace
without qualifying them. For example:
namespace Aviation
{
double convertToSM(double nm);
double convertToNM(double sm);
}
...
using namespace Aviation;
...

Namespaces | 67
 double sm = convertToSM(20.0);
double nm = convertToNM(40.0);

A using directive does not add a name to the current scope; it

only makes a name accessible from it. This means the follow-
ing:
• If a name is declared elsewhere within a local scope, the
name in the namespace is hidden once the local declara-
tion occurs.
• A name in the namespace hides the same name declared
in an enclosing scope.
• A compilation error occurs when the same name is made
visible from multiple namespaces, or if a name in the
namespace hides a name in the global namespace.

Unnamed Namespaces
A namespace is unnamed when you omit the name for it in
its declaration. For example:
namespace
{
...
}

The names in an unnamed namespace have file scope. Con-

sidering this, they are like variables or functions declared as
static without declaring each explicitly using the static key-
word. Since there is no way to refer to an unnamed
namespace, a using directive is implied after the namespace
declaration. Unnamed namespaces are the preferred means
of declaring variables and functions with file scope.

Classes, Structs, and Unions

Classes are types that group data and functionality together
into encapsulated, cohesive units. Structs and unions are simi-
lar to classes, but differ in the ways outlined later under

68 | C++ Pocket Reference

“Structs” and “Unions.” Classes, structs, and unions are col-
lectively called class types. You define a class by declaring a set
of data members and member functions for it. For example:
class Account
{
public:
Account(double b);

void deposit(double amt);

void withdraw(double amt);
double getBalance( ) const;

private:
double balance;
};

This class declares one data member (balance) and three

member functions (deposit, withdraw, and getBalance). One
constructor is declared to perform initialization. The key-
words public and private control how various parts of the
class can be accessed (see “Access Levels for Members”).

Declaring Objects
Objects are specific instances of a class. The following
declares an object that is an instance of Account:
Account account(100.0);

This declaration is also a definition because storage is allo-

cated for the object. The rules for declaring and defining
objects essentially follow the rules presented under “Declar-
ing Variables,” except when it comes to initialization (see
“Constructors”).

Accessing Members
To access a member of an object, you use the dot form of the
selection operator (.). For example, the following assigns 500
to the data member i of an object called object:
object.i = 500;

Classes, Structs, and Unions | 69

The following invokes the member function f of object,
passing it two arguments, a and b:
object.f(a, b);

To access a member of an object using a pointer to the

object, you use the arrow form of selection (->). For exam-
ple, the following assigns 500 to the data member i of an
object referenced by objptr:
objptr->i = 500;

The following invokes the member function f of the object

referenced by objptr passing it two arguments, a and b:
objptr->f(a, b);

To access a member from within the class in which it is

defined, or from within a derived class (see “Inheritance”),
usually you specify only the name of the member, unless
accessing the member through the this pointer. By whatever
means you access a member, the compiler either allows or
denies access based on the member’s access level (see
“Access Levels for Members”).

Declaring Data Members

You declare the data members for a class in the class’s defini-
tion. For example:
class Account
{
...
private:
double balance;
};

The rules for declaring and defining data members in a class

are similar to those for variables (see “Declaring Variables”),
with the following exceptions:
• You cannot use an initializer with a nonstatic data mem-
ber where the member is defined. (Static members are
defined outside of the class.)

70 | C++ Pocket Reference

 • The only storage classes allowed for data members are
static and mutable (see “Storage Classes”).
Each instantiation of a class gets its own copy of data mem-
bers, except those declared using the keyword static. The
declaration of a nonstatic data member also defines the data
member.

Static data members

Static data members are shared by all instances of a class.
You declare a static data member using the keyword static.
For example:
class Account
{
public:
...
void deposit(double amt)
{
...
deposits++;
}
...
private:
...
static int deposits;
};

Unlike the declaration of a nonstatic data member, the decla-

ration of a static data member does not define it. You must
define a static data member outside of the class, at which
point you usually initialize it as well. For example:
int Account::deposits = 0;

You qualify a static data member outside of its class with the
name of the class using the scope operator (::). You can refer
to a static data member without an instantiation of the class.
The access level for a static data member does not apply when
defining the member (see “Access Levels for Members”).

Classes, Structs, and Unions | 71

 NOTE
Definitions of static data members belong in the source
file for the class, not in the class’s header file. Otherwise,
the data member ends up being defined more than once
when the header file is included multiple times.

Constant data members

Constant data members cannot be modified at any point in
an object’s lifetime. You declare a constant data member
with the keyword const. For example:
class Account
{
...
private:
...
const double minBalance;
};

Because constant data members cannot be modified, they

must be initialized using an initializer in the constructor for
the class (see “Member initializers”). However, data mem-
bers declared using both the keywords const and static (see
below) are initialized as described under “Static data mem-
bers.”
class Account
{
...
private:
...
static const double minBalance;
};

You declare a constant data member as static when its value

is to be the same for all instances of the class. This saves
memory because all instances of the class share the same
member.

72 | C++ Pocket Reference

Mutable data members
Mutable data members are modifiable even when the object
that contains them has been declared using the keyword
const. You declare a mutable data member using the key-
word mutable. For example:
class Account
{
public:
Account(double b)
{
balance = b;
}

double getBalance( ) const

{
counter++;
return balance;
}
...
private:
mutable int counter;
double balance;
};

In the following example, even though a is declared using

const, its mutable data member counter can be modified (via
getBalance) to count the number of times the balance is
accessed for the object:
const Account a(100.0);

a.getBalance( ); // Increments counter.

Volatile data members

The volatile qualifier informs the compiler that the data
member may be modified unexpectedly, be it by another pro-
cess, the hardware, or something else. You declare a volatile
data member by preceding it with volatile (similar to const).

Classes, Structs, and Unions | 73

Declaring Member Functions
Member functions are functions declared as members of a
class. You declare the member functions for a class in the
class’s definition. For example:
class Account
{
public:
...
void deposit(double amt);
void withdraw(double amt);
double getBalance( ) const;
...
};

The rules for declaring and defining member functions are

similar to those for functions that are not class members (see
“Declaring Functions”), with the following exceptions:
• You can declare member functions using the keyword
virtual to support polymorphism (see “Virtual Member
Functions”).
• The only storage class allowed for member functions is
static.
When you declare a member function, you can either define
it or wait to provide a definition outside of the class. To
define a member function where it is declared, you follow the
signature of the function with its body. For example:
class Account
{
public:
...
void deposit(double amt)
{
balance += amt;
}
...
private:
double balance;
};

74 | C++ Pocket Reference

Member functions defined within a class definition are
treated as inline (see “Inline functions”) unless they are also
declared using the keyword virtual (see “Virtual Member
Functions”). In a definition outside of the class, you qualify
the member function with the name of its class using the
scope operator (::). For example:
void Account::deposit(double amt)
{
balance += amt;
}

You can also request that a member function defined out-

side of its class be treated as inline by preceding the defini-
tion with the keyword inline.

NOTE
Definitions of member functions outside of their class be-
long in the source file for the class, not in the class’s
header file (see “Header Files”). Otherwise, the member
function ends up being defined more than once when the
header file is included multiple times. The exceptions to
this are definitions of member functions preceded by the
keyword inline, and those implementing template class-
es; these go in header files.

The this pointer

All member functions not declared using the keyword static
(see “Static member functions”) have a special pointer, called
this, that points to the instantiating object. When you access
members within the class where they are defined, the this
pointer is implied, so usually you omit it. For example:
class Account
{
public:
...
void deposit(double amt)
{
balance += amt;

Classes, Structs, and Unions | 75

 }
...
private:
double balance;
};

implies:
class Account
{
public:
...
void deposit(double amt)
{
this->balance += amt;
}
...
private:
double balance;
};

Sometimes you need to specify the this pointer explicitly, as

in the following example of overloading the assignment oper-
ator for the Account class:
Account &Account::operator=(const Account
&a)
{
// Don't assign an account to itself.
if (this != &a)
{
// Copy each member, as necessary.
}

// Returning a reference to an Account

// makes chained assignments possible.
return *this;
}

Static member functions

Static member functions can be invoked with or without an
instance of the class. You declare a static member function
using the keyword static. For example:
class Account
{

76 | C++ Pocket Reference

 public:
...
static int getDeposits( )
{
return deposits;
}
...
private:
...
static int deposits;
};

To invoke a static member function from outside of its class

without using an object, you qualify it with the name of the
class using the scope operator (::). For example:
int total = Account::getDeposits( );

Because static member functions do not have to be invoked

through an instance of the class, they do not have a this
pointer. As a result, they can only directly access data mem-
bers or member functions of the class that are static them-
selves.

Constant member functions

Constant member functions are the only nonstatic member
functions that can be invoked from objects declared using
the keyword const. You declare a constant member function
by appending const to its signature. For example:
class Account
{
public:
...
double getBalance( ) const;
...
};

A const member function is not allowed to modify nonstatic

data members of the class except those declared using the
keyword mutable, or to call member functions not declared
using const themselves.

Classes, Structs, and Unions | 77

Volatile member functions
Volatile member functions are the only nonstatic member
functions that can be invoked from objects declared using
the keyword volatile. You declare a volatile member func-
tion by appending volatile to its signature (similar to const).
See “volatile” under “Qualifiers” for more information.

Access Levels for Members

The access level of a member determines from where in a
program you are allowed to use it. You assign an access level
to a member using one of three access specifiers:
public
Public members can be accessed anywhere, including
outside of the class itself.
protected
Protected members can be accessed within the class in
which they are declared and within derived classes.
private
Private members can be accessed only within the class in
which they are declared.
Access specifiers are always followed by a colon (:). For
example:
class Account
{
public:
Account(double b);

void deposit(double amt);

void withdraw(double amt);
void getBalance( ) const;

private:
double balance;
};

Access specifiers are applied according to the following rules:

78 | C++ Pocket Reference

 • If no access specifier is specified at the start of a class, the
members in that portion of the class are private by
default.
• The default access level for structs and unions is public.
• The access level of a member is defined by the most
recent access specifier appearing before it in the class def-
inition.
• An access specifier can be used any number of times
within a class definition.
• The access level of a member does not apply when defin-
ing the member outside of its class (see “Static data
members”).

Friends
Friends of classes are granted access to all members of the
class. You can declare functions (including member func-
tions of other classes) or entire classes as friends of a class.
For example:
class Account
{
friend class AccountManager;

friend void doDefaultDebit(Account

&a);

friend void ATM::deductFee(Account

&a, double fee);

public:
Account(double b);

void deposit(double amt);

void withdraw(double amt);
void getBalance( ) const;

private:
double balance;
};

Classes, Structs, and Unions | 79

The following rules apply to friends:
• Declarations of friends can appear anywhere within a
class definition, but usually appear at the beginning.
• The access specifier under which friends appear makes
no difference.
• Friendship is not inherited, nor do friends of nested
classes have any special access rights to the members of
their enclosing class.
• Functions first declared as friends have external linkage;
otherwise, they retain their linkage.

Constructors
Constructors are special member functions used to initialize
instances of a class. They give you the opportunity to per-
form initialization (e.g., allocating dynamic storage for data
members, opening files, etc.) before an object is ever used.
Constructors are called whenever storage is allocated for an
object, whether automatically by the compiler or when you
use new or new[] (see “Memory Management”). See “Con-
structors and Inheritance” for information about construc-
tors and base classes.
Constructors have the same name as the class and never
return a value. For example:
class Account
{
public:
Account(double b);
...
};

You can provide the definition for a constructor either where

you declare the constructor or outside of the class, as shown
below:
Account::Account(double b)
{
balance = b;
}

80 | C++ Pocket Reference

Often constructors are overloaded, so there is more than one
way to initialize instances of a class. When allocating an
array of objects using new[], there must be a default con-
structor available, either provided by you or synthesized by
the compiler.

Default constructors
Default constructors are constructors that take no argu-
ments or that have default values specified for all arguments.
To instantiate an object using its default constructor, use one
of the following approaches:
Account account;
Account *p = new Account;

If you do not define any constructors for a class, the com-

piler synthesizes a default constructor for you. The default
constructor supplied by the compiler calls the default con-
structor for each member of the class.

Copy constructors
Copy constructors are special constructors that accept a ref-
erence to an instance of their own class (usually a const
instance). For example:
class Account
{
public:
Account(const Account &a);
...
};

Copy constructors copy the instance of the class supplied as

an argument into the instantiating object. They can accept
additional arguments, provided default values are specified
for them (see “Default arguments”). Copy constructors are
invoked by the compiler in the following situations:
• An instance of a class is passed by value to a function.
• An instance of a class is returned from a function.

Classes, Structs, and Unions | 81

 • An instance of a class is initialized to another instance
using an initializer.
• An instance of a class is provided explicitly as a single
argument to the constructor.
If you do not provide a copy constructor for a class, the com-
piler synthesizes one that does a member-by-member copy.
When you define a copy constructor for a class, often you
also overload the assignment operator (see “Overloading
Operators”).

Explicit constructors
Constructors declared as explicit do not take part in implicit
conversions (see “Type Conversions”). You use the keyword
explicit to declare a constructor as explicit. For example:
class Account
{
public:
explicit Account(double b)
{
balance = b;
}
...
private:
double balance;
};

Because this constructor is declared as explicit, the compiler

disallows the following, which requires an implicit conver-
sion from double to Account:
Account account = 100.0;

You declare only constructors that take one argument as

explicit, since these are the only constructors used in implicit
conversions.

Member initializers
Member initializers stipulate how to initialize data members
even before the constructor is executed. Member initializers

82 | C++ Pocket Reference

are specified with a constructor’s definition. They are placed
in a comma-delimited list between the constructor’s signature
and its body. The list begins with a colon (:). For example:
class Account
{
public:
Account(double b) :
balance(b),
minBalance(25.0)
{
}
...
private:
double balance;
const double minBalance;
};

The following rules apply:

• Nonstatic data members declared as const or references
must be initialized using a member initializer because
they cannot be modified at any point in the object’s life-
time.
• You initialize data members that are objects by provid-
ing a comma-delimited list of arguments within the
parentheses of the member initializer. The arguments
must be suited to a constructor of the object.
• Initialization is performed in the order in which the data
members are declared in the class, not the order in which
members appear in the initializer list.
• Static data members cannot be initialized using member
initializers; you initialize them where they are defined
outside of the class.

Destructors
Destructors are special member functions invoked when an
instance of a class is about to be destroyed. They give you the
opportunity to clean up (e.g., reclaim dynamic storage used
by data members, close files, etc.) before an object goes

Classes, Structs, and Unions | 83

away. Destructors are called just before an object goes out of
scope and when you destroy objects explicitly via a pointer
to them using delete or delete[] (see “Memory Manage-
ment”). See “Destructors and Inheritance” for information
about destructors and base classes, including the importance
of virtual destructors.
Destructors have the same name as their class, but preceded
by a tilde (~). They take no arguments and return no value.
The following example declares a constructor named
~Account:
class Account
{
public:
~Account( );
...
};

You can provide the definition for a destructor either where

you declare the destructor or outside of the class, as shown
below:
Account::~Account( )
{
...
}

On rare occasions, you may have a reason to call a destruc-

tor explicitly. The following invokes the destructor of
Account via an Account pointer p:
p->~Account( );

So that you don’t have to be concerned about whether a

destructor exists for a class, C++ allows such a call regard-
less of whether or not a destructor has been defined. If no
destructor is defined for the class or a base class, the call has
no effect. This mechanism does not enable you to call a
destructor that is defined but that has an access level making
it unreachable (e.g., private).

84 | C++ Pocket Reference

Nested Declarations
Anything declared inside of a class has class scope (see
“Scope”). In addition to data members and member func-
tions, a class definition may contain declarations for enumer-
ations, namespaces, and even other classes, as well as type
definitions using typedef. For example:
class Account
{
public:
enum Status
{
Premier,
Valued,
Standard
};

enum
{
LargePIN = 8,
SmallPIN = 4
};
...
void setStatus(Status s);
void setMinPIN(int n);
...
};

Outside of the class, you qualify a name declared within the

class using the class name with the scope operator (::), as
follows:
Account *p = new Account;

p->setMinPIN(Account::LargePIN);
p->setStatus(Account::Status::Valued);

Whenever you use a name declared within a class, the com-

piler either grants or denies you access based on the access
level under which the name was declared (see “Access Levels
for Members”).

Classes, Structs, and Unions | 85

Forward Declarations
You can declare a class without providing a definition for it.
You do that using what is called a forward declaration, which
declares the class name without specifying any other details
about the class. For example:
class Account;

This informs the compiler that you plan to define the class
later but are going to use its name now without referring to
any of the class’s members. For example, forward declara-
tions are needed when two classes refer to each other:
class Account;

class Bank
{
...
private:
Account *accounts;
};

class Account
{
...
private:
Bank *bank;
};

Alternatively, you could define Account first and provide a

forward declaration for Bank.

Structs
Structs are functionally identical to classes except that the
default access level for their members is public, not private.
To define a struct, you use the keyword struct in place of the
keyword class.

Unions
Unions are similar to classes; however, they can hold a value
for only one data member at a time. As a result, a union

86 | C++ Pocket Reference

occupies only as much space as its largest data member
requires. Other differences between unions and classes are:
• The default access level for unions is public; the default
access level for classes is private.
• Unions cannot have member functions that are declared
using the keyword virtual.
• Unions cannot inherit from anything, nor can anything
inherit from them.
• The members of unions cannot be objects that define
constructors or destructors, or that overload the assign-
ment operator.
Unions can be anonymous (unnamed). This form is used
when nesting a union inside of a struct or class that contains
an extra data member to indicate what the union contains.
For example:
struct AccountInfo
{
enum
{
NameInfo,
BalanceInfo
};

int type;

union
{
char name[20];
double balance;
};
};

When setting a value in the union, you record how the union
is being used. For example:
AccountInfo info;

info.type = AccountInfo::BalanceInfo;
info.balance = 100.0;

Classes, Structs, and Unions | 87

Whenever you need to access the data member, you check
what the union contains. For example:
if (info.type == AccountInfo::BalanceInfo)
{
// Use the balance.
}

Inheritance
When you derive one class from another, the derived class
inherits the data members and member functions that the
other class defines (subject to access controls) while adding
its own. Aside from the benefits that inheritance offers stem-
ming from the reuse of functionality provided by the base
class, inheritance is fundamental to supporting polymor-
phism (see “Virtual Member Functions”), an essential part of
object-oriented programming. Consider the version of
Account below:
class Account
{
public:
Account(double b);

void deposit(double amt);

void withdraw(double amt);
double getBalance( ) const;

protected:
double balance;
};

To derive a new class called BankAccount from Account, you

do the following:
class BankAccount : public Account
{
public:
BankAccount(double r);

void addInterest( );
void chargeFee(double c);

88 | C++ Pocket Reference

 private:
double interestRate;
};

Account is called the base class (or superclass). BankAccount is

called the derived class (or subclass). A BankAccount object
gives you the functionality of both BankAccount and Account:
BankAccount bankAccount(2.25);

bankAccount.deposit(50.0);
bankAccount.addInterest( );

Access to members of the base class depends on two criteria:

the access level of the member in the base class (see “Access
Levels for Members”) and the access level for inheritance (see
“Access Levels for Inheritance”).

Constructors and Inheritance

Whenever you instantiate an object of a class derived from
another class, multiple constructors are called so that each
class in the derivation chain can initialize itself (see “Con-
structors”).

Order of construction
The constructor for each class in the derivation chain is
called beginning with the base class at the top of the deriva-
tion chain and ending with the most derived class.

Base class initializers

Base class initializers stipulate the data to pass to the con-
structors of base classes. They are specified with a construc-
tor’s definition. Base class initializers are placed in a comma-
delimited list between the constructor’s signature and its
body. The list begins with a colon (:). For example:
class BankAccount : public Account
{
public:
BankAccount(double r) :

Inheritance | 89
 Account(0.0),
interestRate(r)
{
}
...
private:
double interestRate;
};

This class also contains a member initializer for

interestRate. Here’s some other information you need to
know about using base class initializers:
• If a base class does not have a default constructor, the
derived class must provide a base class initializer for it.
• Base class initializers frequently appear alongside mem-
ber initializers, which use a similar syntax.
• If more than one argument is required by a base class
constructor, the arguments are separated by commas.

Destructors and Inheritance

Whenever an object of a class derived from another class is
about to be destroyed, the destructor for each class in the
derivation chain, if defined, is called (see “Destructors”).

Order of destruction
The destructor for each class in the derivation chain is called
beginning with the most derived class and ending with the
base class at the top of the derivation chain.

Virtual destructors
When deleting an object using a base class pointer or refer-
ence, it is essential that destructors for each of the classes in
the derivation chain get the chance to run:
BankAccount *bptr;
Account *aptr;

bptr = new BankAccount(2.25);

90 | C++ Pocket Reference

 aptr = bptr;
...
delete aptr;

To ensure that the destructors of all classes in the derivation

chain are called, the destructor in the base class must be
declared using the keyword virtual (see “Virtual Member
Functions”). For example:
class Account
{
public:
virtual ~Account( )
{
...
}
...
};

NOTE
Not declaring a destructor using the keyword virtual in a
class from which other classes are later derived is a com-
mon source of memory leaks and other unexpected be-
havior. This is because only the destructor of the
pointer’s class is called; no polymorphic behavior occurs.

Virtual Member Functions

To give a member function from a base class new behavior in
a derived class, you override it. This must not be confused
with the process by which you overload a member function
(see “Overloading Functions”).
To allow a member function in a base class to be overridden,
you declare the member function using the keyword virtual
in the base class. For example:
class Account
{
public:
...
virtual void withdraw(double amt)

Inheritance | 91
 {
balance -= amt;
}

protected:
double balance;
};

The member function in the derived class is declared like any

other member function, although it is common to declare the
member function using the keyword virtual in the derived
class as well for purposes of documentation. For example:
class BankAccount : public Account
{
public:
...
virtual void withdraw(double amt)
{
if (balance - amt < 0.0)
{
// Do nothing.
}
else
{
// Balance OK.
balance -= amt;
}
}
...
};

When you invoke a virtual member function via a base class

pointer or reference to an object of a derived class, the mem-
ber function of the derived class is called instead of the mem-
ber function of the base class. For example:
BankAccount bankAccount(2.25);
Account *aptr = &bankAccount;

aptr->withdraw(50.0);

The last line in this example calls withdraw of the BankAccount

class. To determine which member function to call, C++
uses polymorphism, or dynamic binding. This allows the

92 | C++ Pocket Reference

determination to be made based on the actual type of the
object addressed by the base class pointer or reference at
runtime.
The following rules apply to virtual member functions:
• Constructors cannot be virtual, but destructors can be.
In fact, there are good reasons to almost always make
destructors virtual (see “Virtual destructors”).
• Declaring a member function using the keyword virtual
in a base class does not require that you override it in the
derived class unless the member function is declared
pure virtual (see “Abstract Base Classes”) in the base
class and you plan to instantiate the derived class.
• Once a member function has been declared using the
keyword virtual in a base class, it is virtual from that
point on in the derivation chain.
• The parameter list and return type for the member func-
tion in the derived class must match those of the mem-
ber function in the base class. Otherwise, the member
function of the derived class hides the member function
in the base class, and no polymorphic behavior will
occur.
• If you don’t declare a member function using the key-
word virtual in the base class, the same member func-
tion in the derived class hides the member function in the
base class (see “Enclosing Scopes”), and no polymorphic
behavior will occur.
• A virtual member function that is private in the base class
can be overridden in a derived class even though it can-
not be accessed via the derived class. A call to the mem-
ber function from within the base class uses
polymorphism to invoke the member function of the
appropriate derived class.
The decision of whether to allow a member function to be
called based on its access level is made at compile time. Since
the decision of which member function to call using poly-

Inheritance | 93
morphism is made at runtime, it is possible that a member
function of a derived class with a less restrictive access level
in the base class could be invoked from a point that seem-
ingly defies its access level.

Abstract Base Classes

An abstract base class, not to be confused with a virtual base
class, is a class that contains one or more pure virtual mem-
ber functions. You make a member function pure virtual
using the keyword virtual and appending the pure specifier,
= 0, to the signature of the member function in the class defi-
nition. For example:
class Account
{
public:
...
virtual double estimateReturn( ) = 0;
...
};

The following rules apply:

• A class containing one or more pure virtual member
functions cannot be instantiated.
• A derived class must provide definitions for the pure vir-
tual member functions of its base class, or declare them
as pure virtual itself.
• A pure virtual member function can be defined in the
class that declares it pure virtual.
• If you declare the destructor of a class pure virtual, you
must provide a definition for it in the class in which you
declare it pure virtual.

Access Levels for Inheritance

Access levels for inheritance influence the extent to which
members inherited from base classes are accessible in derived
classes. You assign an access level for inheritance using one

94 | C++ Pocket Reference

of three access specifiers public, protected, or private before
the name of the base class. For example:
class BankAccount : public Account
{
...
};

In this context, the access specifiers have the following mean-

ings:
public
Public members in the base class remain public in the
derived class. Protected members in the base class remain
protected in the derived class. Public inheritance yields
an “is a” relationship between the derived class and the
base class.
protected
Public and protected members in the base class are pro-
tected in the derived class. Protected inheritance yields
an “implemented in terms of” relationship between the
derived class and the base class.
private
Public and protected members in the base class become
private in the derived class. Private inheritance yields a
stricter “implemented in terms of” relationship between
the derived class and the base class.

Multiple Inheritance
To derive a class from several base classes at once, you specify
the name of each class in a comma-delimited list that begins
with a colon (:) after the name of the derived class. You pre-
cede each class in the list with an access level for inheritance
(see “Access Levels for Inheritance”). For example, the
CheckingAccount class below uses multiple inheritance:
class Account
{
public:

Inheritance | 95
 ...
double getBalance( ) const
{
return balance;
}
...
protected:
double balance;
};

class BankAccount : public Account

{
...
};

class WireAccount : public Account

{
...
};

class CheckingAccount : public

BankAccount, public WireAccount
{
...
};

With multiple inheritance, when any of the base classes from

which the class is derived have a base class in common them-
selves, multiple instances of that common base class occur in
objects instantiated from the derived class. For example,
CheckingAccount objects get two balance data members and
two getBalance methods. As a result, this invocation of
getBalance is ambiguous and results in a compilation error:
CheckingAccount c(100.0);

// This results in a compilation error.

double b0 = c.getBalance( );

Instead, you must qualify the method with the specific

instance of the base class that you desire by explicitly specify-
ing a path to it in the derivation hierarchy, as follows:
double b1 = c.BankAccount::getBalance( );
double b2 = c.WireAccount::getBalance( );

96 | C++ Pocket Reference

If this duplication of base class members is not what you
desire, you can make use of a virtual base class.

Virtual Base Classes

A virtual base class, not to be confused with an abstract base
class, remedies the situation in multiple inheritance whereby
multiple instances of a base class are included in objects of a
derived class. To make a base class virtual, you include the
keyword virtual where the class is specified as a base class.
For example, the following makes Account a virtual base
class:
class Account
{
public:
...
double getBalance( ) const
{
return balance;
}
...
protected:
double balance;
};

class BankAccount : virtual public

Account
{
...
};

class WireAccount : virtual public

Account
{
...
};

class CheckingAccount : public

BankAccount, public WireAccount
{
...
};

Inheritance | 97
Instances of CheckingAccount get a single balance data mem-
ber and a single getBalance method. As a result, the invoca-
tion of getBalance presented under “Multiple Inheritance” is
no longer ambiguous.

Templates
Templates are blueprints from which versions of a class or
function are generated automatically by the compiler based
on a set of parameters. Each time you use a template with a
different set of parameters, a new version of the class or func-
tion is generated to suit how you are trying to use it. Each
new version of the class or function is called a specialization
of the template.

Template Classes
To parameterize a template class, you precede its definition
with the keyword template. Template parameters, which can
be type or nontype parameters, are placed in a comma-delim-
ited list enclosed by angle brackets. Type parameters are pre-
ceded by the keyword class or typename. You use type
parameters in place of specific types within the class defini-
tion. For example:
template <class Type>
class Array
{
public:
...
bool insert(const Type
&element, int pos);
...
private:
Type *bufferForElements;
};

This template class uses a single type parameter called Type.

The template serves as a blueprint for arrays that contain ele-
ments of any one type. See “Nontype parameters in template
classes” for information on nontype parameters.

98 | C++ Pocket Reference

 NOTE
Whenever templates are nested such that angle brackets
are next to one another (e.g., >>), you must insert a space
between the angle brackets to differentiate them from the
shift operator. For example, use > > rather than >>.

Instantiation of template classes

To instantiate a template class, you specify arguments for the
template along with the name of the class itself wherever the
class is used. You specify arguments for a template class as a
comma-delimited list enclosed by angle brackets after the
class name. For example, the following creates an array of
integers and an array of Account objects:
Array<int> counters;
Array<Account> accounts;

Types used as arguments cannot be classes with local scope.

Once a template class has been instantiated, you use it like
any other class.

Member functions in template classes

Member functions defined outside of the template class in
which they are declared must be parameterized as well. For
example:
template <class Type>
bool Array<Type>::insert(const Type
&element, int pos)
{
...
}

This defines the insert member function for the template

class Array. You can define a constructor for the class as
follows:
template <class Type>
Array<Type>::Array<Type>( )
{

Templates | 99
 ...
}

Explicit specialization of template classes

An explicit specialization of a template class defines a specific
form of the template to use for a certain set of parameters.
To define an explicit specialization, you use the keyword
template with an empty pair of angle brackets. After the class
name, you specify a comma-delimited list of parameters, in
angle brackets, to which the specialization applies. For
example:
template<>
class Array<void *>
{
public:
Array(int elmtSize);

bool insert(const void *

element, int pos);
...
private:
void *bufferForElements;
};

This specialization stipulates that an instantiation of the

Array template class with a void pointer should use this form
of the template as opposed to any other. For example:
Array<void *> counters(sizeof(int));

You can define any number of explicit specializations to

override the primary template for a class.

Nontype parameters in template classes

Nontype parameters in template classes are constant values
that you can use in the class definition. Nontype parameters
consist of a type followed by a name. For example, the fol-
lowing template class has s as a nontype parameter:
template <class Type, int s>
class Array
{

100 | C++ Pocket Reference

 public:
...
bool insert(const Type
&element, int pos);
...
private:
Type elements[s];
};

You specify a value for a nontype parameter where the tem-

plate class is instantiated, as shown below:
Array<float, 50> temperatureValues;

This instantiates an array containing 50 elements of type

float. Nontype parameters must be one of the following
types: an integral type, enumeration, pointer, reference, or
pointer to member. Arguments specified for them must be
constant. Pointers and references must address objects or
functions with external linkage (see “Storage Classes”).

Default arguments for template classes

Template classes, unlike templates for functions, can be
defined with default arguments, except with explicit special-
ization. You set a type parameter equal to its default type, or
a nontype parameter equal to its default value, where the
template class is defined. If you omit the argument when you
instantiate the template, the default is used.

Template Functions
To parameterize a template function, you precede its defini-
tion with the keyword template. Template parameters, which
can be type or nontype parameters, are placed in a comma-
delimited list enclosed by angle brackets. Type parameters
are preceded by the keyword class or typename. You use type
parameters in place of specific types within the function defi-
nition. For example:
template <class Type>
void xchg(Type &x, Type &y)
{

Templates | 101
 Type t = x;
x = y;
y = t;
};

This template function uses a single type parameter called

Type. The template serves as a blueprint for any function that
swaps two values of like type. See “Nontype parameters in
template functions” for information on nontype parameters.

Instantiation of template functions

The compiler generates an appropriate instance of a tem-
plate function based on the arguments you use when calling
it. You call a template function just as you do any other. For
example:
int i = 10, j = 20;
Account a(50.0), b(75.0);

xchg(i, j);
xchg(a, b);

Arguments to template functions

Template functions often take template classes as argu-
ments, as do member functions of template classes. For
example, assuming Array is defined as shown under “Tem-
plate Classes,” you can do the following:
template <class Type>
bool insert(Array<Type> &a, const Type
&element, int pos)
{
return a.insert(element, pos);
}

Explicit specialization of template functions

An explicit specialization of a template function defines a
specific form of the template to use for a certain set of
parameters. To define an explicit specialization, you use the
keyword template with an empty pair of angle brackets. After
the function name, you specify a comma-delimited list of

102 | C++ Pocket Reference

parameters, in angle brackets, to which the specialization
applies. For example:
template<>
void xchg<string>(string &s, string &t)
{
s.swap(t);
};

This specialization stipulates that the invocation of the xchg

template function, like that shown below, should use this
form of the template as opposed to any other:
string s1("abc"), s2("xyz");

xchg<string>(s1, s2);

You can define any number of explicit specializations to

override the primary template for a function.

Nontype parameters in template functions

Nontype parameters in template functions are constant val-
ues that you can use in the function definition. Nontype
parameters consist of a type followed by a name. For exam-
ple, the following template function has n as a nontype
parameter:
template <class Type, int n>
void sort(Array<Type> &a)
{
// Only sort up to position n.
}

To specify a nontype parameter where a template function

that has a nontype parameter is invoked, do the following:
Array<int> integerArray;
...
// Sort only the first 10 values.
sort<int, 10>(integerArray);

Arguments specified for nontype parameters must be con-

stant. See “Nontype parameters in template classes” for the
types of values that are allowed for nontype parameters.

Templates | 103
Overloading
Overloading allows you to provide more than one definition
for a function within the same scope. It also lets you define
additional behaviors for most operators.

Overloading Functions
To overload a function, you give it several definitions, each
uniquely identifiable by the arguments it accepts; return type
is not considered. The compiler decides the definition to use
based on the arguments provided when the function is
called. For example:
char *min(char *s, char *t)
{
return (strcmp(s, t) < 0) ? s : t;
}

float min(float x, float y)

{
return (x < y) ? x : y;
}

The first definition is used when min is called with two char-
acter pointers. The second is used when min is called with
two floating-point values. For example, you could call min in
either of the ways below:
char *s = min("abc", "xyz");
float f = min(4.56F, 1.23F);

To choose the definition to use, the compiler first searches

for a definition with parameters that match an invocation
exactly. If an exact match is not found, the compiler tries to
find a match by promoting integers and converting types
where possible (e.g., a double to an int, a derived class
pointer to a base class pointer, etc.). If a suitable definition
cannot be found, a compilation error occurs. The following
rules also apply:

104 | C++ Pocket Reference

 • When considering the parameters of a function, the com-
piler ignores the presence of const or volatile and
whether a type is a reference type, except when const or
volatile is buried within types. For example, const int
*T and int *T are not treated as distinct, but int *const T
and int *T are (see the discussion of const in pointer dec-
larations under “Qualifiers”).
• Member functions can be overloaded based on whether
they are constant, volatile, or both. The appropriate
member function is called based on whether the invok-
ing object is constant or volatile.
• Default arguments make it possible for a single instance
of a function to be called using multiple sets of argu-
ments. No overloaded instance of such a function can
accept one of those sets of arguments.

Overloading Operators
The operators of C++ have defined behaviors with certain
intrinsic types. These behaviors cannot be changed. How-
ever, you can define additional behaviors for types of your
own. For example, one common practice is to define the <<
operator so that it can be used with cout to display objects of
your own class types (see “I/O Streams”).
You overload an operator by defining a function called
operatorX, where X is the operator you want to overload. The
number of arguments for the function depends on the num-
ber of operands an operator requires, and whether you are
overloading the operator using a member function or a func-
tion that is not a member of a class.
When you overload an operator using a member function,
the member function must be nonstatic. The invoking object
is the first operand. If the operator requires a second oper-
and, it is passed as an argument to the member function. For
example, the following class overloads += in two ways.

Overloading | 105
 class Account
{
public:

Account(double b)
{
balance = b;
}

Account &operator+=(double b)
{
balance += b;
return *this;
}

Account &operator+=(const
Account &a)
{
balance += a.balance;
return *this;
}
...
private:
double balance;
}

Based on the way that the += operator is overloaded in this

class, you can do the following:
Account a(50.0);
Account b(75.0);

a += b; // a now contains 125.0.

a += 100.0; // a now contains 225.0.

When you overload an operator using a function that is not a

member of a class, each operand is passed as an argument.
For example:
Account &operator+=(Account &a, const
Account &b)
{
a.balance += b.balance;
return a;
}

106 | C++ Pocket Reference

When you use a nonmember function to overload an opera-
tor, often you declare it as a friend of its arguments’ class
types. This is necessary if the function needs to access non-
public members of its arguments (as shown above).
The following rules also apply:
• The following operators cannot be overloaded: ::, ., .*,
?:, sizeof, typeid, and the C++ cast operators (see
“Casts in C++”).
• The associativity and precedence of an operator cannot
be changed through overloading.
• Derived classes inherit functions that overload opera-
tors, except those functions that overload the assign-
ment operator.
• Functions that overload operators cannot have default
arguments, except for ( ). This operator can be declared to
accept default arguments and can have any number of
parameters. Objects of classes that overload the ( ) opera-
tor are called function objects.
• You can invoke operator functions explicitly (e.g., a.
operator+=(b)). If an operator function is a virtual mem-
ber function and is invoked via a base class pointer or ref-
erence, polymorphic behavior occurs.

Assignment operator
Generally, you overload the assignment operator (=) for a
class when you define a copy constructor. The assignment
operator can be overloaded using a member function only. If
you do not overload the assignment operator for a class,
member-by-member assignment is performed by default. The
function that overloads the assignment operator for a class is
not inherited by derived classes.

Overloading | 107
 NOTE
For polymorphic classes, which are those containing at
least one virtual member function, it is common to de-
clare a member function called clone rather than using
the assignment operator to copy objects. Derived classes
override clone as needed.

Memory management operators

The standard header file <new> outlines the various ways to
overload new, new[], delete, and delete[] (see “Memory
Allocation” and “Memory Reclamation”). You can also over-
load new and new[] for use with placement new. A nonmem-
ber function that you provide for overloading a memory
management operator does not get called for instances of
classes for which you overload the same operator using a
member function.

Memory Management
C++ provides intrinsic support for allocating and reclaiming
memory dynamically as a program runs. Dynamic memory is
memory that you allocate and reclaim (i.e., manage) your-
self, as opposed to storage that is managed automatically by
the compiler on the stack, for example.

Memory Allocation
The operators used to dynamically allocate memory are new
and new[].

new
To dynamically allocate storage for a single instance of a
type, you use the new operator. For example:
int *i = new int;
double *x = new double(10.0);

108 | C++ Pocket Reference

 Circle *c = new Circle;
Pt *p = new Pt(1.0, 2.0);

You provide a comma-delimited list of arguments inside

parentheses for initialization, if needed. The result of new is a
pointer of the appropriate type.
Initialization takes places after storage is allocated. The num-
ber of arguments for initialization must suit the type speci-
fied for the allocation. The intrinsic types of C++ take a
single argument. The number of arguments and their types
for classes depend on the constructors that have been defined
by the class. The appropriate constructors are called for an
object. Multiple constructors are called if the class of the
allocated object is a derived class (see “Constructors and
Inheritance”).

new[ ]
To dynamically allocate storage for an array, you use the
new[] operator. For example:
double *da = new double[5];
Circle *ca = new Circle[8];

The class for the objects being allocated must have a default
constructor. This constructor is called for each object.

Placement new
Placement new is used to pass additional arguments to a func-
tion that overloads new or new[]. You can specify any num-
ber of arguments that an operator function you have written
accepts (see “Overloading Operators”). For example:
Account *a = new(3, x) Account;

yields a call like the following:

operator new(sizeof(Account), 3, x);

The first argument passed to the operator function is the

amount of storage to be allocated (specified as a size_t); the
remaining arguments are the ones explicitly specified where

Memory Management | 109

new is used, in order. Similar rules apply for new[]. For
example:
Account *b = new(x) Account[5];

yields a call like the following (n is array overhead):

operator new[]((sizeof(Account) * 5) + n, x);

Failed allocation
If an allocation fails, new and new[] throw a bad_alloc excep-
tion. You can use the placement version of new if you prefer a
null pointer on a failure:
char *c = new(nothrow) char[10];

You can install your own handler for dealing with failed allo-
cations by calling set_new_handler (include <new>). This
function takes a pointer to a handler function with the signa-
ture below:
void new_handler_function( );

set_new_handler returns the previously installed handler. If

you return from the handler you provide, the allocation is
attempted again.

NOTE
I did not use new_handler as the name for my function to
deal with failed allocations, because that’s a type defined
by the standard for new handlers.

Memory Reclamation
The operators used to reclaim dynamically allocated mem-
ory are delete and delete[].

delete
To reclaim memory previously allocated using new, you use
the delete operator. For example:

110 | C++ Pocket Reference

 Circle *c = new Circle;
...
delete c;

The appropriate destructors are called for the object. Multi-

ple destructors may be called if the class of a deleted object is
a derived class (see “Destructors and Inheritance”).
Once memory has been reclaimed, you must not access it
again, although the pointer to it can be pointed to some-
thing else. No assumptions can be made about the pointer’s
value. Using delete with a null pointer is guaranteed to be
safe.

WARNING
To ensure that the proper destructors are called, the
pointer used with delete must have the same type as the
pointer used with new when the memory was allocated, or
be a base class pointer.

delete[ ]
To reclaim the memory for an array previously allocated
using new[], you use the delete[] operator. For example:
Circle *ca = new Circle[8];
...
delete[] ca;

Using delete[], as opposed to delete, causes destructors to

be called for each object in an array of objects as memory is
reclaimed.

WARNING
As with delete, the pointer used with delete[ ] must have
the same type as the pointer used with new[ ] when the
memory was allocated, or be a base class pointer.

Memory Management | 111

Casts and Runtime Type
Information
Casts are used to explicitly convert an expression’s value to a
different type. Runtime type information is type data embed-
ded by the compiler for use at runtime.

C-Style Casts
C-style casting is the form of casting inherited from C. The
target type is placed in parentheses immediately preceding
the expression to be converted, as shown in the assignment
to c below:
void *v = new Circle(5.0);
Circle *c = (Circle *)v;

With a C-style cast, no check is performed at runtime to

ensure that the cast is reasonable. C-style casting can be used
for the following:
• Casting a pointer or arithmetic type to an integer type.
• Casting any arithmetic type to a floating-point type.
• Casting a pointer or arithmetic type to another pointer
type.

Casts in C++
C++ provides additional forms of casting, which are gener-
ally safer than C-style casts.

dynamic_cast
The dynamic_cast operator casts a pointer of one class type to
a pointer of another within a derivation chain. It is allowed
only with pointers and references to polymorphic types,
which are types that have at least one virtual member func-
tion (see “Virtual Member Functions”). For the examples
that follow, consider the following derivation chain:

112 | C++ Pocket Reference

 class Account
{
// At least one virtual member
// function.
};

class BankAccount : virtual public

Account
{
...
};

class WireAccount : virtual public

Account
{
...
};

class CheckingAccount : public

BankAccount, public WireAccount
{
...
};

class SavingsAccount : public

BankAccount, public WireAccount
{
...
};

To use the dynamic_cast operator, you place the target type

inside angle brackets preceding the expression to be
converted; the expression is enclosed in parentheses.
Account *a;
BankAccount *b;
WireAccount *w;

CheckingAccount c;

// Perform an upcast.
a = dynamic_cast<Account *>(&c);

// Do the same thing again to show that

// no cast is required to do an upcast.
a = &c;

Casts and Runtime Type Information | 113

 // Perform a downcast.
b = dynamic_cast<BankAccount *>(a);

// Perform a cross cast.

w = dynamic_cast<WireAccount *>(b);

This example shows the three operations you are able to per-
form with the dynamic_cast operator:
upcasting
The pointer is moved up the derivation chain to a base
class.
downcasting
The pointer is moved down the derivation chain to a
derived class.
cross casting
The pointer is moved to a sibling class within the deriva-
tion chain.

NOTE
Since some compilers do not enable runtime type infor-
mation by default, ensure that runtime type information
is turned on for dynamic_cast to work properly.

The following rules also apply to dynamic casts:

• The compiler detects whatever errors it can with
dynamic_cast (e.g., a target class type that is not in the
derivation chain at all).
• Runtime type information is used to determine whether a
cast is legal at runtime. If the cast is determined not to be
legal, the result of dynamic_cast is a null pointer. This
would be the case in the previous example, for instance,
if you were to attempt to downcast an Account pointer to
a CheckingAccount pointer when the Account pointer was
pointing to a SavingsAccount object.
• When you perform an illegal cast with a reference type, a
bad_cast exception is thrown.

114 | C++ Pocket Reference

static_cast
The static_cast operator is used to cast a pointer of one
class type to a pointer of another within a derivation chain
while avoiding the runtime checks done with dynamic_cast.
As a result, you can use the static_cast operator with point-
ers to nonpolymorphic types, which are types that have no
virtual member functions. You also can use it to carry out
some of the conversions performed using C-style casts, gen-
erally conversions between related types. The static_cast
operator has the same syntax as dynamic_cast.

const_cast
You use the const_cast operator to cast away the const and
volatile qualifiers. It has the same syntax as dynamic_cast.
Between the angle brackets, you specify the same type as the
original, without the const or volatile qualifier. Using the
result is assured to be safe only if the data to which the
pointer points was not declared as const or volatile when it
was first declared in the program.

reinterpret_cast
The reinterpret_cast allows you to convert a pointer to any
other pointer type. It also allows you to convert any integral
type to a pointer and back. It uses a syntax like the other
forms of casting specific to C++. It is typically used
sparingly.

Runtime Type Information

Runtime type information (RTTI) is type data embedded in a
program by the compiler so that you can use it as the pro-
gram runs. In languages that are statically typed (like C++),
type information is normally not available after compilation
unless explicitly preserved.

Casts and Runtime Type Information | 115

 NOTE
Runtime type information often is not embedded by de-
fault; you use a compiler option to enable it.

typeid
To get type information about a variable or a type itself, you
use the typeid operator. For example:
Circle c(5.0);
const type_info &t = typeid(c);

This assigns information about the type of c to t. The oper-

and for typeid may be an expression or type. The result of
the typeid operator is a constant reference to an object of
type type_info (see “type_info”).
To use the type_info object, you include the standard header
file <typeinfo>. The following rules apply:
• When the operand for typeid is a reference or a derefer-
enced pointer to a polymorphic class (a class with at least
one virtual member function), the result is type informa-
tion for the dynamic type of the operand.
• When typeid is applied to a dereferenced null pointer, a
bad_typeid exception is thrown.

type_info
The standard type_info class encapsulates support for work-
ing with type information. It overloads the == and != opera-
tors so that you can easily compare type_info objects. For
example:
if (typeid(a) == typeid(b))
{
// a and b have the same type.
}

116 | C++ Pocket Reference

The type_info class also contains a name member function for
getting a type’s name as a string. The name that is returned is
implementation-defined. For example:
cout << typeid(c).name( ) << endl;

Exception Handling
Exception handling is performed using try and catch blocks.
For example:
try
{
// Watch out for a bad file name
// or no file handles available.
}
catch (BadFileName &e)
{
// Handle BadFileName exceptions.
}
catch (HandlesGone &e)
{
// Handle HandlesGone exceptions.
}

try
A try block delineates a context in which exceptions may be
raised, or thrown. When an exception is thrown within a try
block, execution immediately jumps to the start of a catch
block responsible for dealing with the exception, if such a
block exists.

throw
You throw an exception inside a try block using the throw
operator. For example:
throw e;

The type of the exception is used to determine which catch

block to execute. The exception itself is passed as an

Exception Handling | 117

argument to the catch block so that it can be used in han-
dling the exception. Within a catch block, you can re-throw
an exception using throw with no operand. The following
rules also apply:
• Exceptions can be of intrinsic types or classes that you
define yourself. Exception classes do not have to be
derived from any particular class.
• Some standard exceptions are defined for use by the lan-
guage (e.g., bad_cast thrown by the dynamic_cast opera-
tor) and facilities in the C++ Standard Library.
• Standard exceptions are all derived from the exception
class, which is defined in the standard header file
<exception>. The what member function of exception
gets a standard exception’s name.

catch
One or more catch blocks follow a try block to define how
specific types of exceptions should be handled. catch blocks
are tried in the order they appear. The first catch block found
to match the exception’s type or its base class is passed the
exception to handle. Therefore, if you catch exceptions of
both a derived class and its base class, the catch block for the
derived class needs to appear first.

NOTE
Exceptions are often declared as references in catch
blocks so that polymorphic behavior is possible when ac-
cessing the objects to handle the exceptions.

An ellipsis (...) can be used to indicate that any exception

type should be caught. For example:
try
{
// Watch out for a bad file name.
}
catch (BadFileName &e)

118 | C++ Pocket Reference

 {
// Handle BadFileName exceptions.
}
catch (...)
{
// Handle exceptions not covered.
}

If no catch block is suitable, the stack is unwound to deter-

mine whether a suitable catch block appears earlier in the
calling chain. If the stack unwinds completely without find-
ing a suitable catch block, the standard function terminate is
called. The default behavior for this function is to terminate
the program. You can install your own handler that termi-
nates exception processing by calling set_terminate (include
<exception>). This function takes a pointer to a handler
function with the signature below:
void terminate_handler_function( );

set_terminate returns the previously installed handler. The

terminate handler you provide should take any actions neces-
sary for unhandled exceptions, then terminate execution of
the program.

Exception Specifications
An exception specification is a guarantee to the caller of a
function that only certain exceptions can be thrown within
it. For example:
void fetch(char *name, char *&data)
throw (BadFileName, HandlesGone);

If the function throws an exception of a type not listed in the

exception specification or of a type not derived from one of
the listed types, the standard function unexpected is called.
The default behavior for this function is to terminate the pro-
gram. You can install your own handler for this by calling
set_unexpected (include <exception>). This function takes a
pointer to a handler function with the signature below:
void unexpected_handler_function( );

Exception Handling | 119

set_unexpected returns the previously installed handler. The
handler you provide should take any actions necessary for
unexpected exceptions, then terminate execution of the pro-
gram. It can also throw exceptions of its own.
The following rules also apply to exception specifications:
• An empty set of parentheses specifies that a function
throws no exceptions.
• If you omit the exception specification altogether, there
is no limit on the types or number of exceptions that can
be thrown.
• A definition for a function must list the same exceptions
as in its declaration.
• If a virtual member function has an exception specifica-
tion (see “Virtual Member Functions”), the member
function of a derived class that overrides the function
must have an exception specification that is either the
same or more limited than the one in the base class.

The C++ Standard Library

The C++ Standard Library consists of facilities used to per-
form tasks common to many programs. In addition to facili-
ties specific to C++, such as stream I/O and facilities in the
Standard Template Library (STL), the C++ Standard Library
provides support for the C Standard Library. To use a fea-
ture of the C++ Standard Library, you include the appropri-
ate header file, which contains prototypes, type definitions,
and other code required to use certain features.
Header files for the C++ Standard Library do not have a .h
extension. This enigmatic naming convention was estab-
lished to avoid compatibility issues with .h versions of the
header files already in use prior to standardization. Most
implementations still provide .h (pre-standard) versions of
header files for backward compatibility.

120 | C++ Pocket Reference

The std Namespace
To keep standardized versions of facilities in the C++ Stan-
dard Library from conflicting with pre-standard ones, header
files without .h extensions place their contents within the
namespace std. Therefore, to use the standardized facilities,
you must qualify them with the std namespace. For example:
#include <iostream>
...
std::cout << "Hello" << std::endl;

Alternatively, you can use a using directive to alleviate hav-

ing to specify the namespace. For example:
using namespace std;
...
cout << "Hello" << endl;

Throughout this reference, a using directive like the one

shown here is assumed for examples in which one or more
header files from the C++ Standard Library are presumed to
have been included.

C Standard Library Support

The C++ Standard Library contains its own versions of stan-
dard header files from the C programming language. The
names of these header files are similar to their analogs in the
C Standard Library; however, they have no .h extension and
are given the prefix “c” (e.g., <cstdlib>). To differentiate the
facilities in each header file from their counterparts in the
header files of the C Standard Library, each facility is declared
in the std namespace. The list below presents the standard
header files of C as named in the C++ Standard Library:
<cassert> <cctype> <cerrno>
<cfloat> <ciso646> <climits>
<clocale> <cmath> <csetjmp>
<csignal> <cstdarg> <cstddef>
<cstdio> <cstdlib> <cstring>
<ctime> <cwchar> <cwctype>

The C++ Standard Library | 121

C++ Standard Header Files
The C++ Standard Library has facilities for language sup-
port, diagnostics, general utilities, strings, locales, contain-
ers, iterators, algorithms, numerics, and I/O. The header files
for these facilities are listed below. Header files that are part
of the Standard Template Library (STL) are marked with an
asterisk.
<algorithm>* <bitset> <complex>
<deque>* <exception> <fstream>
<functional>* <iomanip> <ios>
<iosfwd> <iostream> <istream>
<iterator>* <limits> <list>*
<locale> <map>* <memory>*
<new> <numeric>* <ostream>
<queue>* <set>* <sstream>
<stack>* <stdexcept> <streambuf>
<string> <typeinfo> <utility>*
<valarray> <vector>*

I/O Streams
I/O streams are the preferred means of performing input and
output via standard I/O and files in C++. Four streams are
predefined for standard I/O: cin, cout, cerr, and clog. When
writing output, it is common to use the endl manipulator for
newlines.

cin
The cin object controls input from a stream buffer associ-
ated with the C stream stdin. To use cin, you include the
header file <iostream>. The type of cin is istream. The
istream class overloads the >> operator so that you can use it
to read values into variables of intrinsic types. For example:
double value;

cin >> value;

122 | C++ Pocket Reference

You overload the >> operator yourself to read values into
objects of types that you define. For example:
istream &operator>>(ostream &is,
Account &a)
{
return is >> a.balance;
}

Assuming this operator function has access to the balance

member of Account (usually because you declare such I/O
functions as friends of the class), this definition lets you read
input directly into Account objects. For example:
Account account;

cin >> account;

Because the operator function returns a reference to the

istream object passed to the function, you can chain input
operations together. For example:
Account a1, a2;

cin >> a1 >> a2;

The wcin object is the analog to cin for working with wide-
character streams (see “char and wchar_t”).

cout
The cout object controls output to a stream buffer associ-
ated with the C stream stdout. To use cout, you include the
header file <iostream>. The type of cout is ostream. The
ostream class overloads the << operator so that you can use it
to write variables of intrinsic types as output. For example:
char s[] = "Hello";

cout << s << endl;

You overload the << operator yourself to write out objects of

types that you define. For example:
ostream &operator<<(ostream &os, const
Account &a)

The C++ Standard Library | 123

 {
return os << a.balance;
}

Assuming this operator function has access to the balance

member of Account (usually because you declare such I/O
functions as friends of the class), this definition lets you write
Account objects directly as output. For example:
Account account;
...
cout << account << endl;

Because the operator function returns a reference to the

ostream object passed to the function, you can chain output
operations together. For example:
Account a1, a2;
...
cout << a1 << ", " << a2 << endl;

The wcout object is the analog to cout for working with wide-
character streams (see “char and wchar_t”).

cerr
The cerr object controls output to a stream buffer associ-
ated with the C stream stderr. To use cerr, you include the
header file <iostream>. Aside from working with stderr,
which is unbuffered, the mechanics of cerr are similar to
those described previously for cout. The wcerr object is the
analog to cerr for working with wide-character streams (see
“char and wchar_t”).

clog
Like cerr, the clog object controls output to a stream buffer
associated with the C stream stderr; however, output is buff-
ered. To use clog, you include the header file <iostream>.
The wclog object is the analog to clog for working with wide-
character streams (see “char and wchar_t”).

124 | C++ Pocket Reference

 Index

Symbols << (left shift operator), 43

-- (postfix decrement <= operator, 43
operator), 39 <climits> header file, 13
-- (prefix decrement == operator, 43
operator), 40 -> (member selection
- (subtraction operator), 41 operator), 38
! (logical NOT operator), 40 > operator, 43
!= operator, 43 –>* operator, 42
& (address-of operator), 41 >= operator, 43
& (bitwise AND operator), 43 >> (right shift operator), 43
&& (logical AND operator), 44 [ ] (array subscript operator), 38
( ) (value construction \ (backslash), 8
operator), 39 ^ (bitwise XOR operator), 43
* (indirection operator), 41 | (bitwise OR operator), 43
+ (plus operator), 41 || (logical OR operator), 44
++ (postfix increment ˜ (bitwise NOT operator), 40
operator), 39
++ (prefix increment A
operator), 40 abstract base classes, 94
. (member selection access levels for members, 78
operator), 38 access specifiers, 78
.* operator, 42 addition operator (+), 42
:: (scope resolution operator), 37 address-of operator (&), 41
:? (conditional expression arithmetic operators, 42
operator), 45 array subscript operator ([ ]), 38
< operator, 43

We’d like to hear your suggestions for improving our indexes. Send email to
[email protected].

125
arrays, 19–22 conditional expression operator
initializer list for, 21 (:?), 45
initializing with an array, 22 const qualifier, 57
multidimensional, 20 const_cast operator, 39, 115
passing to functions, 21 constant data members, 72
assignment operators, 34, 45 constant member functions, 77
overloading, 107 constructors, 80–83
atof( ) function, 5 converting, 30
auto storage class, 56 copy, 81
default, 81
B explicit, 82
inheritance and, 89
backslash (\), 8 member initializers, 83
base classes, 89 continue statement, 64
virtual, 97 copy constructors, 81
bitwise AND operator (&), 43 cout object, 123
bitwise NOT operator (˜), 40 _ _cplusplus macro, 12
bitwise OR operator (|), 43 .cpp files, 7
bitwise XOR operator (^), 43 cross casting, 114
break statement, 64 C-style cast operator, 42
C-style casting, 112
C
C Standard Library, 121 D
C++ programs data members
startup, 3–5 constant, 72
structure, 3–12 declaring, 70–73
termination, 5 mutable, 73
C++ Standard Library, 120–124 static, 71
C, compatibility with, 2 volatile, 73
catch blocks, 118 _ _DATE_ _ macro, 11
cerr object, 124 declarations, 50–58
character escape sequences, 14 forward, 86
character literals, 14 nested, 85
cin object, 122 default arguments, 53
class scope, 47 #define directive, 6, 8
class types, 28 delete operator, 42, 110
classes, 69–86 delete[ ] operator, 42, 111
storage, 55–56 dereferencing pointers, 24, 41
clog object, 124 derived classes, 89
clone member function, 108 destructors, 83
comments, 32 inheritance and, 90
compound statements, 59 virtual, 90
compound types, 17–28 directives, preprocessor, 8–11

126 | Index
division operator (/), 42 functions
do loop, 60 declaring, 52
double type, 16 definitions, 53
downcasting, 114 inline, 54
dynamic_cast operator, 39, overloading, 104
112–114 parameters, 53
passing arrays to, 21
E fundamental types, 12–17
#elif directive, 9
ellipsis (...) and exception G
handling, 118 global namespaces, 66
#else directive, 9 goto statement, 65
enclosing scopes, 49
#endif directive, 6, 9 H
enum keyword, 18
enumerations, 18 header files, 5–7
#error directive, 10 C++ Standard Library, 120
escape sequences, 14 wrapping, 6
exception handling, 117–120
ellipsis (...) and, 118 I
exception specifications, 119 I/O streams, 122
exit function, 5 identifiers, 32
explicit specialization rules, 33
of template classes, 100 #if directive, 9
of template functions, 102 if statement, 62
expression statements, 59 #ifdef directive, 9
expressions, 46 #ifndef directive, 6, 9
extern storage class, 56 implicit conversions, 29
#include directive, 6, 10
F indirection operator (*), 41
file scope, 48 inheritance, 88–98
_ _FILE_ _ macro, 11 access levels for, 94
float type, 16 constructors and, 89
floating points, 16 destructors and, 90
for loops, 61 multiple, 95
break statements and, 64 initializer list for arrays, 21
forward declarations, 86 inline functions, 54
friends, 79 inline keyword, 54
function call operator, 38 int type, 15
function pointers, 25 integers, 15
iteration statements, 60–62

Index | 127
J mutable data members, 73
jump statements, 64 mutable storage class, 56

L N
left shift operator (<<), 43 namespace scope, 48
#line directive, 10 namespaces, 66–68
_ _LINE_ _ macro, 11 global, 66
literals, 34 unnamed, 68
local scope, 47 nested declarations, 85
logical AND operator (&&), 44 new operator, 42, 108
logical NOT operator (!), 40 new[ ] operator, 42, 109
logical OR operator (||), 44 null pointers, 25
long double type, 16 null statements, 59
long type, 15
loops, 60–62 O
l-values, 28 .o files, 7
references as, 28 .obj files, 7
objects
M accessing members, 69
main( ) function, 3 declaring, 69
member access levels, 78 operators, 34–46
member functions, 74–78 list of, 35–37
constant, 77 overloading, 105–108
static, 76 precedence, 35
this pointer and, 75 overloading
virtual, 91–94 defined, 104
volatile, 78 functions, 104
member functions and volatile operators, 105–108
qualifiers, 58
member initializers, 82 P
member selection operator parameters, function, 53
(. and ->), 38 plus operator (+), 41
memory allocation failure, 110 pointer arithmetic, 24
memory management, 108–111 pointer variables, 51
operators, 108 pointers, 24–27
memory reclamation, 110 const declaration, 57
message directive, 11 dereferencing, 24, 41
minus operator (-), 41 function, 25
modulus operator (%), 42 null, 25
multidimensional arrays, 20 of type void, 25
multiple inheritance, 95 this, 75
multiplication operator (*), 42

128 | Index
pointers to members, 26 file, 48
pointer-to-member selection local, 47
operators (.* and –>*), 42 namespace, 48
postfix increment and decrement selection statements, 62
operators (++, --), 39 sequence operator (,), 46
#pragma directive, 11 set_new_handler function, 110
precedence, operator, 35 set_terminate function, 119
prefix increment and decrement shift operators, 43
operators, 40 short type, 15
preprocessor directives, 8–11 signed integers, 15
preprocessor macros, 11 sizeof operator, 40
private members, 78 source files, 7
inheritance and, 95 Standard Template Library
protected members, 78 (STL), 122
inheritance and, 95 statements, 59–65
prototypes, 53 static data members, 71
public members, 78 static member functions, 76
inheritance and, 95 static storage class, 55
static_cast operator, 39, 115
Q std namespace, 121
_ _STDC_ _ macro, 12
qualifiers, 57 STL (Standard Template
Library), 122
R storage classes, 55–56
reference parameters, 27 string literals, 23
references, 27 strings, 22
as l-values, 28 strlen function, 23
register storage class, 56 structs, 86
reinterpret_cast operator, 39, subtraction operator (-), 42
115 switch statement, 63
relational operators, 43
reserved words, 33 T
return statement, 65 template classes, 98–101
right shift operator (>>), 43 default arguments for, 101
RTTI (runtime type explicit specialization of, 100
information), 115 member functions in, 99
nontype parameters in, 100
S template functions, 101–103
scope resolution operator (::), 37 arguments to, 102
scopes, 47–50 explicit specialization of, 102
class, 47 instantiation of, 102
enclosing, 49 nontype parameters in, 103
templates, 98–103

Index | 129
this pointer, 75 V
throw operator, 46, 117 value construction operator, 39
_ _TIME_ _ macro, 11 variables
_ _TIMESTAMP_ _ macro, 11 declarations, 51
tokens, 31 initializing, 52
try block, 117 virtual base classes, 97
type cast operators, 39 virtual destructors, 90
type conversions, 28–30 virtual member functions, 91–94
type definitions, 31 void pointers, 25
type_info class, 116 void type, 17
typedef keyword, 28, 31 volatile data members, 73
typeid operator, 39, 116 volatile member functions, 78
types volatile qualifier, 58
compound, 17–28 member functions and, 58
fundamental, 12–17

W
U
while loop, 60
unary minus and plus operators wide characters, 23
(-, +), 41 wrapping header files, 6
#undef directive, 9
unexpected function, 119
unions, 86
unnamed namespaces, 68
unsigned integers, 15
upcasting, 114
user-defined conversions, 29
using declaration, 67
using directives, 67

130 | Index

Download more eBooks here: http://avaxhm.com/blogs/ChrisRedfield